Qweak Main Detector Status

Des Ramsay, Dave Mack, Michael Gericke

Main Detector Project Overview

The Main Detector WBS has spent 85% of our $468.5K budget.

$$ All custom PMT’s are at JLab and tested.

$ All magnetic shields are at JLab (and actually fit!)

$ All TRIUMF low-noise preamplifiers are at JLab and tested.

$$$ Quartz bar shipments are complete. (some QA remaining)

$$ Quartz lightguide shipments are complete. (lots of QA remaining)

Remaining funds are for digital integrators, voltage dividers (bases), detector housing, and support structure.

Several person-years of design, testing, and assembly remain.

Irradiation Tests

Glue Joints in the Optical Assembly

There will be 5 glue joints per optical assembly which all need to be strong and UV transparent down to 250 nm.

The central joint also has to be rad-hard: 100kRad nominal, (1 MRad for plan B)

Glue joints

e-Two glue candidates.

Elastomer Adherence to Quartz

Working Viscosity

Strength

SES 403No

(peels off easily)

Low (veg. oil)

poor (like

tough gelatin)

SES 406 Yes

High (cold

engine oil)

excellent

Optical Transmission Apparatus

Spectrophotometer and software by Carl Zorn of JLab Detector Group.

Monochromator

Integrating sphere with PMT

Sample on

translation stage

out in

monochromator

Integrating sphere + PMT

Due to lamp instabilities, we squat at one λ, normalize the beam with sample out, then measure with sample in, and repeat to estimate the random error.

Samples

Two glued samples and experimental controls.

One of our original

100 cm x 2.5 cm x 12 cm

prototype bars of Spectrosil 2000 was cut lengthwise into 15 slides of about

6 cm x 2.5 cm x 12 cm.

Except for a few avoidable new scratches, the 6cm x 12 cm faces remained in excellent condition.

Analysis

The correction for Fresnel reflection is typically 8%.

Using precision n(λ) from Melles-Griot and a next higher order expression, the error on this correction is below 0.1%.

Random errors are dominated by short-term lamp instability of about 0.2%.

T(λ) = Corr(λ) x

(Iin(λ) – Idark)/(Iout(λ) – Idark)

Corr(λ)

Corr(λ) = 1/(T(2) + T(4))where

T(2) =TF2

T(4) = TF2(1-TF)2

TF = 4n1n2/(n1+n2)2

Pre-irradiation Baselines2.5 cm Spectrosil 2000

(non-glued control)5.0 cm Spectrosil 2000+ glue joint

Below 300 nm, • SES403 glue joint reduces transmission 0.25%.• SES406 glue joint reduces transmission 0.75%. • With SES406, total light losses from all glue joints

will be less than 3%.

The control data are almost consistent with 100% transmission as expected for undamaged Spectrosil 2000.

(The small dip near 275 nm is repeatable but not understood.)

Irradiation

Irradiations with a 60Co source were done by Nuclear Services at North Carolina State University.

Initial 100 KRad irradiation

(nominal Qweak dose):

controls SQ2, Air-Gap, and

glued samples SES403, SES406.

Final additional 1 MRad irradiation

(plan B dose):

control SQ2, and

glued sample SES406

from Nuclear Services website:

Before and After 100 KRad

Our central glue joint will not suffer any significant rad damage during the Qweak experiment.

The 1 MRad data are still under analysis by Katie Kinsley (Ohio U.), but the preliminary results suggest no detectable damage.

