alternate test procedure (atp) validation study report for
TRANSCRIPT
Alternate Test Procedure (ATP) Validation Study Report
For
The Measurement of Drinking Water Turbidity up to 10
NTU Using the Candidate Lovibond Turbidity Methods,
Represented by the PTV 1000, PTV 2000, and PTV 6000
Turbidimeters
Candidate ATP Turbidity Methods:
1. The Continuous Measurement of Drinking Water Turbidity using the
Lovibond White Light LED Method
2. The Continuous Measurement of Drinking Water Turbidity using the
Lovibond 660-nm LED Method
3. The Continuous Measurement of Drinking Water Turbidity using the
Lovibond 6000 Laser Method
December 20, 2016
Tintometer Incorporated
6456 Parkland Drive
Sarasota, FL 34243
All Correspondence Should be Directed to:
Michael Sadar
Tintometer Incorporated
2108 Midpoint Drive, STE 1
Fort Collins, CO 80525
970-682-8148
1
Table of Contents
1.0 Background ................................................................................................................................... 2
1.1 Method Justification .................................................................................................................. 3
1.2 Method Equivalency ................................................................................................................. 5
1.3 Analytes .................................................................................................................................... 7
2.0 Study Implementation .................................................................................................................. 7
2.1 Study Schedule .......................................................................................................................... 9
2.2 Sample Collection ..................................................................................................................... 9
2.3 Types of Analysis Performed .................................................................................................. 10
2.4 Study Plan Deviations ............................................................................................................. 10
3.0 Method Procedure and Data ........................................................................................................ 11
3.1 Validation Study Demonstration Data .................................................................................... 14
3.1.1 Calibration ........................................................................................................................ 15
3.1.2 Initial Demonstration of Capability ................................................................................. 16
The Fort Collins Test Site ..................................................................................................... 16
Precision and Accuracy Data for the Three Candidates and Reference Methods ................. 19
Accuracy (Bias) of the Lovibond WL Led Method. ............................................................. 22
Accuracy (Bias) of the Lovibond 660-nm LED Method ...................................................... 24
Accuracy (Bias) of the Lovibond 6000 Laser Method .......................................................... 25
The Binney South Platte Comparability Test Site ................................................................ 26
The San Patricio MWD Comparability Test Site .................................................................. 31
Limit of Detection (LOD) ..................................................................................................... 36
3.1.3 Quality Controls ............................................................................................................... 37
3.1.4 Precision and Accuracy .................................................................................................... 40
3.2 Holding Time / Storage Stability ............................................................................................ 40
4.0 Data Analysis and Discussion ..................................................................................................... 40
Data Analysis ........................................................................................................................ 41
Data comparison between the Lovibond 660-nm LED method and the EPA approved
Method 10133 ....................................................................................................................... 44
Data comparison between the Lovibond 6000 Laser Method and the EPA approved
Method 10133 ....................................................................................................................... 45
5.0 Conclusions ................................................................................................................................. 48
Appendix A Validation Study Plan……………………………………………………..……...49-69
2
Appendix B.1 Lovibond White Light LED Turbidimeter Method………………………..…70-82
Appendix B.2 Lovibond 660-nm LED Method………………………………………………83-94
Appendix B.3 Lovibond 6000 Laser Method……………………………………..….……..95-107
1.0 Background
These three candidate methods are for the on-line (continuous) measurement of turbidity. Specifically, these
methods have been designed to measure low-level turbidity in water, between 0.01 and 10 NTU. The methods are
designed to measure the turbidity of the sample either in a static state or within a defined sample flow rate between
30 and 150 ml/minute. These candidate methods address the common interferences in turbidity, which included
stray light, bubbles, optical surface condensation, and particle settling; accomplished through the designed
instrumentation described in the candidate methods.
Tintometer Incorporated is a water analytics company that operates under the brand Lovibond. Tintometer
developed new on-line turbidimeters with the specific application to deliver accurate turbidity measurements in the
area of drinking water production, and more specifically filtration performance within a drinking water plant.
Specifically, the turbidimeters were designed for the monitoring of waters with turbidity less than 10 Nephelometric
Turbidity Units (NTU’s). Tintometer has worked diligently to ensure that these turbidimeters will reliably deliver
accurate turbidity assays over time.
This ATP validation study covered three new candidate methods for low level turbidity measurement.
1. The Continuous Measurement of Drinking Water Turbidity using the Lovibond White Light LED
Method. The representative instrument discussed herein is the PTV 1000 Turbidimeter.
2. The Continuous Measurement of Drinking Water Turbidity using the Lovibond 660-nm LED Method.
The representative instrument discussed herein is the PTV 2000 Turbidimeter.
3. The Continuous Measurement of Drinking Water Turbidity using the Lovibond 6000 Laser Method.
The representative instrument discussed herein is the PTV 6000 Turbidimeter.
The candidate methods are listed separately, and differ only with respect to the incident light source. This is noted
in the title of each candidate method. These three methods use solid state light sources, two of which are light
emitting diode (LED) sources and the third which is a laser diode light source. These light sources have extended
lifetimes and provide optical stability when compared to historical methods (e.g. USEPA Method 180.1) that use
incandescent light sources. This translates into improved measurement reliability and reduced instrument
maintenance over the lifetime of the instrument.
All other aspects of these three candidate methods and their representative instrumentation are identical. This
includes measurement geometry (detector angle, detector view angle), sample volume, sample flow rates, sample
transport through the instrument, measurement view volume, calibration, quality control verification, turbidity
measurement determination (i.e. measurement algorithms), and data handling. Denoted collectively (where
appropriate), they are referred to as the PTV 1000/2000/6000. This helps to simplify the understanding of the
differences between these technologies and other approved historical turbidity methods.
The ATP study involved the comparison of measurements between the three candidate PTV 1000/2000/6000
instruments and the EPA approved Method 10133. Method 10133 is a laser-based turbidity method, designed for
measurements between 0.010 and 5.00 NTU. The fully compliant instrument to Method 10133 is the Hach
FilterTrak 660 Laser Nephelometer. This is referred to herein as the reference instrument and the Method 10133 is
referred to as the reference method.
3
The study involved the measurement of three different waters with the candidates and reference instruments. Two
of the waters were filter effluent waters. A third water was tap water that was filtered to remove any residual
particles above 0.05 micron. The two effluent waters were monitored by the candidate methods and test instruments
to collect data to for method comparability. The filtered tap water was spiked with formazin to generate several
defined turbidity levels that covered the range of turbidity in this ATP. This spike data was used to determine the
precision, bias and linearity for the candidate and reference methods over the denoted range from about 0.01 up to
10 NTU.
1.1 Method Justification
The candidate ATP turbidity methods were compared to the approved Hach Method 10133, which is also known as
the FilterTrak 660 method (denoted herein as Method 10133). The Hach Filtertrak 660 laser nephelometer is cited
as the instrument in full compliance with Method 10133 and is the reference method throughout this study.
These candidate turbidimeters utilize numerous state-of-the art technological advancements coupled with proven
design criteria that have been utilized by the Drinking Water Plant (DWP) community and its regulatory partners
over the past several decades. Ultimately, these technologies provide the ability to rapidly interpret turbidity
measurement data which is critical to the mitigation of any risks associated with the breakdown in the water
treatment process.
As noted above, the turbidimeter versions are identical in their design with the exception of the light source. One
version, the PTV 1000 utilizes a white light (WL) LED, a source that is essentially the same as the incident light
source used in the EPA approved Swan Turbidiwell Method and is designed to mimic the light source used in
Method 180.1 with respect to spectral output. The second version, the PTV 2000 contains a red light emitting LED
that emits light at peak intensity within the visible spectrum at a wavelength between 650 and 670nm, (typically at
660 nm). The spectral output from this source is comparable to the Mitchell Method M5271 and Hach Method
10133. The third light source is a 685-nm laser diode source with a spectral output that is comparable to Hach
method 10133. These three light sources provide several advantages over the historical tungsten filament light
sources used in Method 180.1, which are discussed in detail later in this document.
Table 1 provides a summary of the advantages these candidate methods have with respect to the reference method,
and also EPA Method 180.1. When comparing to Method 180.1, this is with respect to the light source only and was
included in this table because is the most common monitoring method in drinking water. Additional candidate
methods’ (and instruments) advantages separate from the incident light sources are expressed at the bottom part of
the table and are compared to the reference method to support their respective justification.
Table 1 – Summary of the design features for the proposed EPA methods on turbidity and the advantages over the
reference Method 10133 and EPA Method 180.1
Feature Advantage over Method 10133 Advantage over Method 180.1
White Light (WL)
LED (Incident Light
Optics) and the
660-nm Red LED
WL LED is more sensitive to small particles than
the reference - Peak response between 400-600
nm; Source was sensitive to a broad range of
particle sizes.
Solid State – Low drift; low output temperature
dependence of LED light source. Source will last
life of the instrument
Long Life (10 years life expectancy). 180.1
sources last 1-2 years and will have spectral shifts
as they age.
Heated optics for all light sources in the
candidate methods. Eliminates instability and
erroneous measurements due to condensation.
Solid state light sources such as LED’s and laser
diodes do not generate heat and are more
susceptible to condensation effects. Reference
was susceptible to condensation effects on the
180.1 beam was divergent and results in higher
stray light.
4
incident light optics, thereby increasing chance
of measurement error.
Sensitivity was near equivalence to 180.1. Peak
response between 400-600 nm; sensitive to a broad
range of particle sizes.
Heated optics; Eliminates instability and erroneous
measurements due to condensation. Some 180.1
designs generate heat that was sufficient to prevent
condensation effects on the incident light optics.
Others designs can experience condensation
problems.
Incident light source monitor detector;
compensates for LED drift over time and
temperature. Reference instrument has no known
drift compensation for its light source.
Incident light source monitor detector;
Compensates for LED drift over time and
temperature. Existing Method 180.1 instruments
have no known drift compensation.
The peak response of the 660-nm reduces
interference due to dissolved organics and sample
color.
The 660-nm spectral bandwidth was narrowed,
thereby reducing stray light and improving the
limit of detection.
The 660-nm LED spectral bandwidth was narrowed
when compared to Method 180.1, thereby reducing
stray light and improving the limit of detection of
the candidate method.
685-nm Laser
The incident light is a highly collimated beam of
high energy light and its small diameter reduces
stray light and improves the limit of detection when
compared to Method 180.1.
The small diameter incident light beam in the
candidate method was sensitive to individual
particles that may penetrate a filter as its run time
progresses toward backwash.
Advantages of the PTV 1000/2000/6000 over the Reference Turbidimeter (Method 10133)
Beam Dump
A stray light trap that absorbs incident beam energy after passing through measurement
chamber and reduces stray light. The beam dump of this design is not a feature in the
reference method.
Scattered Light Detector
The candidate methods detector incorporates a controlled aperture angle. This improves
intra-instrument consistency in detection. The reference method does not have this feature.
The candidate detector was heated and does not have air spaces between the sample and
detector surface. This eliminated instability and error due to condensation. The reference
instrument does not have this feature.
Short total path length of 5.5 cm yields a highly linear response over the regulatory range of
interest (0-10 NTU) and simplifies the calibration protocols. This allows the calibration to
be performed using a turbidity standard that is higher in value (e.g. 5.0 NTU) and thus,
easier to prepare and administer. The reference method requires the preparation of a lower
turbidity standard (0.80 NTU) which is more difficult to prepare and administer. The
published range of the candidate’s methods was up to 10 NTU and up to 5 NTU for the
reference method.
Large collection angle; improves the limit of detection. The candidate’s methods have a
large defined collection angle. The reference instrument does not have this feature.
The Turbidimeter body
Reduced Volume – The candidate instruments have a volume of 275 ml versus the
reference instrument’s volume of 1100 ml. This improves response time to a turbidity
event, reduces the possibility of sample settling within the instrument, minimizes sample
usage and minimizes calibrant usage. The reduced volume of the candidate’s instruments
allow for a reduction in the volume of calibrant and verification standards required by a
factor of at least 3. This reduces the amount of cost to calibrate, verify and with any
associated disposal fees.
Integrated bubble trap – The candidate instruments has an integrated bubble trap that is
removable without tools, is front accessible, is easy to clean (less time and no special
cleaning tools) and has low sample retention. The reference instrument has a larger bubble
trap with numerous tight corners, making it difficult and time consuming to clean.
The candidate turbidimeter body is designed not to have any tight corners. This facilitates
5
efficient draining of the body and bubble trap and allows for more rapid cleaning of the
instrument. The reference instrument has many tight corners, takes a long time to drain and
is difficult to clean. Poorly cleaned instruments will lead to calibration, verification, and
measurement error.
The candidate instrument bodies have polished fluid handling surfaces which are designed
to reduce scale, fouling and bubble formation. The reference instrument does not have this
feature.
The candidate measurement chamber has a “V” shaped form factor that is designed to direct
light reflections away from the detector, thereby reducing stray light. This shape also
reduces particle settling that could result in positive measurement interference. The
reference instrument does not have this “V” shaped feature.
The candidate instrument has an integrated sample flow indicator. The indicator confirms
sample flow through the measurement chamber of the instrument. Alarms and warnings
can be set for either high or low flow conditions. The reference instrument does not have
this feature.
The candidate’s instruments contain multiple internal weirs. These ensure consistent
sample throughput and homogeneity and increased instrument robustness. The reference
instrument has a single overflow weir.
The candidate’s measurement chamber was designed for optimized sample handling. The
sample temperature does not change as it passes through the instrument, thereby
minimizing the chance that the sample composition changes and this reduces bubble
formation.
Complete Measurement System
(Body and Measurement
module)
The linear response of the candidate instruments covers the range of 0-10 NTU. The
published linear response range of the reference instrument is 0-5000 mNTU (0 -5 NTU).
This linear range allows for robust calibration protocols.
The candidate instruments have a self-alignment feature comprised of a magnetic
positioning system that ensures proper alignment of the measurement module onto the flow
body. The reference instrument does not have this feature and can easily become
misaligned, thereby causing a non-measurement condition or erroneous measurements.
The candidate instruments contain redundant user interfaces (smart device and
touchscreen). This accelerates the ability of an operator to respond to a turbidity event and
facilitated the ability to perform maintenance and quality control. The reference instrument
can only be interfaced using a controller that was directly connected to the instrument.
The candidate instruments contain both redundant data and meta data storage. The
reference instrument does not have meta-data storage capability.
The candidate instruments contained both wired (USB) and secure wireless (Bluetooth)
communication for operation of the instrument (User interface) and data transmission. The
reference instrument does not have these features.
The candidate instruments possess multiple outputs for both digital and analog
communication streams (i.e. Modbus, 4-20) mA). The reference instrument requires the
use of a controller to perform communications.
The candidate instruments utilize the latest available component technologies that increase
reliability and robustness when compared to aged technologies such as the reference
instrument.
1.2 Method Equivalency
The approved reference Method 10133 quality control (QC) acceptance criteria was only defined in the linear
calibration range (section 9.2.2) in which two calibration verification standards were prepared and measured. The
criteria were for the instrument to measure within ±0.025 NTU of each of these standards. One standard was in the
0.050 to 0.200 NTU range and the other was in the 0.700 to 0.900 NTU range. Although the range of the method
extends to 5.00 NTU, no other QC data was provided in Method 10133. The candidate methods required the
measurement of a QCS standard in the measurement range of interest. The pass/fail criteria were to measure within
10 percent of the theoretical value of the QCS or within 0.04 NTU for the Lovibond White Light LED Method or
0.030 for the Lovibond 660-nm LED or 6000 Laser Methods.
6
The pass/fail criteria for the candidate methods are based on the capability to prepare a calibration or verification
standard at an assigned value below 1 NTU. Below this turbidity level, it is typical for the manufacturer of
calibration standards to assign error as a fixed value. This error value is typically ± of 0.03 NTU to ± 0.04 NTU.
For example, a standard that is commercially prepared to 0.30 NTU nominal can have an accuracy spec of ± 0.03
NTU (which is ±10% in this example). In the range below 1 NTU, the combination of preparing a standard and then
administrating that standard in a calibration or verification can be very difficult. This is due to the impact of
interferences that result from the turbidity of the dilution water, particle contamination, bubble interferences, the
stability over time, and the technique of preparation of the standard for measurement, which can propagate together.
This is why most manufacturers who prepare standards will express the uncertainty of a given standard using a fixed
error value at turbidities below 1 NTU. Further, the preparation of standards below the value of about 0.10 NTU
should be avoided due to the increasing magnitudes of these interferences. Turbidity standards above the value of
about 1 NTU have better stability and the interferences are not as severe and therefore will have an uncertainty that
is typically a percentage of the stated value and is usually in the range of ±2 to ±10 percent of the nominal value of
the standard.
When performing QC on a method, the challenge is to be able to accurately prepare the standards. The error of the
standard typically dictates the pass/fail criteria for a given instrument, which was the case in the demonstration of
capability of these methods. Using the 0.025 NTU criteria from the reference method, the demonstration of linearity,
and ultimately the linear calibration range was assessed. The candidate methods delivered results on 8 turbidities in
the range of 0 to 1.00 NTU and for all candidate instruments, the difference between the delivered results and the
theoretical values were less than the ±0.025 criteria stated by the reference method. Above 1 NTU, a comparison
between the reference and candidate instruments showed differences that exceeded the ±0.025 NTU criteria for both,
but none of these instruments showed a difference that was greater than 7 percent versus the theoretical value of the
turbidity level. If using a percentage based pass/fail criteria of 10%, which is common practice for regulatory
verifications and is applicable above 1 NTU, the instruments passed the verification criteria. Additional linearity
discussions that support method equivalency are in the data and results sections of this report.
In the method performance section of Method 10133, percent recovery and precision data was provided on six
different turbidity levels that covered the range of 0 to 1.00 NTU. The percent recoveries ranged from 98 to 125
percent. The three candidate instruments were each tested on 8 different turbidities that covered this 0 to 1 NTU
range. Their combined percent recoveries ranged from 99 to 104 percent over this same range (Note all information
in this section data was separated out with respect to each method, which was contained in the data and results
sections). Between 1 and 10 NTU, the percent recoveries for the three candidate instruments ranged from 93 to 100
percent and the reference instrument ranged from 97 to 101 percent. Analysis of variance (ANOVA) statistics were
run on the percent recovery data between the three candidates and reference instruments for all the formazin spikes
used for precision and bias. These statistics showed F values were less than the F-critical values, indicating
statistically that the differences in the percent recoveries were not statistically significant.
In the method performance section of Method 10133, the precision was measured on water samples at 0.108, 0.027
and 0.0213 NTU and delivered standard deviations that were 0.025, 0.0011, and 0.0056 NTU respectively. It was
not known if these statistics were derived under flowing conditions. This ATP study derived standard deviations for
the spiked turbidity levels using the on-line or flowing condition, which arguably was a more valid approach. In this
ATP study, standard deviations for the candidate and reference instruments were derived and averaged within the
ranges of 0 to 0.100 NTU, 0.100 to 1.00 NTU and 1.00 to 10 NTU. The 0 to 0.100 NTU range reported standard
deviations that were below 0.001 NTU for both the candidate and reference methods. The 0.10 to 1.0 NTU range
reported standard deviations that were between 0.004 and 0.005 NTU for the candidate methods and 0.003 NTU for
the reference instruments. Above 1.0 NTU, the candidate methods delivered higher standard deviations that ranged
between 0.016 and 0.048 NTU and the reference instrument delivered an averaged standard deviation of 0.014 NTU.
7
Quality control samples were run on prior to and at the completion of data collection for the precision and bias data.
All instruments passed the QC criteria, which was based on QCS samples with theoretical turbidities of 0.31 and 1.0
NTU.
In summary the three candidate methods showed equivalency to the approved reference Method 10133 over the
range of 0.00 to 1.0 NTU, which was supported by the passing of the QCS samples and the ANOVA statistics on the
percent recovery data. The reference method does not contain performance data above 1 NTU, but this ATP study
did collect additional data on all methods up to 10 NTU.
1.3 Analytes
Turbidity is essentially the measurement of aggregate light scatter that was caused by all materials within a sample
that are capable of scattering light in a direction that reached the 90-degree detector. The nature of this “analyte”
turbidity does not lend itself to the CAS registry. However, turbidity calibrations are all traced to formazin
standards which have proved to be universally accepted standards across the world for many decades. Historically,
formazin has been used as the accepted surrogate for turbidity, and historically used to derive performance
specifications for instruments and to evaluate method performance. Formazin was used in this study to derive
precision, bias, and linearity data for the three candidates and reference methods. The formazin polymer itself does
not have a CAS registration number, but the components that are used to synthesize formazin stock standards do.
These components are listed below:
• Hydrazine Sulfate – CAS 10034-93-2
• Hexamethylenetetramine CAS 100-97-0
• Water, Filtered and Deionized CAS 7732-18-5
2.0 Study Implementation
This ATP study was conducted at three different test sites. The study was organized and managed by Tintometer
Incorporated’s research and development center that was located in Fort Collins, Colorado. The coordinator of the
study was Mike Sadar. He was assisted by the Tintometer R&D staff and by staff at the different drinking water test
sites in setting up the test sites and with data collection activities.
The test sites involved two drinking water plants (DWP) and the Fort Collins R&D facility. These sites are referred
to herein as the Fort Collins Filtered Tap Water Site (Fort Collins), the Aurora Binney South Platte Site (Binney
South Platte), and the San Patricio Municipal Water District (San Pat MWD) site. The source water from all three
sites was surface water that was treated by a variety of techniques prior to the filtration step. These details of each
were described separately.
The Fort Collins site was at the Tintometer Inc. R&D facility and it had access to city tap water. The tap water was
produced by the City of Fort Collins Water Treatment Plant. The surface water source originates from mountain
snowmelt runoff and was stored in a nearby reservoir (Horsetooth Reservoir). This raw water was treated using a
conventional treatment train that included flocculation, sedimentation, and dual-media anthracite/sand filtration.
This tap water at the Fort Collins site was further filtered through a size exclusion membrane with nominal pore size
of 0.05 um prior to entering the test panel. The purpose of this secondary filtration step was to ensure the removal
of all particulate materials and ensure a consistent particle-free baseline that would be suitable for multiple spikes.
This allowed the precision and bias (spike recovery) data to be collected at this site. The details the test setup is
provided in Section 6 of the appended study plan (Appendix A).
8
The Fort Collins site was selected to perform the precision and bias for the ATP study for several reasons. The three
candidate methods were continuous analysis methods and required adequate sample flow and pressure to ensure
consistent flow rates for the duration of the study. The tap water provided the required flow and pressures to ensure
all turbidity spikes would be accurate and consistent over the duration of the testing. In addition to the candidate and
reference instrumentation, a high precision peristaltic pump and an analytical balance were needed to carry out the
spikes (please refer to the appended study plan). NIST traceable weights were used to validate the accuracy of the
balance. The theoretical values of each spike ultimately trace to the accuracy of this balance. The testing was
conducted by the coordinator of the study, and assisted by other Tintometer R&D personnel.
The Binney South Platte site was located southeast of the Denver, Colorado metropolitan area. The raw water was
surface water from the South Platte River drainage and was characterized with high levels of total dissolved solids
and micro-pollutants such as pharmaceutical residuals. The raw water was treated through a multibarrier process that
included riverbank filtration, aquifer recharge and recovery, precipitative softening, ultraviolet light coupled with
advanced peroxide oxidation, biological activated carbon filtration and activated carbon adsorption. The filtration
step was through dual media filtration and the sample tap for this study was on the filter effluent water. The water
plant was commissioned in 2010 and had a treatment capacity of 50 million gallons per day. Mr. Kevin Linder,
Supervisor of Operations and Maintenance for the Binney water plant was the main contact person at this plant. Mr.
Linder delegated his instrumentation engineer to assist Tintometer in setting up the test site. This included
providing the sample taps and drains, power and the work area for the study. Tintometer also signed a memorandum
of understanding with the Binney facility that there would be no financial gain or loss to either entity as a result of
this testing. Thus, no compensation was paid to the Binney facility in relation to this ATP. Tintometer promised to
deliver a copy of the study results to Binney for their review.
The Binney South Platte site was selected for several reasons. First, the raw water was highly compromised with
high total dissolved solids. A large percentage of the raw water was made up of discharge from a permitted
wastewater plant upstream of its intakes. Second, the plant was a member of the Partnership for Safe Drinking
Water and the plant’s management consistently evaluates new technologies. This provided the basis for the
invitation to test at this facility. Last, the treatment train had been offline for several months to install two new filter
units. The plant had recently come on-line but was not in an optimized state of operation. The plant contacts
encouraged the study coordinator to move up the test timeline for the method comparison evaluation to ensure
observance of changes in the filter effluent turbidity levels. Study data covered more than three compete filtration
cycles (ripening, run and backwash), and variation in the filter effluent turbidity levels were observed.
The San Pat MWD site was located on the Southeast Texas coast north of Corpus Christi, Texas. The raw water was
surface water from the Nueces River Basin. The water was treated with flocculation and followed by sedimentation.
The settled water was then forced through Pall Microza™ membranes, which served as physical barrier to particles,
with a nominal pore size less than 0.1 micron. The membrane effluent was subject to regulatory monitoring for
turbidity.
The San Pat MWD site was selected primarily because it would subject the candidate and reference methods to
challenging environmental and measurement conditions. At the time this study was conducted, the raw water
temperature was 32 C (90F). The humidity was 100 percent which were prime conditions for condensation
formation. The membrane filtration process involves the use of pressure to force the water through the membrane.
Sample that passed through the membrane was subject to a significant decrease in pressure that resulted in
outgassing. This challenged the ability of the candidate turbidimeters to remove bubbles from the samples.
The primary contact at San Pat MWD was Mr. Jake Krumnow, who is supervisor of operations and maintenance.
Mr. Krumnow delegated resources to help set up the test site, they provided the sample tap and drain, power and
space for the equipment. The staff oversaw the study and San Pat MWD management has requested a copy of the
study data after the completion of this study. Participation in the study by San Pat MWD was wholly voluntary and
9
Tintometer Inc. was able to provide the staff with a detailed overview of the three candidate instruments. No
compensation was provided to the San Pat MWD facility for hosting this portion of the study.
The membrane effluent from each membrane rack was collected into a combined effluent line. This line was tapped
and served as the sample line for the three candidates and reference instruments in this study. Because the sample
was membrane effluent, there were no expected turbidity excursions. To generate some movement in the turbidity,
the study plan called for the injection of settled water into the filtrate water for a defined period of time. This
required the precision peristaltic pump (used at the Fort Collins site) to inject the sample. The turbidity of the settled
water was about 5 NTU, which was measured on a 2100AN turbidimeter. Two different injection rates were
performed to generate two turbidity spikes. The spikes were of a short duration, but long enough to provide
evidence of response and comparability by the three candidate and reference instruments.
The two drinking water plant matrices were used to provide comparability with respect to each of the candidate
instrument’s ability to track the reference turbidimeter. The goal of each site was to collect at least 8 hours of
continuous run time, which was exceeded. Data from these sites was broken down into approximately 8-hour blocks
of time and also into a separate data block that was dedicated to evaluate the injected turbidity spikes. Last, the
entire data set was analyzed and was ultimately reported with respect to comparability. The comparability between
each of the candidate instruments to the reference instrument was expressed as the net difference in turbidity
between each. This was discussed in detail in the data and results sections.
2.1 Study Schedule
The study plan commenced in April 2016 with an initial visit to EPA office of water ATP group to better understand
the ATP process. This was followed by method and study plan development. The data collection study took place
between July 15 and July 28, 2016. A breakdown of this phase was provided in Table 2.
Table 2 – The Lovibond Turbidity ATP – Details on Data Collection at the Different Study Sites
Dates Location Details
July 15-18, 2016 Binney South Platte
Site
July 15: Set up the instruments, calibrate, run QC. July 15-18 – Data
collection. July 18: Run short term spikes, final QC, take down instruments.
July 18-20, 2016 Fort Collins Site July 18: Set Up Instruments and allow to run. July19: calibrate and run QC.
July 19: Run 10 spikes to collect precision, bias, and linearity data. This
cumulated with final QC after all spikes were complete.
July 26-28, 2016 San Pat MWD Site July 26 – Travel and set up instruments and allow to run overnight. July 27:
Calibrate and run QC. July 27-28 collect data and run short term spikes. July
28: Final QC and take down instruments.
In Section 4 of the study plan (Appendix A) the initial schedule was to collect the precision and bias data first.
However a pump failure near the beginning of the spiking occurred and several of the spikes had to be terminated.
It was decided that this test would have to be repeated after the Binney South Platte site testing was completed. In
order to meet the Binney South Platte scheduled plan to host their portion of this ATP comparability study, we
moved to this site which became the first site of data collection. At the completion of this study, the instrumentation
was promptly relocated back to the Fort Collins Site and the formazin turbidity spike study was performed. This
was then followed by testing at the San Pat MWD site, the final site for this study.
2.2 Sample Collection
10
This ATP study was for on-line turbidity instrumentation. Therefore, samples continuously flow through the
instruments and do not require collection or holding times in the traditional sense. However, sampling was a very
important part of the study to ensure representative tracking and response times between the three candidate and
reference instruments. To ensure this was correct, the sample flowed from its tap to a manifold that split it into
equal streams, with one stream leading to each instrument in the study. The length of each sample line from the
manifold to each instrument was the same. Sample flow rates were then set to the middle of the manufacturer’s
suggested sample flow range. The candidate’s flow rates were about 70 ml/minute and the reference’s flow rate was
about 330 ml/minute. The overall flow rate to all the instruments had to exceed the sum of the flow rates to
individual instruments to ensure no instrument was starved of sample. The study plan contains renderings of the test
setup for all the sites, which included the splitting of the sample to the test instruments.
