ASSESSMENT OF FORCE SENSING RESISTORS: CONTROL AND DESIGN FOR E-BRAILLE DEVICE
FINAL PROPOSAL
Timothy D. CarambatAdvisor: Dr. Mohammad Saadeh
Dr. Cris KoutsougerasET 493 Senior Design Project I
Fall 2016
ABSTRACT
The objective of this proposal is to continue the design and control of the E-braille device that
has been under research in previous semesters. In particular, this proposal will deal less with the
research and theory of the E-braille device-as it has been performed, and will instead focus on
the finalization and fruition of the E-braille device and control systems. The E-braille device is
an assistive technology for the visually impaired, allowing them to simulate tactile sensation in
the form of the Braille language. The E-braille device will be worn on the dorsal portion of the
user’s finger. Currently, the manner in which tactility is simulated is via an electronic tactile
display. The pressure in which the finger is subject to is controlled by a selected force sensing
resistor (FSR). The feedback from this sensor drives a miniature DC motor which in turn
controls the vertical movement of the top of the device via rack and pinion gears. The sensor
allows real time reactive force of the device on the finger pad to maintain a custom comfortable
pressure for the user. This proposal focuses on identifying several FSRs for use in the E-Braille
device in terms of the clamping mechanism. Also, it aims at investigating control methods, the
tactile display currently in use, and other design facets that improve device efficiency and user
experience.
INTRODUCTION
The main goal of the E-braille device is to provide a suitable, low-cost and low-maintenance
assistive technology for the aid of visually impaired persons. The E-braille is novel in its nature
as its technology is not expensive, an unfavorable characteristic that other devices have yet to
overcome, as well as providing a platform that accommodates a large percentage of users. The E-
braille device, in final form will be compact in nature, cheap to produce, accessible by the largest
available population of the target market, and powered externally using normally available
power sources. These physical traits in conjunction with necessary criteria for use by an impaired
user do present obstacles in design and control.
2
The E-braille device, in full functionality, will provide a novel way for visually impaired users to
have information read to them without the need for audio or voice synthetization of their reading
material. Providing a more natural and private method of information exchange. The device
hopes to fulfill a personal and market need for such an assistive technology.
Additionally the underlying technology that I will be researching in terms of the FSR can be
easily utilized in other applications aside from this device. In its most simplistic operation the
FSR should operate as an accurate load cell. This technological advancement would be critical
in applications where load detection is required but due to space or cost restraints a typical load
cell is not feasible.
COMPONENTS AND CONSTRUCTION
Currently, as the device has been research and worked on previously, progress has been made in
regards to the physical manifestation of the device. The E-braille device is composed of several
necessary components working in harmony to provide the ideal user experience. The E-braille
device is primarily composed of a controller with necessary coding and software, the physical
clasping device with rack and pinion, a driving motor and gear to adjust clasp, tactile display
board, and force sensing resistor (FSR). Currently the prototype appears as follows:
3
Figure 1-Prototype Figure 2- Prototype with hand
Figure 3-Assembly
4
The prototype currently features larger than final-product componets for ease of data-acquisition
and continuing adjustments. The prototpye currently is controlled via a programmable
microcontroller, called an Arduino Uno, seen below:
Figure 4-Arduinio Uno
The Arduino Uno is a microcontroller board based on the ATmega328. It has 14 digital
input/output pins (of which 6 can be used as PWM outputs), 6 analog inputs, a 16 MHz ceramic
resonator, a USB connection, a power jack, an ICSP header, and a reset button. It contains
everything needed to support the microcontroller; simply connect it to a computer with a USB
cable or power it with an AC-to-DC adapter or battery to get started.
Paired to this Arduino is an accessory for the controller, commonly called a “shield”. This shield
gives us the ability to communicate with the E-braille controlling motor easily, as well as
granting us the ability to utilize forward and reverse motion.
5
Figure 5-MegaMoto Shield
Next, the critical feedback device, the force sensing resistor. FSR's a resistor that changes its
resistive value (in ohms Ω) depending on how much it has been compressed. The greater the
force the lower the resistance. These sensors are fairly low cost, and easy to use but they're
rarely accurate.
