MarkRivers
commented
4 years ago Luca Bombelli from XGLab agreed that the noise seen in the ADC trace is the source of the problems. He suggested increasing the Fast threshold to eliminate zero energy the peak. I found that I needed to increase it to 18 keV (284 MCA bins) to eliminate it. Once I did that the ICR and OCR were nearly the same and the zero energy peak was drastically reduced.
I repeated the measurements. For logistical reasons I used older Cd109 and Co57 sources (2015 vs 2019) so the count rates are less.
I also increased the Peaking time and Max Peaking Time on the Dante from 2 microseconds each to 4 and 6 microseconds respectively.
This is the Dante configuration reported by the driver:
epics> asynReport 1 DANTE1
DANTE1 multiDevice:Yes canBlock:Yes autoConnect:Yes
enabled:Yes connected:Yes numberConnects 1
nDevices 1 nQueued 0 blocked:No
asynManagerLock:No synchronousLock:No
exceptionActive:No exceptionUsers 0 exceptionNotifys 0
traceMask:0x1 traceIOMask:0x0 traceInfoMask:0x1
addr 0 autoConnect Yes enabled Yes connected Yes exceptionActive No
exceptionActive No exceptionUsers 1 exceptionNotifys 0
blocked No
traceMask:0x1 traceIOMask:0x0 traceInfoMask:0x1
Configuration 0:
fast_filter_thr: 284
energy_filter_thr: 63
energy_baseline_thr: 63
max_risetime: 87.000000
gain: 4.400000
peaking_time: 125
max_peaking_time: 187
flat_top: 6
edge_peaking_time: 31
edge_flat_top: 6
reset_recovery_time: 625
zero_peak_freq: 0.010000
baseline_samples: 64
inverted_input: 0
time_constant: 0.000000
base_offset: 0
overflow_recovery: 0
reset_threshold: 0
tail_coefficient: 0.000000
Statistics 0:
real_time: 60000000.000000
live_time: 57654776.000000
ICR: 4.664349
OCR: 4.482033
last_timestamp: 60000000.000000
detected: 282617
measured: 268922
edge_dt: 27545
filt1_dt: 586306
zerocounts: 572
baselines_value: 26999628
pup_value: 888
pup_f1_value: 13723
pup_notf1_value: 0
reset_counter_value: 1954This is the Canberra MCA plot now with the weaker sources.
This is the Dante plot
This is the fit of the Canberra spectrum:
*******************************************************
Fit of 13BMD:aim_adc1
Real time (seconds): 60.00
Live time (seconds): 56.98
Initial FWHM offset, slope: 0.150000 0.000000
Optimized FWHM offset, slope: 0.150000 0.000000
Initial energy offset, slope: 1.141311 0.062424
Optimized energy offset, slope: 1.141765 0.062402
# Iterations, function evals: 0 71
Chi squared: 1.145288e+007
Status code: 2
Error message:
Peak Energy FWHM Area Background Area/MDL Area/Bkg
ag ka 22.104 0.8118* 166995.6 2870.0 1039.1 58.2
ag kb 24.987 0.8565* 35357.2 1854.0 273.7 19.1
ag pu1 43.656* 2.1363* 207.5 409.0 3.4 0.5
ag pu2 47.122* 1.8737* 46.2 150.0 1.3 0.3
pb ka1 74.956 0.8503* 543.3 329.0 10.0 1.7
pb ka2 72.794 0.7645* 295.8 287.0 5.8 1.0
pb kb1 84.922 1.0052* 153.7 374.0 2.6 0.4
cd109 g1 88.040 0.8371* 5198.7 362.0 91.1 14.4
co57 g2 122.061 0.8462* 4807.0 241.0 103.2 19.9This is the fit for the Dante:
*******************************************************
Fit of 13BMD_Dante1:mca1
Real time (seconds): 60.00
Live time (seconds): 57.65
Initial FWHM offset, slope: 0.150000 0.000000
Optimized FWHM offset, slope: 0.150000 0.000000
Initial energy offset, slope: -2.789470 0.063168
Optimized energy offset, slope: -2.988031 0.063529
# Iterations, function evals: 0 141
Chi squared: 2.109271e+007
Status code: 2
Error message:
Peak Energy FWHM Area Background Area/MDL Area/Bkg
ag ka 22.104 0.6661* 169465.2 2182.0 1209.3 77.7
ag kb 24.987 0.7257* 36069.6 1352.0 327.0 26.7
ag pu1 44.259* 0.4358* 52.2 29.0 3.2 1.8
ag pu2 47.294* 0.9748* 26.4 57.0 1.2 0.5
pb ka1 74.956 0.6361* 548.0 203.0 12.8 2.7
pb ka2 72.794 0.6727* 288.7 215.0 6.6 1.3
pb kb1 84.922 0.9198* 211.6 226.0 4.7 0.9
cd109 g1 88.040 0.6806* 5204.6 252.0 109.3 20.7
co57 g2 122.061 0.7301* 5077.5 115.0 157.8 44.2Now the Dante has slightly better throughput than the Canberra, significantly better Area/Bkg, and significantly better FWHM.
