[Sbs_gems] On APV25 polarity flip

Wed Feb 23 16:55:22 EST 2022

Ah. Thanks for clarifying. That is still not great, but considerably less scary.

Andrew

From: Gnanvo, Kondo (kg6cq) <kg6cq at virginia.edu>
Date: Wednesday, February 23, 2022 at 4:53 PM
To: Andrew Puckett <puckett at jlab.org>, Holly Szumila-Vance <hszumila at jlab.org>, Sbs_gems at jlab.org <Sbs_gems at jlab.org>
Subject: [EXTERNAL] RE: On APV25 polarity flip
Hi Andrew,
No I am not saying that it affects all strips in an APV. In the Hall D data, it affects only a few channels but these strips are the one carrying the signal ==> I lost these data because it last for the full time window of the triggered event. It is not a long lasting flip, the next triggered event, that will no longer be the case for the same channels

But the situation might be different for SBS GMn case because my hypothesis is that these flips are caused by background events long before the triggered event, so we lose only hits that happens at the given strip but even then, it is still a pretty bad situation

Yes, in principle, you can set APV25 baseline pedestal in the middle of the ADC range, I can do that with APV25-SRS, but I don’t know how well it works with signal of both polarity. I would not expect to be great or balance though

Best regards
Kondo

From: Andrew Puckett <puckett at jlab.org>
Sent: Wednesday, February 23, 2022 3:25 PM
To: Gnanvo, Kondo (kg6cq) <kg6cq at virginia.edu>; Holly Szumila-Vance <hszumila at jlab.org>; Sbs_gems at jlab.org
Subject: Re: On APV25 polarity flip

PS—If the analysis Sean is working on shows a high probability of “negative hits” in the area where positive hits are missing from good tracks, then that would be pretty close to a smoking gun, and would have significant implications for the analysis of the GMN data set.

In light of this possible failure mode, I am glad we didn’t run GMN at the proposed luminosity or we might have inadvertently killed even the reduced physics output of the experiment that we got.

From: Sbs_gems <sbs_gems-bounces at jlab.org<mailto:sbs_gems-bounces at jlab.org>> on behalf of Andrew Puckett <puckett at jlab.org<mailto:puckett at jlab.org>>
Date: Wednesday, February 23, 2022 at 3:20 PM
To: Gnanvo, Kondo (kg6cq) <kg6cq at virginia.edu<mailto:kg6cq at virginia.edu>>, Holly Szumila-Vance <hszumila at jlab.org<mailto:hszumila at jlab.org>>, Sbs_gems at jlab.org<mailto:Sbs_gems at jlab.org> <Sbs_gems at jlab.org<mailto:Sbs_gems at jlab.org>>
Subject: Re: [Sbs_gems] On APV25 polarity flip
Hi Kondo, looks pretty convincing for these few examples. If it’s what we are seeing in Hall A, this could be a more serious and fundamental limitation than simple bias of the common-mode calculation or gain drop, if the “negative occupancy” is high enough and the “polarity flip” probability is high enough, and if there is no way to eliminate/fix this problem.

A question that I have is, you seem to be implying that if this happens, it affects ALL strips on any given APV affected by this issue. Is that correct? If so, then the frightening implication is that the good signal would be not even in principle recoverable for any strips on that APV, when this condition exists.

If the “negative occupancy” as calculated by Sean is proportional to the “positive occupancy” and comparable in magnitude, and if this quantity is a reasonable proxy for the “polarity flip” probability, then you might imagine that any given APV in any given event, including the one containing the good signal in that event, would suffer an efficiency loss of similar magnitude as the “negative occupancy”.

A good study might be to see if the observed reduction in elastic yield with beam current is consistent with this hypothesis.

I can imagine a possible “hack” where we set the baseline at the midpoint of the ADC dynamic range (I don’t know what is possible in the electronics configuration), and then somehow lower the gain to fit the signal within the reduced dynamic range of the ADC, and then detect the polarity of any given time sample in any given event, and do analysis using the absolute value of the ADC. But that’s just spitballing.