Prior to the study, testing was successfully performed on both the candidate and reference instruments to confirm
that variations in flow did not have an impact on the turbidity results. Both designs measure at atmosphere and are
of an overflow weir design. The overflow weir design compensates for any fluctuations or changes in flow, as it
continuously maintains the same volume of sample within the measurement chamber.
2.3 Types of Analysis Performed
The analysis of turbidity was continuously performed by the candidate and reference instruments, of which all were
on-line turbidity instruments. The turbidity data was collected concurrently for all instruments, which included the
three candidate method instruments and the reference method instrument. All instruments were capable of
performing their measurements at a frequency of once per second. However, this could challenge the data logging
capabilities of the reference instrument, so the data logging frequency was set to once every 15 seconds.
Turbidity data was collected on three waters. The study plan required a minimum of 8-hours of data collection at
each DWP site. The two DWP sites primarily had data collected for much longer than this period and the data was
statistically analyzed to 8-hour segments, and the entire data set.
The Fort Collins site included a shorter period of data collection time. The analysis started with the defining of the
baseline for the filtered tap water, which was the blank, since the tap water was pushed through a filter was through
a pore size smaller than 0.05-micron, essentially producing particle free water. This baseline establishment was
followed by ten successive spikes of turbidity using formazin. Formazin was the primary standard for all turbidity
methods and was generally used in the industry and in the regulatory community as the standard to calibrate and
evaluate method performance. The entire time to perform the baseline and all spikes was 7 hours.
The candidate turbidimeters were identical with the exception of the light source. Their operation, maintenance,
calibration and QC protocols were identical throughout this study. This presented uniqueness of the study overall
because the same sample was simultaneously analyzed by four different light sources. The differences in response
with respect to these light sources could be scrutinized with respect of their ability to detect and quantify turbidity
levels in the range of 0-10 NTU and more specifically, at turbidity levels below 1 NTU.
2.4 Study Plan Deviations
The intent of the data collection phase of this ATP was to follow the validation plan and for the majority of this plan,
it was extensively followed. There were three exceptions of deviations from the validation plan. The first deviation
was during the spike study at the Fort Collins Site. When the first spike was initiated, we noticed that the
characteristic ramp up to the theoretical value was very slow and unstable. The cause was traced to the wrong
connection to the injection port for the spiking of the turbidity standards. In order to generate accurate turbidities of
11
spikes; the entire filtered flow stream must be injected with the prepared turbidity standard. The test stand had two
injection ports. One injection port impacted only a portion of the sample stream. If this port was used, there was no
means available to accurately measure the flow through that portion of the system. The correct port injected the
entire filtered water stream and it could be measured accurately, which was necessary to be able to calculate the
theoretical value of the formazin spikes. This error was immediately noticed and was corrected. The data, which
was for a very short span of time (July 20, 2016 from 950-1011) was identified and then ignored. It was called out
in Figure 1. The spike data, which was graphed, contained a section that was circled and was omitted from
calculations. Upon correction, the first spike was repeated. Subsequent spikes beyond this were as per the
validation plan.
The validation plan called for the use of the FilterTrak 660 as the reference instrument. At the Binney South Platte
Site, we also had installed a 1720E as an additional sensor on the water. Data from that instrument was also logged
and was plotted graphically in Figures 3 and 4. However, the data was not used for any of the calculations or
analysis and was included in these two graphs to illustrate its comparative response to the three candidates and
reference instruments in this study. The installation of this instrument at the Binney South Platte Site had no impact
on the results that were generated at this site.
The timeline with respect to the order of the validation plan changed. The plan had the precision and bias data being
collected first, followed by the Binney South Platte and then the San Pat MWD sites. However, the Binney South
Platte was moved to the front of the schedule at their request. The change in the schedule was within a few days of
the original plan and thus, the impact from environmental conditions was minimal; and likely had no impact on the
overall scope or outcome of this ATP validation study.
Last, the validation plan called for the use of one QCS sample after calibration and at the end of data collection and
its value would be 1.0 NTU. Instead two QCS samples were prepared and run after each calibration and at the end
of data collection at each test site. One QCS sample was at 1.0 NTU and the other was either at 0.3 or 0.6 NTU. This
had no impact on the overall scope or outcome of this ATP validation study.
3.0 Method Procedure and Data
The Lovibond Process Turbidity ATP validation plan was designed to concurrently collect comparability data for
the three candidate turbidity methods and compare to the approved reference method (Method 10133). It was
mentioned that the difference in the three candidate methods was only with the type of light source used and all
other sections of these methods (and their representative instruments) was identical, with one exception. That was
with respect to the data in the method performance section (Section 13). The three candidate methods are appended
in this report as Appendix B.1, B.2 and B.3; for the Lovibond White Light LED Method first, followed by the
Lovibond 660-nm LED Method, and ending Lovibond 6000 Laser Method respectively. The data that was collected
from this validation study was implemented into these three proposed methods. All data was highlighted in purple
text and was primarily located in Section 13.
Section 1.1 of this report, Method Justification, provided comparison and contrasts between each of the candidate
methods (and representative instrumentation) and reference Method 10133 (and the FilterTrak 660). The comparison
was from a component and feature perspective, and included the mandated design criteria. This is summarized in
Table 1 of this report. It was important to note that some of the comparisons utilize certain technologies that
ultimately become features that may have otherwise been restricted in previous instrument designs. For example,
when a collimated light source such as an LED or laser was used, its characteristics allowed for a reduction in
sample measurement space because it required less optics to retain beam columniation through the measurement
12
cell. This for example, can lead to smaller sample volumes, which translates to reduced sample usage, reduced
calibrant usage and reduced generated reagent waste. It was common to see secondary and tertiary impacts from a
simple difference in between methods. Table 3 separately compares each of the three candidate and reference
methods to each other section by section. For each section, a brief comparison and contrast to the reference method
was provided. If the cells are merged, this indicated the comparability was across all of the candidate methods. The
most significant differences between the methods will be with respect to Section 6, Equipment.
Table 3 – Compare and Contrast Between the Candidate and Reference Methods in this Turbidity ATP.
Section Reference Description (or
specification)
Lovibond White
Light LED
Method
Lovibond 660 nm
LED Method
Lovibond 6000 Laser
Method
1.0 Scope and
Application
Range was 0-5.0 NTU Range was 0-10 NTU
2.0 Summary of
Method
Method was based on light scattered by the sample under defined conditions to intensity of light scattered
by a reference standard solution. The reference standards are formazin or a stabilized version of formazin
for calibration and verification.
2.0 Summary of
Method – Dry
Verification
Modules
A dry verification device,
specific for the instrument
was cited in the method.
The candidate methods do not reference a dry verification device for
verification. However, such devices are available and have been tested.
3.0 Definitions
Definitions were the same for all three candidate methods. Definitions were essentially the same for both
the reference and the candidate methods except for the terms Calibration Verification Standards
(CALVER) and MSDS, both in the reference are synonymous to the Secondary Calibration Standards
(SCAL) and SDS respectively in the candidate methods.
4.0 Interferences. The candidate methods incorporate all the stated interferences in the reference method, plus the candidate
methods include the interference from internal surface reflections (i.e. stray light).
5.0 Safety The candidate methods incorporate all the stated safety concerns that are in the reference method, plus
additional safety concerns on Hydrazine sulfate.
6.0 Equipment
The Turbidimeter shall consist of a nephelometer with a light source for illuminating the sample and one or
more photo-detectors to measure the amount of scattered light at a right angle to the incident beam. This
was the same for the three candidate and the reference methods.
All methods correlate scattered light signal, detected at a 90-degree angle, to turbidity, which was defined
by standard reference turbidity suspensions.
6. Optics – Incident
Light Source
Laser diode Operated at
660±30 nm. This will
provide sensitivity to the
presence of individual
particles that flow within
the analysis volume of the
instrument.
LED emitting white
light in the visible
spectrum between 380
and 780-nm. This
method may show
enhanced sensitivity to
the smallest particle
size when compared to
other methods.
LED with a peak
emitting wavelength
between 650 and
670-nm. This
method has the most
stringent
wavelength criteria,
which can be
reflected in intra-
instrument
comparability.
Laser Diode with a peak
emitting center wavelength
between 650 and 690-nm.
Like the reference method,
this method will provide
sensitivity to the presence
of individual particles that
flow with the analysis
volume of the instrument.
6. Peak spectral
Response of System
The detector must
encompass the entire
spectral output of the
incident light source,
which is optimally at 660-
nm
The spectral peak
response shall be
between 400 and 600
nm.
The spectral peak
response shall be
between 600 and
700 nm.
The spectral peak response
shall be between 600 and
700 nm.
6. Distance
traversed by the
incident and
scattered light
This distance was not to exceed 10-cm in all methods.
6. Parallelism of
incident light path
within the
measurement
volume.
No divergence in
parallelism and
convergence must not
exceed 1.5 degrees.
No divergence in parallelism and divergence was not to exceed 1 degree
13
6. Detector position
Centered at 90 and not to
exceed ±2.5 degrees from
90.
Centered at 90 degrees ± 1.5 degrees relative to the incident light beam.
6. Detector receive
angle
None was specified in the
reference method.
Scattered light detector/receiver shall be at a subtended angle between 20 and
30 degrees from the center point of the measurement volume. The provided a
defined view volume for the detector.
6. Detector type
Shall be a photomultiplier
tube, which is very
expensive.
Detector is a silicone photodiode with a spectral sensitivity that encompasses
the spectral output of the incident light source. The detector signal is
amplified and converted to a turbidity signal. Only the performance
characteristics that include: 1) no airgap between the sample and the surface
of the detector, which minimizes stray light; 2) spectral sensitivity must
encompass the spectral output of the incident light source.
6. Stray Light
Design
All the methods essentially state that the overall optical/instrument design shall be to minimize stray light
so it will not cause significant error in the determination of turbidity in the sample.
6. Stray Light Trap None was specified in the
reference method
All candidate methods state that un-scattered incident shall pass into a light
trap that encompasses the entire diameter of this incident light beam.
6. Fiber Optic
Cables
Method allows the use of
fiber optic cables to
transport scattered light.
There was not a reference to fiber optic cables in this method.
6. Algorithm Usage
No algorithm type is
specified in the reference
method.
All candidate methods require the algorithm that converts light scatter signal
to turbidity readings to be a linear-based algorithm (e.g. y=mx+b). This
guarantees a linear calibration range for the instrument, which was possible
by the incorporation of the optical configuration of the instrument.
6. Sample deaerator No aerator is specified in
the reference method
The candidate methods incorporate the required use of a sample deaerator to
remove entrained air from the system.
6. Instrument Drift
All methods include a specification that the turbidimeter shall be free from significant drift after a short
warm-up period. The use of solid state light sources help to reduce drift. The candidate instruments all use
a monitor detector and feedback mechanism to correct for any drift that would occur over time.
6. Instrument
Sensitivity
Capable of detecting
turbidity difference of
0.001 NTU or less.
Capable of detecting
turbidity difference of
0.010 NTU or less.
Capable of detecting
turbidity difference
of 0.010 NTU or
less.
Capable of detecting
turbidity difference of
0.005 NTU or less.
6. Instrument
Specification
The FT660 is an
instrument that meets the
criteria of Method 10133
the reference method.
The Lovibond PTV
1000 is an instrument
that meets the criteria
of the Lovibond White
Light LED Method
The Lovibond PTV
2000 is an
instrument that
meets the criteria of
the Lovibond 660-
nm LED Method
The Lovibond PTV 6000 is
an instrument that meets the
criteria of the Lovibond
6600 Laser Method
7.0 Reagents and
Standards
Instructions were provided to produce reagent water that would have a residual turbidity less than 0.03
NTU.
7.2 Stock Standard
Suspension
The reference method
prepares 400 NTU
formazin Solution
The three candidate methods prepare a 4000 NTU formazin stock solution.
This method of preparation was in alignment with the methods used in the
preparation of commercially available formazin stock solutions.
Gravimetrically, the masses of the raw materials used to prepare the stock
suspensions are equivalent for the candidate and reference methods.
7.3 Instrument
calibration
Standards
The reference method
provides instructions on
the preparation of
dilutions to prepare the
calibration standard. The
reference method
recommends the use of
stabilized formazin
standards.
The candidate methods provide instructions to prepare a working stock
solution at 40 NTU. The methods then instruct to dilute this solution with
Class A glassware and reagent turbidity free water to the final standard
value(S). The candidate methods also recommend the use of stabilized
formazin standards
Section 8.1 Sample
Collection
All methods state that the methods for on-line analysis and sample collection, cooling, and preservation
does not apply.
Section 8.2
Instrument Setup
The reference method
references the
manufacturer’s
installation manual. It
provides some additional
The candidate methods reference the manufacturer’s instructions for
installation and startup. This was kept open in case new technologies or
techniques can be implemented over time.
14
guidance that pertains to
individual instrument
features.
Section 9 Quality
Control
The reference and candidate methods have very similar quality control programs. They have the same
introductory section and very similar initial demonstration of performance. The reference the use of quality
control samples (QCS) to verify the linear calibration range (LCR). The reference method instructs to
prepare two standards and have 0.025 NTU criteria as pass/fail. The candidate methods have a requirement
to use a blank and three standards for the LCR determination, with a pass/fail criterion of ±10 percent.
Section 9.2.3
Quality Control
Sample
The reference and candidate methods are very similar. Both use a single QCS to confirm the LCR on at
least a quarterly basis. The pass fail criteria on the candidate methods are stated at ±10% or ±0.040 NTU.
There was no pass fail criteria statement for the QCS in the reference method.
Section 10
Calibration
The reference methods
state to calibrate
according to the
manufacturer’s
instructions.
The candidate methods state to calibrate according to the manufacturer’s
instructions. In addition, these methods require cleaning and maintenance
prior to calibration. These methods require calibration under the same
ambient conditions as measurement. The instructions also require calibration
in the instrument itself (no calibration cylinders), which insures proper
maintenance of the instrument.
Section 10
Calibration
Verification
The candidate and reference methods instruct to perform verification in the measurement range of interest.
The candidate and reference methods reference wet or solid standards (dry verification standards) for use in
verification.
Section 11
Procedure
The reference method
references the instrument
manual with emphasis to
maintain consistent
sample flow.
The candidate methods reference the manufacturer’s instructions and
emphasize the importance to maintain consistent sample flow. These
methods recommended the installation of a rotometer to eliminate
fluctuations in sample flow rate. The candidate methods also emphasize the
need to maintain constant sample temperature.
Section 12 Data
Analysis and
Calculations
The reference and candidate methods instruct to report to the nearest 0.01 NTU or 10 mNTU. Above 1
NTU, the candidate methods provide additional reporting criteria.
Section 13 Method
Performance
The reference method
reports precision and bias
on 5 spike levels between
0 and 0.65 NTU.
The candidate methods provide precision and bias on 10 spike levels between
0 and 10 NTU, with 8 of the spikes below 1 NTU. The results in section 13
for each of the candidate methods are discussed in the Data Analysis section
of this report. The candidate methods also have a final section to check the
precision and bias on a routine basis at the frequency required for regulatory
compliance. The reference method does not include this section.
Section 14 Pollution
Prevention
The reference and candidate methods are the same. In addition, the reference method includes a citation on
waste reduction.
Section 15 Waste
Management
The reference and candidate methods essentially have the same section with regards to waste management.
Section 16
References
The reference method has
its own set of references.
The candidate methods have the same set of references that are different from
the reference method.
Section 17 Tables
and Validation Data
This section does not exist
in the reference method.
The validation data
wholly contained in
section 13.
The candidate methods contain their unique precision and percent recovery
tables for each method.
3.1 Validation Study Demonstration Data
This validation study was designed to concurrently demonstrate the performance of the three candidate methods to
the reference method. This performance was in concurrence across the three study sites. The study called for
precision and bias determination to be performed at the Fort Collins site and this will be discussed first. The other
two waters, the Binney South Platte and the San Patricio MWD site will follow respectively.
For each test site, a graph of the data was presented. These graphs represented all data from the start to the stop
times of the study. No points were excluded from any graphs. The two drinking water plant studies also had a short
15
spiking session, in which water upstream from the filtration step was spiked to into the effluent to generate turbidity
movement. For these spiking sessions, a second graph was generated that zoomed in on the spikes.
The data presented in this section will be summary data copied from Microsoft Office Excel (Excel) workbooks. All
the calculations are found in worksheets (spreadsheets) that are part of a given workbook. Each test site has a
dedicated workbook and an explanation for each spreadsheet inside a respective workbook was provided. The
explanation of each spreadsheet was intended to provide the needed guidance to navigate through the calculations
that were applied to deliver the final results for each test site. These explanations will be further explained in the
discussion of each demonstration data section (sections 3.1.2 through 3.1.4).
In order to maintain consistency within this section, the validation demonstration data was presented in tables, along
with a discussion of that table. The demonstration data was ordered accordingly.
1. Presentation of the data graphically, that represented the three candidate methods and the reference
method.
2. Reference to the spreadsheets in the Excel workbook used to generate the precision, bias and linearity data.
(Table 4)
3. Summary of the spike response data for all the methods in this validation study. (Table 5)
4. Summary of the precision for all the methods in this validation study (Table 6).
5. Data and respective summary on the accuracy or bias (Percent Recovery) for the Lovibond White Light
LED Method (Table 7).
6. Data and summary on the accuracy or bias (Percent Recovery) for the Lovibond 660-nm LED Method
(Table 8)
7. Data and summary on the accuracy or bias (Percent Recovery) for the Lovibond 6000 Laser Method (Table
9).
8. Comparability results for the three candidate methods and the reference method at the Binney South Platte
test site (Table11)
9. Comparability results for the different candidate methods and the reference method at the San Pat MWD
membrane test site (Table13)
10. Limit of Detection discussion and analysis (Table 14).
3.1.1 Calibration
Calibration was performed at each test site after the instruments were installed and running for several hours. The
same procedures were performed for calibration at each site. Calibration was performed using the instructions
provided by the instrument manuals for the three candidates and the reference methods. An overview of each is
provided herein.
The calibration procedure for the three candidate methods required the preparation of a 5.01 NTU calibration
standard. The standard was set to this value because of the turbidity of the dilution water (water filtered through a
0.02 um filter yields a turbidity of about 0.01 NTU). A 1-liter volume of standard was prepared by diluting 1.25 ml
of a 4000 NTU standard to 1.00 liters total, using a Class A glass volumetric flask. The 1-liter volume was
sufficient to calibrate the three candidate instruments concurrently.
The 1.25 ml pipette was a high resolution micro-pipette with manufacturer-certified calibration accuracy to within
0.5%. The dispensation volume of the pipette was further checked by pipetting 1.25 ml of water onto a high
resolution balance (measured to the nearest 0.0001 g). The temperature correction for the density of water was
applied. The approach balance confirmed the accuracy of the pipette’s dispensation.
16
Once prepared, a small aliquot of sample was taken from the flask and measured on the laboratory turbidimeter, a
2100AN. This instrument had been calibrated prior to the study. The sample measurement had to be within 2
percent of the theoretical value, which was the error budget for the laboratory turbidimeter.
Immediately prior to calibration each candidate instrument was drained and then rinsed with a liter of turbidity free
water. It was then allowed to drain before commencing with the calibration. The standard was directly introduced
into the body of each of the candidate instruments. After introduction of the standard, the measurement module was
then replaced onto the respective turbidimeter. The instruments then counted down one minute to allow for any
bubbles to dissipate from the standard. The live measurements were displayed during this time, which became
stable as the time passed. Once the displayed values were stable, the calibration was confirmed and written into the
firmware of each (candidate) instrument.
The calibration of the reference instrument was performed according to the instructions in the reference instrument
manual. The calibration standard was an 810 mNTU standard, prepared by diluting 200-ul of 4000 NTU stock
standard solution to a final volume of 1.00 Liter in a Class A volumetric flask. Consistent with the preparation the
calibration standards for the candidate instruments, the pipet accuracy was certified to 0.5 percent and was
confirmed through the dispensation of water at that volume onto a high resolution balance. In addition, a small
aliquot of the standard was taken and measured the same 2100AN turbidimeter used to measure the other calibration
standards. The standard measured within 2% of the theoretical value of 810 mNTU.
The calibration was then performed on the reference turbidimeter according to the manufacturer’s instructions. We
did make an effort to thoroughly clean the instrument and then rinse it was several liters of turbidity free water prior
to the calibration. The calibrated instrument gain was within pre-established limits for the reference instrument (this
was controlled by its software).
To confirm calibration accuracy, two QCS standards were prepared and run on the three candidates and the
reference instruments. The standards were prepared at 1.0 NTU and 0.31 NTU for two of the test sites and a 0.61
NTU standard was prepared instead of the 0.3 NTU at one site (Binney South Platte Site). Taking into the account
of the dilution water, the standard was adjusted accordingly, making the 1.0 standard a 1.01 and the 0.3 standard a
theoretical value of 0.31 NTU. The instruments must measure within 10% or 0.030 NTU of the theoretical value of
each standard to pass QC for the PTV 2000 and 6000 turbidimeters and within 0.040 NTU for the PTV 1000
turbidimeter. These QCS standards were prepared before and at the conclusion of data collection at each site.
Each Excel workbook contains a spreadsheet titled “Quality Assurance”. The information on the preparation and
measurement of calibration standards and QCS standards and the results respectively was included on the “Quality
Assurance” page.
3.1.2 Initial Demonstration of Capability
The Fort Collins Test Site
The initial demonstration of capability, in terms of precision, bias, and linearity was performed at the Fort Collins
site. The validation plan was designed to test the three candidates and the reference methods concurrently on
surrogate turbidity spikes of formazin. The study plan was designed to perform all spikes in an on-line condition so
the demonstration of capability was under realistic operating conditions. Figure 1 was a summary graph of the
spikes involved in this study.
17
Figure 1- Spike response graph for the three candidate method technologies and the reference method over the turbidity
range of 0-10 NTU.
In Figure 1 the legend is at the bottom of the graph, and it identifies the reference and candidate turbidimeters. The
response to each spike is represented on the y-axis and was a log scale. The log scale is used because the spikes
cover three decades of measurement. The time scale is on the x-axis and as the study plan describes, the study starts
with an establishment of the turbidity free baseline, which is then followed by 10 successive spikes of increasing
turbidity. The portion of the data run that was subject to having the wrong injection port is circled and this data was
omitted from the study (please see section 2.4 Study Plan Deviations). Figure 1 provides an overview demonstrating
the instruments, at their set flow rates respond with near equivalence to the reference on each of the spikes. This is
consistent across all of the turbidity spikes at this site.
Figure 2 is a summary graph for all of the spike responses for the three candidate and reference instruments. The
legend at the bottom identifies the candidate and reference instruments. The y-axis is the theoretical turbidity of
each spike. The x-axis is the instrument response to each spike. Both axes cover three decades of turbidity response
and are on a log-log scale to best illustrate the comparability at each turbidity level. The importance of this graph
was that it summarizes the responses over the entire test range of 0-10 NTU for the three candidate and reference
instruments and it illustrates the impact of stray light at the lowest turbidity levels tested. It is a visual comparison
of the significance of stray light at these low levels.
18
Figure 2 - Spike Response charts, on log-log scale for the candidate and reference technologies. The deviation from
linearity at the lower part of the range is the stray light of each technology.
In reference to Figure 2, the stray light was indicated by a positive deviation from linearity at the bottom of the
measurement range. The reference instrument had very little stray light. The PTV 1000, which was the White Light
LED method had the highest stray light, which would be expected since it was the polychromatic light source. The
PTV 2000 and PTV 6000, which had near monochromatic light sources that have very low stray light, which were
within 2-3 mNTU of the reference instrument.
To address the high stray light of the PTV 1000 white light method, this method has a procedure that will allow the
subtraction of the blank value for the turbidity free sample water, which will be applicable for the measurement of
samples below 0.10 NTU.
The Excel workbook that contain the precision and bias data is titled “Ft Collins Filt Tap P&B Test Site.xlsx”. This
workbook is used to deliver the precision and bias tables that are listed later in this section. The derivation of all the
data is through the formulas contained in the spreadsheets from this workbook. A breakdown of the Excel
workbook is for each spreadsheet. From left to right, the spreadsheets are sorted as follows: Exec Summary,
Reporting, Checklist, Quality Assurance, Injection Summary, Raw Data, Fzn Spike Resp & Recovery Data, Spike
Response Graph, Response Summary Graph, ANOVA Single Factor, and Photos. Table 4 contains a description of
each of these spreadsheets and how they were related to the data in this validation report. Note: with respect to these
workbooks, a spreadsheet, worksheet or page is synonymous.
19
Table 4 – A Description of the Excel Workbook From the Fort Collins Test Site for Precision and Bias
Determination
Excel Page
Description
Exec Summary This is the Executive Summary of the Study. It briefly describes the study, the results and the
conclusions. The results and conclusions are mirrored into this validation report.
Reporting The reporting page contains the final tables that provide the side-by-side comparison between each
of the three candidate methods and the reference method. The data in these tables originate in the
Fzn Spike Resp & Recovery Data spreadsheet. The active cells on this spreadsheet are at the
bottom to categorize the data into specific turbidity ranges (e.g. 0 to 0.100 NTU, 0.100 to 1.00 NTU,
and 1.00 to 10 NTU). At the bottom of this page is a summary that includes a description on how
the tables are generated. The limit of detection calculations are on the reporting page.
Checklist This is a checklist to help the site coordinators insure all the details that pertain to the setup,
collection of data and QC are performed. Most of the details are discussed in Section 7 of the
validation plan.
Quality
Assurance
This spreadsheet contains information and data that pertains to the calibration and QC for the study
site.
Injection
Summary
This spreadsheet contains the details that relate to each injection. It includes the start and stop times
of the injection, the beginning and end masses of the standards that were injected, and flow volume
measurements. This sheet contains the calculations that are used to derive the theoretical turbidities
for each of the spikes, which are used for percent recovery (accuracy) calculations.
Raw Data Raw Data – This spreadsheet contains the original raw data from the study. The date/time are in the
first column, followed by the data from the reference instrument in column B, the PTV 2000 Red
LED in Column C, the PTV 1000 WL LED in Column D, followed by the PTV 6000 Laser in
Column E. This raw data overlap the actual study data from a time perspective in that it contains
data prior to calibration.
Fzn Spike Resp &
Recovery Data
This is the most important spreadsheet in the Excel workbook with respect to the precision
and bias derivations. Raw data that is from the applicable blocks of time that is logged from the
background and spikes is copied into this page (Columns A-E are the same headings). This data is
then segmented into each spike, which is statistically analyzed to generate the average, standard
deviation and RSD. These statistics are highlighted in green. Simply scroll down Columns A-E to
see the segments from where these statistical operations are applied. These calculated values are
then copied and pasted(as values) into the tables that are just to the right of column E (Columns H-
O, Rows 4-38) The tables on this page are labelled as follows: Table 1 contains the averaged value
of the response for each instrument on each spike. The bottom of Table 1 contains the least squares
analysis for linearity and the assessment of accuracy over the turbidity range for the three candidate
and reference instruments. Table 2 contains the precision data for each of the spikes. Tables 3
through 7 contain the percent recovery calculations for the reference instrument (Table 3), the PTV
2000 Red LED (Table 4), the PTV 1000 WL LED (Table 5), and the PTV 6000 Laser (Table 6).
Tables 4-7 contain both blank corrected and non-blank corrected percent recovery tables. The data
in Tables 4-7 is the data that are in the tables on the Reporting spreadsheet.
Spike Response
Graph
This is the summary graph that covers the initial baseline (blank) derivation and the responses to
each of the ten spikes. It is pasted into the validation report as Figure 1.
Response
Summary Graph
This summary graph is derived from Table 1 on the “Fzn Spike Resp & Recovery Data” page and is
pasted into the validation report as Figure 2.
ANOVA Single
Factor
This page contains the ANOVA analysis of the different spike recovery tables that are generated on
the “Fzn Spike Resp & Recovery Data” page. The ANOVA analysis is based on columns of data.
The ANOVA calculations are on this page.
Photos This page contains embedded photographs that were taken during this portion of the study. Most of
the photographs are with respect to the calibration and QC portion of the validation study.
Precision and Accuracy Data for the Three Candidates and Reference Methods
For each instrument in this study, the accuracy data was generated through a calculation of the instrument turbidity
response relative to the theoretical turbidity for a given spike. The theoretical value of the spike takes into the
account the combination of molecular scatter (turbidity) and fluorescence effects of the blank solution, so all data
was blank corrected.
20
The data collection started with the determination of the turbidity free baseline (the blank) and followed with
collection of data from each of the turbidity spikes over the test range of 0-10 NTU. A total of 10 spikes were
performed, from the lowest turbidity to the highest turbidity. A total of 3 spikes were in the range of 0 to 0.100
NTU, 4 spikes between 0.100 and 1.00 NTU, and three spikes between 1.00 and 10.0 NTU. For each spike, the
instruments were first allowed to respond to the increase in turbidity until a stabilized condition was observed. Once
stabilized, the instruments continued to collect measurement data on that spike until a total of 60 consecutive
measurements were logged. Statistics were then performed on these 60 measurements for each instrument, which
included the average, standard deviation and relative standard deviation. The averages are provided in Table 5 and
the standard deviations are provided in Table 6. The data from Table 5 was used to generate the percent recoveries
for each of the three candidate methods and the reference method. The data in Table 6, the precision data, was used
to provide the summarized precision statements that were provided in each of the three candidates and reference
methods.