Figure 6-FSR Diagram and FSR
Finally, the motor to actuate the upper portion of the E-braille device is of low-cost and high
power. With a cross section measuring only 10×12 mm (0.39″×0.47″), these small brushed DC
gear motors have a gear ratios—from 5:1 up to 1000:1—and offer a choice between three
different motors: high-power (HP), medium-power (MP), and standard. (Figure 6).
6
Figure 7-DC motor
Other componets currently are used as needed. Such as Date Aquisiton devices and software,
strain guages, amplifiers, filters, and other circuity devices. These items are used as needed for
analysis or testing purposes and may not be present in the final version of the E-braille device.
CURRENT STATUS:
The current status of the E-braille device is “semi-functional” in the manner that physical
componets preform their desired actions, but the control of such motions and actions do not
preform as desired from inputs. Currently, system control is not written and executed through the
Arduino development environment. This is due to its lack of interface and data-aquisiton in real
time. Using Arduino as a “slave”, all commands and code are run through computer desktop
software know as LabView. LabView is currently used for real-time data input, real-time data
acquisition and its ease of use and powerful analtyical libiraries while the device is running.
Actions that cannot be preformed with other softwares.
Previous reseach and devlopment of this deivce largrly hinged on the selection of the FSR. The
FSR is a system-critical componet as it is the feedback mechanism for the clamping action of the
device on the user’s phalange. It is important to note that an inherent property of all FSR’s is that
when compressed the output voltage and resitivity of an FSR is not linear, especially under light
compression.
7
Figure 8-FSR linearity
Seeing as though the E-braille device will not be operating under high compressive force it can
be seen that a model will need to be used to measure and predict the output voltage and what
force that corresponds to, so that adjusts can be made if needed.
In order to attempt to linearize the FSR it would need to have some basis of measurement. To
accomplish this during testing a traditonal load cell was used. A load cell was used for their
predicitabilty and accuracy, normally. The load cell used in past testing was extremely senstive
and would have to be continually calibrated after powering off, which caused small measurement
errors during testing. Overcoming the load cell error and sensetivities extended testing time
considerably. Pictured below is the old Omega LC201 load cell as well as the new FUTEK
lrf400 load cell.
Figure 9-(Left) Omega LC201 and FUTEK lrf400 (Right)
8
-1 0 1 2 3 4 510
1
102
103
104
105
Force (N)
Res
ista
nce
(koh
m)
Testing of the FSR:
Given that due to human error it is impossible and impractical to provide a continous and
accurate force; a cam actuating mechansim was developed and implemetned into testing.
Pictured below is the first model of the cam actuator. This machine would be used to simulate
sinusodial, triangular, square, and other traditonal signals.
Figure 10 &11- Cam Mechanism
Modeling, Prediction, and PID control:
In addition to the testing of the FSR, this data was collected and implemented into a
mathematical model called the Hammerstein-Wiener model. The model actually represents an
amalgamation of two non-linear system identification models, Hammerstein and Wiener. For
more information on the workings on the model and identification process used, the reader is
deferred to (Saadeh and Trabia, 2012).
9
An example of system-identification would be shown as below.
Figure 12-Hammer-Wiener System Identification
It can be noted that the Hammer-Weiner model is fairly accurate in the identification of the
system as compared to each model independently and linear models. For future testing it may be
of better value to use the individual Hammer or Wiener models. This will be determined from
data. Above is simply a single case example and does not mean this particular system type will
be the best overall for all models of FSR’s
The reasoning for use of this model was to identify the FSR signal and then feed this data into
the Proportional, Integral, Derivative (PID) controller to control the motor and clamping force.
The PID takes the error from actual measurement and the system-prediction model and uses this
difference to adjust the PID controller to bring the motor on target more quickly without
oscillation or overshoot of the set force. The tuning of this control as well as effectively pushing
data into the model and making adjusts in real-time without delay to the system is where
progress was halted due to time constraints.
ADVANCEMENTS MADE
Setup new load cell and configure the device.
Make adjustments to cam mechanism to accommodate new load cell.
Identify FSRs using new load cell to calibrate them using cam mechanism.
Signal mapping to Arduino around arbitrary set point from user.
Data Acquisition without slowing system in use.
BREAKDOWN OF ADVANCEMENTS MADE:
10
Setup new load cell:
The new FUTEK LRF400 load cell arrived in my possession mostly disassembled. The
physical load unit, was however, a solid component that came pre-calibrated from the factory.