2.5 cm Spectrosil 2000

(non-glued control)

5.0 cm Spectrosil 2000

with glue joint

PMT Nonlinearity

PMT Nonlinearity Measurements

Our goal is to keep the detector chain nonlinearity below 0.1%. (See D. Mack at http://qweak.jlab.org/doc-public/ShowDocument?docid=172)

Measuring linearity to better than 1% requires special techniques:

1. We previously developed a 2-LED method with sensitivity at the few times 10-4 level.

(See M. Geicke at http://qweak.jlab.org/doc-public/ShowDocument?docid=575)

But Riad Suleiman wasn’t happy. If the nonlinearity is frequency dependent, this method may measure it only in the DC limit. So …

2. In the last few weeks, summer student M. Andersen (U. Manitoba) has

demonstrated a new 3-LED technique which measures nonlinearity near the reversal frequency with sensitivity at the 10-6 level.

cathode current

anod

e si

gnal

IAC

VAC

IDC

quadratic nonlinearity

linear

Testing PMT linearity

Definitions

The PMT transfer function from cathode to anode can be written

Vanode = G Ik (1 + βIk)

Vanode is the effective voltage at the anodeG is the gain x 50 Ohmsβ is a small nonlinearity parameterIk is the cathode current or signal

Differentiating,

ΔVanode = GΔIk (1 + 2βIk)

The result depends on the load, so β alone isn’t very useful. We will call the dimensionless quantity 2βIk “the nonlinearity”: the relative error made when making a measurement with a bent ruler.

It may help to think of nonlinearity as a load-dependent gain:

G’ = G(1+βIk)

New 3-LED Technique

• Measures nonlinearity away from DC limit.• Self-normalizing: insensitive to drifts • Great sensitivityWe have only demonstrated proof of principle. No reliable numbers yet.

• Nonlinearity is equivalent to self-multiplication. Multiplying two frequencies yields sum and differences. The appearance of mixing peaks at f1+-f2 therefore gives access to the nonlinearity.

Technique requires:

1. one DC LED to provide the load

2. one AC LED at f1 to mimic a small signal

3. one AC LED at f2 (near f1) to induce the mixing.

Inserting

Ik(f) = IDC + I1 + I2

into

Vanode = GIk(1 +βIk)

the nonlinearity in terms of easily measurable quantities is

2 IDC 50Ω |V(f1+-f2)| / (|V(f1)|| V(f2)| )

f1 f2 f1+f2f2-f1

Low Gain Base Tests

Michael Gericke

PMT Low Gain Base Testing StatusWe went through several generations of low gain dividers since summer 2006

We need a nominal gain of 2000.

We want to be able to go up to a “contingency” gain of 16000.

Last year we had 3 base generations with 4 and 5 active stages respectively:

Voltage required to get to a gain of 16000 was still too high for 5 stages.

Go to more stages :

7 active stages with 6 141 kOhm Resistors and 2 Zeners

Small setup at University of Manitoba:

As before: Used reference PMT 128 and 2 280 nm UV LEDs

Assembly and measurements done bysummer student Charles Koop

Currently at JLab

Dark current increased too much withvoltage with 5 stage. 7 stage is good.

Several tests indicate the dark currentis mostly coming from the PMT (not leakage in the base)

The dark current is now only a 0.05% dilution for a 6 A nominal signal.

Measured gain vs voltage for 5 and 7stage

Contingency gain range can be obtainedwith 1000 to 1250 Volts of bias.

Everything looks good so far, but morestages means less stability and linearity measurements are in progress

MiscellaneousDave Mack

High gain divider tests

Panel stiffness measurement

Magnetic field sensitivity

High Gain Divider Tests• Mitchell Andersen also built our first high gain divider to look at pulses. (gain = few x 106)• Large pulses OK, but found unacceptable baseline noise for spe at 1-2 mV.• Problem tracked down, with great difficulty, to noisy zeners. • For now, we are using all-resistive dividers which give acceptable pulses and quiet baselines for

upcoming cosmic tests. • Lower noise zeners and external amplifiers were ordered and have arrived. • Tests continuing.

Panel Stiffness Measurement

• The glued quartz bars will be supported in front by a low radiation length (1.7%) composite panel.• The prototype panel from Composiflex has a core of IG-71 Rohacell (0.075 g/cm3) wrapped in 7 layers of

epoxy-impregnated Carbon fiber.

The deflection was measured with a Mitutoyo Dial Caliper BS-74 under a load representing the 10 kG weight of a 200 cm long quartz bar.

Measurements by Mkrtchyan et al.