The baseline collection was performed after the QC when the instruments became stable, which indicated that all the
QC standards were completely flushed from the instruments. The instruments then ran until a minimum of 60
measurements were collected, and actually 80 measurements were collected and used to determine the baseline
values. In process analytics, baseline establishment is extremely important that it is determined accurately, so
additional values improves the probability that this is correct. During the spikes, we set a more rigorous collection
protocol in which the last 60 measurements of a given spike were used to determine the average response for each
instrument.
At the bottom of Table 5 least squares correlations are performed between the theoretical turbidity and the given
instrument response for each of the instruments in this study. A perfect correlation, characterized with a slope and r-
squared value equal to exactly 1.000 would result when all the measured values are equal to their corresponding
theoretical values for each data pair. The slopes for the three candidate methods are all within 1.5 % of this
theoretical value of 1.000 and are also within 2 percent of the slope from the reference instrument. The correlation
coefficients are derived between the measured and theoretical responses for each of the three candidates and the
reference instrument. In all cases, the r-squared values are better than 0.999. This demonstrated high and
acceptable linearity over the turbidity test range of 0 to 10 NTU.
This analysis of the slope and correlation coefficient for each instrument is important as it demonstrates the
calibration for each candidate method is to adjust a linear based algorithm. This confirms that a 1-point calibration
is valid over the stated measurement range of 0 -10 NTU. Further, this demonstrates the high linearity of the three
candidates’ methods (and instruments) allows for QC validation using a single standard within this stated range of
measurement. Note that the three candidate methods still utilized two QCS standards (instead of one) to validate the
linear calibration range (LCR) of the method as insurance for data validity.
Data from Table 5 is used to generate the percent recoveries for the candidate methods separately. This data is
contained in Tables 7-9.
Table 5: Averaged Response for Each Turbidity Spike and the Analysis of Linearity for the Three Candidates and Reference
Methods
Spike
Description
Stable
Measurement
Time Range
Theoretical
Turbidity
in Sample
in NTU
Theoretical
Turbidity of
Spike in NTU
FT660
Reference -
in NTU
PTV2000
Red 660 Test
- NTU
PTV 1000
WL Test -
NTU
PTV 6000 685
Laser Test -
NTU
Initial
Baseline 911 - 947 0.007 0.000 0.010 0.013 0.024 0.012
2.01 NTU at
4 RPM 948 - 1033 0.021 0.014 0.025 0.027 0.040 0.027
21
2.01 NTU at
8 RPM 1034 - 1111 0.035 0.028 0.039 0.042 0.054 0.041
2.01 NTU at
16 RPM 1112 - 1146 0.067 0.060 0.070 0.074 0.086 0.072
16.0 NTU at
4 RPM 1147- 1218 0.117 0.110 0.123 0.124 0.138 0.123
16.0 NTU at
8 RPM 1219 - 1251 0.232 0.225 0.243 0.245 0.259 0.242
134 NTU at
2 RPM 1252 - 1324 0.561 0.554 0.569 0.565 0.588 0.566
134 NTU at
4 RPM 1325 - 1411 0.928 0.921 0.941 0.931 0.964 0.927
134 NTU at
8 RPM 1412 - 1455 1.902 1.895 1.938 1.870 1.953 1.897
800 NTU at
2 RPM 1456 - 1530 3.575 3.568 3.473 3.356 3.468 3.378
800 NTU at
6 RPM 1531 - 1602 9.383 9.376 9.482 9.317 9.460 9.431
Slope vs
Theoretical 0.9933 1.0129 0.9974 1.0006
% Accuracy -0.6723 1.2893 -0.2567 0.0624
Correlation
Coefficient 0.99972 0.99969 0.99963 0.99963
% Linearity
(100% is
Perfect) 99.97 99.97 99.96 99.96
In turbidity it is common to determine accuracy over a stated range using the approach provided in Table 5. The
reason is that the chance of a given standard being prepared with a significant level of uncertainty is high at low
turbidity levels (<1.0 NTU). This becomes more significant at even lower levels. Using a least squares fit between
theoretical and measured, an accuracy value over a given range can be provided. This approach works if it is known
that the measurement algorithm is linear based. In technologies where the algorithm is not linear based, there may
be a better approach to estimate accuracy.
Table 6 contains the precision data that was generated at the Fort Collins test site. All precision values in Table 6
are as the standard deviations and the units are in NTU. The table contains the precision for each instrument that
represents its respective candidate method and also contains the precision for the for the reference method
instrument. The precision was calculated from the same 60 measurements collected on each turbidity spike that was
used to generate the averaged response.
The bottom of Table 6 categorizes the precision data into specific turbidity ranges. These ranges are 0-0.100, 0.100-
1.00, and 1.00 to 10.0 NTU. The spikes that fell into a range for a given method are averaged together to deliver the
stated precision for the given range. These stated precision values for each given range are reported into Section 13
22
of the respective candidate method. Below 1 NTU, the averaged precision for each candidate method is within
0.002 NTU of the precision reported by the reference instrument. Between 1.0-10 NTU, the precision for each of
the candidate methods is within 0.034 NTU of the precision for the reference instrument. This demonstrates
comparability with respect to precision between each of the candidate methods and the reference method.
It is important to note that a decrease in precision was observed in the Lovibond 6000 Laser instrument. Precision is
normally expected to increase in comparison to other methods because of the narrow view volume of the laser
instrument. Similar to a particle counter, the passing of an individual particle into and out of the view volume can
lead to this higher variation. Published literature has shown this is useful information with respect to the
optimization of filter performance.
Table 6 - Precision for Each Spike (Standard Deviation) for Each of the Candidate and Reference methods
Spike
Stable
Measuremen
t Time
Range
Theoretical
Turbidity in
Sample in
NTU
Theoretical
Turbidity of
Spike in
NTU
FT660
Reference -
in NTU
PTV2000
Red 660 Test
- NTU
PTV 1000
WL Test -
NTU
PTV 6000
685 Laser
Test - NTU
Baseline
(Blank) 911 - 947 0.007 0.000 0.0000 0.0002 0.0008 0.0002
2.01 NTU
at 4 RPM 948 - 1033 0.021 0.014 0.0001 0.0002 0.0009 0.0002
2.01 NTU
at 8 RPM 1034 - 1111 0.035 0.028 0.0002 0.0003 0.0002 0.0002
2.01 NTU
at 16 RPM 1112 - 1146 0.067 0.060 0.0002 0.0003 0.0007 0.0008
16.0 NTU
at 4 RPM 1147- 1218 0.117 0.110 0.0005 0.0012 0.0014 0.0007
16.0 NTU
at 8 RPM 1219 - 1251 0.232 0.225 0.0014 0.0020 0.0021 0.0015
134 NTU
at 2 RPM 1252 - 1324 0.561 0.554 0.0036 0.0040 0.0051 0.0055
134 NTU
at 4 RPM 1325 - 1411 0.928 0.921 0.0079 0.0088 0.0094 0.0100
134 NTU
at 8 RPM 1412 - 1455 1.902 1.895 0.0069 0.0077 0.0081 0.0070
800 NTU
at 2 RPM 1456 - 1530 3.575 3.568 0.0077 0.0140 0.0267 0.0566
800 NTU
at 6 RPM 1531 - 1602 9.383 9.376 0.0277 0.0251 0.0313 0.0804
Averaged precision for each range (in bold
text) to be reported in the respective
candidate method.
Average 0-
0.100
0.0002 0.0003 0.0006 0.0004
Average
0.101 - 1.00
0.0034 0.0040 0.0045 0.0044
Average 1.01
to 10
0.0141 0.0156 0.0220 0.0480
Average 0-10
0.0056 0.0064 0.0086 0.0163
Accuracy (Bias) of the Lovibond WL LED Method.
Table 7 contains the calculated percent recoveries for each instrument regarding each turbidity spike. Table 7 is
split into two sections, with the left section containing the percent recovery data for the candidate Lovibond White
Light LED method and the right side of the table containing the percent recovery data for the reference method. The
data is generated from the last 60 measurements of a given spike, which is stated in the far right column of Table 7.
23
The percent recovery data is blank corrected for the highly filtered particle-free water. Both the reference and the
candidate technologies offer this blank subtraction feature in their respective instruments. The candidate instrument
will allow for a blank correction up to 0.05 NTU. The blank correction value for the PTV 1000 was 0.024 NTU in
this study. Historically, the estimated turbidity of turbidity free water was approximately 0.012 NTU when using a
polychromatic “white light” source, thus delivering the stray light error (which includes the combination of
molecular scattering and stray light) of approximately 0.012 NTU for the candidate method. The technical
specification on accuracy for the PTV 1000 instrument was up to 0.015 NTU for any measurement below 0.5 NTU.
In turbidity measurement, there is no true theoretical means to determine the true turbidity of a particle free sample.
Historically, only estimates have been made, much of which was from empirical measurements. True turbidity is a
function of the light source used, and with white light, that has been around 0.012 NTU. However, observations
with 660-nm wavelengths have seen turbidity measurements as low as 0.007 NTU when a linear algorithm is used
(this is the “b” value for a linear curve). This is what was used as the theoretical turbidity of the water in this
validation study. If using the same approach for a linear algorithm, it is probably not possible for a white light LED
system to read that low without some additional subtraction that would have to be applied.
The value for turbidity free water, when measured on the reference instrument is also 0.007 NTU, and the blank
value measured 0.010 NTU. The difference, of these two values was 0.003 NTU, which was the estimated stray
light of FilterTrak 660 Laser nephelometer.
The percent recovery data for each turbidity spike for this candidate instrument range between 96 and 111 percent,
with an average recovery being 99.6 percent over the 0-10 NTU test range. The reference instrument deliver a
slightly tighter recovery range of 97 and 103 percent, with an average recovery of 101 percent. The difference
between the two instruments fall within the accuracy specification for the candidate and the reference instrument.
This is further supported by an ANOVA analysis, which deliver an F value that is less than the F-critical value,
indicating there is no statistical significant difference between the two data sets. The percent recovery data summary
and same percent recovery data that is in Table 7 is presented in sections 13 and 17, respectively, of the candidate
Lovibond White Light LED method.
Table 7 – Results Table for the Percent Recovery and Precision with Respect to Turbidity Spikes for the PTV1000 WL LED
Turbidimeter
PTV 1000 WL Reading (NTU) Reference Turbidimeter Reading (EPA approved
Method 10133)
Spike
#
Baseline
(Blank) in
NTU
Theoretical
Value of
Spike in
NTU
Response
(blank
Corrected)
in NTU
Recovery
(%)
Baseline
(Blank) in
NTU
Theoretical
Value of
Spike in
NTU
Response
(blank
Corrected)
in NTU
Recovery
(%) N
1 0.024 0.014 0.016 110.8 0.010 0.014 0.014 102.3 60
2 0.024 0.028 0.030 104.9 0.010 0.028 0.029 102.6 60
3 0.024 0.060 0.062 102.4 0.010 0.060 0.060 99.4 60
4 0.024 0.110 0.114 103.4 0.010 0.110 0.112 102.0 60
5 0.024 0.225 0.235 104.5 0.010 0.225 0.232 103.3 60
6 0.024 0.554 0.564 101.7 0.010 0.554 0.559 100.8 60
7 0.024 0.921 0.940 102.1 0.010 0.921 0.930 101.0 60
8 0.024 1.895 1.929 101.8 0.010 1.895 1.928 101.8 60
9 0.024 3.568 3.444 96.5 0.010 3.568 3.463 97.1 60
10 0.024 9.376 9.437 100.6 0.010 9.376 9.472 101.0 60
24
Accuracy Results - PTV 1000 Test Accuracy Results - FT660 Reference
Average Recovery Entire Range (0-10
NTU)
102.87
Average Recovery Entire Range (0-
10 NTU) 101.13
Average Percent Recovery up to 0.10 NTU
Range
106.01
Average Percent Recovery up to
0.10 NTU Range 101.46
Average Percent Recovery 0.1 to 1.0 NTU
Range
102.93
Average Percent Recovery 0.1 to
1.0 NTU Range 101.77
Average Percent Recovery 1.0 to 10.0 NTU
range
99.65
Average Percent Recovery 1.0 to
10.0 NTU range 99.94
Accuracy (Bias) of the Lovibond 660-nm LED Method
Table 8 contains the calculated percent recovery data for each instrument regarding each turbidity spike. Table 8 is
split into two sections, with the left section containing the percent recovery data for the candidate Lovibond 660-nm
LED method and the right side of the table containing the percent recovery data for the reference method. The data
for each spike is generated from the last 60 measurements of a given spike, which is stated in the far right column of
Table 8.
The percent recovery data is blank corrected for turbidity free water. Both the reference and the candidate
instruments offer the instrument feature for blank correction. The candidate instrument will allow for a blank
correction up to 0.05 NTU. The blank correction value for the PTV 2000 was 0.013 NTU in this study. Using the
value of 0.007 NTU as the turbidity of the blank (the Fort Collins filtered water), the difference between this
turbidity and the PTV 2000 measurement of 0.013 NTU can be attributed to primarily to stray light. This would
calculate to be 0.006 NTU. Thus, the blank value is the sum of the molecular scatter caused turbidity (0.007 NTU)
plus the stray light (0.006) NTU to generate a total measurement value of 0.013 NTU. In summary, the blank
subtraction does not subtract out the instrument stray light, but only the amount of theoretical turbidity in the particle
free blank.
The value for turbidity free water, when measured on the reference instrument was also 0.007 NTU, and the blank
value measured 0.010 NTU. The difference, of these two values was 0.003 NTU, which was the estimated stray
light of FilterTrak 660 Laser nephelometer.
The percent recovery data for each spike for the candidate instrument range between 93.7 and 104 percent among
the 10 different turbidity spikes, with an average recovery being 101 percent over the 0-10 NTU test range. The
reference instrument delivers a slightly tighter recovery range of 97 and 103 percent over the same ten turbidity
spikes, with an average recovery of 101 percent. The difference between the two instruments falls within the
accuracy specifications for both the candidate and reference instrument (2 percent of reading or 0.010 NTU for both
the candidate and the reference instrument). This minor difference is further supported by an ANOVA analysis,
which delivers an F value that is less than the F-critical value, indicating there is no statistical significant difference
between the two data sets. The percent recovery data summary and the data for the candidate method that is in Table
8 is presented in sections 13 and 17, respectively in the candidate Lovibond 660-nm LED method as the bias data.
Table 8 – Results Table for the Percent Recovery to Turbidity Spikes for the PTV2000 660-nm LED Turbidimeter
PTV 2000 660-nm LED Reading (NTU) Reference Turbidimeter Reading (EPA approved
Method 10133)
Spike #
Baseline
(Blank) in
NTU
Theoretical
Value of
Spike in
NTU
Response
(blank
Corrected)
in NTU
%
Recovery
Baseline
(Blank) in
NTU
Theoretical
Value of
Spike in
NTU
Response
(blank
Corrected)
in NTU
%
Recovery N
25
1 0.013 0.014 0.015 104.5 0.010 0.014 0.014 102.3 60
2 0.013 0.028 0.029 104.3
0.010 0.028 0.029 102.6 60
3 0.013 0.060 0.061 101.6 0.010 0.060 0.060 99.4 60
4 0.013 0.110 0.112 101.4 0.010 0.110 0.112 102.0 60
5 0.013 0.225 0.232 103.3 0.010 0.225 0.232 103.3 60
6 0.013 0.554 0.553 99.7 0.010 0.554 0.559 100.8 60
7 0.013 0.921 0.918 99.7 0.010 0.921 0.930 101.0 60
8 0.013 1.895 1.858 98.1 0.010 1.895 1.928 101.8 60
9 0.013 3.568 3.344 93.7 0.010 3.568 3.463 97.1 60
10 0.013 9.376 9.304 99.2 0.010 9.376 9.472 101.0 60
Accuracy Results - PTV 2000 660-nm Test Accuracy Results - FT660 Reference
Average Recovery Entire Range (0-10
NTU)
100.55
Average Recovery Entire Range (0-
10 NTU)
101.13
Average Percent Recovery up to 0.10 NTU
Range
103.47
Average Percent Recovery up to 0.10
NTU Range
101.46
Average Percent Recovery 0.1 to 1.0 NTU
Range
101.03
Average Percent Recovery 0.1 to 1.0
NTU Range
101.77
Average Percent Recovery 1.0 to 10.0 NTU
range
97.00
Average Percent Recovery 1.0 to
10.0 NTU range
99.94
Accuracy (Bias) of the Lovibond 6000 Laser Method
Table 9 contains the calculated percent recoveries for each instrument regarding each turbidity spike. Table 9 is
split into two sections, with the left section containing the percent recovery data for the candidate Lovibond 6000
Laser method, which is denoted as the PTV6000 685-nm Laser turbidimeter in this table. The right side of this table
contains the percent recovery data for the reference method. The data for each spike is generated from the last 60
measurements of a given turbidity spike, which is stated in the far right column of Table 9.
The percent recovery data is blank corrected for turbidity free water. Both the reference and the candidate
instruments have a blank correction feature. The candidate instrument will allow for a blank correction up to 0.05
NTU. The blank value for the PTV 6000 was 0.012 NTU in this study. Historically, the estimated turbidity of
turbidity free water was approximately 0.007 NTU when using a 660-nm monochromatic source, thus delivering the
stray light error of approximately 0.005 NTU for this candidate method. The technical specification of this PTV
6000 instrument was up to within 0.010 NTU for any measurement below 0.5 NTU.
The value for turbidity free water, when measured on the reference instrument was also 0.007 NTU, and the blank
value measured 0.010 NTU. The difference, of these two values was 0.003 NTU, which is the estimated stray light
of FilterTrak 660 Laser nephelometer.
This information helps explain how the different candidate instruments compare at the bottom end of the turbidity
range as well as provide for a comparison to the reference instrument. As can be seen with the comparison between
the PTV 6000 and the FilterTrak 660 is 0.002 NTU at the bottom of the range. From a practicality perspective, this
value is so small it is very difficult to quantify and essentially provides equivalency to each between these two
instruments.
The percent recovery data for each spike for the candidate instrument range between 94.3 and 103 percent among
the 10 different turbidity spikes, with an average recovery being 100 percent over the 0-10 NTU test range. The
reference instrument deliver a slightly tighter recovery range of 97 and 103 percent over the same ten turbidity
26
spikes, with an average recovery of 101 percent. The difference between the two instruments falls within the
accuracy specifications for both the candidate and reference instrument (2 percent of reading or 0.010 NTU for both
the candidate and the reference instrument). This minor difference is further supported by an ANOVA analysis,
which delivers an F value that is less than the F-critical value, indicating there is no statistical significant difference
between the two data sets. The percent recovery data summary and the percent recovery data for the candidate
method found in Table 9 is presented in sections 13 and 17, respectively in the Lovibond 6000 Laser candidate
method as the percent recovery data.
Table 9 – Results Table for the Percent Recovery to Turbidity Spikes for the PTV6000 685-nm Laser Turbidimeter
PTV 6000 685-nm Laser Reading (NTU) Reference Turbidimeter Reading (EPA approved
Method 10133)
Spike #
Baseline
(Blank) in
NTU
Theoretical
Value of
Spike in
NTU
Response
(blank
Corrected)
in NTU
%
Recovery
Baseline
(Blank) in
NTU
Theoretical
Value of
Spike in NTU
Response
(blank
Corrected)
in NTU
%
Recovery N
1 0.012 0.014 0.015 103.4
0.010 0.014 0.014 102.3 60
2 0.012 0.028 0.029 101.8
0.010 0.028 0.029 102.6 60
3 0.012 0.060 0.060 99.2
0.010 0.060 0.060 99.4 60
4 0.012 0.1100 0.111 100.6
0.010 0.110 0.112 102.0 60
5 0.012 0.225 0.230 102.3
0.010 0.225 0.232 103.3 60
6 0.012 0.554 0.554 99.9
0.010 0.554 0.559 100.8 60
7 0.012 0.921 0.915 99.4
0.010 0.921 0.930 101.0 60
8 0.012 1.895 1.885 99.5
0.010 1.895 1.928 101.8 60
9 0.012 3.568 3.366 94.3
0.010 3.568 3.463 97.1 60
10 .012 9,377 9.419 100.4 .010 9.376 9.472 101.0 60
Accuracy Results - PTV 6000 685-nm Laser Test Accuracy Results - FT660 Reference
Average Recovery Entire Range (0-10
NTU)
100.08
Average Recovery Entire Range (0-
10 NTU)
101.13
Average Percent Recovery up to 0.10 NTU
Range
101.43
Average Percent Recovery up to
0.10 NTU Range
101.46
Average Percent Recovery 0.1 to 1.0 NTU
Range
100.55
Average Percent Recovery 0.1 to
1.0 NTU Range
101.77
Average Percent Recovery 1.0 to 10.0 NTU
range
98.09
Average Percent Recovery 1.0 to
10.0 NTU range
99.94
Tables 5 through 9 contain data that demonstrate initial comparability between each of the three candidate test
methods and the reference method with respect to precision and accuracy on defined turbidity spikes that were
derived at the Fort Collins site. The other two sites will demonstrate comparability with respect to filter runs at the
Binney South Platte and the San Patricio MWD test sites.
The Binney South Platte Comparability Test Site
The purpose of this portion of the study was to collect EPA comparability data for three candidate turbidity methods
on three different waters. The Binney South Platte site represents the second water type. The three candidate
methods were identical with the exception of the incident light source. The reference method was Hach Method
27
10133, which offers the FilterTrack 660 sc laser nephelometer as an instrument that is fully compliant to this test
method. This study compared newly developed Lovibond PTV 1000/2000/6000 turbidimeters that were
representative of the proposed turbidity methods to the EPA accepted reference method. A detailed description of
the test setup is found in the document "Alternative Test Procedure Method and Validation Plan, Revision 1.0", for
the PTV Series of turbidimeters evaluation. The validation plan is identified as Appendix A of this validation report.
This study plan required the collection of a sample from a common tap of filter effluent water. Sample was split
into five parallel streams at a manifold that fed a total of five instruments involved in this study. The fifth
instrument was added to the test panel and as a second EPA approved instrument that was very popular throughout
the drinking water industry. This instrument was compliant to EPA Method 180.1 and was known as the 1720E
turbidimeter. Due to its popularity in filter effluent monitoring, the 1720E was included in the monitoring and any
information presented herein is simply for additional information. Data collected from this instrument is completely
excluded from all calculations that are used in this ATP study.
The length for each sample line was the same for each instrument. The flow rate of sample to each candidate
instrument was the same, at about 70 ml/minute. This was within the manufacture's recommendations. The flow
rate for the reference instrument was about 350-ml/minute, which was within that manufacturer's recommendations.
After calibration and QC was performed, the instruments measured the turbidity of the sample for approximately 3
days. Data was automatically logged from each instrument at 15-second intervals into a common data logger for all
the instruments. This data is contained in a Microsoft Excel Workbook titled Aurora Binney Filtration SP Treat
Train Test Site.xls, which was provided external of this validation report. This workbook contains several
worksheets and graphs which are described in Table 10. Note: spreadsheets, pages, and worksheets were
synonymous with respect to the workbook discussion in Table 10.
Table 10 – A Description of the Excel Workbook for the Aurora Binney South Platte Phase of this Method
Comparability Study
Excel Page
Description
Exec Summary This is the Executive Summary for the test site. It briefly describes the study, the results and the
conclusions. The results and conclusions are mirrored into this validation report.
Reporting The reporting spreadsheet contains the final tables that directly compare the three candidate
methods to the reference method. The basis for the comparison is on the average value and standard
deviation for each of the methods for a given block of data. A final table at the bottom of this
spreadsheet lists the net difference in turbidity between each of the candidate methods and the
reference for each of the data blocks and the data over the entire study.
Checklist This is a checklist to help the site coordinators insure that all the details that pertain to the setup,
collection of data and QC are addressed. Most of the details are discussed in Section 7 of the
validation plan.
Quality
Assurance
This spreadsheet contains information that pertains to the calibration and QC for the study site.
Raw Data Raw Data – This worksheet contains the original raw data from the study. The date/time is in the
first column, followed by the data from the 1720E turbidimeter in column B, reference FT660
instrument data in mNTU in column C, rescaled reference FT660 instrument data in NTU in column
D, the PTV 2000 Red LED in Column F, the PTV 1000 WL LED in Column G, followed by the
PTV 6000 Laser in Column H. This raw data overlaps the actual study data from a time perspective
in that it contains data prior to calibration.
Data EPA This is the most important page of the Excel spreadsheet with respect to the calculations. The
Raw Data columns are resorted, with columns A-F containing: Date and Time, FT660 in NTU, PTV
2000 Red LED in NTU, PTV 1000 WL LED in NTU, PTV 6000 Laser in NTU and the 1720E in
NTU respectively. To the right of this data are the tables that contain the summarized average and
standard deviations for these instruments. Measurement data is separated into 8-hour blocks with
the exception of the last block of data which contain the remaining measurements. There is also a
data table that contains all the data from the study and another data table that incorporates the small
spike of turbidity that took place near the end of the study. The final table on this spreadsheet is a
compilation of the net differences in the averages between different instruments for each of the data
28
blocks.
Turbidity Plot This is the summary graph that shows the start and stop times of valid data that was collected and
analyzed. It is pasted into this validation report as Figure 3.
Graph of
Turbidity Spike
This graph is generated to cover the spiking of raw water into the sample prior to the manifold that
split and fed it to the test instruments. The graph illustrates instrument responses and subsequent
recoveries relative to these spikes. This is pasted into this validation report at Figure 4.
Photos This page contains the photographs that were taken during this portion of the study. Most of the
photographs are with respect to the calibration and QC portion of the validation study.
Figure 3 provides a graphical illustration of all the data logged at this test site. The instrument response to turbidity
is on the y-axis and the date and time is on the x-axis. The y-axis is scaled to a maximum of 0.3 NTU, which is the
reporting limit for filter effluent turbidity. The legend for this graph is at the bottom and contains traces for the three
candidate methods, the FT660 reference and the 1720E instrument. The 1720E trace was only provided for
empirical informational purposes only and was not included in any data or calculations. The start and stop times of
data collection is identified by two vertical red lines on the graph. Data outside of these lines is not part of the
comparability analysis.
The run time of this study incorporates three complete filter runs, which included ripening, the production run and
backwash phases. The turbidity ranged from less than 0.05 NTU up to approximately 0.1 NTU. It was worth noting
that within a given filter run, the turbidity variance for the laser instruments (both the Reference and the PTV 6000)
increased as the run progressed. This increase was thought to be caused by very low numbers of particles that
eventually pass through the filter and are detected within the small, but high incident beam density view volumes
that are characteristic with laser based turbidimeters.
Figure 3 illustrates that the three candidate instruments tracked the reference instrument throughout the study. There
are slight differences between the technologies, primarily with the Lovibond White Light LED method (represented
by the PTV 1000 WL turbidimeter, which is the green trace). The measurements are consistently between 0.01 and
0.02 NTU above the reference value, which is primarily due to the inherent stray light that is characteristic of a
polychromatic light source. The Lovibond 660-nm LED method and the Lovibond 6000 Laser methods tracked
very closely to the reference turbidimeter, but as each filter run progressed, these two instruments exhibited
increased sensitivity to turbidity. Table 11 provided a summary of the net differences between each of the candidate
methods and the reference method at this test site.
With respect to tracking the turbidity events, all instruments track each other in the detection and trending of all
events. No instrument is out of sync of the other methods by more than 1-minute. Considering the reporting
requirement of 15-minute intervals, this difference in responsiveness to turbidity changes was very minor.
29
Figure 3 - Graph of on-line turbidity data for the three candidate methods and the reference method. Data was logged at
15-second intervals. QC validated study data is between the vertical red lines.
The last day of the study involved the spiking of raw water into the effluent sample line that led to the three
candidate and reference instruments. The injection was before the manifold that splits the sample. This yielded the
turbidity spike. The spike was analyzed to insure the three candidate technologies detected the excursion and then
the recovery back to the baseline. Figure 4 provides a zoomed in portion of the on-line trending graph that
illustrates the response of the three candidates and reference instruments to this turbidity excursion. The graph
illustrates the response and recoveries of all the instruments that measured this water. The candidate PTV 6000
turbidimeter shows the greatest absolute response, which is expected from a laser based light source. This is
followed by the candidate PTV 2000 turbidimeter. The PTV 1000 WL turbidimeter shows the highest turbidity, but
it also has the highest baseline. However, the net difference is comparable to the reference instrument. The graph
also shows the 1720E instrument. This instrument measures very close to the PTV 1000 white light instrument, but
it shows a slower response time, because its flow rate was at 100-ml per minute, which was below the recommended
250 ml/minute flow range of that instrument.
30
Figure 4 - Graph of the candidate and reference method instruments’ response to a turbidity spike of raw water into the
filter effluent sample line.
Table 11 contains the net difference in turbidity between the candidate methods and the reference method
(Difference = Candidate turbidity-Reference turbidity) for each of the data sets. The greatest difference is with the
data set for the spike of turbidity, which shows an increase in the net turbidity difference for each of the candidate
methods. The differences demonstrate the sensitivity of these methods to a turbidity spike, such as the one that was
performed at this study. The other 8 data blocks display relatively small net differences relative to the reference
instrument. The entire data set, 16157 measurements is used for final comparisons between each of the three
candidate instruments and the reference instrument.