The LRF400 load cell came with peripheral devices for data acquisition such as an
amplifier/filter, a serial connection assembly kit, and a cable to allow communication of the
LRF400 with a data acquisition device if needed. A photo below shows the LRF400 that is
hooked to an OMEGA data acquisition device. It should be noted that in this setup I have both
the old Omega load cell and the new Futek cell being analyzed by the DAQ device so that I may
compare the steady state signal of both load cells.
Figure 14-DAQ setup of Futek and Omega Load Cell
A sample of data can be seen below. The noise frequency can be seen by each load cell in
the graph, the new Futek cell in red and the Omega in yellow. Both cells are unloaded in this
experiment as well as zeroed. It was noted that after some time the noise of the Omega load cell
was excessive in comparison to the Futek cell. This test, as well as a hysteresis experiment,
proved the Futek cell to be far superior to the Omega load cell for uses of fine data acquisition.
11
Figure 15- Comparison of Futek Cell (red) and Omega Cell(yellow) unloaded noise
Make adjustments to cam mechanism to accommodate new load cell:
It is evident that the two load cells for our analysis have dramatically different
dimensions. Given that the Futek cell appears to be a more stable platform for testing; the cam
mechanism initially used to test the Omega load cell will have to modified extensively so that it
may support the new Futek cell without difficulty or testing error. Modeling of the Futek cell in
the old cam mechanism is seen below.
Figure 16-Full cam assembly with Futek Cell structure accommodations.
12
Figure 17-Futek LRF400 Cradle
Figure 18-Cam mechanism riser to adjust for height of Futek Cell
13
Identify FSRs using new load cell to calibrate them using cam mechanism:
When undergoing the retrofitting of the new cam mechanism it was evident that the new
load cell would necessitate the need for revisions to the cam mechanism that was designed prior
to the spring 2016 semester. These parts were designed with the intention of allowing the new
FUTEK load cell to be centered beneath the cam. The parts design can be seen in Figures 17-18.
These parts would simply allow the FUTEK cell to be center-aligned to the cam mechanism and
would rise the entire cam mechanism by the difference in height of the Omega and FUTEK load
cells. Functionally, the old system and new system would be identical.
Prior to the actual printing of these new components, testing was performed using the sinusoidal
cam. Immediately, there were issues with testing. The returning spring force on the cam seemed
to prevent the motor from rotating the cam. For example, at the bottom of eccentricity of the
cam, where the returning spring would be compressed, the return force of the spring would lock
the motor-voiding all data in the test. This was remedied by modifying the returning spring.
After adjusting the spring, it was discovered that during revolutions-right after the bottom of the
eccentricity- the cam would accelerate and spin faster than the motor speed. Additionally,
sometimes the cam would not compress the spring and would lock. Usage with other springs also
displayed that the cam follower was making off-center contact with the cam, making contact
difficult in continuous rotation.
14
Figure 19-Bad contact of cam with follower. Using older designed for Omega load cell
It became evident that the motor was under-powered for our required torque needs. This was
simply a real-world testing error. Replacement of the motor was imperative, this would prove to
be an issue as the existing cam structure was 3D-printed and will not be easily modifiable and
revisions cannot be undone once preformed. After advisement, it was decided that constructing a
similarly functioning system would need to be done, but this new system would need to be
modular and have revisions be able to be done easily with minimum modification to the system.
It was decided to use an assembly kit. The kit would need to be modeled first and custom
components designed afterwards so that the assembly could come together.
15
Figure 20-Isometric view of new cam assembly
Figure 21-Front view of new cam structure
16
Figure 22-Exploded view of new cam structure
Figure 23-Solidworks model of DC cradle
17
Figure 24-Solidworks model of FUTEK cradle
Figure 25-New printed parts for new cam structure
18
Figure 26-Futek Cell in cradle
Figure 27-High Torque DC motor in cradle
19
Figure 28-New assembly pre-construction
As it can be seen the new the system is easily pieced together and the motor can be replaced with the mounting bracket to accommodate it. Otherwise, the system is functionally the same as the old mechanism, but should be able to allow easy modification for use of new motors as to obtain our speed and torque requirements.