This issue describes issues we are having with the Dante MCA on the Canberra Ge detector in 13-BM-D. The loaner Dante we tested in August 2019 worked very well with this detector. However, the performance now is poorer.
In both August 2019 and today I compared the Dante to the Canberra AIM MCA electronics. The AIM performance has not changed, but the Dante is worse.
The comparisons were done under the following conditions.
The following screen shots show how the Canberra system was configured. These are the settings that the LVP experiments normally use except that I set the ADC offset to 0 and decreased the fine gain slightly to put the 122 keV Co57 peak a bit further below the end of the spectrum. Pileup reject is enabled, and the shaping time is 2 microsec.
This is the pre-amp trace with the loaner system in August. The analog offset was set to 230 which put the entire pre-amp ramp within the range of the ADC, as shown in this waveform trace plot. The range is about 4000 to 56000 ADC units, or about 52000 peak to valley.
The following screen shows how the Dante system was configured in August 2019. The peaking time and maximum peaking time are both 2 microsec. The gain was set to 2.7 to put the 122 keV Co57 gamma line just below the maximum energy. The firmware is 3.3.13.
This is the pre-amp trace from the system today. The range is about 25000 to 56000, or about 31000 peak to valley. Compared to the loaner system the front-end gain is reduced by 31000/52000 = 0.60
The following shows the Dante configuration today.
The major differences in the settings from the loaner system in February are:
This is the IDL MCA plot for the Canberra system (yellow), along with the fitted background and peaks (green). This is the plot from August.
This is the plot from today. The main qualitative difference is that the pileup peaks at 44 keV (about 1/3 of maximum energy) are much less sharp than they were in August. Could this be because pileup reject is not working, perhaps a cable is disconnected?
This is the IDL MCA plot for the Dante system (yellow), along with the fitted background and peaks (green). This is the plot from August.
This is the plot from today. The main differences are that the peak near 0 energy is now much larger than the Ag Ka peak (first blue peak), while in August it was smaller than the Ag Ka peak. The spectrum background is also significantly noiser because there are fewer counts in the 60 seconds of real time.
To quantitatively compare the spectra the background and peaks were fit using the following parameters.
The peaks fit were:
The energies of all of the peaks except the pileup peaks were fixed, only the 2 global energy calibration coefficients were optimized. For the pile-up peaks the energies were also optimized. The FWHM of each peak was optimized, rather than using a global FWHM fit. The amplitude of each peak was optimized.
The following are the results of the peak fit for the Canberra system measured in August.
The following are the results of the peak fit for the Canberra system measured today. The peak areas, FWHM, and peak/bkg ratios are very similar between August and today except for the pileup peaks.
The following are the results of the peak fit for the Dante system measured in August. The peak areas and peak/bkg are significantly better than the Canberra system.
The following are the results of the peak fit for the Dante system measured today. The peak areas are 3-4 times smaller than in August and are now about 2X less than the Canberra. The peak/bkg ratios are 2X less than measured in August. The live time is 3X less.
To try to understand why the zero energy peak and ICR are so much larger now than in August I collected an ADC trace with the fastest sampling time, and zoomed in on a single 22 keV x-ray pulse.
There is clearly a noise source present. Unfortunately I did not save an ADC trace from August to see if this noise source was also present then. The noise has a peak to peak amplitude of about 50 ADC units. It repeats 6 times in 30 microseconds, so the period is about 5 microseconds, which is about 200 kHz.
Is this what is causing the large 0 energy peak and large ICR?
Would that be reducing the throughput so it is now 2X less than the Canberra, and 3-4 times less than the Dante measured in August?