The candidate Lovibond White Light LED method instrument measured 0.019 NTU higher than the reference
method instrument over the entire study. As mentioned previously, this is primarily due to the inherent stray light of
a polychromatic light source that was in the Lovibond instrument. It could also be due to this light source having
greater sensitivity to smaller particles, since a portion of the emitted light was in the 400-600 nm range. In addition,
dissolved compounds can potentially cause fluorescence effects upon the interaction of the lower wavelengths of
light from the white light LED source. However, this candidate method still tracks the reference method throughout
the study that involved several complete filter runs and the spike of turbidity.
The candidate Lovibond 660-nm LED method measured 0.009 NTU higher than the reference method over the
16157 data points that were logged at this test site. The difference was very small and was likely attributed to stray
31
light. This candidate method tracked the reference method throughout the study that included several filter runs and
the spike of turbidity.
The candidate Lovibond 6000 Laser method measured 0.009 NTU higher than the reference method at this test site.
This instrument has essentially the same stray light, but as each filter run progressed, it showed additional response
to the trended increases in turbidity. This was further demonstrated on the turbidity spike, where the candidate
method demonstrated the greatest response to this spike. The candidate method tracked the reference method
throughout the study that included several filter runs and the spike of turbidity.
Table 11 - Net Difference Relative to the Reference (FT660) at the Binney South Platte
Treatment Train. Data in Red Was Reported in the Respective Candidate Methods.
Data Set
PTV 2000 660-
nm Test in NTU
PTV 1000
Test in NTU
PTV 6000 685
nm Laser in
NTU N
Data Block
1 0.006 0.018 0.005 1918 Data Block
2 0.016 0.023 0.018 1918 Data Block
3 0.006 0.014 0.003 1918 Data Block
4 0.010 0.017 0.009 1918 Data Block
5 0.008 0.017 0.009 1918 Data Block
6 0.010 0.020 0.009 1918 Data Block
7 0.009 0.019 0.009 1918 Data Block
8 0.010 0.021 0.009 1918 Data Block
9 0.008 0.020 0.010 802 Entire Data
Set 0.010 0.019 0.009 16157 Data from
Raw Water
Spike 0.023 0.027 0.034 143
The three Lovibond candidate methods demonstrated excellent comparability to the reference turbidimeter at the
Binney South Platte Test Site. The site provided an excellent opportunity to observe filter performance as the plant
had just come on-line after several months of construction and maintenance. The plant was not in an optimized state
at the time of data collection. The plant management is very interested in this data and understood its value to help
optimize their treatment processes.
The San Patricio MWD Comparability Test Site
The San Patricio MWD membrane facility was the third water to be analyzed by the three candidate Lovibond
turbidity test methods and the reference method. The three candidate methods were identical with the exception of
the incident light source. The reference method was Hach Method 10133, which offers the FilterTrack 660 sc laser
nephelometer as an instrument that was fully compliant to this test method. This study compared newly developed
Lovibond PTV 1000/2000/6000 turbidimeters that are representative of the proposed turbidity methods to the EPA
accepted reference method. A detailed description of the test setup is found in the document "Alternative Test
Procedure Method and Validation Plan, Revision 1.0", for the PTV Series of turbidimeters evaluation that is
appended to this validation report.
32
This study plan required the collection of a sample from a common tap of membrane effluent water. Sample was
split into four parallel streams that fed the four instruments involved in this study. The length of each sample line
was the same to each instrument. The flow rate of sample to each of the three candidate instruments was the same,
at about 70 ml/minute. This was within the manufacture's recommendations. The flow rate for the reference
instrument was about 350-ml/minute, which was within that manufacturer's recommendations.
After calibration and QC was completed, the instruments measured the turbidity of the sample for about 18 hours.
Data was automatically logged from each instrument at 15-second intervals into a common data logger for all the
instruments. This data is contained in the Microsoft Excel Workbook titled San Pat MWD Test Site.xls, which was
provided external of this validation report. This workbook contains several worksheets and graphs which are
described in Table 12. Note: spreadsheet, worksheet, and page are synonymous with respect to Table 12.
Table 12 – A Description of the Excel Workbook From the San Pat MWD Phase of this ATP
Comparability Study
Excel Page
Description
Exec Summary This is the Executive Summary of the Study. It briefly describes the study, the results and
the conclusions. The results and conclusions are mirrored into this validation report.
Reporting The reporting spreadsheet contains the final tables that directly compare the three
candidate methods to the reference method. The basis for the comparison is on the
average value and standard deviation for each of the methods for a given block of data. A
final table at the bottom of this page lists the net difference in turbidity between each of
the candidate methods and the reference method for each of the data blocks and the data
over the entire study.
Checklist This is a checklist to help the site coordinators insure tall the details that pertain to the
setup, collection of data and QC were completed. Most of the details are discussed in
Section 7 of the validation plan.
Quality
Assurance
This spreadsheet contains information that pertained to the calibration and QC for the
study site.
Raw Data Raw Data – This spreadsheet contains the original raw data from the study. The date/time
is in the first column, followed by the reference FT660 instrument data in mNTU in
column C, rescaled reference FT660 instrument data in NTU in column D, the PTV 2000
Red LED in Column F, the PTV 1000 WL LED in Column G, followed by the PTV 6000
Laser in Column H.
Data EPA This is the most important spreadsheet in the Excel workbook with respect to the
calculations. The Raw Data columns are resorted, with columns A-E containing: Date
and Time, FT660 in NTU, PTV 2000 Red LED in NTU, PTV 1000 WL LED in NTU, and
the PTV 6000 Laser in NTU respectively. To the right of this data are the tables that
contain the summarized average and standard deviations for these instruments. Data is
separated into four Tables. Tables 1 and 2 contains data from the two turbidity spikes
(settled water), Table 3 contains the first 8 hours of run time, Table 4 contains the second
8 hours of run time, Table 5 contains remaining run-time data, and Table 6 contains the
entire data run (after calibration and verification). The final table, Table 7 contains the net
difference between each of the three candidate methods and the reference method for each
of the blocks of data. The cells in these tables contain the formulas for all these results.
Graph of CFE
Turbidity
This is the summary graph that illustrates the start and stop times of data that are
analyzed. It is pasted into the Validation Report as Figure 5.
Graph of
Turbidity Spike
This graph is generated to cover the spiking of settled water into the sample prior to the
manifold that split and fed sample to the three candidate and reference instruments. The
graph illustrates the response and recoveries for these instruments relative to these spikes
of settled water. This is pasted into this validation report at Figure 6.
Photos This spreadsheet contains the photographs that were taken during this portion of the study.
Most of the photographs are with respect to the calibration and QC portion of the
validation study.
33
The data that was collected at the San Patricio MWD Site is graphically illustrated in Figure 5. The instrument
response to turbidity is on the y-axis and the date and time is on the x-axis. The y-axis is scaled to a maximum value
of 0.3 NTU, which is the reporting limit for filter effluent turbidity. The legend for this graph is at the bottom and
contains the traces for the three candidate methods and the FT660 reference instrument. The start time for data
collection is identified by the red vertical line on the left part of the graph. Data that is to the left of this line was not
part of this study.
This study included two spikes of turbidity that were conducted early in the study. A sample of settled water, that
was approximately 5 NTU was collected and injected at two different fixed rates into the sample line prior to the
manifold that split it into the parallel streams. The spikes were intended to be in the 0.1 and then 0.2 NTU range.
After the completion of the spiking, the instruments were allowed to run for another 18 additional hours before data
collection was terminated.
In figure 5, the reference trace (in blue) is overlapped by the two candidate instrument traces for the Lovibond 660-
nm LED and the Lovibond Laser LED. These two candidate instruments demonstrated comparability throughout
the entire data set, including the turbidity spikes. The candidate Lovibond White Light LED method delivered an
overall higher response throughout the study, but still correlated closely to all the turbidity events, including the
turbidity spikes. The net difference between the candidate white light LED method was greater on this water, which
was also observed when grab samples were taken and measured on a benchtop instrument that was an EPA 180.1
compliant instrument (the Hach 2100N turbidimeter). While some of the difference was likely attributed to stray
light, the additional difference may be due to some compounds in the sample that exhibit a fluorescence effect was
not detected by the other instruments in the study.
Grab samples were taken from this site and examined back at the Fort Collins R&D facility. The examination was
to determine if light in the 410-410-nm range does cause a molecular fluorescence event. The results from this test
confirmed these effects do occur on this sample. Since the white light LED does emit light as low as 400-nm, this
fluorescence effect possibly does contribute to the higher turbidity on the San Patricio membrane filtered sample.
34
.
Figure 5 - Graph of the three candidates and the reference method for on-line measurement of turbidity on membrane
effluent. Data right of the vertical red line was QC validated for this test site.
The two turbidity spike levels that were conducted at the San Pat MWD test site is the focus of Figure 6. The graph
illustrates the response times for the three candidate and reference instruments. The graph illustrates that these are
comparable with respect to both spikes and their respective recoveries. The PTV 1000 LED net response is
equivalent on the first spike, but is slightly reduced on the second (higher) spike. This decrease may be due to some
absorbance of a portion of the incident white light at the higher turbidity level. The candidate Lovibond 660-nm
LED and the candidate Lovibond 6000 laser turbidimeter methods show near equivalent response relative to the
reference method instrument on these two spikes.
35
Figure 6 - Graph of the three candidates and the reference instrument for two different levels of spiked turbidity (settled
water) into the membrane filtrate stream.
The actual comparability data from the San Pat MWD test site is summarized into Table 13. This contains the net
difference in turbidity between each of the candidate method’s respective instrument and the reference method
(Difference = Candidate turbidity-Reference turbidity) for each of the data sets. The greatest difference is with the
data set for the two turbidity spikes, which shows an increase in the net turbidity difference for each of the candidate
methods. The greatest difference was with the candidate Lovibond White Light LED Method which measured
0.045 NTU higher than the reference. This was followed by the candidate Lovibond 660-nm LED Method which
measured within 0.010 NTU for both spikes. The candidate Lovibond 6000 Laser Method compared the closest to
the reference method with an average difference of 0.007 for the two spikes. All instruments were capable of
detecting the spikes with comparable response and recovery times.
The candidate Lovibond White Light LED method measured 0.045 NTU higher than the reference method over the
entire study. As mentioned previously, this is primarily due to the inherent stray light of a polychromatic light
source as one factor. It could also be that this light source exhibited greater sensitivity to smaller particles, since a
portion of the emitted light was in the 400-600 nm range. The higher turbidity values obtained in this method were
also observed on grab samples on a 180.1 compliant benchtop turbidimeter. Thus, there could be some additional
fluorescence component in this membrane effluent that caused the increased response that a polychromatic light
source can detect but was impacted by the other methods in this study. In general, this candidate method still tracks
the reference method throughout the study that involved several filter runs and the spikes of turbidity.
36
The candidate Lovibond 660-nm LED method measured 0.002 NTU higher than the reference method over the 4692
data points that were logged at this test site. The difference is very small and was likely attributed to stray light.
This candidate method tracks the reference method throughout the study that included several filter runs and the
spike of turbidity. This candidate method demonstrated comparability to the reference method for the entire
monitoring phase at this test site.
The candidate Lovibond 6000 Laser method measured 0.0003 NTU lower than the reference method at this test site.
This instrument has essentially the same stray light as the reference, and showed a very slight heightened response
to the spikes. Overall, this method shows comparable but slightly higher sensitivity to the turbidity events on this
membrane effluent.
Table 13 - Net Difference Relative to the FT660 at the San Pat MWD Membrane Plant
Data Set PTV 2000 660-
nm Test in NTU
PTV 1000 Test in
NTU
PTV 6000 685
nm Laser in
NTU
N
Data Block 1 0.011 0.042 0.007 23
Data Block 2 0.007 0.022 0.009 59
Data Block 3 0.002 0.045 -0.001 1923
Data Block 4 0.002 0.045 -0.001 1918
Data Block 5 0.002 0.047 -0.001 358
Entire Data Set 0.002 0.045 -0.000 4692
The San Patricio MWD test site challenged the three candidate methods to collect turbidity near the bottom of their
ranges in an extreme environment that included a very warm sample. All instruments responded to the turbidity
spikes. This membrane facility uses significant quantities of air to prevent membrane fouling and it was a
challenging application to be able to eliminate any interference from entrained air. In all cases, the test methods
demonstrated the ability to eliminate this interference and deliver comparability data that correlated strongly to the
reference method.
Limit of Detection (LOD)
Turbidity methods typically do not report detection limits due to challenges associated with maintaining stable low-
level turbidity standards and the ability to separate out stray light and the turbidity of pure water from any low-level
measurement. However, the limit of detection can be estimated and was calculated for the three candidate methods
and the reference method.
The precision data from the three lowest turbidity spikes that were performed at the Fort Collins test site were used
to determine the LOD for each of the methods. The precision value was multiplied by a factor of three to deliver an
estimated LOD based on the each of the three lowest turbidity levels. These three estimated LOD calculations were
then averaged to deliver the estimated LOD for the respective method. Table 14 provided the calculated estimated
LOD for each turbidity level. The bottom line in this table provided the averaged and reported estimated LOD value
for the each of the three candidates and reference method (in bold).
Table 14 – The Estimated Limit of Detection Estimate for the Candidate and Reference methods
Theoretical
Turbidity of Spike in
NTU
FT660 Reference
3* Precision
PTV 2000 Reference
3* Precision
PTV 1000
Reference 3*
Precision
PTV6000
Reference 3*
Precision
0.0141 0.0004 0.0006 0.0028 0.0007
37
0.0282 0.0006 0.0009 0.0007 0.0007
0.0603 0.0007 0.0010 0.0020 0.0025
LOD 0.0006 0.0008 0.0018 0.0013
The estimated LOD for the candidate Lovibond White Light LED was determined to be 0.0018 NTU. This method
stated the reporting criterion to the nearest 0.010 NTU at the lowest turbidity range, which this calculation
supported.
The estimated LOD for the candidate Lovibond 660-nm LOD was determined to be 0.0008 NTU. This method
stated the reporting criterion to 0.01 NTU for the lowest turbidity range, which this calculation supported. However,
the low estimated LOD for this method provides capability to support a water plant in its optimization of its
treatment and filtration processes.
The estimated LOD for the Lovibond 6000 Laser Method was determined to be 0.0013 NTU. This method stated
the reporting criterion to the 0.01 NTU for the lowest range, which this calculation supported. The low estimated
LOD and increased sensitivity that was demonstrated at the filter plant comparability studies provides capability of
the method’s application for the optimization of the treatment and filtration processes in drinking water plants.
3.1.3 Quality Controls
The quality controls are described in Section 6 of the validation plan. The plan required the use of one quality
control sample (QCS) that was a fresh prepared formazin standard. The QCS would be run on each instrument after
calibration. A second QCS sample would be run at the completion of data collection. The QCS samples would be
checked on a benchtop turbidimeter (the Hach 2100N or 2100 AN) to insure its preparation was correct. This
approach was to be performed at each of the three test sites.
One of changes to the validation plan was to run an additional QCS sample after calibration and to run an additional
QCS sample after the data run. Thus, a total of two QCS samples were prepared and run before and after data
collection at each site. The reason for this is with the difficulty in preparing low turbidity standards. This can be
very difficult at values below about 1 NTU. Difficulties with contamination and bubbles often lead to erroneous
measurements. Thus, we opted for a higher value at 0.6 NTU thinking that this would be easier to prepare in the
field. However, once we prepared that accurately, we decided to prepare a second one at 0.3 NTU. This was
selected as it was closer to the true operational limit for regulatory. Thus, instead of eliminating the 0.6 NTU we
just added the second standard. The demonstration that we could prepare these in the field was very important to
demonstrating the method performance where it is practiced.
It was mentioned in the descriptions of the three Microsoft Excel workbooks (one for each test site); each contains a
Quality Assurance spreadsheet. The pass fail criteria for the QCS samples was either 10% of the value of the
sample or 0.04 NTU for the Candidate Lovibond White Light LED method, or 0.03 NTU for both the candidate
Lovibond 660-nm LED method and the candidate Lovibond Laser Turbidimeter Method. If the instrument does not
pass the pass/fail criteria, then the standard should be checked on a calibrated benchtop turbidimeter to insure it is
correctly prepared. If it is and it fails, then the method is unacceptable for reporting purposes. The reference method
did not state a pass fail criteria in its QCS section, but did state that the linear calibration range (LCR) standards
should read within 0.025 NTU. This was assumed to be the pass fail criteria for the reference method.
As was mentioned above, low level standards are hard to prepare and even the commercially available stabilized
versions do change after preparation. Providing absolute P/F criteria allows for prepared sealed standards to meet
such as specification at low levels up to their respective expiration date. Preparation of these standards is very
38
susceptible to technique and makes the use of a P/F percentage difficult to achieve. Because this is a continuous
monitoring method, the combination of the instrument measurement error plus the preparation techniques (in
preparing the standard itself and using the standard) result in propagated error. At low levels, these errors will
typically bias high and will exceed percentage criteria. For example, at 0.1 NTU, a 10% limit is nearly impossible
to meet because of the turbidity of the diluent, contamination, bubbles generated during transfer, when combined
will contribute to a high positive bias. Thus, the absolute criteria only apply to low level solutions. Once the value
is above the level where the percent is greater than the absolute value then the specification is exclusively as a
percentage.
The difference in the absolute values between the WL LED and the 660-nm methods was the additional stray light.
The additional stray light that is inherent in a WL instrument will have difficulty passing if this pass/fail criterial is
not opened up to a larger level. Note if a higher standard was used, such at 5 NTU, this would not be an issue
because the effects of stray light are not present.
The QC data for the test sites is on the “Quality Assurance” spreadsheet in each of the Microsoft Excel workbooks.
Table 15 provides a summary of QC data for the Fort Collins Test site. These QCS values were adjusted for the
small additional turbidity of the dilution water, which was 0.010 NTU. Thus, the pass fail was relative to this
adjusted value. All instruments passed their respective QCS criteria at this site.
Table 15 - Fort Collins Filtered Tap Water Test Site - Wet Validation QCS - 1.01 NTU Formazin - Must be less than ± 10% to
Pass or ±0.04 NTU (WL LED) or ±0.03 NTU (660 LED or Laser)
Event (date)
FT660
Ref in
NTU % Err
PTV2000
Red 660
Test -
NTU % Err
PTV
1000
WL
Test -
NTU % Err
PTV 6000
685 Laser
Test -
NTU % Err
2100AN
Reading
in NTU
7/19/16 After Calibration 1.009 -0.09 1.025 1.49 1.044 3.37 1.023 1.29 1.01
7/20/16 Before formazin
Turbidity Spikes 0.999 -1.00 1.028 1.78 1.042 3.17 1.024 1.39 1.01
7/20/2016 Conclusion of
Formazin Turbidity Spikes 0.973 -3.71 1.051 4.06 1.026 1.58 1.012 0.20 1.00
Wet Validation QCS - 0. 3 NTU Formazin - Must be less than ± 10% to Pass or ±0.04 NTU (WL LED) or ±0.03 NTU (660 LED
or Laser)
7/19/16 After Calibration 0.308 -0.73 0.317 2.26 0.33 5.81 0.309 -0.32 0.34
7/20/16 Before formazin
Turbidity Spikes 0.305 -1.56 0.315 1.62 0.33 7.10 0.317 2.26 0.33
7/20/2016 Conclusion of
Formazin Turbidity Spikes 0.315 1.58 0.352 13.55 0.34 8.71 0.321 3.55 0.33
The quality control data from the Aurora Binney South Platte site is summarized in Table 16. Two standards were
run before and after data collection. One standard was prepared at 1.02 NTU and the second standard was prepared
at 0.62 NTU. These standards were prepared with water that had sat in their glass containers for several days and
measured a slightly elevated turbidity value when measured on the same 2100AN benchtop turbidimeter. All
instruments passed their respective method’s QC criteria.
Table 16 – Binney South Plate Comparability Site - Wet Validation QCS - 1.01 NTU Formazin - Must be less than ± 10% to Pass
or ±0.04 NTU (WL LED) or ±0.03 NTU (660 LED or Laser)
39
Event (date)
FT660
Ref in
NTU % Err
PTV2000
Red 660
Test -
NTU % Err
PTV
1000
WL
Test -
NTU % Err
PTV 6000
685 Laser
Test -
NTU % Err
2100AN
Reading
in NTU
7/15/16 After Calibration,
Before Data 10.30 1.96 1.033 2.28 1.047 3.66 1.062 5.15 1.02
7/18/16 After Comparison
Run on South Platte
Treatment Train Combined
Filter Effluent 1.029 1.89 1.022 1.19 1.050 3.96 1.063 5.25 1.02
Wet Validation QCS - 0. 6 NTU Formazin - Must be less than ± 10% to Pass or ±0.04 NTU (WL LED) or ±0.03 NTU (660 LED
or Laser)
7/15/16 After Calibration,
Before Data 0.608 -0.30 0.648 6.23 0.645 5.74 0.642 5.25 0.626
7/18/16 After Comparison
Run on Aurora Reservoir
Treatment Train Combined
Filter Effluent 0.656 7.52 0.638 4.59 0.657 7.70 0.640 4.92 0.658
Table 17 provided a summary of the quality control data from the San Patrico MWD water sample. This QCS
samples were prepared through the dilution of stock formazin with membrane effluent water. The membrane
effluent measured 0.015 NTU on their regulatory monitoring turbidimeters, which were the FT660 laser
nephelometers. We also measured the turbidity of the QCS samples on their laboratory turbidimeter, a 2100N. The
2100N measured a higher turbidity of this water. However, this benchtop turbidimeter did not have available QC
data, so instead the reporting turbidimeter (FilterTrak 660) was used to estimate the turbidity of the water. This was
factored into the theoretical value of the of the QCS samples.
The three candidate and the reference instruments passed both QCS samples before and after the data collection at
this test site. The candidate instrumentation comply with the QCS pass/fail criteria for their respective methods.
The reference instrument did measure the QCS sample slightly lower than expected, but it was within the 10 percent
of the P/F value of the QCS.
Table 17 – San Patricio MWD Comparability Site - Wet Validation QCS - 1.01 NTU Formazin - Must be less than ± 10% to
Pass or ±0.04 NTU (WL LED) or ±0.03 NTU (660 LED or Laser)
Event (date)
FT660
Ref in
NTU % Err
PTV2000
Red 660
Test -
NTU % Err
PTV
1000
WL
Test -
NTU % Err
PTV 6000
685 Laser
Test -
NTU % Err
2100AN
Reading
in NTU
7/27/16 After Calibration 0.950 5.90 1.009 0.10 1.038 -2.77 0.998 1.19 1.08
7/28/16 After Comparison
Run on MF Treatment
Train Combined Filter
Effluent 0.992 1.77 1.012 -0.20 1.025 -1.49 0.954 5.54 1.07
40
Wet Validation QCS - 0. 32 NTU Formazin - Must be less than ± 10% to Pass or ±0.04 NTU (WL LED) or ±0.03 NTU (660 LED
or Laser)
7/27/16 After Calibration 0.291 6.05 0.304 1.94 0.347 -11.94 0.304 1.94 0.34
7/28/16 After Comparison
Run on MF Treatment
Train Combined Filter
Effluent 0.301 2.75 0.308 0.65 0.350 -12.90 0.308 0.650 0.35
The passing of the QCS criteria for the three candidate methods provided validity of the data that was collected at
the three different test sites. Because the measurements data was close to theoretical values across all of the testing
and the error in preparation of each standard was not considered, this further insured the QC approach was
appropriate for each of the candidate methods.
3.1.4 Precision and Accuracy
The precision and accuracy data was initially presented in Section 3.1.2 of this report. Table 5 provides the response
data to each of the 10 turbidity spikes for the three candidate methods. Least squares analysis was performed on the
linear relationship between each instrument’s response and the theoretical turbidity for each turbidity level. The
three candidate methods demonstrated high linearity with r-squared values that exceeded 0.999 over the test range of
0 to 10 NTU. The reference instrument also demonstrated comparable performance, which was indicative of a
sound test approach for precision and bias determination.
Table 6 provides a summary of the precision for the three candidate methods and the reference method. The
precision is comparable between the methods and was further summarized for each of three ranges of turbidity: 0 to
0.100 NTU, 0.100 to 1.00 NTU and 1.00 NTU to 10 NTU. This summary is at the bottom of Table 6.
Tables 7 through 9 provide a side-by-side comparison between each of the three candidate methods and the
reference method. Table 7 provides comparison data between the Lovibond White Light LED method and the
reference method. Table 8 provides comparison data between the Lovibond 660-nm LED Method and the reference
method. Table 9 provides the comparison data between the Lovibond Laser Method and the reference method.
Table 14 provides limit of detection information to further support the defined reporting limits that were stated in
Section 12 of each respective candidate method (see Appendices B.1, B.2 and B.3).
3.2 Holding Time / Storage Stability
The candidate and reference methods are on-line or continuous monitoring methods. Therefore, holding times are
not applicable. Further, turbidity samples are typically not stable and should be analyzed upon collection. With
respect to turbidity standards and QCS, they were prepared immediately before use and measured within 1 hour of
preparation to insure no degradation of the standards.
4.0 Data Analysis and Discussion
41
This section is broken down into two sections. The first section discusses how the data was analyzed and references
the tables that were provided in the demonstration of capability section. The second part discusses the comparison
between the alternate method data for each of the candidate methods to the reference method. Each candidate
method is individually compared to the reference method.
Data Analysis
The precision and bias data was analyzed from the ten different formazin turbidity levels that were generated at the
Fort Collins site. Prior to spiking the formazin, the filtered water sample was analyzed for approximately 30-
minutes prior to commencing the first spike. This baseline served as the blank for that respective instrument. The
filtered water was then spiked with known quantities of a defined formazin standard solution into a sample stream
with a known flow rate. This allowed for the calculation of the theoretical turbidity value. As each turbidity spike
was perfumed, the data was plotted (trended) graphically in real time. The initial ramp up of the turbidity
measurements (trace) was observed for each of the test instruments. Once all these traces were stable (trending
horizontally), the data collection count was noted. At this point of a stable turbidity level in the sample stream, the
injection continued at the constant rate until at least 60 measurements were logged for each of the instruments.
These 60 measurements were then used in the precision and bias calculations for each of the different turbidity
levels (i.e. spikes).
The derivation of the theoretical value for each spike was from the calculations that were provided in Section 6 of
the validation plan, under the section titled “Spike Injection Method”. This section described how a defined
turbidity standard was spiked into the flowing sample of filtered tap water. Equations 1 through 3 provide the
calculation of the dilutions that are performed on the spiked standard to ultimately yield a stable level of known
turbidity that ultimately flowed through each instrument. The Microsoft Excel workbook for the Fort Collins test
site contains a spreadsheet titled “Injection Summary” (see Table 4) that includes the embedded formulas used to
yield the theoretical turbidity of each spike.
A given spike, that was stable, contained 60 measurements. The mean, standard deviation and percent relative
standard deviation were determined from these measurements. These results were used as the reported value for the
given spike. This was performed for a total for 10 spikes (i.e. turbidities) that covered the range of 0.01 to 10 NTU.
The majority of the spikes were in the 0.010 to 1 NTU range, as that range of turbidity was most critical for
reporting and filtration optimization purposes. Each spiked turbidity level had the calculated theoretical value that
would be used for percent recovery calculations.
Percent recovery calculations were determined for each candidate and the reference instrument. The results are
provided in Tables 7, 8 and 9 for the Lovibond White light LED method, the Lovibond 660-nm LED method, and
the Lovibond Laser Method respectively. The percent recovery calculations were presented in Section 6 of the
validation plan (last paragraph). The analysis was performed by dividing the test instrument response by the
theoretical value of the response for each spike. This was performed on blank corrected and non-blank corrected
responses. However, both the candidate and test instruments have blank correction features and thus, the blank
corrected values were used in the reported percent recoveries. Percent recoveries were analyzed for discrete
turbidity ranges that were between 0.00 and 0.100 NTU, 0.100 and 1.00 NTU and between 1.0 and 10 NTU. For
each of these ranges, the averaged recovery was calculated and reported for in the respective method for each of the
three candidates.
Precision was calculated through the derivation of the standard deviation for each of the spiked turbidity levels.
Table 6 provided the calculated precision for the three candidates and reference methods. Consistent with the
percent recovery data, the precision was categorized in the ranges of 0.00 and 0.100 NTU, 0.100 and 1.00 NTU and
between 1.0 and 10 NTU. For each of these ranges, the averaged precision was calculated and reported for in the
respective method for each of the three candidates.
42
The analysis of linearity was determined for the three candidates and reference instruments (see Table 5). For each
instrument, a least squares regression line was derived between the theoretical turbidity and the averaged response
for the ten spikes. The r-squared value could then be used to determine if the respective response instrument’s
response was compliant to the linear response over the test range of 0 to 10 NTU.
In calculating the correlation coefficient from 0-1 NTU, all the instruments had a value of 0.9999 or better. The
accuracy degraded from less than 1 percent to less than 2 percent over this range on the PTV 1000. The PTV 2000
and 6000 retained accuracy that was better than 1 percent over this range, and is slightly better than the reference.
The calculations are in Table 1B on the tab “Fzn Spike and Recovery Data” that is in the Microsoft Excel Work
Book titled Ft Collins Filt Tap P&B Test Site.xls.
An additional analysis was performed on the percent recovery data. A statistical ANOVA analysis was performed
on 1) the entire averaged response data set for all instruments; 2) the blank corrected and the non-blank corrected
percent recovery data for all instruments; and 3) the separated data sets between each of the candidate and reference
instruments. With respect to the ANOVA analysis, if the F value is less than the F-critical value, the data sets are
not statistically different relative to each other. In other words, the differences between the data sets are likely to be
by chance. ANOVA results for these different groups of data were consistently the same. The F values were always
less than the respective F-critical value, indicating there was no statistical significance between any of the three
candidate methods and the reference method.