The motor currently used on the old cam mechanism was, as discussed, not capable of torque requirements to spin the cam while force was applied on the cam. We have decided to move to a multitude of motors that will spin much slower (typically around 2Hz, or 120rpm). Currently supplied are two geared DC motors and three geared stepper motors. The DC motors seem to spin at a usable rpm with great torque output. Currently, a DC motor was assumed to be utilized with the new cam structure. The stepper motors do not seem to reach an operable rpm, this seems to be a gearing issue, as the stepper motor cannot rotate the main shaft quickly enough with the gearbox attached. From testing it appears to be limited to approximately 1HZ, which will be used in lower frequency testing.
20
Figure 29- Available Testing motors. DC left and steppers right
Signal mapping to Arduino around arbitrary set point from user:
Using the LabVIEW software that is currently being utilized to analyze the FSR input of
the E-braille device, this problem has been solved. When I began working on this issue it was
evident that the motor controller had issues discerning which direction to drive the rack and how
slowly it should do so depending on both the input of the FSR and the user’s custom set point of
comfortability. It should be noted that the LabVIEW software does not have a map function for
this specific utilization. I was able to accomplish this by utilizing the Arduino map function
mathematics and transferring this into MATLAB code. I was then able to make a MATLAB
function called “MAPV” that took exactly the same parameters as the function would in
Arduino. Using a continuous loop function in LabVIEW I was able to continuously input the
FSR value and user set point into the system that would then rotate the E-braille motor in the
right direction and magnitude to make adjustments on the load experienced by the FSR. Upon
solving this I then found the motor was constantly making adjustments, causing excessive heat.
This was solved by simply making a range of acceptable values around the user set point.
21
Allowing the motor to rest and cool while also maintaining the correct pressure on the user’s
finger.
Figure 30-LabView Interface and Code. Lower blue window is MATLAB script that performs MAPV function.
Data Acquisition without slowing system in use:
In order to identify the FSR signal and correlate this into a usable force via the
Hammerstein-Wiener model, we must first collect the FSR data to determine this signal. Initially
the system had issues during data collection that while attempting to save the data and record it
simultaneously the system would run extremely slow, running any data being collected. I was
able to get around the system performance lag during collection by locally saving the data into a
simple array when “Save” was enabled. This data was stored locally in temporary memory and
allowed collection to continue uninhibited, where each millisecond of runtime correlated into a
sample of data. After ending the testing the data is then saved in a folder on the testing computer
as a formatted Excel spreadsheet. Allowing the data to be reviewed as well as read back into a
LabVIEW code for analysis later by the Hammerstein-Wiener model. Currently the issue with
this prediction model seems to be the coefficients. These coefficients and input of this data real
time will be the next obstacle.
22
Figure 31- LabVIEW display and code that solves data acquisition performance issues.
Design/tune a proportional-integral-derivative (PID) control system to drive a DC-motor:
This objective remains unfinished at the writing of this report. The PID system will be
integral to the use of the E-Braille device motor actuation. The PID system is used after the
Hammerstein-Weiner model prediction. The input of a value from the model into the PID will
likely be easy to configure, but the coefficients of the PID controller may have to be fine-tuned
later in this projects development. The coefficients will likely have to be determine empirically
through testing, which should be easily achieved through LabVIEW.
CURRENT OBJECTIVES (DELIVERABLES):
Finish new cam construction and begin testing.
Identify best system model for each model FSR.
Identify FSR from selection for best use in E-Braille application.
Have model work with real time system.
Integrate PID control to adjust user comfort via force input from FSR.
23
Finish new cam construction and begin testing:
As discussed from previous efforts, the cam mechanism originally designed had to be initially modified to accommodate the newer FUTEK load cell. It was then later discovered that testing with the older cam mechanism was not reliable enough to produce usable data for identification of the FSR’s. This unreliability lead to the need for an entire re-design of the cam system. The new system was going to take a modular approach so that adjustments could be made easily or new parts could be integrated into the system.
During the interim semester work was done to complete the cam mechanism. Structurally the system was rigid and was promising to be a good platform in which to start obtaining FSR’s signals.
