Difference between revisions of "Test stand for measurements of the NOvA electronics"

From AstroNuWiki
Jump to: navigation, search
(Results)
(Results)
 
Line 179: Line 179:
  
 
[[Image:pmt_photons.jpeg | 540 px]] [[Image:pmt_gains.jpeg | 540 px]]
 
[[Image:pmt_photons.jpeg | 540 px]] [[Image:pmt_gains.jpeg | 540 px]]
 +
 +
'''PMT''' gain increases with signal width (slightly decreases on the very end). This '''PMT''' behavior may connect with dynode system. It is a personal feature of photosensor. Gain pattern allows us to make the correction for previous results (put the right '''N_pe''' number to the denominator).
 +
 +
''Final results''
 +
 +
[[Image:Ch3_A_Npe_SP_abs_betas.jpeg | 540 px]] [[Image:Ch19_A_Npe_SP_abs_betas.jpeg | 540 px]]
 +
 +
[[Image:Ch3_A_Npe_SP_rel_betas.jpeg | 540 px]] [[Image:Ch3_A_Npe_SP_rel_betas_zoomed.jpeg | 540 px]]
 +
 +
[[Image:Ch19_A_Npe_SP_rel_betas.jpeg | 540 px]] [[Image:Ch19_A_Npe_SP_rel_betas_zoomed.jpeg | 540 px]]
 +
 +
''Double check with SIPM''
 +
 +
We tested several relative time points using '''SIPM''' to make sure that everything is correct right now. APDs were tested using the same generator parameters but with '''SIPM''' like monitoring photosensor. Output behavior well matches with '''PMT''' one.
 +
 +
[[Image:Ch3_A_Npe_SP_sipm.jpeg | 540 px]] [[Image:Ch19_A_Npe_SP_sipm.jpeg | 540 px]]
  
 
===Technical notes===
 
===Technical notes===
  
'''!Extended information and technical details about long signals measurements one can find [https://nova-docdb.fnal.gov/cgi-bin/private/ShowDocument?docid=40216 here]!.'''
+
'''!Extended information and all technical details about long signals measurements one can find [https://nova-docdb.fnal.gov/cgi-bin/private/ShowDocument?docid=40216 here]!.'''
  
 
'''!General technical note describing the analysis algorithm of the slow monopole triggered data one can find [https://nova-docdb.fnal.gov/cgi-bin/private/ShowDocument?docid=22610 here]!.'''
 
'''!General technical note describing the analysis algorithm of the slow monopole triggered data one can find [https://nova-docdb.fnal.gov/cgi-bin/private/ShowDocument?docid=22610 here]!.'''

Latest revision as of 17:01, 11 October 2019

The NOvA test stand was made at JINR for studying native electronic responses through Avalanche PhotoDiode and Front-End board chain. The bench was used to perform investigations for physical signals in the detectors also. Main activities of the NOvA-JINR stand are connected with signal shaping studies, simulation of the electronic response on hypothetical magnetic monopole signals, sag measurements and other required to do simulations and precise energy reconstruction.

Primal goal

One can see the interesting hit structures in the picture below. They were called Flashes. The main feature of the flashes is that most of the FEB channels in the same FEB produce signals over thresholds simultaneously but these events take place only in high energy region. It makes them really important for high energy deposition and may take effect on background estimation (cosmic muons) or some Exotics (monopoles etc).

NOvA-flashes.jpeg

That's why one of the main reasons to have test bench was the investigation of hit structures but of course it is not only one. It is possible to measure the shaping parameters of the electronics and electronics' response to different initial light signals. It is very important for the MC simulation and right understanding of the electronics behaviour.

General technical information

NOvA test bench at JINR consists of a few parts. First is the native NOvA electronics — Avalanche PhotoDiode and Front-End Board. Second is the special hardware — download cable, DCM-emulator, LED, Low and High Voltage Sources, Pulse generator and cooling system. Third is a PC with necessary software.

All sensitive devices were put into a black metal box. The black box allows to perform all the measurements with photodetectors like APD and PMT and screens all external electromagnetic noises. Since APD in NOvA operate at -15 degrees we employed the cooling system based on Nitrogen evaporation flow.

Pulse generator waits for the FEB trigger and sends the electric pulse to the LED. After that the light pulse from the LED comes through the fibers to the APD and FEB is reading at this moment.


General hardware bench view

Bench general view.jpg


Inside the black box

Inside-black-box.jpg


General bench and data taking schemes

Data-taking-scheme.jpg

Results

Cross-talk measurements and dynamic range extension

When test bench had been built we started with APD response on different light intensities measurements. It was very important for understanding the detectors performance for high energy dissipations. By sending very high intensity light into a single APD pixel which saturates ADC we found out that the same small inverted signals occur in most of channels. First we checked FEB by injecting a huge charge into a single FEB channel and we observed normal cross-talk in neighbouring channels that drops exponentially. The next step was to check APD feeding chain. We performed measurement of APD gain with different temperatures to study APD operation voltage at different temperatures. In NOvA detectors all the APDs operate with the gain about 100. We employed different capacitors into FEB PCB and obtained different value of negative cross-talk pulse. The suggestion was that capacitance blocks a voltage drop which was named signal sag.