Maximum deflection is 0.5 mm. We’re pretty happy with this, but still need to check that the glue joint won’t pop during transport.

Blue = more realistic loading

Application of weights

Magnetic Field SensitivityStray fields from QTOR will be < 0.1 Gauss. (W. Falk) Earth’s field dominates.

10% variation < 1% variation

Mkrtchyan et al.

Sensitivity of our 5” PMT’s with Vk-d1 = 280 V is negligible with shields.

Current Mode Electronics Update

Des Ramsay

Overall Tally of Current Mode Electronics

14Octal

integrators

14 main

14 Lumi28

Dual preamps

TOTAL MODULES

610SPARE MODULES

818MINIMUM MODULES

63362736TOTALS

2020Essential Beamline Monitors

44Target BPM (in scattering chamber)

22“Fake BPM” isolation monitor

11“Fake BCM” isolation monitor

2Lumi1 + 122Soft background detector

2main1 + 122Real-time isolation detector

16Lumi8 + 41616Lumi Monitors: 2 monitors x 8 tubes

16main8 + 41616Main Detectors: 8 bars x 2 tubes

VME

integrator

channels

type

I-V

modules +

spares

I-V

channels

voltage

signals

current

signalsITEM

Current Mode Electronics Summary

10 dual preamplifiers (20 channels) with transimpedance selection 0.5, 1, 2, 4 M are already at JLab for the main detectors.

4 More main detector style preamps are finished at TRIUMF. 14 Lumi-style preamps with gain selection 0.5, 1, 25, 50 M are also complete.

The testing is almost finished at TRIUMF.

Paul king is testing the prototype VME octal integrator. We have made a couple of firmware upgrades and the module appears to be working properly. We now need more detailed tests.

Preliminary designs are ready for a TRIUMF test source that will give us a realistic current of ~5 A, upon which a small simulated parity violating signal can be superimposed.

Modulated Current source

Reference current of 5 A DC

Choice of 16 modulations from ~10-6 to ~10-9

Unmodulated reference channel available

Responds to external spin state signals, or can run in stand-alone mode.

Modulated Current Source

Modulated Current Source Block Diagram

Voltage ramp on small capacitor

~10-15 A

Some Comments on Our Frequency Acceptance

switching function -- 18 ms quartet

4 ms

4 ms 4 ms

4 ms

0.5 ms0.5 ms

0.5 ms 0.5 ms

Switching function in time domain = ten regular 18 ms quartets.

Fast Fourier Transform (FFT)

Odd multiples of 55.5 Hz Hzms

5.5518

1

FFT essentially assumes waveform goes on forever

Simulation for finite run times

The FFT does not properly account for finite run times

For this I took a test sinusoid, multiplied by the switching function and integrated over the run time

I stepped the frequency and integrated each frequency for the run time

The simulation shows the same “acceptance” frequencies as the FFT,but shows a sensitivity to “off resonance” frequencies for finite run times.

For very long run times, only signals coherent with the switching function remain

100 random 18 ms quartets = 1.8 s run

1000750500250

222.2 444.4 666.6 888.8

111.1

333.3 555.5 777.7

Exactly equal + and – rejects DC The 4 ms spin state rejects multiples of 250 Hz The quartet structure rejects multiples of 111.1 Hz

200 random or 9 ms doublets = 1.8 s run

1000 Hz750 Hz500 Hz250 Hz

222.2 Hz 444.4 Hz 666.6 Hz 888.8

Exactly equal + and – rejects DC The 4 ms spin state rejects multiples of 250 Hz The doublet structure rejects multiples of 222.2 Hz

400 random or 4.5 ms singlets = 1.8 s run

250 Hz 500 Hz 750 Hz

1000 Hz

Each spin state is integrated for 4 ms

1/(4ms) = 250 Hz, so multiples of 250 Hz are rejected

States are randomly chosen, so in general there will notbe exactly the same number of + and -, and there will besome sensitivity to DC.