Figure 32- Newly Designed Cam Mechanism Setup
24
For the cams used in these experiments they had to be outfitted to fit with the new motors that would be used for testing. It was decided that for low-frequency tests (~1Hz) a geared stepper motor would be used. For higher frequency (~10Hz) tests a geared DC motor would be used. For each FSR, cam tests would be performed with a sinusoidal and triangular cam profile.
Figure 33- Cam Attached to Stepper and Hub Configurations for mounting. DC motor is circular hub.
Currently, there are 5 FSR models that will be tested. They are all functionally similar but through testing, they appear to output various voltages at varying applied forces.
Figure 34- All FSR’s. A-E from left to right
25
For each FSR the following tests were run, usually multiple times so that consistent results could be determined.
Sinusoidal Chirp Signal (.5Hz to 10Hz) Sinusoidal Signal (1Hz Constant) Triangular Signal (.5Hz) Triangular Signal (1Hz) Triangular Signal (2Hz)
This objective remains current as until testing and all FSR models are recognized testing is done on a case-by-case basis if results seem to be inconclusive or the data is not usable for analysis.
Identify best system model for each model FSR:
This is the most system-critical component of not only the purpose of this project, but also the key functionality of the E-Braille device. The system model used for each FSR will be the key component that allows it to function as a load cell. If this functionality is not achieved this technology will have no more use that it does now in force-sensitive applications, let alone application in the E-Braille device as the feedback for the clamping mechanism.
It should be clarified that from the possible systems (linear, Hammer, Weiner, Hammer-Weiner) that system should already be as optimized as possible in its configuration so that it performs as accurately as possible under all conditions of use.
Identify FSR from selection for best use in E-Braille application:
After identify the system models for each FSR it must then be determined with FSR to integrate into the E-Braille assembly. This is imperative because from past testing and results so far, each FSR is sensitive at different regions of force application. Being that the E-Braille device will not be clamping the user’s finger under high forces, the FSR should be most sensitive under light loads up to approximately 2N.
Have model work with real time system:
After integration of the FSR into the E-Braille device it must then utilize the design system. Under regular operation the FSR in the E-Braille device should input a voltage into the system and return a force. With the force value other operations can be done. The system should do this operation easily and without notable lag. If the system does not perform the process quickly enough enhancements will need to be made to either the computations or the system.
Integrate PID control to adjust user comfort via force input from FSR:
26
Lastly the entire system should work under PID control so that motor adjustments are not made suddenly and that the user’s comfort level is obtained quickly, but without overshoot and oscillation around the set point. Competition of this step would then have a fully functional system where the clamping action of the E-Braille device is a direct result from the technology enhanced from the identification of the FSR behavior-a practical application of this new development in FSR technology.
27
References
1. Arduino - ArduinoBoardUno. (n.d.). Retrieved February 28, 2016, from https://www.arduino.cc/en/Main/ArduinoBoardUno
2. Brushless DC-Servomotors with analog Hall sensors. Retrieved November 15, 2015 . https://www.faulhaber.com/en/global/
3. Florez, J. and Velasquez, A., 2010, “Calibration of Force Sensing Resistors (FSR) For Static and Dynamic Applications,” 2010 IEEE ANDESCON, pp.1-6.
4. FUTEK Advanced Sensor Technology. (n.d.). Retrieved February 28, 2016, from
http://www.futek.com/product.aspx?stock=FSH00264
http://www.futek.com/product.aspx?t=instrument&m=csg110
5. Interlink Electronics, FSR 402. (2011). Retrieved November 12, 2015, from http://www.interlinkelectronics.com/FSR402.php#
6. Nakamura, N; Fukui, Y; , "Development of Fingertip Type Non-grounding Force Feedback Display," EuroHaptics Conference, 2007 and Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems. World Haptics 2007. Second Joint , vol., no., pp.582-583, 22-24 March 2007
7. OMEGA Engineering, DAQ and Load cell. Retrieved November 12, 2015
http://www.omega.com/pptst/OMB-DAQ-2408.html
http://www.omega.com/pptst/LC201.html
8. Pololu - Micro Metal Gearmotors. (n.d.). Retrieved February 28, 2016, from https://www.pololu.com/category/60/micro-metal-gearmotors
9. Saadeh, M. & Trabia, M. (2012). "Identification of a Force Sensing Resistor
for Tactile Applications," Journal of Intelligent Material Systems and
Structures, JIMSS, 24(7): 813-827.
28