Sag is a value of the amplitude of the inverted signal in the neighbouring channels in case of huge amplitude in the primary one. We put different amplitudes to APD and measured the value of the sag. The first result was that relative value of the sag doesn't depend on the amplitude. It is equal to 1.89 %. Sag doesn't depend on the temperature and only thing that can define it is bypass capacitor. If we replace original capacitor by 9.4 nF we could reduce the effect sevenfold. This effect could be very helpful for extending of the ADC dynamic range. Sag can be measured in many channels (up to 31) which gains precision. As the result we can increase the statistics and expand the dynamic range by 10 times.

NOvA-sag.jpeg NOvA-cap.jpeg

Monopole simulation and Response to long signals

The approximate scheme of the special analog simulation of the electronic's response on hypothetical magnetic monopole signals. "Ordinary" response is the standard behaviour of our electronics for the external signal. Also we have a theoretically predicted monopole signal. Since monopole leaves a huge energy and travels slowly through the detector the light pulse duration may reach level of a few microseconds. After the convolution with electronic shaping it was anticipated to obtain signal like it shown in the picture below. During the simulation we noticed that the tail of the signal was longer than it's expected and it depends on the amplitude. The behaviour of this effect is linear.

Special kind of experimental setup was made for measurements with long signals. In addition to the ordinary setup PMT and Digitizer were added. PMT was used to monitor the shape of the initial light pulse and ADC with fast sampling to digitize PMT's signals.

The main idea was to send light with the same integral intensity which corresponds to the constant charge. We tried to generate rectangular-like light pulses. The ASIC shaper integrates the APD pulse width and converts into a Rise time and shaping parameters vary on pulse amplitude. We thank Dr. Martin Frank for useful discussions and help during this measurement.

Sig-conv.png

NOvA-APD-PMT.jpeg

Improved Monopole simulation and Response to long signals

Main parameters of the previous modeling were:

1) Rectangular shape electric pulse (this shape went from the suggestion of high ionization power of the monopole) – it converts into rectangular light pulse and this light pulse goes to PMT and APD – this assumption is correct.

2) For the first time point (20 ns) we chose the signal with amplitude that didn’t saturate the APD and had 20 ns width. We calculated charge and kept it constant for all other time points – increased the width and decreased the amplitude – for improved modeling we have to select primary light (and then electric) pulse amplitude according to NOvA soft simulation output.

NOvA soft simulation has several features:

1) Several possible dE/dx for monopoles (high, nominal and low). They are connected with monopole energy loss uncertainties.

Monopole energy loss.png

2) Two possible APD gains – low (100) and high (140). They are connected with two sets of NOvA data taking.

Sim.jpg

We have to extract primary light pulse after the scintillator and fiber, put it inside the bench and investigate what is the lowest possible signal that NOvA readout chain can detect.

Results for the high dE/dx and high gain (highest energy deposition) and low dE/dx and low gain (lowest energy deposition) are as follows (beta = 10-4):

High dE/dx and high gain EVD and planes signals (planes 7 and 21, cell 192 – it is the beginning of the detector along the monopole track):

High dEdx high gain.png

Plane7 192 hh.png Plane21 192 hh.png

Low dE/dx and low gain EVD and planes signals (planes 7 and 21, cell 192 – it is the beginning of the detector along the monopole track):

Low dEdx low gain.png

Plane7 192 ll.png Plane21 192 ll.png

Light signal extraction

On this step we want to find number of photoelectrons for every monopole beta point.

First of all we have to dump photon signals to a textfile. We have a special module (PhotonDump) for this (great thanks to Andrey Sheshukov for his help). It works with PhotonSignals class — inside one can find the number of photoelectrons for a cell binned in 2 ns chunks. Our module extracts these numbers for every plane and cell along the track.

Then I have a script that sorts photoelectrons in time and can draw these distributions for every cell and plane and also draw general distribution along the monopole track.

Resulting plots for one cell with low and high statistics (beta = 0.316 * 10-3) are shown below:

0 316 10 3.jpeg 0 316 10 3 hs.jpeg

At final step I made plots that look like Npe (photoelectrons number) vs Time for all cells along the monopole track (sum) and average number of Npe for all cells. Second one is used like input parameter for the light signal in test stand modeling.

Beta.jpg

Preparatory measurements

1) PMT and SiPM comparison - Quantum efficiency check.

PMT SIPM.jpg

2) Fibers splitter check.

I used PMT and DRS4 Evaluation Board ADC - I’ve made several runs to test fibers brightness.

PMT-before.jpg

3) PMT gain measurement.