A-B (Lumi-BCM), 25 A, LH2, 2mm square raster, normal target cooling and pump speed

60

120

180

240

300

360

(A-B)/(A+B), 10 A, LH2, 2mm square raster, normal target cooling and pump speed

Next 6 Months

• Delivery of last preamplifiers to JLab (D. Ramsay)

• Continue testing prototype sampling ADCs (P. King)

• Support structure design

• Full-scale glue-up (Yerevan, Mack)

• Complete scintillation measurements (K. Kinsley)

• Complete low gain divider design (Gericke)

• Complete high gain divider design (Mack)

• Procure dividers (Mack, Gericke)

• Full-scale prototype (Yerevan, Mack)

Main Detector Summary

• We’re making adequate progress. Need more designer help.

• Full-scale optical assembly and cosmic tests by end of summer ’07.

• Full-scale prototype module fall ’07.

• Procurement of production module parts in early ’08.

• Complete gluing and module assembly in summer ’08.

• Production modules complete by Sept. 1, ’08.

… followed by more QA and detailing until installation

END

Supplementary Slides Follow

Expected Performance (updated 7/12/07)

Isolates the elastic e+p e+p channel good elastic focus with few hard inelastics.

Operates close to counting statistics 3.5% excess noise from sum of: showering,

pe-statistics, and nonuniformity of light collection at a tilt angle of 0 degrees

(where main faces of 12 o’clock bar are vertical)

Insensitive to backgrounds

(weighted by f(1-Abkg/Ael) )

Elastic channel bkg: O(1%) electrons + photons

photons from primary channel unknown

photons from neutron capture unknown

Insensitive to 100 kRad radiation damage (assuming 250nm cutoff)

Fused silica, brand name “Spectrosil 2000”,

Glue joint tested to 1 MRad

Modest Q2 bias 2.5%

Might worsen with radiation damage

Nonlinearity less than 1% < 1x10-4 with gain = 2000 base

Switchable between

current-mode (PV production) and

pulsed-mode (background studies)

PV production – 2x103 gain

Bkg studies – 2x106 gain

Our 2-LED Technique

Issue: Measures nonlinearity in the DC limit O(min-1)Annoyance: Precision is limited by the stability of the AC LED during the

measurement, so serious measurements require 24 hours of burn-in.

It works! Here, 2βIk= few x 10-4

Using early 5-stage prototype:Changes in small AC signal due to shifts in DC load give

access to the nonlinearity. Technique calls for:

•one DC LED to provide the load

•one AC LED to mimic a small signal

Doing the math:

Vanode = GIk(1 +βIk)

Ik(f) = IDC + IAC

Then to non-mixing order,

Vanode(f) = G(1+βIDC)IDC + G(1+2βIDC)IAC

The non-linearity is

(ΔVAC/VAC) / (ΔIDC/IDC)

Bias on <Q2> Due to Detector Response

The PV asymmetry is proportional to Q2, so we need to understand <Q2> with an error << 2%.

Our earlier estimates of <Q2> included the acceptance, cross section, and radiation, but neglected the detector bias.

Simulations indicate that our current

mode experiment will give events far from the

center of the bar about 5% higher weight.

Because our higher Q2 events have a wider distribution, the weak correlation with yield increases the detected <Q2>.

The new <Q2> = 0.02754.

5% bias

pe

Y

X

The estimated detector bias on Q2 is +2.5%.

This could worsen with radiation damage, and will be measured with wire chambers.

Y

Use of a Pre-radiator: Trade-offs A shower-max preradiator could increase Signal/Background.

But at what cost?

Potential for increasing S/B > 30 Excess noise increases to 12%.

A 2 cm Lead sheet in front of our quartz bars would increase S/B by > 30, but would require 390 additional hours, and increase the radiation dose to 3 MRad. Won’t need this if backgrounds are only 1%.

Photo-electron Count

PDG formula predict ~900 photons

above 250 nm cutoff:

Simulation gives ~ 1000 photons

on average for arbitrary path lengths.

~250 photons get to the cathode the rest is lost

The photon is counted only if it makes a volume

transition from the PMT window to the cathode.