The analysis of the data from the two drinking water plants was broken down into 8-hour blocks of time (when
appropriate), which represented approximately 1920 data points per instrument per 8-hour block. Any remaining
data that did not add up to a full 8 hours were also analyzed. Additionally, any spikes of unfiltered water into the
effluent stream were analyzed separately. Finally, the entire run at the test site was analyzed as a whole. For each
data set, the average (mean), standard deviation and percent relative standard deviation was determined. This data
was then pasted into a table so the different blocks of data could be compared with respect to the average and
standard deviation. After review of the blocks of data, the means compared closely to each other for a given site.
Thus, for reporting purposes, the entire run was reported for each test site. When comparing the three candidates to
the reference methods, the net difference in turbidity between each candidate and the reference method was
calculated and reported. This analysis approach was performed for both the Binney South Platte and the San Pat
MWD waters. This data are presented in Tables 11 and 13 respectively.
Data comparison between the Lovibond White Light LED method and the EPA approved Method
10133
Table 18 provides a comparison between these two methods over the test range of the candidate method. The key
performance parameters for these methods include the range of percent recoveries for the test range of 0 to 10 NTU,
the average precision over the test range of 0-10 NTU, the linearity, limit of detection, and the reporting limit for
each method. The two right columns compare the net measurement difference between the two methods for each of
the two different drinking water plant filter effluent samples.
The range of percent recoveries is very close between the two methods with the candidate White Light LED method
exhibiting a slightly higher percent recovery at the lower turbidity spikes. The cause of this deviation is primarily
due to the additional stray light that was present in this candidate method. However, the instrument does have a
blank subtraction feature (which was not used during testing); the stray light impacts could be minimized. An
ANOVA analysis was performed between these two methods on the percent recovery data and showed there is no
statistical significance between them (F was less than F-critical). The linearity of each method was very high, with
43
correlation coefficients both exceeding 0.999. The precision of the two methods is within 0.003 NTU, with the
candidate method having slightly poorer precision.
Although not stated in the candidate or reference method, an estimated limit of detection was performed by taking
the averaged precision for the lowest three turbidity spikes for each method. This value was multiplied by a factor
of 3 to generate a conservative estimated LOD. The estimated limit of quantitation (LOQ) could be determined
using the same precision and it was below the method reporting level at its lowest turbidity range. It was calculated
as 10 times this average precision from these lowest three turbidity spikes and provides an estimated LOQ.
The robustness of the candidate and reference method was compared at the two drinking water sites. At both sites
the overall turbidity response of the approved reference method was lower than the candidate method. The
additional response by the candidate method was likely due to three factors. The first was additional stray light
exhibited in the candidate method and second, the candidate method does utilize a light source with lower-
wavelengths of light that scatter light more efficiently by small particles. Third, the lower wavelengths of incident
light may exhibit fluorescence effects from dissolved residual compounds that may be present in the sample. At the
San Patricio test site, the filtration of the sample is through an absolute barrier that passed integrity testing. The
sample has virtually no turbidity. Thus, the elevated readings that were only observed with the white light LED
instruments indicate this fluorescence effects may the major contributor to the elevated readings, along with some
elevated stray light. At the San Patricio test site, the reference instrument was measuring at approximately 10
mNTU during the studies which is indicative of no turbidity In addition, there could be some fluorescent effects
from dissolved materials that result with the with light sources at the lower wavelengths that are near 400 NTU. The
concern with the white light method is that stray light is very difficult to quantify and the contribution of stray light,
dissolved particle fluorescence, it was the combination of these factors that resulted in the slightly elevated turbidity
levels in the candidate method’s instrument. In practice, the positive bias of candidate method relative to the
reference method provided a conservative approach relative to reporting limits in that it was favorable to have a
false positive bias versus a false negative bias. The positive bias is an offset value that was not impacted by the
spike events that were directed in these studies in that the absolute response was observed for all spikes and was
comparable to the reference method.
The three test sites provided the opportunity for the methods to demonstrate their respective responsiveness to
turbidity spikes and recoveries from turbidity spikes. The candidate method detected all turbidity spikes that were
detected by the reference method, which was indicative of the robustness of the candidate method.
The quality controls that were followed in this ATP demonstrated that they were appropriate for the candidate test
method as there were no data deviations or events that could be traced to any inadequacy of the QC approach that
was utilized.
Table 18 – Summary of Method Performance between the Lovibond White Light LED and the Reference method 10133
Method % Recovery
Range (0-10
NTU)
Average
Precision (0 to
10 NTU)
Linearity
(0-10 NTU
Limit of
Detection
(Estimate)
Method
Reporting
(lowest
range)
Net
Difference
Binney (All
Data)
Net
Difference
San Pat
MWD (All
Data
Lovibond WL LED
Candidate
96.5 – 110.8 0.0086 0.99962 0.0018 0.01 +0.027 +0.045
10133 (Approved) 97.1 – 102.3 0.0056 0.99972 0.0006 0.01 -0.027 -0.045
At the time of this ATP, the reference method was commonly regarded as the state-of-the art turbidity method in the
industry. The FilterTrak 660 Laser Nephelometer was commonly used to measure low-level turbidity samples and
44
was considered to be an effective tool that could be used in the optimization of filtration processes. However, the
technology does have drawbacks that have prevented its broader use in drinking water filter effluent applications.
This was primarily due to expensive optical components (and instrumentation), difficulties in maintenance, and
difficulties in performing calibrations. These are unique challenges that do not always translate back to the
statistical analysis of the method but should be considered when a method (and its representative instrumentation)
was considered for use. The candidate Lovibond White Light LED method has been designed to deliver
representative instrumentation that resolves many of the challenges that were mentioned with the reference method’s
instrumentation. The white light source was well known and users have requested a better, more stable light source,
which candidate method provided. Although this candidate method does not statistically meet to the high level of
performance of the reference method in all statistical categories, it does deliver the combination of features and
performance that will allow users to adopt its respective instrumentation to ultimately deliver reliable turbidity
results over time.
Due to the uniqueness of the candidate method to be susceptible to the combination of: fluorescence effects from
dissolved compounds within some samples, higher scatter efficiency from small particles, and additional stray light
from a polychromatic light source, provisions have been incorporated into this method to address this interference.
These include the instructions on how to determine the blank value of the sample that can then be subtracted from
the measurements, which would be applicable to the measurement of samples below 0.100 NTU.
Data comparison between the Lovibond 660-nm LED method and the EPA approved Method 10133
Table 19 provides a comparison between these two methods over the range of the candidate method. The key
performance parameters for these methods include the range of percent recoveries for the test range of 0 to 10 NTU,
the averaged precision over the test range of 0-10 NTU, the linearity, limit of detection, and the reporting limit for
each method. The two right columns compare the net measurement difference between the two method’s
representative instrumentation for each of the two different drinking water plant filter effluent waters.
The range of percent recoveries are very close between the two methods with the candidate Lovibond 660-nm LED
method exhibiting a slightly higher (2%) percent recovery at the lower turbidity spikes. The cause of this deviation
was primarily due to the slightly higher stray light that was present in the candidate method’s instrument. However,
the candidate and reference instruments have a blank subtraction feature (which was not used during testing), and
the stray light impacts can be minimized. An ANOVA analysis was performed between these two methods on the
percent recovery data and it concluded there was no statistical significance between them (F was less than F-
critical). The linearity both the candidate and reference methods were high and essentially equivalent, with their
respective correlation coefficients exceeding 0.999. The precision of these two methods was within 0.0008 NTU
which was essentially equivalent with respect to each other.
Although not stated in either the candidate or the reference method, an estimated limit of detection was performed
by taking the averaged precision for the lowest three turbidity spikes for each method. This value was multiplied by
a factor of 3 to generate a conservative LOD. The LOD of the Lovibond 660-nm LED method was 0.0003 NTU
higher than the reference method. The limit of quantitation (LOQ) was determined using these same precision
values as the LOD. The LOQ was under the method reporting level at its lowest turbidity range.
The robustness between the candidate and reference methods was compared at the two drinking water sites. Both
sites exhibited the overall turbidity response of the approved reference method was slightly lower than the turbidity
response of the candidate method by less than 0.010 NTU. The higher response by the candidate method was likely
due to slightly elevated stray light, but this error would be captured in the instrument specification for the candidate
method’s representative instrument. In practice, the slight positive bias of candidate method over the reference
method provides a conservative approach relative to reporting limits in that it was more favorable to have a false
45
positive bias versus a false negative bias. This is because in turbidity there is always positive bias due to the
interferences discussed in this report. Some manufacturers attempt to subtract out the interference through the use of
measuring a very low turbidity calibration standard, which is both difficult to prepare measure. This approach
adjusts the calibration gain and not simply the offset. A more conservative approach which is practiced by these
methods is to design a highly linear measurement system where the dependence on the preparation of low turbidity
standards is not require to perform a successful and accurate calibration.
The three test sites provided the opportunity for these two methods to demonstrate their respective instrument’s
responsiveness to turbidity spikes and recoveries from these turbidity spikes. The candidate method’s instrument
detected all turbidity spikes that were detected by the reference method’s instrument, which was indicative of the
robustness of the method.
The quality controls that were followed in this ATP demonstrated that they were appropriate for the candidate test
method as there were no data deviations or events that could be traced to any inadequacy of the QC approach that
was followed.
Table 19 – Summary of Method Performance between the Lovibond 660-nm LED and Reference method 10133
Method % Recovery
Range (0-10
NTU)
Average
Precision (0 to
10 NTU)
Linearity
(0-10 NTU
Limit of
Detection
(Estimate)
Method
Reporting
(lowest
range)
Net
Difference
Binney (All
Data)
Net
Difference
San Pat
MWD (All
Data
Lovibond 660-nm
LED (Candidate)
93.7 – 104.5 0.006 0.99970 0.0008 0.010 +0.009 +0.002
10133 (Approved) 97.1 – 102.6 0.006 0.99972 0.0006 0.010 -0.009 -0.002
At the time of this ATP, the reference method was commonly regarded as the state-of-the art turbidity method in the
industry. The FilterTrak 660 Laser Nephelometer was commonly used to measure low-level turbidity samples and
was considered to be a tool that could be used in the optimization of filtration processes. However, the technology
does have drawbacks that have prevented its broader use in drinking water filter effluent applications, which was
primarily due to expensive optical components (and instrumentation), difficulties in maintenance, and difficulties in
performing calibrations. These are unique challenges that do not always translate back to the statistical analysis of
the method but should be considered when a method is considered for use. The candidate Lovibond 660-nm LED
method was designed to deliver instrumentation that resolves many of the challenges that were mentioned with the
reference method. The 660-nm LED source was very comparable to the laser nephelometer with respect to all of the
statistical evaluations in this study. This candidate method was statistically equivalent to the reference method and
can deliver a high level of measurement performance, with reduction in instrument cost, maintenance and effort to
perform calibration and QC. Ultimately, the Lovibond 660-nm LED Method delivers the combination of equivalent
performance and improved features relative to the reference instrumentation to ultimately deliver reliable turbidity
results over time.
Data comparison between the Lovibond 6000 Laser Method and the EPA approved Method 10133
Table 20 provides a comparison between these two methods over the range of the candidate method. The key
performance parameters for these methods included the range of percent recoveries for the test range of 0 to 10
NTU, the averaged precision over the test range of 0-10 NTU, the linearity, limit of detection, and the reporting limit
for each method. The two right columns compare the net measurement difference between the two methods for each
of the two different drinking water plant filter effluent waters.
46
The range of percent recoveries between these two methods was nearly equivalent. The candidate Lovibond 6000
Laser Method exhibited a slightly higher (0.6%) percent recovery at the lower turbidity spikes. The small difference
in recovery was likely due to experimental errors. An ANOVA analysis was performed between these two methods
on the percent recovery data and showed there was no statistical significance between them (F was less than F-
critical). Essentially the differences were likely due to chance. The linearity of each method was very high and
essentially equivalent, with correlation coefficients for both methods exceeding 0.999. The precision of the
Lovibond 6000 Laser Method was slightly poorer than the Reference Method. However, the higher variability could
be due to individual particle influence when in the measurement view volume. This variability can be used as an
analysis parameter when optimizing filter performance.
Although not stated in either method, an estimated limit of detection was performed by taking the averaged
precision for the lowest three spikes for each method. This value was multiplied by a factor of 3 to generate a
conservative LOD. The LOD of the Lovibond 6000 Laser Method was approximately double the LOD of the
reference instrument. This could be due to the influence of single particle variability that was better observed on
this candidate method. The limit of quantitation (LOQ) was determined using the same precision values as the
LOD and was under the method reporting level at the lowest turbidity range.
The robustness of these two methods was compared at the two drinking water sites. Both sites showed the overall
turbidity response of the approved reference method’s instrument was lower than the candidate method’s instrument
by an average of 0.009 NTU at the Binney South Platte Site, but was then higher than the candidate instrument by
0.0003 NTU at the San Pat MWD membrane site. The two water plant sites showed that the Lovibond Laser 6000
method did exhibit a higher response to turbidity spikes than the reference turbidimeter. During the filter runs at the
Binney South Platte Site (Figure 4), the two laser methods showed additional sensitivity to particle penetration
through the filter as the run progressed. This was observed as the increase in the measurement baseline as the filter
run progressed. However, the Lovibond 6000 method exhibited increased sensitivity to turbidity as the run
progressed when compared to the reference method. Both methods’ representative instruments exhibited the
enhanced sensitivity that would be expected from a laser-based method and would be ideal for reporting and process
optimization at low turbidity levels.
In Table 20, the precision is averaged across the entire range of 0 to 10 NTU tested. However, this average is biased
high from the spikes in the highest range which was from 1 to 10 NTU. For measurements below 1 NTU, the
precision was 0.0004 NTU up to 0.100 NTU and 0.0044 NTU from 0.100 to 1.00 NTU. Between 1.0 and 10 NTU,
the precision averaged 0.048 NTU. Thus, the method will have two reporting limits. Below 1 NTU, the reporting
limit will be to the nearest 0.010 NTU and at 1.0 and greater NTU it will be 0.050.
The quality controls that were followed in this ATP demonstrated that they were appropriate for the candidate test
method as there were no data deviations or events could be traced to any inadequacy of the QC approach that was
followed.
Table 20 – Summary of Method Performance between the Lovibond 6000 Laser Method and Reference method 10133
Method % Recovery
Range (0-10
NTU)
Average
Precision (0 to
10 NTU)
Linearity
(0-10 NTU
Limit of
Detection
(Estimate)
Method
Reporting
(lowest
range)
Net
Difference
Binney (All
Data)
Net
Difference
San Pat
MWD (All
Data
Lovibond 685-nm
Laser (Candidate)
94.3 – 103.4 0.0163 0.99963 0.0013 0.010 +0.009 -0.0003
10133 (Approved) 97.1 – 102.6 0.0056 0.99972 0.0006 0.010 -0.009 +0.0003
47
At the time of this ATP, the reference method was commonly regarded as the state-of-the art turbidity method in the
industry. The FilterTrak 660 Laser Nephelometer was commonly used to measure low-level turbidity samples and
was considered to be a tool that could be used in the optimization of filtration processes. However, the technology
does have drawbacks that have prevented its broader use in drinking water filter effluent applications, which was
primarily due to expensive optical components (and instrumentation), difficulties in maintenance, and difficulties in
performing calibrations. These are unique challenges that do not always translate back to the statistical analysis of
the method but should be considered when a method was considered for use. The candidate Lovibond 6000 Laser
Method was designed to deliver instrumentation that resolved many of the challenges that were mentioned with the
reference method’s instrument. The 685-nm Laser source was very comparable to the laser nephelometer with
respect to all of the statistical evaluations in this study. This candidate method was statistically equivalent to the
reference method and can deliver a similar high level of measurement performance as the reference method, with
reduction in instrument cost, maintenance and effort to perform calibration and QC. Ultimately, the Lovibond 6000
Laser Method delivered the combination of equivalent performance and slightly improved sensitivity to the presence
of low numbers of particles that eventually penetrate a filter as its run progressed. Similar to the reference method,
this candidate method can serve the purpose of being adequate for regulatory monitoring and as a tool for
optimization of filtration.
5.0 Conclusions
This Turbidity ATP test plan was designed for the synchronous evaluation of the three candidate test methods that
were presented by Lovibond, a Trademark of Tintometer Inc. The three methods were defined as the following in
this validation report:
1. The Continuous Measurement of Turbidity using the Lovibond White Light LED Method. The
representative instrument discussed herein was the PTV 1000 Turbidimeter. This method was successfully
tested with the other two proposed ATP methods using the appended validation plan.
2. The Continuous Measurement of Turbidity using the Lovibond 660-nm LED Method. The representative
instrument discussed herein was the PTV 2000 Turbidimeter. This method was successfully tested with the
other two proposed ATP methods using the appended validation plan.
3. The Continuous Measurement of Turbidity using the Lovibond 6000 Laser Method. The representative
instrument discussed herein was the PTV 6000 Turbidimeter. This method was successfully tested with the
other two proposed ATP methods using the appended validation plan.
This validation study had three key objectives. The objectives were to effectively demonstrate method equivalency
between the EPA approved reference Method 10133 and each of the proposed ATP candidate methods.
First was to determine comparability for each of the candidate methods to the reference Method 10133 with respect
to precision and bias in the turbidity range of about 0.010 to 10 NTU. Both precision and bias was determined for
each candidate method and for the reference method. Statistical analysis of the precision and bias data for each
candidate method confirmed equivalency of performance with respect to the reference method.
The second objective of this study was to demonstrate linearity of each candidate method over the range of 0-10
NTU. This required a carefully designed test plan that could generate on-line samples that were characterized with a
stable and known turbidity level(s) over time. This was necessary in order to collect run-time data on each turbidity
spike of a defined value, and allowed for an adequate amount of data to be collected to perform statistical analysis.
The success in testing did deliver the required stable turbidity values for each of the three candidate methods and
least squares analysis between measured and theoretical turbidities for each spike was calculated. This analysis
demonstrated high linearity for the three candidate methods with respective correlation coefficients that exceeded
0.999 over the range of the testing.
48
The last objective of this study was to derive comparison data between the candidate methods and the reference
methods at two drinking water plants. These plants treated very different source waters and used different filtration
processes. The study succeeded in procuring on-line analysis time at two different plants that allowed concurrent
collection of comparability data for the three candidate and the reference methods. All data that was collected
between an initial and a final QC (at each site) was used to demonstrate method comparability and equivalency
between the candidate and reference methods.
This ATP study exceeded the requirements and expectations of the study objectives and provided the necessary
evidence to demonstrate method comparability and equivalency between these three candidate methods and the
reference method for the measurement of turbidity between 0.010 and 10 NTU.
49
Appendix A
Confidential
Alternate Test Procedure Method Validation Plan
Revision 1.0
Proposed ATP Methods:
1. The Continuous Measurement of Turbidity using the Lovibond White Light
LED Method
2. The Continuous Measurement of Turbidity using the Lovibond 660-nm LED
Method
3. The Continuous Measurement of Turbidity using the Lovibond 6000 Laser
Method
April 22, 2016
Tintometer Incorporated
6456 Parkland Drive
Sarasota, FL 34243
Michael Sadar
Tintometer Incorporated
2108 Midpoint Drive, STE 1
Fort Collins, CO 80525
970-682-8148
50
Table of Contents
1.0 ..................................................................................................................................... Introduction
.......................................................................................................................................................... 51
2.0 ..................................................................................................................................... Background
.......................................................................................................................................................... 51
3.0 .................................................................................................................... Scope of this Test plan
.......................................................................................................................................................... 52
4.0 ............................................................................................................................... Study Timelines
.......................................................................................................................................................... 53
5.0 .................................................................................................................... Materials and Methods
.......................................................................................................................................................... 54
6.0 ............................................................ Test Protocols – The Technology Comparability Test Plan
.......................................................................................................................................................... 56
7.0 ................................................................................................................................ Data Collection
.......................................................................................................................................................... 59
8.0 ............................................................................................................................................. Results
.......................................................................................................................................................... 63
9.0 ..................................................................................................................................... Conclusions
.......................................................................................................................................................... 64
10.0 ..................................................................................................................................... References
.......................................................................................................................................................... 65
11.0 ........................................................................................................................................... Figures
.......................................................................................................................................................... 65
51
1.0 Introduction
This study plan was for the continuous measurement of turbidity between 0 and 10 NTU.
The study plan was designed for application of the monitoring of turbidity from drinking
water plant filter systems and from combined filter effluent systems. Specifically, this
study plan is designed to determine the precision and bias of the candidate ATP methods
over the range of turbidity up to 10 Nephelometric Turbidity Units (NTU’s).
This test plan is designed for the synchronous evaluation of the following candidate test
methods:
1. The Continuous Measurement of Turbidity using the Lovibond White Light LED
Method. The representative instrument discussed herein is the PTV 1000
Turbidimeter.
2. The Continuous Measurement of Turbidity using the Lovibond 660-nm LED
Method. The representative instrument discussed herein is the PTV 2000
Turbidimeter.
3. The Continuous Measurement of Turbidity using the Lovibond 6000 Laser
Method. The representative instrument discussed herein is the PTV 6000
Turbidimeter.
These three methods are identical with respect to design, fluid hydraulics, software,
operational procedures, data generation, and results reporting. The difference in the three
methods was with the type of light source used. Otherwise, the optical geometry that
includes the incident light source, the detection system, and the pathlength are identical.
The candidate methods will be compared to a reference instrument that complies with the
EPA accepted method 10133. The instrument is the Hach FilterTrak 660.
The proposed methods are listed separately, but they are discussed collectively herein to
simplify the understanding of the differences each has relative to the other methods and
to historical methods. When denoted collectively, they are referred to as the PTV
1000/2000/6000.
2.0 Background
These new methods utilize numerous state-of-the art technological advancements that
have been coupled with proven design criteria that have been utilized by the Drinking
Water Plant (DWP) community and its regulatory partners over the past several decades.
52
Ultimately, these technologies provide the generation of and rapid interpretation of
turbidity measurement data is critical to the mitigation of any risks associated with the
breakdown in the water treatment process.
The turbidimeter versions are identical in their design with the exception of the light
source. One version, the PTV 1000 utilizes a white light LED, a source that is essentially
the same as the incident light source used in the EPA approved Swan Turbidiwell1
Method. The second version, the PTV 2000 contains a red light emitting LED that emits
light at peak intensity within the visible spectrum at a wavelength between 650 and
670nm, (typically at 660 nm). The spectral output from this source is comparable to the
Mitchell Method M52712 and Hach Method 101333. The third light source is a 685-nm
laser diode source that is comparable to Hach method 10133. These three light sources
provide several advantages over the historical tungsten filament light sources which are
discussed in detail later in this document.
The sample hydraulics through these instruments is identical. This includes
manufacturer’s recommended flow rate ranges, sampling requirements and instrument
settings.
3.0 Scope of this Test plan
It is understood that technologies (i.e. new methods) that differ from EPA Method 180.1
require a performance-based approval4. This performance comparison will be using the
accepted EPA approved reference method, Hach Method 10133. This is the FilterTrak
660 measurement technology that complies with Method 10133.
This test plan is designed to collectively deliver the following data:
1. Precision and bias data at several turbidity levels up to about 10 NTU. Bias is
determined as percent recoveries to theoretical spikes of turbidity over a baseline.
2. Direction comparison between the candidate method and reference method on
real-world drinking water plant filter effluent samples.
3. Demonstration of linearity over the range of about 0.010 and 10 NTU.
The purpose of this test plan is to generate and deliver comparability data between the
PTV 1000, 2000, and 6000 with the white light LED, 660-nm red LED, and 685-nm laser
diode light sources respectively (i.e. the test instruments) and a reference instrument that
is considered to be the regulatory benchmark in the drinking water plant (DWP) industry,
the FilterTrak 660. Comparability data will be on three different water sources in which
water from the same source will flow in parallel to each of the test instruments and the
reference instrument. In addition, one of the waters will be spiked with known levels of
turbidity (i.e. formazin) across the operational range of the instrument (from 0-10 NTU)
and percent recoveries (bias) and precision data will be generated. All test instruments
53
will be set up and then calibrated according to the respective manufacturers
specifications.
This testing has been designed to collect data such as that which is presented in the most
recent EPA approvals for turbidimeters, i.e. the Swan Turbiwell and Mitchell Method
M5271.
4.0 Study Timelines
Once study plan approval is received, testing will commence within 30 days. The
proposed schedule is as follows:
1. Precision and bias testing to be completed first.
2. Water treatment plant comparability testing at Aurora Colorado second.
3. Water treatment plant comparability testing at San Patricio MWD, Ingleside,
Texas.
4. Validation Report to be generated within 45-days after the completion of the field
testing.
Table 1 provides the details of the proposed plants that have committed to be test sites at
the time this plan was proposed.
Table 1 – Proposed test sites for the EPA ATP involving the Lovibond PTV
1000/2000/6000Turbidimeters.
Test Site Type of Water Type of Treatment Contact Name
and Title
Status
Tintometer
Inc., Fort
Collins Co
City Tap Water
that originates as
surface water
from snowmelt in
the Colorado
Rocky Mountains.
Conventional Dual Media
Filtration to generate the “tap”
water. This is followed with
filtration through a size
exclusion membrane with
nominal pore size of 0.05 um
prior to entering the test panel.
Mike Sadar,
Manager of
Research and
Development
Confirmed
participant
San Patricio
WTP, Ingleside
Texas
Surface water
with elevated
ambient
temperatures.
Conventional
flocculation/sedimentation
followed by microfiltration
Jake Krumnow,
Supervisor of
Operations and
Maintenance
Confirmed
participant
Binney
Filtration Plant,
Aurora
Colorado (a
Partnership for
Safe Drinking
Water Plant)
South Platt River,
taken from that
includes discharge
from Denver
Metro Wastewater
Plant
Advanced UV peroxide
Oxidation, followed by Dual
Media Filtration
Kevin Linder,
Supervisor of
Operations and
Maintenance
Confirmed
participant
54
5.0 Materials and Methods
The new methods and representative instrumentation are described in full detail in the
document titled “Justification for a New ATP for the Continuous Monitoring of
Turbidity.” An abbreviated summary of this document is provided below:
Figure 1 is a Picture of the PTV 1000 Turbidimeter, along with its power supply and
sample fluidics module. As mentioned previously, the PTV 2000 and PTV 6000 are
identical in design with the exception of the incident light source. Figure 5 provides
additional detail to the features of these instruments.
The construction details and features of the turbidimeter system enable an
environmentally sound, fast and easy approach to calibration. The performance
characteristics of the turbidimeter, (the combination of low stray light and a highly linear
conversion of scattered light to turbidity over the applicable regulatory range), enable the
device to be calibrated using a single turbidity standard. The calibration standard is
easily prepared in an easy-to-use, single dose package. Exceptional linearity enables the
calibration standard can be prepared at an elevated turbidity value (such as at 5.0 NTU)
which minimizes dilution errors due to residual sample within the fluidic body. This
approach has been proven to be a key factor for reliable and accurate turbidity
measurement.
Table 2 provides a summary of the advantages that this proposed method will have over
existing methods. In addition, advantages of this design that extend beyond the actual
turbidity measurement are also included.
Table 2– Summary of the Design features for the proposed EPA methods on turbidity and the
advantages over EPA Method 180.1
Feature Advantage
White Light LED (Incident
Light Optics)
Solid State – Low drift; low output temperature dependence of LED light
source.
Long Life (10 years life expectancy)
Collimated beam– Uniform and parallel light rays as they pass through the
sample to minimize stray light
Peak response between 400-600 nm; sensitive to a broad range of particle
sizes.
Heated optics; Eliminates instability and erroneous measurements due to
condensation
ILS Monitor Detector; Compensates for LED drift over time and
temperature.
660-nm Red LED All advantages above except, the peak response is between 630 and 660
nm, which reduces interference due to dissolved organics and sample color
Reduced spectral bandwidth; reduces stray light and increases the limit of
detection.
685-nm Laser Highly collimated beam of high energy and small diameter reduces stray
light and improves the limit of detection
Narrow beam of high energy is sensitive particles in very low
55
concentrations that can be pre-cursers to a larger turbidity event.
Beam Dump Absorbs beam energy after passing through measurement chamber and
reduces stray light.
Scattered Light Detector Orthogonal to the Incident light Beam – Sensitive to scatter from a range of
particle sizes and shapes.
Controlled aperture angle – improves intra-instrument consistency in
detection.
Heated optics with no air spaces between the sample and detector –
eliminates instability due to condensation.
Responsivity overlaps the spectral output of the incident light source for
improved sensitivity.
Short total path length of 5.5 cm yields a highly linear response over the
regulatory range of interest (0-10 NTU) and simplifies the calibration
protocols.
Large collection angle; improves the limit of detection.
The turbidimeter body Reduced Volume – improves response time, reduces sample settling,
minimizes sample usage and minimizes calibrant usage.
Integrated bubble trap – removable without tools, front accessible, easy to
clean and low sample retention.
Void of tight corners – easy to clean
Polished fluid handling surfaces reduce scale, fouling and bubble
formation.
Comprised of absorptive (black) material – reduces internal light reflections
(interferences that contribute to stray light).
Measures at Atmosphere – eliminates any sensitivity to flow and pressure
changes in the sample line (i.e. water hammers).