PMT gain was measured on low light using Poisson distribution of PMT response.

PMT gain elb.jpg

4) Light adjustment (Pulse generator).

One has to “search” for light pulse with such a number of photons inside using monitoring PMT.

Light-adj.jpg

5) APD pixel by pixel gain influence.

We decided to check the influence of APD pixels gains variations on one PCB board on our readout chain output. We found that our tested channels are super close to each other and APD pixels gain variations can’t take any effect on our further measurements.

Absolute and relative long signals measurements

For absolute APD measurements – we have to prepare the special light pulse that corresponds to correct Npe from the simulation on APD. Monitor number of photons during all data taking using PMT. In case of Low Light (less then 200 photons) it is better to use x9.75 CAEN amplifier.

For relative APD measurements – choose one Npe value (87 p.e.), make the shortest possible light signal with charge that corresponds to this Npe (the lowest possible signal width on our generator is 6.25 ns) and try to keep this charge constant for all time points (one has to increase the signal width and decrease the voltage on LED). Monitor number of photons during all data taking using PMT.

All the measurements were made at the room temperature +25 oC.

Amplitudes/Npe vs betas (Absolute measurements) — Preliminary result

A-Npe-abs.jpg

Amplitudes/Npe to Start Point vs betas (Relative measurements) — Preliminary result

A-Npe-SP-rel.jpg

C-scripts for both FEB operation modes output files

I used both FEB operation modes and extracted amplitude values from both output file types - in case of DSO amplitude is Max value inside the array minus Pedestal, in case of DCS it is the difference between last and first points (DCS algorithm has 4 points like an output). Wide distribution for DCS FEB mode is connected with sampling clock that is not synchronized to our pulsing trigger. This means that on a trigger by trigger basis we have a t0 offset between the first sample and when the pulser actually flashes. This is what is giving our DCS envelope such width. If one doesn't take out this jitter he will end up with the wrong rise time. This feature has no effect on current measurement but I plan to add fit-function to my C script in future.

Resulting plots of APDs outputs for two FEB operation modes (Oscope (left) and DCS MP (right)) are:

Oscope output monopole.jpeg DCS output monopole.jpeg

Results validation

Preliminary results show very interesting results but it is really hard to find an easy explanation for such NOvA readout chain behavior. Before NOvA readout chain investigation we decided to check LED and PMT stability.

We placed SiPM into the Black Box and connected fiber 1 that previously was connected to PMT to him.

SIPM Black Box.jpeg

Photoelectron number extraction

Here we use statistical method of extracting of photonelectrons for photosensors (please see formula 3 here {Precise analysis of the metal package photomultiplier spectra). We can connect photoelectron numbers with sigma of the signal and signal itself.

We used SIPM to check LED and PMT stability. First we have to check SIPM itself. We put the same time points (signal width) and generator parameters like for the previous measurements and checked what photoelectron numbers ratio and gains do we have for each point.

SIPM gains.jpeg

We can say that SIPM gain is pretty stable for all time points. This also means that LED is stable too. But also SIPM shows that we have less light for longer signal width then we thought. So it is possible that PMT gain can vary with signal width. To check this we applied statistical method for PMT output files.

PMT photon numbers ratio and gains

Pmt photons.jpeg Pmt gains.jpeg

PMT gain increases with signal width (slightly decreases on the very end). This PMT behavior may connect with dynode system. It is a personal feature of photosensor. Gain pattern allows us to make the correction for previous results (put the right N_pe number to the denominator).

Final results

Ch3 A Npe SP abs betas.jpeg Ch19 A Npe SP abs betas.jpeg

Ch3 A Npe SP rel betas.jpeg Ch3 A Npe SP rel betas zoomed.jpeg

Ch19 A Npe SP rel betas.jpeg Ch19 A Npe SP rel betas zoomed.jpeg

Double check with SIPM

We tested several relative time points using SIPM to make sure that everything is correct right now. APDs were tested using the same generator parameters but with SIPM like monitoring photosensor. Output behavior well matches with PMT one.

Ch3 A Npe SP sipm.jpeg Ch19 A Npe SP sipm.jpeg

Technical notes

!Extended information and all technical details about long signals measurements one can find here!.

!General technical note describing the analysis algorithm of the slow monopole triggered data one can find here!.

Shaping parameters

For simulation of the detectors performance NOvA collaborators asked us to find all the shaping parameters for both detectors. It was necessary for computer modelling and simulations. The concept of the measurement was the same that aforementioned. The fall time isn't fit to the table value. It linearly depends on the amplitude. The Rise time doesn't fit to the table value either but it doesn't vary or the variation is negligible.

NOvA-shapping.jpeg

We develop the powerful tool for direct manipulation and measurements with NOvA electronics. A lot of issues have been solved and JINR-NOvA stand can solve many problems or clarify some misunderstood features in the future.

AlexanderAntoshkin (talk) 21:09, 21 August 2018 (MSK)