The mean number of photoelectrons per event:

No Wrapping: ~40 Pes Millipore : ~50 Pes

Original design criteria: > 10 PEs

Target

QTORMini-torus

R-3 Chambers & Rotation System

R-2 HDCsPb Shielding

Beam

Experiment Component Details

GEMs

R-3 VDC Main Detectors

Lumis

Detector Design

Elastic envelope on bar of Spectrosil 2000.

Dimensions are 200 cm x 18 cm x 1.25 cm.

Modulated Current Source -- I/O

integrates for 4 ms stored as four 1 ms integrals Tsettle as short as 50 s allowed

Anticipated DAQ pattern

one spin state – (1/250) second

1 ms

t

next spin state

200 s settling time(not to scale)

NIM gate NIM gate

Rapid spin flip reduces noise from target boiling

switching function -- 18 ms quartet

4 ms

4 ms 4 ms

4 ms

0.5 ms0.5 ms

0.5 ms 0.5 ms

switching function -- 18 ms quartettest signal -- 9 ms period sinusoid

9 msf =111.1 Hz

(18 ms quartet) x (9 ms period test sinusoid)

product

integral

any multiple of

will integrate to zero regardless

of phase

Hzms

1.1119

1

The 18 ms quartet rejects multiples of 111.1 Hz

switching function test signal

product

integral of product

time

current

3.6 pA(0.6 ppm p-p)( ppm)

6 A

helicity- + - + - + -

Size of Qweak Signal

• figure shows regular spin flip; in practice use + - - + or - + + -

• for 50 kHz noise bandwidth, rms shot noise is 70 nA

• on a scope the noise band would be 100,000 x the signal !

3.0zA

• sample at the center of each interval (500 samples)• first sample 1 s after gate• Q = (sum of samples) x (t)• band limit signal to small fraction of sampling frequency to eliminate the wiggles and kinks.

Integral From Samples (rectangular rule)

= 1 ms

2 s

1 s

NIM gate

• sample at the sides of each interval (n+1 samples)• Q = (average of first and last samples plus sum of others) x (t)• band limit signal to small fraction of sampling frequency to eliminate the wiggles and kinks.• we impose an analog cutoff at 1/10 the sampling frequency

Integral From Samples (trapezoidal rule)

Prototype TRIUMF VME integrator details

FPGA

FPGA Prog/Debug Ports

VME Module SelectSwitches

Status LEDsVME AccessExt Clock EnbExt Gate Enb

Ext NIM Gate

Ext NIM Clock

DC-DCConverter

ADC

8 inputs

Existing Gzero Ion Source Signals

• signals derived from 20 MHz crystal clock• Qweak integrator should use this clock as well• Integration triggered by MPS (is present form OK?)

charge

counts

Q0

- helicity+ helicity

charge

ADCerror

+s

-s

• ADC reads S channels low below Q0 and jumps to S channels high above Q0

• This causes the measured asymmetry to depart from the real asymmetry, A0, by an amount , where is in channels.

• The DNL won’t introduce an asymmetry when none is there, it only changes an existing one.

)(0 sAA

Differential Nonlinearity (DNL) Example

Switching function in time domain = one 18 ms quartet.

Fast Fourier Transform (FFT)

B (BCM only), 10 A, LH2, 2mm square raster,normal target cooling and pump speed

A (Lumi sum only), 10 A, LH2, 2mm square raster, normal target cooling and pump speed

5 inch S20 Cathodes have a specified thermionic dark current of ≤ 0.1 pA at room temperature = 0.2 nA (Gain 2000 Low-Gain Base)

Several tests indicate the dark currentis mostly coming from the PMT (not the base)

The dark current is a 0.05% dilution for a 6 A nominal signal.

Contingency gain range can be obtained in the range of 1000 to 1250 Volts of bias.

The cathode currents were measured with the diode base.

LED intensity was monitored andremained constant over the measurements.

Everything looks good so far, but morestages means less stability and linearity measurements are in progress