“V” measurement chamber form factor – Prevents particulate settling and
reduces stray light.
Flow monitor – Confirms sample flow through the measurement chamber
(alarms and warnings can be set for both high and low flow conditions). To
ensure measurement integrity.
Multiple internal weirs – ensure consistent sample throughput and increases
instrument robustness
Optimized sample handling – sample temperature does not change as it
passes through the instrument, thereby minimizing changes in sample
composition and reduces bubble formation.
Complete Measurement
System (Body and
Measurement module)
Reduced stray light – leads to simplified and robust calibration protocols
Enhanced detection limit (0.002 NTU for White Light LED, 0.0005 NTU
for 660-nm Red LED, and 0.0003 NTU for the 685-nm Laser Diode)
Highly linear response over the range 0-10 NTU. This allows for robust
and simplified calibration protocols.
Elimination of common low-level interferences (bubbles, stray light,
particle settling.
Self-aligning magnetic positioning of the measurement module on the body
– ensures the proper alignment of the measurement module to the fluidic
body.
Redundant interfaces (Smart Device and Touchscreen).
Redundant data and meta data storage.
Both wired (USB) and secure wireless (Bluetooth) communication for
operation of the instrument (User interface)
Multiple outputs for both digital and analog communication streams (i.e.
Modbus, 4-20 mA)
56
6.0 Test Protocols – The Technology Comparability Test Plan
The purpose of this test plan is to generate comparability turbidity data between the
proposed PTV 1000/2000/6000 technologies and referenced approved EPA technologies.
The plan involves the testing of the PTV 1000/2000/6000 and PTV 2000 Turbidimeters,
one device equipped with a White LED, the second device equipped with a 660-nm red
LED, and the third device equipped with a 685-nm laser diode respectively. A reference
turbidimeter is used for comparability of results. The reference turbidimeter will be the
Hach FT660sc Laser Nephelometer. The reference turbidimeter was purchased directly
from Hach Company. All devices will be run on the same sample waters and exposed to
the same spiking protocols throughout the testing. This setup will be referred to as the
test panel.
The comparability testing will involve the monitoring three different waters with on-line
turbidimeters. The plan will be designed to include an alternative test site. This site will
be used in lieu of unforeseen circumstances that render a site or its data invalid. Table 1
provides a list of the proposed sites for this study. Two of the sites will be simple
measurement comparability studies on during a typical filter run. The third study will
involve the spiking of a filter effluent stream with turbidity standards to generate percent
recovery data over the specified test range.
The test panel (with two test devices and one reference device) will be installed to
monitor filter effluent water, for minimum of 24 hours at each site. The sample water
will be split into four tap streams, one feed line branching to each turbidimeter. The flow
for each stream will be adjusted to meet the manufacturer’s requirement for the
respective instrument. Ideally, this will be the middle of the flow range for each
instrument. After the sample flows through the instruments, the discharge will flow to
drain. Figure 2 shows the proposed hydraulic plumbing for the test and reference
instruments.
Calibration of each device on the test panel will take place after setup and all pluming
connections are completed at each of the respective test sites. Calibration will be
performed after each device has been cleaned according to the respective manufacturer’s
instructions. Freshly prepared, EPA approved formazin standards will be used as the
calibrant. If dilutions are needed, this will be made with filtered water that has been
processed through a 0.2 um or smaller filter. The calibrations will be performed within
the actual device bodies, i.e. no calibration cylinders will be used.
The test devices will be setup according to operating instructions per the respective
instrument manuals. The setups will include the recommended signal averaging times,
bubble rejection algorithms, and displayed measurement resolution. The measurement
(data) log rate will be set to 15-second intervals for the duration of the study. Data will
be logged into each of the respective instrument data log files until the measurement run
57
for the site has been completed. This data, along with date and time stamp (to the nearest
second) will then be exported into an Excel spreadsheet for statistical analysis.
After calibration and prior to data collection, a quality control sample (QCS) will be
prepared and run through each instrument. The QCS will be a fresh prepared formazin
standard that will be checked on a laboratory turbidimeter (e.g. a Hach 2100N). A
second QCS will be run after the completion of the data run to confirm system
performance.
The Spike Injection Method
The Fort Collins test site will be used for the turbidity spike recovery portion of this test
plan. This will quantatively generate the percent recovery (i.e. accuracy) of each
intended turbidity spike. This spike data is necessary to demonstrate performance
throughout the test range of 1 to 10 NTU. This portion of the study will also be used to
generate the precision data.
The spike injection method involves a minor modification of the test panel used for the
comparability testing at the different sites. The modification is the incorporation of a
peristaltic pump. This pump will continuously inject a prepared turbidity standard into a
flowing sample stream. The turbidity standard is a freshly prepared formazin standard
and is of a volume that will allow for an adequate stabilization of each spike that is
necessary for the percent recovery calculation.
Figure 3 is the schematic of this modified test panel for the spike injection method (SIM).
The theory of the SIM is detailed as follows. First, the incoming water (the sample
stream) flows at a constant turbidity and flow rate. This is the baseline or blank water. A
series of filters and valves are utilized to ensure the consistency of this blank over the
duration of the testing. Second, a peristaltic pump is used to inject a defined turbidity
standard at a constant flow rate into the sample stream. The flow rate of both the sample
stream and the injected, turbidity standard are known and flow at a constant rate. The
blending of this spiked standard with the sample stream results is a level of turbidity
within the sample stream. Third, the spiked turbidity sample stream flows through a
mixing coil and is then split into four branches originating from the single feed line, each
branch leading to either a device under test or the reference instrument. Excess sample
simply runs to the overflow drain.
A combination of different turbidity standards and different injection rates will deliver
stable turbidity values of different concentrations. This approach allows for the delivery
of a continuous stream of sample at varied turbidity concentrations into the devices under
test and reference instrument under normal operating conditions.
The specific details of these key elements of this spike injection methodology are
discussed in more detail.
1. The incoming tap water (the sample stream) is filtered through a set of size
exclusion filters that removes all particles of a size greater than 0.03 microns.
58
This system virtually removes all turbidity from the system but does not remove
any dissolved solids such as residual hardness. The system is sized to allow
continuous and adequate flow through all instruments on the test panel. The
turbidity level of this water is also constant and provides the baseline turbidity for
this study. The baseline turbidity is typically in the range of 0.015 to 0.020 NTU,
depending on the instrument that is used to measure it. As mentioned previously,
this is the blank.
The turbidity of the blank will be derived from the measurement of this filtered
tap water prior the start and then at the conclusion of the spike injections. It will
be derived for each instrument. Approximately 30 minutes of data will be taken
prior to the first spike. At the completion of the spikes and after the sample
stream has become stable with respect to the reference instrument, an additional
30 minutes of monitoring and data collection will occur. The combined data will
then be averaged to calculate the turbidity value of the blank. As mentioned
above, this value will be in the 0.015 to 0.020 NTU range.
2. The flow rate of the incoming sample stream (Qss) is measured prior to the point
of injection of the turbidity standard. The measurement is in grams of sample per
minute and a gravimetric balance with resolution to the nearest 0.01 grams is used
(equation 1). The measurement can be converted to ml/minute by measuring the
temperature of the sample and then adjusting for its density. It should be noted
that the density adjustment is typically negligible.
Qss = grams of sample dispensed/minute (1)
3. The injection rate of the defined turbidity standard utilizes a freshly prepared
formazin standard. The standard is prepared by gravimetric dilution and the
resultant value can also be confirmed through the measurement on a calibrated
laboratory or portable turbidimeter. The flow of the (Qstd) is calculated as the net
mass of standard delivered over a given period of time: the prepared standard is
first weighed on a gravimetric balance (M0) at the start time of the injection (T0).
After the injection is complete, the standard is re-weighed on the balance (MFinal),
and the time that the injection was also completed (TFinal) is also recorded. The
net mass of the standard injected (MFinal- M0) is divided by the net time of the
injection (TFinal - T0) to deliver the injection rate (Qstd) in g/minute is given by
equation 2.
Qstd = (MFinal- M0) / (TFinal - T0) (2)
Where:
Qstd = Injection rate of the defined turbidity standard;
M0 = the mass of the defined turbidity standard at the beginning of the
injection;
59
MFinal = the mass of the defined turbidity standard at the end of the
injection;
T0 = Time at the beginning of the injection in minutes
TFinal = Time at the end of the injection in minutes
4. The resultant spike of turbidity (NTUSpike) into the sample stream is then
calculated by multiplying the value of the turbidity standard (NTU) by the
injection rate (Qstd) and dividing this by the total flow rate (Qstd + Qss). This is the
theoretical increase in turbidity over the turbidity of the sample stream, expressed
by equation 3.
NTUSpike = NTUstd * Qstd / (Qstd + Qss) (3)
Where:
NTUspike = the spike of turbidity in NTU’s above the turbidity of the
filtered sample stream (i.e. the blank);
NTUstd = the turbidity of the formazin standard that is injected with the
peristaltic pump;
Qstd = the flow rate of the formazin standard into the sample stream;
Qss = the combined flow rate of the sample stream and the formazin
standard.
5. The response of the test instrument (s) to this theoretical increase can be measured
and calculated as a percent recovery. The measured response of the test
instrument minus the turbidity of the baseline prior to injection is first calculated.
This is then divided by the theoretical value of the spike and expressed as a
percentage (multiplied by 100%).
Percent recovery of the spike = 100*(Instrument response to the spike – blank
value of the sample stream) /Theoretical Value of the spike
7.0 Data Collection
Table 3 provides an example of the range of spikes that can be generated for the test
panel. In this table, each horizontal row contains the test parameters and conditions
collected during the test which are used to generate a stable value of turbidity. This
includes entering the overall sample flow at the beginning and end of each run, the
turbidity of the standard to be spiked, the start and end times of the injection of the
60
respective formazin standard, and the start and end mass of the formazin standard that is
spiked (via the peristaltic pump) into the sample stream. The cells that require data to be
entered are shaded in light green. The other cells are calculations (as explained in
equations 1 through 4 above). These equations are entered into an Excel spreadsheet and
will automatically calculate the theoretical turbidity value of each spike. The farthest row
to the right in the table contains the calculated theoretical turbidity value of each spike.
This value is then the basis for the percent recoveries for each spike that is generated in
this study.
Through the measurement of the sample flow at the beginning and end of each run, and
using a gravimetric balance for all mass measurements, the method accurately
compensates for any unforeseen changes in sample flow or injection flow rates over the
duration of the test. In doing so, the method has been demonstrated (internally) to
produce low experimental error (including that of particulate contamination) than
alternative methods of producing large volumes of turbidity standards at very low
turbidity values.
In Table 3, the top and bottom row are the filtered sample stream baselines (used to
determine the blank value). A baseline is run to ensure the turbidity values are stable,
which confirms the performance of the filtration system. In addition, a post spike
baseline is also run. The post spike baseline run typically takes a couple of hours for to
reach its steady state condition after high turbidity values are flowed through the devices.
This is because additional time is needed instruments on the test panel to flush out the
high levels of turbidity from the highest spike.
The second row from the top in Table 3 is the lowest turbidity spike and each progressive
row down the table, turbidity spike is of a higher turbidity value. Spikes are run from the
lowest turbidity value to the highest turbidity value and are arranged within Table 3 from
the top to the bottom of the table in the order of study. By spiking the sample stream
from low to high values the transition time from one turbidity value to the next, higher
turbidity value is fairly rapid. This allows the study to be completed in a relatively short
period of time, (approximately 8 hour’s total).
Table 3 – Sample table that presents the input information to generate the different turbidity spikes for the Spike Injection Method
Spike
Descript
ion
Turbidity
of
Formazin
Spike
standard
(NTU)
Startin
g Mass
(g)
Startin
g Time
(min)
Starting
Sample
Flow
(g/min)
End
Mass
(g)
Change
in Mass
(g)
End
Time
Chan
ge in
Time
(min)
End
Sampl
e Flow
(g/min)
Average
Sample
Flow
(g/min)
Flow
Rate of
Spike
(g/min)
Theoretica
l Turbidity
of Spike
(NTU)
Initial
Baseline (Blank)
0.00 N/A 927
881.4 N/A N/A 1017 60
881.4 881.4 0 0
2.0 NTU
at 2
RPM
2.01 2348.22
855 881.4 2158.4
8 189.74 946 51
881.4 881.4 3.720 0.008
2.0 NTU
at 4 RPM
2.01 2158.4
8 947
881.4 1863 295.48 1031 44
881.4 881.4 6.715 0.015
61
The spike injection method, as proposed, will generate at least two but as many as four
turbidity levels between 0.010 and 0.100; and between two and three turbidity levels
between 0.101 and 1.00 NTU; and between two and three turbidity levels between 1.0
and 10 NTU. The plan will be to produce at least eight different levels of spiked
turbidities, and more (up to a max of 11 spikes) if it is feasible from a time perspective.
The exact turbidity of any of the spike can be adjusted accordingly.
The accuracy of this test plan also involves several tasks that will be accomplished
throughout this testing. These include the following:
• Sample lines shall be equivalent in length, but minimal in distance from the point at
which each line branches to the individual devices under test.
• The formazin standards that are to be spiked into the sample stream shall be measured
on a benchtop turbidimeter, (a Hach® 2100N or equivalent) that has been calibrated
on formazin standards. The bench top device will be used to confirm the value of the
turbidity standard.
• QCS samples will be run at prior to and after the completion of the spiking. QCS
samples will be checked on a benchtop turbidimeter prior to use.
• The volume of the formazin standards will be in the amount which allows for
approximately one hour of injection time. In Table 4, the second to the right column
provides information on the different flow rates for the peristaltic pump that are to be
used in this study.
2.0 NTU
at 8 RPM
2.01 1863 1032
881.4 1230.5
6 632.44 1119 47
881.4 881.4 13.456 0.031
2.0 NTU at 16
RPM
2.01 1230.5
6 1120
881.4 403.58 826.98 1150 30
881.4 881.4 27.566 0.063
16 NTU
at 4
RPM
16.0 1268.44
1151 881.4
859.44 409 1252 61 881.4 881.4
6.705 0.122
16 NTU at 8
RPM
16.0 859.44 1253
881.4 271.73 587.71 1339 46
881.4 881.4 12.776 0.232
134
NTU at
2 RPM
134 1234.77
1346 881.4 1046.4
4 188.33 1435 49
881.4 881.4 3.843 0.584
134
NTU at 4 RPM
134 1046.4
1 1436
881.4 742.7 303.71 1521 45
881.4 881.4 6.749 1.026
134
NTU at
8 RPM
134 740.7 1522
881.4 368.78 371.92 1550 28
881.4 881.4 13.283 2.020
800
NTU at 3 RPM
801 1247.7
4 1551
881.4 1066.5
3 181.21 1624 33
881.4 881.4 5.491 4.991
800 NTU at
6 RPM
801 1066.5
3 1625
881.4 718.29 348.24 1657 32
881.4 881.4 10.883 9.890
Final
Baseline (Blank)
0.000 N/A 1900
881.4 NA N/A 2000 60
881.4 881.4 0.000 0
62
• All formazin standards to be spiked are continuously mixed using a magnetic stirrer
to ensure these solutions remain homogeneous throughout the duration of the test. It
has been confirmed that the magnetic stir bars does not cause measurement error
when weighed on the prescribed balance.
• The data log rates on all instruments will be exactly the same. The proposed time
interval at which data is recorded shall be 15-minutes.
• The clocks on all instruments will be synchronized to the nearest second. This will
ensure the data is consistently time-stamped for all instruments.
• The data will be exported digitally from each instrument using Modbus protocol to a
common computer. This export is in CSV format that is compatible with Microsoft
Excel.
• As a backup, each instrument has a SD card for logging data. If necessary, the data
can be copied from the respective instrument’s SD card into an Excel spreadsheet.
• Microsoft Excel will be the program that all the measurement data will be logged.
All calculations will be performed in Excel.
• After calibration, each instrument will be verified using a separate wet standard. The
value of the verification standard will be at or below 1.0 NTU.
• Prior to each spike, the injection line will be primed with the respective standard.
The injection line is primed up to the valve that allows the standard to be injected into
the sample line. This will minimize the response time to all instruments.
• All instruments will be calibrated on formazin from the same manufacture lot.
Calibration standards will be prepared with Class A glassware and dilution water that
has been filtered through a size exclusion filter with a pore size that is less than 0.2
um.
• The time between spikes will be minimized. This will allow the test and reference
instruments to respond to each increasing step in turbidity, thereby reducing the
stabilization time needed for each turbidity value.
• After the response of each device under test has become stable for a given turbidity
value, the spike will continue for a minimum of 15-minutes. This will allow for the
collection of at least 60 measurements at each turbidity level, (data logged at a rate of
one point every 15 seconds). These measurements will then be used for the statistics
evaluation of each instrument with respect to calculated turbidity.
• For each spike, the 60 measurements will be used to generate the mean, standard
deviation, and relative standard deviation for each instrument.
• For each spike, the calculated mean, based on the 60 measurements will be used in
the percent recovery calculations (accuracy). The standard deviation (SD) will be
used to derive the precision of each instrument on each turbidity level.
• All spike injections will be performed in sequential order from lowest to highest in
the course of a single day.
• All instruments will run a minimum of 8 hours before the commencement of any
portion of the study during which data is collected. This includes calibration. The
intent is to ensure all surfaces that contact the sample are adequately conditioned, (i.e.
wetted).
• At the conclusion of the spike study, the instruments will again be verified using a
wet standard.
63
8.0 Results
Turbidity Spike Results - The results from all the spikes will be provided in tabular
format. Table 4 provides each of the result elements for each spike for the PTV 1000
White Light instrument, Table 5 provides the result elements for the PTV 2000 Red 660-
nm LED, and Table 6 provides the result elements for the PTV 6000 685-nm Laser
instrument. It is expected that these Tables will become part of the study report and
become part the Method Performance section for the respective proposed methods.
Table 4 – Results Table for the Percent Recovery to Turbidity Spikes for the PTV1000 WL LED Turbidimeter
PTV 1000 WL Reading (NTU) Reference Turbidimeter Reading (EPA approved Hach Method
10133)
Spike
#
Baseline
(Blank)
Theoretical
Value
Theoretical Value (blank
Corrected)
Recovery
(%)
Precision
(SD)
Baseline
(Blank)
Theoretical
Value
Theoretical Value (blank
Corrected)
%
Recovery
Precision
(SD) N
1 2 3 4 5 6 7 8
Table 5 – Results Table for the Percent Recovery to Turbidity Spikes for the PTV2000 660-nm LED Turbidimeter
PTV 1000 660 Reading (NTU) Reference Turbidimeter Reading (EPA approved Hach Method
10133)
Spike #
Baseline (Blank)
Theoretical Value
Theoretical
Value (blank
Corrected)
Recovery (%)
Precision (SD)
Baseline (Blank)
Theoretical Value
Theoretical
Value (blank
Corrected)
% Recovery
Precision (SD)
N
1 2 3 4 5 6 7 8
Table 6 – Results Table for the Percent Recovery to Turbidity Spikes for the PTV6000 685-nm Laser Turbidimeter
PTV 6000 660 Reading (NTU) Reference Turbidimeter Reading (EPA approved Hach Method
10133)
Spike
#
Baseline
(Blank)
Theoretical
Value
Theoretical
Value (blank Corrected)
Recovery
(%)
Precision
(SD)
Baseline
(Blank)
Theoretical
Value
Theoretical
Value (blank Corrected)
%
Recovery
Precision
(SD) N
1 2
64
3 4 5 6 7 8
Water Plant Comparability Data - The data from each water plant will be plotted for a
period of at least 8 hours. The plot will have time on the x-axis and the turbidity levels
for the test and reference instruments will be plotted on the left y-axis. PTV
1000/2000/6000 instruments will have separate plots versus the reference instrument. An
example of a plot is provided in Figure 4. Based on an 8-hour run of data collection, each
instrument will generate 1920 time stamped data points that can be used for direct
comparability on each water treatment plant sample.
During the water plant run, data that is collected during the process of filtration will be
valid for this study. Any data that is collected during backwash or when the filter is
offline will be excluded.
During the water plant data collection process, it is possible that the drinking water plant
filter will produce water with little to no turbidity excursions over the time of the study.
Under these conditions, surrogate turbidity events may be created using the pump
apparatus that is described in Figure 3. The surrogate turbidity would be to inject kaolin
into the filter effluent to generate turbidity values up to 1 NTU. This will only be used if
no normal filter excursions are observed.
9.0 Conclusions
This test plan is intended to demonstrate measurement comparability between the three
proposed ATP technologies that are presented by Lovibond Inc., and a well-established
EPA approved reference turbidimeter. The study plan has been designed to demonstrate
the comparability under operating conditions of continuous flowing sample streams for
both the direct comparability and for the spike studies.
This test plan is designed to test the three proposed technologies simultaneously in order
to provide to the EPA and to future users of the technologies a comparison of the impact
of different light sources on turbidity measurement. The authors of this test plan feel its
design is comprehensive to the scope and purpose of the application and use of this
technology in the regulatory arena and it exceeds the criteria that has been accepted by
the agency in past approval studies. However, the authors of this test plan welcome any
suggestions and changes that the agency may request.
65
10.0 References
1. SWAN Analytische Instrumente AG, “Continuous Measurement of Turbidity Using a
SWAN AMI Turbiwell Turbidimeter”, (2009)
2. Mitchell, Leck, Ph.D, “Determination of Turbidity by Laser Nephelometry, Mitchell
Method M5271”, (2009).
3. Hach Company, “Hach Method 10133, the Determination of Turbidity by Laser
Nephelometry, (2000).
4. USEPA, “Method 180.1 Determine of Turbidity by Nephelometry”, 1993.
11.0 Figures
Figure 7- The Lovibond PTV 1000 Low Range Turbidimeter
66
Figure 2 - Proposed Schematic of the test panel for the Lovibond Turbidimeter ATP. This includes the WL LED, Red
660-nm LED, and 685-nm Laser versions. The panel also includes the proposed reference turbidimeter.
67
Figure 3 - Schematic for the Spike Injection approach to generate percent recovery data to turbidity spikes for the
Lovibond PTV 1000/2000/6000 Turbidimeter ATP.
68
Figure 4 - Sample plot of the turbidity measurements of a test and reference instrument for a 24-hour period.
69
Figure 5 - Schematic of the PTV 1000/2000/6000 turbidimeter and optional fluidics module.
70
Appendix B.1
Continuous Measurement of Drinking Water
Turbidity Using a Lovibond PTV 1000 White
Light LED Turbidimeter
The Lovibond White Light LED Method
Revision 1.0
December 20, 2016
Tintometer Inc.
6456 Parkland Drive
Sarasota, FL 34243
71
Continuous Measurement of Drinking Water Turbidity Using a
Lovibond PTV 1000 White Light LED Turbidimeter
1. SCOPE AND APPLICATION
1.1 This method is applicable to any colorless drinking water samples with a turbidity
less than 10 Nephelometric Turbidity Units (NTU).
1.2 The applicable range is from 0 to 10 NTU.
1.3 The method meets the requirements for compliance monitoring and reporting as
demanded under the Safe Drinking Water Act (SDWA).
2. SUMMARY OF METHOD
2.1 The method is based upon a comparison of the intensity of a collimated beam of light
that is generated by a white light emitting diode (LED), that is scattered by the sample
under defined conditions with the intensity of the same white light LED scattered by a
standard reference suspension. The higher the intensity of scattered light, the higher
the turbidity. Readings, in NTU, are made in a nephelometer designed according to
specifications given in section 6.2.
2.2 Formazin, prepared under closely defined conditions, is used as a primary standard
suspension to calibrate the instrument. However, other approved primary standards
may be used with this method.
2.2.1 Examples of standards that can be used to calibrate the instrument include
dilutions from commercially available 4000 NTU formazin, stabilized versions of
formazin with preassigned turbidity values such as T-Cal™, and styrene
divinylbenzene suspensions with preassigned values for the specific make and
model of the instrument to be calibrated.
2.3 The method generates a linear response between scattered incident light that is
detected at 90-degrees over the applicable range. The method defines 0.00 NTU as no
light impinging on the 90-degree detector and requires at least one defined standard to
perform a calibration over the applicable range.
3. DEFINITIONS
3.1 LABORATORY REAGENT BLANK (LRB) – An aliquot of reagent water or other
blank matrices that are treated exactly as a sample including exposure to all
glassware, equipment, solvents, reagents, internal standards, and surrogates that are
72
used with other samples. The LRB is used to determine if method analytes or other
interferences are present in the laboratory environment, the reagents, or the apparatus.
3.2 LINEAR CALIBRATION RANGE (LCR) – The concentration range over which the
instrument response is linear.
3.3 SAFETY DATA SHEET (SDS) – Written information provided by vendors
concerning a chemical/s toxicity, health hazards, physical properties, fire, and
reactivity data including storage, spill, and handling precautions.
3.4 PRIMARY CALIBRATION STANDARD (PCAL) – A suspension prepared from
the primary dilution stock standard suspension. The PCAL suspensions are used to
calibrate the instrument response with respect to analyte concentration.
3.5 QUALITY CONTROL SAMPLE (QCS) - A solution of the method analyte of known
concentrations that is used to fortify an aliquot of LRB matrix. The QCS is obtained
from a source external to the laboratory, and is used to check laboratory performance.
3.6 SECONDARY CALIBRATION STANDARDS (SCAL) – commercially prepared,
stabilized sealed liquid or gel turbidity standards, or other apparatus or mechanism
calibrated against properly prepared and diluted Formazin or styrene divinylbenzene
polymers.
3.7 STOCK STANDARD SUSPENSION (SSS) – A concentrated suspension containing
the analytic solution prepared in the laboratory using assayed reference materials or
purchased from a reputable commercial source. Stock standard suspensions are used
to prepare calibrant suspensions or other needed values of suspensions.
4. INTERFERENCES
4.1 The presence of floating debris and coarse particulate matter within the sample may
settle out of suspension resulting in low turbidity readings.
4.2 Finely divided air bubbles will cause random high spikes in readings.
4.3 The presence of dissolved, light absorbing substances or chemicals in the sample, i.e.
the presence of color, can absorb portions of the incident light spectra, resulting in
low turbidity readings, although this effect is generally not significant for drinking
water.
4.4 Light-absorbing particles in suspension within the sample, such as activated carbon of
significant concentration, can cause low readings.
4.5 Certain dissolved molecules or compounds can impart a fluorescence effect that can
result upon the interaction with shorter wavelengths from the incident light source
73
used in this method. This interference should be considered at turbidities below 0.100
NTU.
4.6 Construction materials of the nephelometric device within the measurement chamber
can result in elevated stray light due to spurious reflections of the incident beam can
cause a false positive bias at the bottom end of the range.
5. SAFETY
5.1 The toxicity or carcinogenicity of each reagent used in this method has not been fully
established. Each chemical should be regarded as a potential health hazard and
exposure should be as low as reasonably achievable.
5.2 Each laboratory is responsible for maintaining a current awareness file of OSHA
regulations regarding the safe handling of the chemicals specified in this method. A
reference file of Safety Data Sheets (SDS) should be made available to all personnel
involved in the chemical analysis. The preparation of a formal safety plan is also
advisable.
5.3 Hydrazine sulfate (7.2.1) may reasonably be anticipated to be a carcinogen1.
Formazin can contain residual hydrazine sulfate. Proper protection should be
employed.
6. EQUIPMENT
6.1 The installation shall be according to the manufacturer’s instructions.
6.2 The turbidimeter shall consist of a nephelometer with a light source for illuminating
the sample, one or more photo-detectors to measure the amount of scattered light at a
right angle to the incident beam, a correlation means to relate the amount of scattered
light to a known turbidity standard, and a communication means to convey the
turbidity value to the plant operator or other responsible water authority.
6.3 Differences in the physical design of the turbidimeter will cause differences in
measured values for turbidity, even though the same suspension is used for
calibration. To minimize such differences, the following design criteria shall be
observed:
6.3.1 The light source shall be a Light Emitting Diode (LED) emitting white light in the
visible spectrum between 380 and 780 nm. The LED, all optical elements, and
detectors shall have a spectral peak response between 400 nm and 600 nm.
6.3.2 The rays comprising the incident beam shall be parallel with no divergence and
not to exceed 1 degree of convergence within the measurement volume.
74
6.3.3 Non-scattered or non-attenuated light of the incident beam after passing through
the sample shall pass into a light trap.
6.3.4 Distance traversed by incident light and scattered light not to exceed 10 cm.
6.3.5 Detector/Light Receiver shall be centered at 90º ± 1 ½ degrees to the incident
beam.
6.3.6 Scattered light shall be received by the detector/light receiver at a subtended angle
between 20 and 30 degrees from the center-point of the measurement volume.
6.3.7 The detector/light receiver shall have a spectral response that encompasses the
peak spectral output of the incident light source.
6.3.8 A linear-based algorithm shall be used to convert detector signal to NTU in
correlation to a known calibration standard.
6.3.9 The turbidimeter shall be free from significant drift after a short warm-up period.
6.3.10 The sensitivity of the instrument shall detect differences in turbidity of 0.01 NTU
below 1.0 NTU.
6.3.11 The instrument shall be capable of measuring from 0 to 10 NTU turbidity units.
Several ranges may be necessary to obtain both adequate coverage and sufficient
sensitivity for low turbidities.
6.3.12 Instrument shall incorporate a sample deaerator for the removal of entrained air
from the sample stream.
6.4 A nephelometric device that meets these specifications is a Lovibond PTV 1000
turbidimeter.