Andrew

From: Sbs_gems <sbs_gems-bounces at jlab.org<mailto:sbs_gems-bounces at jlab.org>> on behalf of Gnanvo, Kondo (kg6cq) <kg6cq at virginia.edu<mailto:kg6cq at virginia.edu>>
Date: Wednesday, February 23, 2022 at 2:53 PM
To: Holly Szumila-Vance <hszumila at jlab.org<mailto:hszumila at jlab.org>>, Sbs_gems at jlab.org<mailto:Sbs_gems at jlab.org> <Sbs_gems at jlab.org<mailto:Sbs_gems at jlab.org>>
Subject: [Sbs_gems] [EXTERNAL] On APV25 polarity flip
Dear all,
I thought I would share again the slides I showed last October when we noticed the “negative pulses” issues for the first time early during the GMn run.

The slides are about some tests I was performing in 2017 in Hall D with a GEM prototype used as transition radiation detector and this is important because in that case the prototype has 21 mm drift so total primary ionization is 7 times higher than our standard GEM tracker and was causing the polarity flip.
Here I am referring to polarity flip because the data is clean enough that we could actually confirm that APV25 flip polarization without too much speculation and I am going to try to explain the reason why.

  1.  First, these were data taken with APV25-SRS electronics (not MPD) and for that, as a convention, the raw APV25 signal points downside (just a software decoding convention) as you can see on slide 1. Also, because it is GEM-TRD which operates in TPC mode, we needed to operate in the full APV25 data width mode (27 time samples) to be able to catch (in time and space) different ionization clusters and TRD photon signal from the same electron hit traversing the detector

  1.  Slide 2 shows one event with clearly polarity flip in y-strips.
     *   The two top frames are x-strips raw data frames and on the right, you see actually signal from at least 2 ionization clusters (waveforms clearly separated in time)
     *   The bottom 2 frames are the corresponding signal in y-plane where the signal flipped early  after 3 time sample and remained in the flip mode for almost all the 24 remaining time sample
     *   The system here triggered on the Hall D Pair spectrometer electron, so it is a clean trigger with on hit per event virtually no background. This can be confirmed by standard detector so no debate here. So we see hits in x-strips , we expect hit in y ==> what we are seeing is an actual polarity flip
  2.  Slide 3 shows 3 different event with polarity flip in y and more than half of the actual data at the time where showing the same behavior and 95% of the time on y strips
  3.  The explanation on why it is always on y-strips is pretty simple, it it cause by pile-up because of  if you look at the 2D profile of Hall D PS electron beam, it is a narrow strips, with the electron normal to the detector in the vertical direction but coming at a large angle ~10 degree in the horizontal direction ==> for a GEM-TRD which produce an ionization 7 times higher than a standard GEM, the signal collected in x-strips will show several waveform separated in time and also in spaces (several different strips collect the signal form the single electron). But the same signal will be hitting a small set of strips in y causing pile-up and I believe that what cause the polarity to flip.
  4.  When the flip happens, it stays on for several time samples > 20 APV25 time sample ==> It is very likely that the negative pulses we are seeing also are due to similar background hits pile-up that happens way before our real triggered signal but remained in flip state for the time window of the triggered signal and in that case, you will not see when the flip happens, only that it is there.
  5.  You can also clearly see the saturation effect because of the ADC dynamic range for the flip signal, depending on where the baseline is set, the flipped signal amplitude will always be small and the waveform shape will be flat ==> It does not mean that it is not real signal.
  6.  In some case, you see a glimpse of flipping back to the correct polarity (circled in magenta) but the signal is never fully recover and with 6 time sample, it is not clear if e will even have this luck

This is the more serious problem we have other than the effect on corrupting the calculation of CM. It is a long email but I hope these few slides shed a little bit of light on the issue we are trying to understand

Best regards
Kondo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/sbs_gems/attachments/20220223/7a5ca5aa/attachment-0001.html>