7. REAGENTS AND STANDARDS
7.1 Reagent water, turbidity-free: Pass deionized distilled water through a 0.2-µm or
smaller pore-size membrane filter if necessary. Water produced by reverse osmosis is
acceptable. Water should have a turbidity that is ≤0.030 NTU. This value should be
considered when preparing standards.
7.2 Stock standard suspension (Formazin) 4000 NTU
7.2.1 Dissolve 5.00 g hydrazine sulfate (CASRN 10034-93-2) into approximately
400 mL of reagent water contained in a cleaned 1-L Class A volumetric flask.
75
7.2.2 Dissolve 50.00 g hexamethylenetetramine (CASRN 100-97-0) in approximately
300 mL of reagent water contained in a 500-mL volumetric flask.
7.2.3 Quantitatively transfer the hexamethylenetetramine solution (7.2.2) into the 1-L
flask that contains the dissolved hydrazine sulfate solution (7.2.1).
7.2.4 Dilute the 1-L flask (7.2.3) to volume with reagent water.
7.2.5 Stopper and mix by inversion for 10 minutes.
7.2.6 Allow to stand 24 hours at 25 ± 2 °C. During this time, the formazin polymer will
develop. The turbidity of this standard is 4000 NTU.
7.2.7 Store this solution in the dark and away from a source of heat. Bring the solution
to room temperature and thoroughly mix before preparing dilutions (7.3).
7.3 Primary calibration standards: Using a pipette with accuracy to 1 percent or better,
first mix and then dilute 25.0 mL of stock standard suspension (7.2) to 0.10 L with
reagent water. The turbidity of this suspension is defined as 1000 NTU. For other
turbidity values, mix and dilute portions of this suspension as required using clean
Class A glassware.
7.3.1 A new stock standard suspension (7.2) should be prepared each quarter. Primary
calibration standards (7.3) should be prepared daily by dilution of the stock
standard suspension.
7.4 Formazin in commercially prepared, certified, concentrated stock standard suspension
(SSS) may be diluted and used as required. Dilute turbidity standards should be
prepared daily.
7.5 Stabilized formazin suspensions or styrene divinylbenzene (SDVB) polymers are
commercially prepared, certified, and ready to use dilutions. Manufacturer’s
instructions should be followed for choosing the appropriate standard values for the
instrument.
7.6 Secondary standards may be acceptable as a calibration check, but must be monitored
on a routine basis for deterioration and replaced as required.
8. SAMPLE COLLECTION AND INSTRUMENT SETUP
8.1 Online instrumentation does not require sample cooling, preservation or storage.
8.2 Install and set up the instrument according to the manufacturer’s instructions.
8.3 Determination of the sample water offset.
76
8.3.1 This step is optional and can be performed if the measured sample measures high
when compared to a EPA approved reference measurement. This can occur when
fluorescence effects are suspected (4.5).
8.3.2 This step should be applied when the expected turbidity is below 0.100 NTU.
8.3.3 Procedure
8.3.3.1 Collect approximately 25 mL of sample. Record the value of the turbidity
on the on-line instrument.
8.3.3.2 Filter the sample through 0.2-µm filter directly into a clean sample vial.
Rinse the vial at least two times with the filtered sample prior to filing
with the filtered aliquot for analysis.
8.3.3.3 Carefully prepare the sample aliquot for measurement in a calibrated
reference benchtop or portable instrument that meets the specifications
outlined in the method or any EPA approved turbidity method.
8.3.3.4 Record the turbidity value on the reference instrument.
8.3.3.5 Enter this value into the process turbidimeter as the blank value for the
sample. The maximum allowable value that should be entered into the on-
line instrument is 0.05 NTU.
8.3.4 The offset does not impact the calibration gain.
9. QUALITY CONTROL
9.1 Each laboratory using this method is required to operate a formal quality control (QC)
program. The minimum requirements of this program consist of an initial
demonstration of the process turbidimeter system’s capability and analysis of
laboratory reagent blanks and other solutions as a continuing check on performance.
The laboratory is required to maintain performance records that define the quality of
data generated.
9.2 Initial demonstration of performance
9.2.1 The initial demonstration of performance is used to characterize instrument
performance as determined by the LCR and QCS analyses.
9.2.2 Linear Calibration Range (LCR) – The LCR must be determined initially and
verified whenever a significant change in instrument response is observed or
expected. The initial demonstration of linearity must use sufficient standards to
insure that the resulting curve is linear. The verification of linearity must use a
77
minimum of a blank and three standards. For example, the standards could be
0.30 NTU, 5.0 NTU and 10.0 NTU. If any verification data exceeds the initial
values by ± 10% or exceeds the stated specifications of the turbidity standard,
whichever is greater, linearity must be reestablished. If any portion of the range is
shown to be nonlinear, sufficient standards must be used to clearly define the
nonlinear portion. The selection of standards should at least cover the range of
turbidity values that are expected from samples.
9.2.3 Quality Control Sample (QCS) – When beginning the use of this method, on a
quarterly basis or as required to meet data-quality needs, verify the calibration
standards and acceptable instrument performance with the preparation and
analysis of a QCS. If the determined concentrations are not within ± 10% or ±
0.040 NTU of the stated QCS values, performance of the determinative step of the
method is unacceptable. The source of the problem must be identified and
corrected before continuing with on-going analyses.
9.3 Accuracy and precision should be checked on a routine basis to monitor the overall
performance of the instrument. A series of reagent blanks and check standards should
be run to validate the quality of sample data. These checks should occur at a
frequency that is required for regulatory compliance.
9.3.1 A QCS sample as described in 9.2.3 must be run at least on a quarterly interval.
The instrument must be checked to insure it has been cleaned and maintained
according to the manufacturers recommendations prior to running the QCS.
9.3.2 Solid standards are an option and can only be used for verification purposes. The
instrument should be checked to insure it has been cleaned and maintained
according to the manufacturers recommendations prior to running a solid
verification standard.
10. CALIBRATION AND STANDARDIZATION
10.1 Turbidity Calibration: The manufacturer’s operating instructions should be followed
for calibration. Perform any cleaning and maintenance prior to calibration as per
manufacturer’s instructions. The turbidimeter measurement chamber should be rinsed
with at least 1 L of water that has been filtered through a 0.45-um filter or smaller
prior to calibration. Calibration should be performed under the same ambient
conditions as sample measurement.
10.2 Measure standards on the turbidimeter covering the range of interest. If the
instrument is already calibrated in standard turbidity units, this procedure will check
the accuracy of the calibration scales.
78
10.3 At least one standard should be run in each instrument range to be used. Some
instruments permit adjustments of sensitivity so that scale values will correspond to
turbidities.
10.4 Solid standards can only be used for verification purposes. If used, they must be
protected from surface scratches which may cause potential changes.
10.5 If a pre-calibrated scale is not supplied, calibration curves should be prepared for each
range of the instrument. Calibration must be performed under identical optical
conditions as operational conditions.
11. PROCEDURE
11.1 A sample from the treatment process is taken and flows through the turbidimeter for
measurement. It is then drained or recycled back into the process after the
measurement has been taken.
11.2 The sample flow rate shall be in accordance with the instrument specifications. The
sample flow to the instrument shall be constant without variations due to pressure
changes or surges. Installation of a flow control device such as a rotameter in the
sample line can eliminate fluctuations of the flow rate.
11.3 The range of the sample temperature should be in accordance with the instrument
specifications. The sample temperature within this range should be constant.
12. DATA ANALYSIS AND CALCULATIONS
12.1 Report results as follows:
NTU Record to Nearest
0.01 - 1.0 0.01
1 - 10 0.025
13. METHOD PERFORMANCE
13.1 Prior to testing, all instruments were calibrated according to manufacturer’s
instructions. This was followed by running a QCS sample to verify the determinative
step for each turbidimeter was within the specified criteria.
13.2 On-line accuracy and precision testing with turbidity spikes – The Lovibond White
LED Method was conducted in a qualified laboratory. Since the test and reference
instruments require continuous sample flow, a pump injection system was used to
introduce spikes of turbidity. Each turbidity spike was of constant and stable turbidity
79
that was generated by the addition of a defined formazin standard pumped into the
sample stream at a constant flow rate. The sample itself passed through a 0.02-µm
pore size filter prior to being spiked. After the filtered sample was spiked with the
formazin, it traveled through a mixing coil that was then split into parallel feed lines
that led to the test and reference instruments. This provided a continuous parallel feed
of sample with a stable turbidity to both the test and reference instrument.
13.2.1 Changing the injection rate of the formazin standard that was spiked into the
filtered sample or changing the actual value of the formazin standard that was
spiked into the filtered sample yielded various stable turbidity values that were
continuously delivered to the test and reference instruments.
13.2.2 The injection rate of the turbidity standard was calculated in grams per minute
and the flow of the filtered sample was measured in grams per minute. This
allowed for the theoretical calculation of each turbidity spike. The instrument
response was calculated as a percent recovery of this spike. This data is presented
in Table 1 (17.1).
13.2.3 The sequence of spikes started with a turbidity free baseline and progressed with
increasing turbidity up to the highest turbidity level.
13.2.4 The pre-filtered water supply was a blend of hot and cold tap water that is
supplied by the city of Fort Collins. The source water was from mountain snow
runoff that was treated with conventional techniques and filtered using dual media
filtration.
13.2.5 Results were used for the initial demonstration of linearity of the measurement
system.
13.3 On-Line testing at public water utilities – The PTV 1000 turbidimeter (Lovibond WL
LED Method) was tested at two public water utilities. One utility was a surface
source water treatment plant that required an additional softening step. The second
plant treated a surface source water that went through an integrated micro-filtration
membrane as the filtration step. Both utilities are members of the Partnership for Safe
Water. The PTV 1000 turbidimeter and a reference turbidimeter that was compliant
with the USEPA laser nephelometry method was connected to the same source water
line for analysis. Both turbidimeters were operated for 24 hours, collecting data once
per 15 seconds. The deviation between the two instruments was 0.019 and 0.045 NTU
for these two plants respectively. Response time to changes in turbidity differed
slightly and was a function of flow rates, but the magnitude of response was
consistent.
13.4 The instruments were calibrated according to manufacturer’s instructions. After
calibration, a QCS was run on each instrument to verify that the determinative step
for each turbidimeter was within the specified criteria.
80
13.5 Accuracy of the Lovibond WL LED Method – Accuracy (bias) was presented as
percent recoveries relative to the theoretical values for the turbidity spikes compared
to results from the reference Hach Method 101333. Refer to Table 1 in section 17 for
a summary of all spike recovery and precision data.
13.5.1 Percent Recoveries for spikes in the 0 to 0.10 NTU range:
The average percent recoveries of turbidity for turbidity spikes in the 0.014 to
0.10 NTU range were:
Lovibond WL LED Method: 106.0%
Turbidimeter reference method 10133: 101.5%.
13.5.2 Percent Recoveries for spikes in the 0.10 to 1.0 NTU range
The average percent recoveries of turbidity for the spikes in the 0.10 to 1.0 NTU
range were:
Lovibond WL LED Method: 102.9%
Turbidimeter reference method 10133: 101.8%.
13.5.3 Percent Recoveries for spikes in the 1.0 to 10.0 NTU range
The average percent recoveries of turbidity for turbidity spikes from the 1.0 to 10
NTU range were:
Lovibond WL LED Method: 99.7%
Turbidimeter reference method 10133: 99.9%.
13.6 Precision Lovibond WL LED Method – Precision was presented here as the standard
deviation for each of the turbidity spikes compared to results from the reference Hach
Method 101333. Refer to Table 1 in section 17 for a summary of all spike recovery
and precision data.
13.6.1 Precision for spikes in the 0 to 0.10 NTU range
The standard deviation for the turbidity for turbidity in the 0.014 to 0.10 NTU
range were:
Lovibond WL LED Method: 0.0006 NTU
Turbidimeter reference method 10133: 0.0002 NTU.
13.6.2 Precision for spikes in the 0.10 to 1.0 NTU range
The standard deviation for the turbidity spikes in the 0.10 to 1.0 NTU range were:
Lovibond WL LED Method: 0.0045 NTU
Turbidimeter reference method 10133: 0.0034 NTU.
13.6.3 Precision for spikes in the 1.0 to 10.0 NTU range
The standard deviation for the turbidity spikes from the 1.0 to 10 NTU range
were:
Lovibond WL LED Method: 0.0220 NTU
Turbidimeter reference method 10133: 0.0141 NTU.
81
14. POLLUTION PREVENTION
14.1 Pollution prevention encompasses any technique that reduces or eliminates the
quantity or toxicity of waste as the point of generation. Numerous opportunities for
pollution prevention exist in laboratory operation. The EPA has established a
preferred hierarchy of environmental management techniques that places pollution
prevention as the management option of first choice. Whenever feasible, laboratory
personnel should use pollution-prevention techniques to address their waste
generation. When wastes cannot be feasibly reduced at the source, the Agency
recommends recycling as the next best option.
14.2 The quantity of chemicals purchased should be based on expected usage during its
shelf life and disposal cost of unused material. Actual reagent preparation volumes
should reflect anticipated usage and reagent stability.
15. WASTE MANAGEMENT
15.1 The U.S. Environmental Protection Agency requires that laboratory waste
management practices be conducted consistent with all applicable rules and
regulations. Excess reagents, samples and method process wastes should be
characterized and disposed of in an acceptable manner. The Agency urges
laboratories to protect air, water, and land by minimizing and controlling all releases
from hoods and bench operations; complying with the letter and spirit of any waste
discharge permit and regulations; and by complying with all solid hazardous waste
regulations, particularly the hazardous waste identification rules and land disposal
restrictions.
16. REFERENCES
1. Fourth Annual Report on Carcinogens (NTP85-002 1985), p115
2. USEPA. “Method 180.1 Determination of Turbidity by Nephelometry,” Revision 2.0,
(1993).
3. Hach Company, “Hach Method 10133, “Determination of Turbidity by Laser
Nephelometry”, (2000).
17. TABLES AND VALIDATION DATA
17.1 Summarized precision and bias (percent recovery) data in tabular form (Table 1).
82
Table 1 – Results Table for the Percent Recovery and Precision with Respect to Turbidity Spikes for the PTV1000 WL LED
Turbidimeter
PTV 1000 WL Reading
Reference Turbidimeter Reading (EPA approved
Method 10133)
Spike #
Baselin
e
(Blank)
Theoretic
al Value of Spike
in NTU
Response (blank
Correcte
d) in NTU
Recovery (%)
Precision (SD)
Baselin
e
(Blank)
Theoretic
al Value of Spike
in NTU
Response (blank
Correcte
d) in NTU
Recovery (%)
Precision (SD)
N (both
Test and Referenc
e)
1 0.024 0.014 0.016 110.8 0.0009
0.010 0.014 0.014 102.3 0.0001 60
2 0.024 0.028 0.030 104.9 0.0002
0.010 0.028 0.029 102.6 0.0002 60
3 0.024 0.060 0.062 102.3 0.0007
0.010 0.060 0.060 99.4 0.0002 60
4 0.024 0.110 0.114 103.4 0.0014
0.010 0.110 0.112 102.0 0.0005 60
5 0.024 0.225 0.235 104.5 0.0021
0.010 0.225 0.232 103.3 0.0014 60
6 0.024 0.554 0.564 101.7 0.0051
0.010 0.554 0.559 100.8 0.0036 60
7 0.024 0.921 0.940 102.1 0.0094
0.010 0.921 0.930 101.0 0.0079 60
8 0.024 1.895 1.929 101.8 0.0081
0.010 1.895 1.928 101.8 0.0069 60
9 0.024 3.568 3.444 96.5 0.0267
0.010 3.568 3.463 97.1 0.0077 60
10 0.024 9.376 9.437 100.6 0.0313
0.010 9.376 9.472 101.0 0.0277 60
83
Appendix B.2
Continuous Measurement of Drinking Water
Turbidity Using a Lovibond PTV 2000 660-nm
LED Turbidimeter
The Lovibond 660-nm LED Method
Revision 1.0
December 20, 2016
Tintometer Inc.
6456 Parkland Drive
Sarasota, FL 34243
84
Continuous Measurement of Drinking Water Turbidity Using a
Lovibond PTV 2000 Red LED Turbidimeter
1. SCOPE AND APPLICATION
1.1 This method is applicable to any colorless drinking water samples with a turbidity
less than 10 Nephelometric Turbidity Units (NTU).
1.2 The applicable range is from 0 to 10 NTU.
1.3 The method meets the requirements for compliance monitoring and reporting as
demanded under the Safe Drinking Water Act (SDWA).
2. SUMMARY OF METHOD
2.1 The method is based upon a comparison of the intensity of a collimated beam of light
that is generated by a 660-nm LED, that is scattered by the sample under defined
conditions with the intensity of the same 660-nm light LED scattered by a standard
reference suspension. The higher the intensity of scattered light, the higher the
turbidity. Readings, in NTU, are made in a nephelometer designed according to
specifications given in section 6.2.
2.2 Formazin, prepared under closely defined conditions, is used as a primary standard
suspension to calibrate the instrument. However, other approved primary standards
may be used with this method.
2.2.1 Examples of standards that can be used to calibrate the instrument include
dilutions from commercially available 4000 NTU formazin, stabilized versions of
formazin with preassigned turbidity values such as T-Cal™, and styrene
divinylbenzene suspensions with preassigned values for the specific make and
model of the instrument to be calibrated.
2.3 The method generates a linear response between scattered incident light that is
detected at 90-degrees over the applicable range. The method defines 0.00 NTU as no
light impinging on the 90-degree detector and requires at least one defined standard to
perform a calibration over the applicable range.
3. DEFINITIONS
3.1 LABORATORY REAGENT BLANK (LRB) – An aliquot of reagent water or other
blank matrices that are treated exactly as a sample including exposure to all
glassware, equipment, solvents, reagents, internal standards, and surrogates that are
85
used with other samples. The LRB is used to determine if method analytes or other
interferences are present in the laboratory environment, the reagents, or the apparatus.
3.2 LINEAR CALIBRATION RANGE (LCR) – The concentration range over which the
instrument response is linear.
3.3 SAFETY DATA SHEET (SDS) – Written information provided by vendors
concerning a chemical/s toxicity, health hazards, physical properties, fire, and
reactivity data including storage, spill, and handling precautions.
3.4 PRIMARY CALIBRATION STANDARD (PCAL) – A suspension prepared from
the primary dilution stock standard suspension. The PCAL suspensions are used to
calibrate the instrument response with respect to analyte concentration.
3.5 QUALITY CONTROL SAMPLE (QCS) - A solution of the method analyte of known
concentrations that is used to fortify an aliquot of LRB matrix. The QCS is obtained
from a source external to the laboratory, and is used to check laboratory performance.
3.6 SECONDARY CALIBRATION STANDARDS (SCAL) – commercially prepared,
stabilized sealed liquid or gel turbidity standards, or other apparatus or mechanism
calibrated against properly prepared and diluted Formazin or styrene divinylbenzene
polymers.
3.7 STOCK STANDARD SUSPENSION (SSS) – A concentrated suspension containing
the analytic solution prepared in the laboratory using assayed reference materials or
purchased from a reputable commercial source. Stock standard suspensions are used
to prepare calibrant suspensions or other needed values of suspensions.
4. INTERFERENCES
4.1 The presence of floating debris and coarse particulate matter within the sample may
settle out of suspension resulting in low turbidity readings.
4.2 Finely divided air bubbles will cause random high spikes in readings.
4.3 The presence of dissolved, light absorbing substances or chemicals in the sample, i.e.
the presence of color, can absorb portions of the incident light spectra, resulting in
low turbidity readings, although this effect is generally not significant for drinking
water.
4.4 Light-absorbing particles in suspension within the sample, such as activated carbon of
significant concentration, can cause low readings.
86
4.5 Construction materials of the nephelometric device within the measurement chamber
can result in elevated stray light due to spurious reflections of the incident beam can
cause a false positive bias at the bottom end of the range.
5. SAFETY
5.1 The toxicity or carcinogenicity of each reagent used in this method has not been fully
established. Each chemical should be regarded as a potential health hazard and
exposure should be as low as reasonably achievable.
5.2 Each laboratory is responsible for maintaining a current awareness file of OSHA
regulations regarding the safe handling of the chemicals specified in this method. A
reference file of Safety Data Sheets (SDS) should be made available to all personnel
involved in the chemical analysis. The preparation of a formal safety plan is also
advisable.
5.3 Hydrazine sulfate (7.2.1) may reasonably be anticipated to be a carcinogen1.
Formazin can contain residual hydrazine sulfate. Proper protection should be
employed.
6. EQUIPMENT
6.1 The installation shall be according to the manufacturer’s instructions.
6.2 The turbidimeter shall consist of a nephelometer with a light source for illuminating
the sample, one or more photo-detectors to measure the amount of scattered light at a
right angle to the incident beam, a correlation means to relate the amount of scattered
light to a known turbidity standard, and a communication means to convey the
turbidity value to the plant operator or other responsible water authority.
6.3 Differences in the physical design of the turbidimeter will cause differences in
measured values for turbidity, even though the same suspension is used for
calibration. To minimize such differences, the following design criteria shall be
observed:
6.3.1 The light source shall be a light emitting diode (LED) with a peak emitting
wavelength between 650 and 670 nm. The optical system, including the emitter,
all optical elements and detector(s), shall have a spectral peak response between
600 nm and 700 nm.
6.3.2 The rays comprising the incident beam shall be parallel with no divergence and
not to exceed 1 degree of convergence within the measurement volume.
87
6.3.3 Non-scattered or non-attenuated light of the incident beam after passing through
the sample shall pass into a light trap.
6.3.4 Distance traversed by incident light and scattered light not to exceed 10 cm.
6.3.5 Detector/Light Receiver shall be centered at 90º ± 1 ½ degrees to the incident
beam.
6.3.6 Scattered light shall be received by the detector/light receiver at a subtended angle
between 20 and 30 degrees from the center-point of the measurement volume.
6.3.7 The detector/light receiver shall have a spectral response that encompasses the
peak spectral output of the incident light source.
6.3.8 A linear-based algorithm shall be used to convert detector signal to NTU in
correlation to a known calibration standard.
6.3.9 The turbidimeter shall be free from significant drift after a short warm-up period.
6.3.10 The sensitivity of the instrument shall detect differences in turbidity of 0.01 NTU.
6.3.11 The instrument shall be capable of measuring from 0 to 10 NTU turbidity units.
Several ranges may be necessary to obtain both adequate coverage and sufficient
sensitivity for low turbidities.
6.3.12 Instrument shall incorporate a sample deaerator for the removal of entrained air
from the sample stream.
6.4 A nephelometric device that meets these specifications is a Lovibond PTV 2000
turbidimeter.
7. REAGENTS AND STANDARDS
7.1 Reagent water, turbidity-free: Pass deionized distilled water through a 0.2-µm or
smaller pore-size membrane filter if necessary. Water produced by reverse osmosis is
acceptable. Water should have a turbidity that is ≤0.030 NTU. This value should be
considered when preparing standards.
7.2 Stock standard suspension (Formazin) 4000 NTU
7.2.1 Dissolve 5.00 g hydrazine sulfate (CASRN 10034-93-2) into approximately 400
mL of reagent water contained in a cleaned 1-L Class A volumetric flask.
88
7.2.2 Dissolve 50.00 g hexamethylenetetramine (CASRN 100-97-0) in approximately
300 mL of reagent water contained in a 500-mL volumetric flask.
7.2.3 Quantitatively transfer the hexamethylenetetramine solution (7.2.2) into the 1-L
flask that contains the dissolved hydrazine sulfate solution (7.2.1).
7.2.4 Dilute the 1-L flask (7.2.3) to volume with reagent water.
7.2.5 Stopper and mix by inversion for 10 minutes.
7.2.6 Allow to stand 24 hours at 25 ± 2 °C. During this time, the formazin polymer will
develop. The turbidity of this standard is 4000 NTU.
7.2.7 Store this solution in the dark and away from a source of heat. Bring the solution
to room temperature and thoroughly mix before preparing dilutions (7.3)
7.3 Primary calibration standards: Using pipetts with an accuracy of 1 percent or better,
first mix and then dilute 25.0 mL of stock standard suspension (7.2) to 0.10 L with
reagent water. The turbidity of this suspension is defined as 1000 NTU. For other
turbidity values, mix and dilute portions of this suspension as required using clean
Class A glassware.
7.3.1 A new stock standard suspension (7.2) should be prepared each quarter. Primary
calibration standards (7.3) should be prepared daily by dilution of the stock
standard suspension.
7.4 Formazin in commercially prepared, certified, concentrated stock standard suspension
(SSS) may be diluted and used as required. Dilute turbidity standards should be
prepared daily.
7.5 Stabilized formazin suspensions or styrene divinylbenzene (SDVB) polymers are
commercially prepared, certified, and ready to use dilutions. Manufacturer’s
instructions should be followed for choosing the appropriate standard values for the
instrument.
7.6 Secondary standards may be acceptable as a calibration check, but must be monitored
on a routine basis for deterioration and replaced as required.
8. SAMPLE COLLECTION AND INSTRUMENT SETUP
8.1 Online instrumentation does not require sample cooling, preservation or storage.
8.2 Install and set up the instrument according to the manufacturer’s instructions.
89
9. QUALITY CONTROL
9.1 Each laboratory using this method is required to operate a formal quality control (QC)
program. The minimum requirements of this program consist of an initial
demonstration of the process turbidimeter system’s capability and analysis of
laboratory reagent blanks and other solutions as a continuing check on performance.
The laboratory is required to maintain performance records that define the quality of
data generated.
9.2 Initial demonstration of performance
9.2.1 The initial demonstration of performance is used to characterize instrument
performance as determined by the LCR and QCS analyses.
9.2.2 Linear Calibration Range (LCR) – The LCR must be determined initially and
verified whenever a significant change in instrument response is observed or
expected. The initial demonstration of linearity must use sufficient standards to
insure that the resulting curve is linear. The verification of linearity must use a
minimum of a blank and three standards. For example, the standards could be
0.30, 5.0 and 10.0 NTU. If any verification data exceeds the initial values by ±
10% or exceeds the stated specifications of the turbidity standard, whichever is
greater, linearity must be reestablished. If any portion of the range is shown to be
nonlinear, sufficient standards must be used to clearly define the nonlinear
portion. The selection of standards should at least cover the range of turbidity
values that are expected from samples.
9.2.3 Quality Control Sample (QCS) – When beginning the use of this method, on a
quarterly basis or as required to meet data-quality needs, verify the calibration
standards and acceptable instrument performance with the preparation and
analysis of a QCS. If the determined concentrations are not within ± 10 % or ±
0.030 NTU of the stated QCS values, performance of the determinative step of the
method is unacceptable. The source of the problem must be identified and
corrected before continuing with on-going analyses.
9.3 Accuracy and precision should be checked on a routine basis to monitor the overall
performance of the instrument. A series of reagent blanks and check standards should
be run to validate the quality of sample data. These checks should occur at a
frequency that is required for regulatory compliance.
9.3.1 A QCS sample as described in 9.2.3 must be run at least on a quarterly interval.
The instrument must be checked to insure it has been cleaned and maintained
according to the manufacturers recommendations prior to running the QCS.
9.3.2 Solid standards are an option and can only be used for verification purposes. The
instrument should be checked to insure it has been cleaned and maintained
90
according to the manufacturers recommendations prior to running a solid
verification standard.
10. CALIBRATION AND STANDARDIZATION
10.1 Turbidity Calibration: The manufacturer’s operating instructions should be followed
for calibration. Perform any cleaning and maintenance prior to calibration as per
manufacturer’s instructions. The turbidimeter measurement chamber should be rinsed
with at least 1-liter of water that has been filtered through a 0.45-um filter or smaller
prior to calibration. Calibration should be performed under the same ambient
conditions as sample measurement.
10.2 Measure standards on the turbidimeter covering the range of interest. If the
instrument is already calibrated in standard turbidity units, this procedure will check
the accuracy of the calibration scales.
10.3 At least one standard should be run in each instrument range to be used. Some
instruments permit adjustments of sensitivity so that scale values will correspond to
turbidities.
10.4 Solid standards can only be used for verification purposes. If used, they must be
protected from surface scratches which may cause potential changes.
10.5 If a pre-calibrated scale is not supplied, calibration curves should be prepared for each
range of the instrument. Calibration must be performed under identical optical
conditions as operational conditions.
11. PROCEDURE
11.1 A sample from the treatment process is taken and flows through the turbidimeter for
measurement. It is then drained or recycled back into the process after the
measurement has been taken.
11.2 The sample flow rate shall be in accordance with the instrument specifications. The
sample flow to the instrument shall be constant without variations due to pressure
changes or surges. Installation of a flow control device such as a rotameter in the
sample line can eliminate fluctuations of the flow rate.\
11.3 The range of the sample temperature should be in accordance with the instrument
specifications. The sample temperature within this range should be constant.
12. DATA ANALYSIS AND CALCULATIONS
91
12.1 Report results as follows:
NTU Record to Nearest
0.01 - 1.0 0.01
1 - 10 0.1
13. METHOD PERFORMANCE
13.1 Prior to testing, all instruments were calibrated according to manufacturer’s
instructions. This was followed by running a QCS sample to verify the determinative
step for each turbidimeter was within the specified criteria.
13.2 On-line accuracy and precision testing with turbidity spikes. The Lovibond 660-nm
LED Method was conducted in a qualified laboratory. Since the test and reference
instruments require continuous sample flow, a pump injection system was used to
introduce spikes of turbidity. Each turbidity spike was of constant and stable turbidity
that was generated by the addition of a defined formazin standard pumped into the
sample stream at a constant flow rate. The sample itself passed through a 0.02-µm
pore size filter prior to being spiked. After the filtered sample was spiked with the
formazin, it traveled through a mixing coil that was then split into parallel feed lines
that led to the test and reference instruments. This provided a continuous parallel feed
of sample with a stable turbidity to both the test and reference instrument.
13.2.1 Changing the injection rate of the formazin standard that was spiked into the
filtered sample or changing the actual value of the formazin standard that was
spiked into the filtered sample yielded various stable turbidity values that were
continuously delivered to the test and reference instruments.
13.2.2 The injection rate of the turbidity standard was calculated in grams per minute
and the flow of the filtered sample was measured in grams per minute. This
allowed for the theoretical calculation of each turbidity spike. The instrument
response was calculated as a percent recovery of this spike. This data is presented
in Table 1 (17.1)
13.2.3 The sequence of spikes started with a turbidity free baseline and progressed with
increasing turbidity up to the highest turbidity level.
13.2.4 The pre-filtered water supply was a blend of hot and cold tap water that is
supplied by the city of Fort Collins. The source water was from mountain snow
runoff that was treated with conventional techniques and filtered using dual media
filtration.
13.2.5 Results were used for the initial demonstration of linearity of the measurement
system.
92
13.3 On-Line testing at public water utilities – The PTV 2000 660-nm LED Method was
tested at two public water utilities. One utility was a surface source water treatment
plant that required an additional softening step. The second plant treated a surface
source water that went through an integrated micro-filtration membrane as the
filtration step. Both utilities are members of the Partnership for Safe Water. The
Lovibond 660-nm LED turbidimeter and a reference turbidimeter that was compliant
with the USEPA laser nephelometry method was connected to the same source water
line for analysis. Both turbidimeters were operated for 24 hours, collecting data once
per 15 seconds. The deviation between the two instruments was 0.002 and 0.009 NTU
for the two plants respectively. Response time to changes in turbidity differed slightly
and was a function of flow rates, but the magnitude of response was consistent.
13.4 The instruments were calibrated according to manufacturer’s instructions. After
calibration a QCS was run on each instrument to verify that the determinative step for
each turbidimeter was within the specified criteria.
13.5 Accuracy of the Lovibond 660-nm LED Method – Accuracy (bias) was presented as
percent recoveries relative to theoretical values for the turbidity spikes compared to
results from the reference Hach Method 101333. Refer to Table 1 in Section 17 for a
summary of all spike recovery and precision data.
13.5.1 Percent Recoveries for spikes in the 0 to 0.10 NTU range:
The average percent recoveries of turbidity for turbidity spikes in the 0.014 to
0.10 NTU range were:
Lovibond 660-nm LED Method: 103.5 %
Turbidimeter reference method 10133: 101.5 %.
13.5.2 Percent Recoveries for spikes in the 0.10 to 1.0 NTU range
The average percent recoveries of turbidity for the spikes in the 0.10 to 1.0 NTU
range were:
Lovibond 660-nm LED Method: 101.0 %
Turbidimeter reference method 10133: 101.8%.
13.5.3 Percent Recoveries for spikes in the 1.0 to 10.0 NTU range
The average percent recoveries of turbidity for turbidity spikes from the 1.0 to 10
NTU range were:
Lovibond 660-nm LED Method: 97.0 %
Turbidimeter reference method 10133: 99.9 %.
13.6 Precision Lovibond 660-nm LED Method Precision was presented here as the
standard deviation for each of the turbidity spikes compared to results from the
reference Hach Method 101333. Refer to Table 1 in Section 17 for a summary of all
spike recovery and precision data.
13.6.1 Precision for spikes in the 0 to 0.10 NTU range
93
The standard deviation for the turbidity for turbidity spikes in the 0.014 to 0.10
NTU range were:
Lovibond 660-nm LED Method: 0.0003 NTU
Turbidimeter reference method 10133: 0.0002 NTU.
13.6.2 Precision for spikes in the 0.10 to 1.0 NTU range
The standard deviation for the turbidity spikes in the 0.10 to 1.0 NTU range were:
Lovibond 660-nm LED Method: 0.0040 NTU
Turbidimeter reference method 10133: 0.0034 NTU.
13.6.3 Precision for spikes in the 1.0 to 10.0 NTU range
The standard deviation for the turbidity spikes from the 1.0 to 10 NTU range
were:
Lovibond 660-nm LED Method: 0.0156 NTU
Turbidimeter reference method 10133: 0.0141 NTU.
14. POLLUTION PREVENTION
14.1 Pollution prevention encompasses any technique that reduces or eliminates the
quantity or toxicity of waste as the point of generation. Numerous opportunities for
pollution prevention exist in laboratory operation. The EPA has established a
preferred hierarchy of environmental management techniques that places pollution
prevention as the management option of first choice. Whenever feasible, laboratory
personnel should use pollution-prevention techniques to address their waste
generation. When wastes cannot be feasibly reduced at the source, the Agency
recommends recycling as the next best option.
14.2 The quantity of chemicals purchased should be based on expected usage during its
shelf life and disposal cost of unused material. Actual reagent preparation volumes
should reflect anticipated usage and reagent stability.
15. WASTE MANAGEMENT
15.1 The U.S. Environmental Protection Agency requires that laboratory waste
management practices be conducted consistent with all applicable rules and
regulations. Excess reagents, samples and method process wastes should be
characterized and disposed of in an acceptable manner. The Agency urges
laboratories to protect air, water, and land by minimizing and controlling all releases
from hoods and bench operations; complying with the letter and spirit of any waste
discharge permit and regulations; and by complying with all solid hazardous waste
regulations, particularly the hazardous waste identification rules and land disposal
restrictions.
94
16. REFERENCES
1. Fourth Annual Report on Carcinogens (NTP85-002 1985), p115
2. USEPA. “Method 180.1 Determination of Turbidity by Nephelometry,” Revision 2.0,
(1993).
3. Hach Company, “Hach Method 10133, “Determination of Turbidity by Laser
Nephelometry”, (2000).
17. TABLES AND VALIDATION DATA
17.1 Summarized precision and bias (percent recovery) data in tabular form (Table 1).
Table 1 – Results Table for the Percent Recovery to Turbidity Spikes for the PTV2000 660-nm LED Turbidimeter
PTV 2000 660-nm LED Reading
Reference Turbidimeter Reading (EPA approved
Method 10133)
Spike #
Baselin
e (Blank)
in NTU
Theoretic
al Value of Spike
in NTU
Response
(blank Correcte
d)
Recover
y
(%)
Precision (SD)
Baselin
e (Blank)
in NTU
Theoretic
al Value of Spike
in NTU
Response
(blank Correcte
d)
Recovery (%)
Precision (SD)
N (both
Test and Referenc
e)
1 0.013 0.014 0.015 104.5 0.0002
0.010 0.014 0.014 102.3 0.0001 60
2 0.013 0.028 0.029 104.3 0.0003
0.010 0.028 0.029 102.6 0.0002 60
3 0.013 0.060 0.061 101.5 0.0003
0.010 0.060 0.060 99.4 0.0002 60
4 0.013 0.110 0.112 101.4 0.0012
0.010 0.110 0.112 102.0 0.0005 60
5 0.013 0.225 0.232 103.3 0.0020
0.010 0.225 0.232 103.3 0.0014 60
6 0.013 0.554 0.553 99.7 0.0040
0.010 0.554 0.559 100.8 0.0036 60
7 0.013 0.921 0.918 99.7 0.0088
0.010 0.921 0.930 101.0 0.0079 60
8 0.013 1.895 1.858 98.1 0.0077
0.010 1.895 1.928 101.8 0.0069 60
9 0.013 3.568 3.344 93.7 0.0140
0.010 3.568 3.463 97.1 0.0077 60
10 0.013 9.376 9.304 99.2 0.0251
0.010 9.376 9.472 101.0 0.0277 60
95
Appendix B.3
Continuous Measurement of Drinking Water
Turbidity Using a Lovibond PTV 6000 Laser
Turbidimeter
The Lovibond 6000 Laser Method
Revision 1.0
December 20, 2016
Tintometer Inc.
6456 Parkland Drive
Sarasota, FL 34243
96
Continuous Measurement of Drinking Water Turbidity Using a
Lovibond PTV 6000 Laser Turbidimeter
1. SCOPE AND APPLICATION
1.1 This method is applicable to any colorless drinking water samples with a
turbidity less than 10 Nephelometric Turbidity Units (NTU).
1.2 The applicable range is from 0 to 10 NTU.
1.3 The method meets the requirements for compliance monitoring and reporting as
demanded under the Safe Drinking Water Act (SDWA).
2. SUMMARY OF METHOD
2.1 The method is based upon a comparison of the intensity of a collimated beam of
light that is generated by a 685-nm laser diode, that is scattered by the sample
under defined conditions with the intensity of the same laser diode light
scattered by a standard reference suspension. The higher the intensity of
scattered light, the higher the turbidity. Readings, in NTU, are made in a
nephelometer designed according to specifications given in section 6.2.
2.2 Formazin, prepared under closely defined conditions, is used as a primary
standard suspension to calibrate the instrument. However, other approved
primary standards may be used with this method.
2.2.1 Examples of standards that can be used to calibrate the instrument include
freshly prepared dilutions from commercially available 4000 NTU
formazin, stabilized versions of formazin with preassigned turbidity values
such as T-Cal™, and styrene divinylbenzene suspensions with preassigned
values for the specific make and model of the instrument to be calibrated.
2.3 The method generates a linear response between scattered incident light that is
detected at 90-degrees over the applicable range. The method defines 0.00 NTU
as no light impinging on the 90-degree detector and requires at least one defined
standard to perform a calibration over the applicable range.
3. DEFINITIONS
3.1 LABORATORY REAGENT BLANK (LRB) – An aliquot of reagent water or
other blank matrices that are treated exactly as a sample including exposure to
97
all glassware, equipment, solvents, reagents, internal standards, and surrogates
that are used with other samples. The LRB is used to determine if method
analytes or other interferences are present in the laboratory environment, the
reagents, or the apparatus.
3.2 LINEAR CALIBRATION RANGE (LCR) – The concentration range over
which the instrument response is linear.
3.3 SAFETY DATA SHEET (SDS) – Written information provided by vendors
concerning a chemical/s toxicity, health hazards, physical properties, fire, and
reactivity data including storage, spill, and handling precautions.
3.4 PRIMARY CALIBRATION STANDARD (PCAL) – A suspension prepared
from the primary dilution stock standard suspension. The PCAL suspensions
are used to calibrate the instrument response with respect to analyte
concentration.
3.5 QUALITY CONTROL SAMPLE (QCS) - A solution of the method analyte of
known concentrations that is used to fortify an aliquot of LRB matrix. The QCS
is obtained from a source external to the laboratory, and is used to check
laboratory performance.
3.6 SECONDARY CALIBRATION STANDARDS (SCAL) – commercially
prepared, stabilized sealed liquid or gel turbidity standards, or other apparatus
or mechanism calibrated against properly prepared and diluted Formazin or
styrene divinylbenzene polymers.
3.7 STOCK STANDARD SUSPENSION (SSS) – A concentrated suspension
containing the analytic solution prepared in the laboratory using assayed
reference materials or purchased from a reputable commercial source. Stock
standard suspensions are used to prepare calibrant suspensions or other needed
values of suspensions.
4. INTERFERENCES
4.1 The presence of floating debris and coarse particulate matter within the sample
may settle out of suspension resulting in low turbidity readings.
4.2 Finely divided air bubbles will cause random high spikes in readings.
4.3 The presence of dissolved, light absorbing substances or chemicals in the
sample, i.e. the presence of color, can absorb portions of the incident light
spectra, resulting in low turbidity readings, although this effect is generally not
significant for drinking water.
98
4.4 Light-absorbing particles in suspension within the sample, such as activated
carbon of significant concentration, can cause low readings.
4.5 Construction materials of the nephelometric device within the measurement
chamber can result in elevated stray light due to spurious reflections of the
incident beam can cause a false positive bias at the bottom end of the range.
5. SAFETY
5.1 The toxicity or carcinogenicity of each reagent used in this method has not been
fully established. Each chemical should be regarded as a potential health hazard
and exposure should be as low as reasonably achievable.
5.2 Each laboratory is responsible for maintaining a current awareness file of
OSHA regulations regarding the safe handling of the chemicals specified in this
method. A reference file of Safety Data Sheets (SDS) should be made available
to all personnel involved in the chemical analysis. The preparation of a formal
safety plan is also advisable.
5.3 Hydrazine sulfate (7.2.1) may reasonably be anticipated to be a carcinogen.
Formazin can contain residual hydrazine sulfate. Proper protection should be
employed.
5.4 This device contains a laser diode light source. The device has been designed
with safety interlocks. Follow instructions for use and never attempt to defeat
the safety interlocks on this device.
6. EQUIPMENT
6.1 The installation shall be according to the manufacturer’s instructions.
6.2 The turbidimeter shall consist of a nephelometer with a light source for
illuminating the sample, one or more photo-detectors to measure the amount of
scattered light at a right angle to the incident beam, a correlation means to relate
the amount of scattered light to a known turbidity standard, and a
communication means to convey the turbidity value to the plant operator or
other responsible water authority.
6.3 Differences in the physical design of the turbidimeter will cause differences in
measured values for turbidity, even though the same suspension is used for
calibration. To minimize such differences, the following design criteria shall be
observed:
99
6.3.1 The incident light source (emitter) shall be a laser diode with a peak
emitting center wavelength between 650 and 690 nm. The optical system,
including the emitter, all optical elements and detector(s), shall have a
spectral peak response between 600 nm and 700 nm.
6.3.2 The rays comprising the incident beam shall be parallel with no divergence
and not to exceed 1 degree of convergence within the measurement volume.
6.3.3 Non-scattered or non-attenuated light of the incident beam after passing
through the sample shall pass into a light.
6.3.4 Distance traversed by incident light and scattered light not to exceed 10 cm.
6.3.5 Detector/Light Receiver shall be centered at 90º ± 1 ½ degrees to the
incident beam.
6.3.6 Scattered light shall be received by the detector/light receiver at a subtended
angle between 20 and 30 degrees from the center-point of the measurement
volume.
6.3.7 The detector/light receiver shall have a spectral response that encompasses
the peak spectral output of the incident light source.
6.3.8 A linear-based algorithm shall be used to convert detector signal to NTU in
correlation to a known calibration standard.
6.3.9 The turbidimeter shall be free from significant drift after a short warm-up
period.
6.3.10 The sensitivity of the instrument shall detect differences in turbidity of 0.01
NTU or less in waters less than 1 NTU.
6.3.11 The instrument shall be capable of measuring from 0 to 10 NTU turbidity
units. Several ranges may be necessary to obtain both adequate coverage
and sufficient sensitivity for low turbidities.
6.3.12 Instrument shall incorporate a sample deaerator for the removal of entrained
air from the sample stream.
6.4 A nephelometric device that meets these specifications is a Lovibond PTV 6000
turbidimeter.
7. REAGENTS AND STANDARDS
100
7.1 Reagent water, turbidity-free: Pass deionized distilled water through a 0.2-µm
or smaller pore-size membrane filter if necessary. Water produced by reverse
osmosis is acceptable. Water should have a turbidity that is ≤0.030 NTU. This
value should be considered when preparing standards.
7.2 Stock standard suspension (Formazin) 4000 NTU
7.2.1 Dissolve 5.00 g hydrazine sulfate (CASRN 10034-93-2) into approximately
400 mL of reagent water contained in a cleaned 1-L Class A volumetric
flask.
7.2.2 Dissolve 50.00 g hexamethylenetetramine (CASRN 100-97-0) in
approximately 300 mL of reagent water contained in a 500-mL volumetric
flask.
7.2.3 Quantitatively transfer the hexamethylenetetramine solution (7.2.2) into the
1-L flask that contains the dissolved hydrazine sulfate solution (7.2.1).
7.2.4 Dilute the 1-L flask (7.2.3) to volume with reagent water.
7.2.5 Stopper and mix by inversion for 10 minutes.
7.2.6 Allow to stand 24 hours at 25 ± 2 °C. During this time, the formazin
polymer will develop. The turbidity of this standard is 4000 NTU.
7.2.7 Store this solution in the dark and away from a source of heat. Bring the
solution to room temperature and thoroughly mix before preparing dilutions
(7.3)
7.3 Primary calibration standards: Using a pipet with accuracy to 1 percent or
better, first mix and then dilute 25.0 mL of stock standard suspension (7.2) to
0.10 L with reagent water. The turbidity of this suspension is defined as 1000
NTU. For other turbidity values, mix and dilute portions of this suspension as
required using clean Class A glassware.
7.3.1 A new stock standard suspension (7.2) should be prepared each quarter.
Primary calibration standards (7.3) should be prepared daily by dilution of
the stock standard suspension.
7.4 Formazin in commercially prepared, certified, concentrated stock standard
suspension (SSS) may be diluted and used as required. Dilute turbidity
standards should be prepared daily.
101
7.5 Stabilized formazin suspensions or styrene divinylbenzene (SDVB) polymers
are commercially prepared, certified, and ready to use dilutions. Manufacturer’s
instructions should be followed for choosing the appropriate standard values for
the instrument.
7.6 Secondary standards may be acceptable as a calibration check, but must be
monitored on a routine basis for deterioration and replaced as required.
8. SAMPLE COLLECTION AND INSTRUMENT SETUP
8.1 Online instrumentation does not require sample cooling, preservation or
storage.
8.2 Install and set up the instrument according to the manufacturer’s instructions.
9. QUALITY CONTROL
9.1 Each laboratory using this method is required to operate a formal quality control
(QC) program. The minimum requirements of this program consist of an initial
demonstration of the process turbidimeter system’s capability and analysis of
laboratory reagent blanks and other solutions as a continuing check on
performance. The laboratory is required to maintain performance records that
define the quality of data generated.
9.2 Initial demonstration of performance
9.2.1 The initial demonstration of performance is used to characterize instrument
performance as determined by the LCR and QCS analyses.
9.2.2 Linear Calibration Range (LCR) – The LCR must be determined initially
and verified whenever a significant change in instrument response is
observed or expected. The initial demonstration of linearity must use
sufficient standards to insure that the resulting curve is linear. The
verification of linearity must use a minimum of a blank and three standards.
For example, standards could be 0.3, 5.0 and 10.0 NTU. If any verification
data exceeds the initial values by ± 10% or exceeds the stated specifications
of the turbidity standard, whichever is greater, linearity must be
reestablished. If any portion of the range is shown to be nonlinear, sufficient
standards must be used to clearly define the nonlinear portion. The selection
of standards should at least cover the range of turbidity values that are
expected from samples.
102
9.2.3 Quality Control Sample (QCS) – When beginning the use of this method, on
a quarterly basis or as required to meet data-quality needs, verify the
calibration standards and acceptable instrument performance with the
preparation and analysis of a QCS. If the determined concentrations are not
within ± 10% or ± 0.030 NTU of the stated QCS values, performance of the
determinative step of the method is unacceptable. The source of the problem
must be identified and corrected before continuing with on-going analyses.
9.3 Accuracy and precision should be checked on a routine basis to monitor the
overall performance of the instrument. A series of reagent blanks and check
standards should be run to validate the quality of sample data. These checks
should occur at a frequency that is required for regulatory compliance.
9.3.1 A QCS sample as described in 9.2.3 must be run at least on a quarterly
interval. The instrument must be checked to insure it has been cleaned and
maintained according to the manufacturers recommendations prior to
running the QCS.
9.3.2 Solid standards are an option and can only be used for verification purposes.
The instrument should be checked to insure it has been cleaned and
maintained according to the manufacturers recommendations prior to
running a solid verification standard.
10. CALIBRATION AND STANDARDIZATION
10.1 Turbidity Calibration: The manufacturer’s operating instructions should be
followed for calibration. Perform any cleaning and maintenance prior to
calibration as per manufacturer’s instructions. The turbidimeter measurement
chamber should be rinsed with at least 1 L of water that is filtered through a
0.45-um filter or smaller prior to calibration. Calibration should be performed
under the same ambient conditions as sample measurement.
10.2 Measure standards on the turbidimeter covering the range of interest. If the
instrument is already calibrated in standard turbidity units, this procedure will
check the accuracy of the calibration scales.
10.3 At least one standard should be run in each instrument range to be used. Some
instruments permit adjustments of sensitivity so that scale values will
correspond to turbidities.
10.4 Solid standards can only be used for verification purposes. If used, they must be
protected from surface scratches which may cause potential changes.
103
10.5 If a pre-calibrated scale is not supplied, calibration curves should be prepared
for each range of the instrument. Calibration must be performed under identical
optical conditions as operational conditions.
11. PROCEDURE
11.1 A sample from the treatment process is taken and flows through the
turbidimeter for measurement. It is then taken to drain or recycled back into the
process after the measurement has been taken.
11.2 The sample flow rate shall be in accordance with the instrument specifications.
The sample flow to the instrument shall be constant without variations due to
pressure changes or surges. Installation of a flow control device such as a
rotameter in the sample line can eliminate fluctuations of the flow rate.
11.3 The range of the sample temperature should be in accordance with the
instrument specifications. The sample temperature within this range should be
constant.
12. DATA ANALYSIS AND CALCULATIONS
12.1 Report results as follows:
NTU Record to Nearest
0.01 - 1.00 0.01
1 - 10 0.05
13. METHOD PERFORMANCE
13.1 Prior to testing, all instruments were calibrated according to manufacturer’s
instructions. This was followed by running a QCS sample to verify the
determinative step for each turbidimeter was within the specified criteria.
13.2 On-Line Accuracy and Precision Testing with turbidity spikes – The Lovibond
6000 Laser Method was conducted in a qualified laboratory. Since the test and
reference instruments require continuous sample flow, a pump injection system
was used to introduce spikes of turbidity. Each turbidity spike was of constant
and stable turbidity that was generated by the addition of a defined formazin
standard that is pumped into the sample stream at a constant flow rate. The
sample itself passed through a 0.02 µm pore size filter prior to being spiked.
After the filtered sample was spiked with the formazin, it traveled through a
mixing coil that was then split into parallel feed lines that led to the test and
104
reference instruments. This provided a continuous parallel feed of sample with a
stable turbidity to both the test and reference instrument.
13.2.1 Changing the injection rate of the formazin standard that was spiked into the
filtered sample or changing the actual value of the formazin standard that
was spiked into the filtered sample yielded various stable turbidity values
that were continuously delivered to the test and reference instruments.
13.2.2 The injection rate of the turbidity standard was calculated in grams/minute
and the flow of the filtered sample was measured in grams per minute. This
allowed for the theoretical calculation of each turbidity spike. The
instrument response was calculated as a percent recovery of this spike. This
data is presented in Table 1 (Section 17.1).
13.2.3 The sequence of spikes started with a turbidity free baseline and progressed
with increasing turbidity up to the highest turbidity level.
13.2.4 The pre-filtered water supply was a blend of hot and cold tap water that is
supplied by the city of Fort Collins. The source water was from mountain
snow runoff that was treated with conventional techniques and filtered using
dual media filtration.
13.2.5 Results were used for provided the initial demonstration of linearity of the
measurement system.
13.3 On-Line testing at public water utilities – The PTV 6000 Laser Method was
tested at two public water utilities. One utility was a surface source water
treatment plant that required an additional softening step. The second plant
treated a surface source water that went through an integrated micro-filtration
membrane step. Both utilities are members of the Partnership for Safe Water.
The Lovibond 6000 laser turbidimeter and a reference turbidimeter that was
compliant with the USEPA laser nephelometry method was connected to the
same source water line for analysis. Both turbidimeters were operated for 24
hours, collecting data once per 15 seconds. The deviation between the two
instruments was 0.0003 and 0.009 NTU for the two plants respectively.
Response time to changes in turbidity differed slightly and was a function of
flow rates, but the magnitude of response was consistent.
13.4 The instruments were calibrated according to manufacturer’s instructions. After
calibration a QCS was run on each instrument to verify that the determinative
step for each turbidimeter was within the specified criteria.
13.5 Accuracy of the Lovibond 6000 Laser Method – Accuracy (bias) was presented
as percent recoveries relative to theoretical values for the turbidity spikes
105
compared to results from the reference Hach Method 101333. Refer to Table 1
in Section 17 for a summary of all spike recovery and precision data.
13.5.1 Percent Recoveries for spikes in the 0 to 0.10 NTU range:
The average percent recoveries of turbidity for turbidity spikes in the 0.014
to 0.10NTU range were:
Lovibond 6000 Laser Method: 101.4 %
Turbidimeter reference method 10133: 101.5%.
13.5.2 Percent Recoveries for spikes in the 0.10 to 1.0 NTU range:
The average percent recoveries of turbidity for the spikes in the 0.10 to 1.0
NTU range were:
Lovibond 6000 Laser Method: 100.6 %
Turbidimeter reference method 10133: 101.8%.
13.5.3 Percent Recoveries for spikes in the 1.0 to 10.0 NTU range:
The average percent recoveries of turbidity for turbidity spikes from the 1.0
to 10 NTU range were:
Lovibond 6000 Laser Method: 98.1 %
Turbidimeter reference method 10133: 99.9%.
13.6 Precision Lovibond 6000 Laser Method. Precision was presented here as the
standard deviation for each of the turbidity spikes. Refer to Table 1 in Section
17 for a summary of all spike recovery and precision data.
13.6.1 Precision for spikes in the 0 to 0.10 NTU range
The standard deviation for the turbidity for turbidity spikes in the 0.014 to
0.10 NTU range were:
Lovibond 6000 Laser Method: 0.0004 NTU
Turbidimeter reference method 10133: 0.0002 NTU.
13.6.2 Precision for spikes in the 0.10 to 1.0 NTU range
The standard deviation for the turbidity spikes in the 0.10 to 1.0 NTU range
were:
Lovibond 6000 Laser Method: 0.0044 NTU.
Turbidimeter reference method 10133: 0.0034 NTU.
13.6.3 Precision for spikes in the 1.0 to 10.0 NTU range
The standard deviation for the turbidity spikes from the 1.0 to 10 NTU
range were:
Lovibond 6000 Laser Method: 0.0480 NTU.
Turbidimeter reference method 10133: 0.0141 NTU.
106
14. POLLUTION PREVENTION
14.1 Pollution prevention encompasses any technique that reduces or eliminates the
quantity or toxicity of waste as the point of generation. Numerous opportunities
for pollution prevention exist in laboratory operation. The EPA has established
a preferred hierarchy of environmental management techniques that places
pollution prevention as the management option of first choice. Whenever
feasible, laboratory personnel should use pollution-prevention techniques to
address their waste generation. When wastes cannot be feasibly reduced at the
source, the Agency recommends recycling as the next best option.
14.2 The quantity of chemicals purchased should be based on expected usage during
its shelf life and disposal cost of unused material. Actual reagent preparation
volumes should reflect anticipated usage and reagent stability.
15. WASTE MANAGEMENT
15.1 The U.S. Environmental Protection Agency requires that laboratory waste
management practices be conducted consistent with all applicable rules and
regulations. Excess reagents, samples and method process wastes should be
characterized and disposed of in an acceptable manner. The Agency urges
laboratories to protect air, water, and land by minimizing and controlling all
releases from hoods and bench operations; complying with the letter and spirit
of any waste discharge permit and regulations; and by complying with all solid
hazardous waste regulations, particularly the hazardous waste identification
rules and land disposal restrictions.
16. REFERENCES
1. Fourth Annual Report on Carcinogens (NTP85-002 1985), p115.
2. USEPA. “Method 180.1 Determination of Turbidity by Nephelometry,” Revision
2.0, (1993).
3. Hach Company, “Hach Method 10133, “Determination of Turbidity by Laser
Nephelometry”, (2000).
17. TABLES AND VALIDATION DATA
17.1 Summarized precision and bias (percent recovery) data in tabular form (Table
1).
107
Table 1 – Results Table for the Percent Recovery to Turbidity Spikes for the PTV6000 685-nm Laser Turbidimeter
PTV 6000 685-nm Laser Reading
Reference Turbidimeter Reading (EPA approved Method
10133)
Spike #
Baseline (Blank)
in NTU
Theoretical
Value of Spike in
NTU
Response (blank
Corrected)
Recovery (%)
Precision (SD)
Baseline (Blank)
in NTU
Theoretical
Value of Spike in
NTU
Response (blank
Corrected)
Recovery (%)
Precision (SD)
N (both Test and
Reference)
1 0.012 0.014 0.015 103.3 0.0002
0.010 0.014 0.014 102.3 0.0001 60
2 0.012 0.028 0.029 101.8 0.0002
0.010 0.028 0.029 102.6 0.0002 60
3 0.012 0.060 0.060 99.2 0.0008
0.010 0.060 0.060 99.4 0.0002 60
4 0.012 0.110 0.111 100.6 0.0007
0.010 0.110 0.112 102.0 0.0005 60
5 0.012 0.225 0.230 102.3 0.0015
0.010 0.225 0.232 103.3 0.0014 60
6 0.012 0.554 0.554 99.9 0.0055
0.010 0.554 0.559 100.8 0.0036 60
7 0.012 0.921 0.915 99.4 0.0100
0.010 0.921 0.930 101.0 0.0079 60
8 0.012 1.895 1.885 99.5 0.0070
0.010 1.895 1.928 101.8 0.0069 60
9 0.012 3.568 3.366 94.3 0.0566
0.010 3.568 3.463 97.1 0.0077 60
10 0.012 9.376 9.419 100.4 0.0804
0.010 9.376 9.472 101.0 0.0277 60