Star-tpc-l mailing list

["Van Buren, Gene" <gene AT bnl.gov>]

Hi, TPC-ers

I took it upon myself to see where hits have flag==2 in two nightly test datasets:

1) 2016 AuAu200 pre-iTPC data

2) 2019 AuAu4.59 FXT iTPC-era data

In the 2016 data, as seen in the first plot showing percentage of hits with flag==2 vs. padrow, we have over a third (34%) of the hits getting this substantial increase in TPC hit errors. The rate climbs at lower radii presumably because occupancy is also climbing, and thus the chance of overlapping hits. I looked at other possible dependencies (across padrows, in z, phi, etc.) and saw no other dependencies worth noting here.

In the 2019 FXT data (second plot), the dependency on padrow is largely gone, but a dependency on iTPC (inner) vs. TPCX (outer) is prominent, at rates of ~25% and ~13% respectively. I again looked at other dependencies and found only one particularly worth noting, shown in the third attached plot of fraction of flag==2 hits vs. padrow and z. This shows that the rate climbs above 50% at large negative z (far away from the target).

The questions these findings raise for me are...

a) While I understand de-convoluted hits should have larger hit errors, is the proposed increase in magnitude reasonable for such a significant fraction of hits?

b) When opening hit errors so much, doesn't that allow more false inclusion of hits on tracks as well?

c) When opening hit errors so much, does that aid in connecting tracks across the central membrane independently of any improvements to TPC alignment?

______________________

....The above led me to take a very quick look at track chi^2, which should depend heavily on the used hit errors.

For the 2016 data, mean global track chi^2 drops from 2.09 to 1.46. This makes sense with the larger hit errors reducing chi^2.

For the 2019 FXT data, mean global track chi^2 drops from 1.221 to 1.061. This may also make sense for the hit error increases.

Similar reductions are seen in mean primary track chi^2.

It isn't clear to me how these observations meld with the UCDavis group seeing large increases in track chi^2 in Yuri's data production.

Thanks,

-Gene

On Oct 5, 2024, at 10:13 AM, Fisyak, Yuri V <fisyak AT bnl.gov> wrote:

Hello, 

• Comparison for Gene’s samples  https://www.star.bnl.gov/~fisyak/star/TbyT/AuAu200_2016_HitError/ shows that increase tpc hit errors for deconvoluted clusters, and only that, increase relative efficiency by ~ 4%.

• This topic is not related to QA from the DAVIS group. 

 

                                                              Yuri Fisyak

STAR                                           Phone: +1 631 344 3913
Brookhaven National Laboratory  Fax:    +1 631 344 4206
510A/1-161
http://www.star.bnl.gov/~fisyak       E-mail: fisyak AT bnl.gov
 
 
 
From: star-tpc-l-request AT lists.bnl.gov <star-tpc-l-request AT lists.bnl.gov> on behalf of videbaek <videbaek AT bnl.gov>
Date: Friday, October 4, 2024 at 7:44 PM
To: Frank Geurts <geurts AT rice.edu>
Cc: Star-tpc L <Star-tpc-l AT lists.bnl.gov>
Subject: Re: [[Star-tpc-l] ] Fwd: [star-bnl/star-sw] 💥DO NOT MERGE💥 ❗ add new TPC alignment, hit errors and slewing corrections for 2024 (PR #702)

I was typing and something happened and send the e-mail before I was 
done with the second part.
recently we were shown some of the QA from the DAVIS group and it was 
pointed out the the chisqaured from trackfitting was about 2 rather than 
one in previous productions. Was this actually resolved. The
change for flag=2 clusters could actually improved this. Deos it?

best F


On 2024-10-04 13:40, videbaek wrote:
> Hi
> 
> thanks for the forward to the tpc list.
> I have two questions/comments
> 
> At the meeting Yuri said the error was increased by factor 4. I assume
> the fudgefactor is a multiplier on somethings like sigma**2 ?
> I recall way back I questioned why de-convoluted had same erros a
> isoltaed clsuter, so this is actually a good change.
> 
> 
> 
> 
> On 2024-10-04 12:38, Frank Geurts wrote:
>>> Begin forwarded message:
>>> From: Gene Van Buren <notifications AT github.com>
>>> 
>>> Subject: Re: [star-bnl/star-sw] 💥DO NOT MERGE💥 ❗ add new TPC
>>> alignment, hit errors and slewing corrections for 2024 (PR #702)
>>> 
>>> Date: October 4, 2024 at 11:05:22 AM CDT
>>> 
>>> To: star-bnl/star-sw <star-sw AT noreply.github.com>
>>> 
>>> Cc: Frank Geurts <geurts AT rice.edu>, Mention
>>> <mention AT noreply.github.com>
>>> 
>>> Reply-To: star-bnl/star-sw
>>> <reply+AEMNVR6PVQ67TU7N4GHG4NWFBPZUFEVBNHHJEYYC7A AT reply.github.com>
>>> 
>>> @genevb commented on this pull request.
>>> -------------------------
>>> 
>>> In StRoot/Sti/StiTrackNodeHelper.cxx [1]:
>>> 
>>>> +    if (tpcHit) {
>>> +      if ((tpcHit->detector() == kTpcId || tpcHit->detector() ==
>>> kiTpcId)) {
>>> +    if (tpcHit->flag() == 2) {
>>> +      fudgeFactor = 16.;
>>> +    }
>>> +      }
>>> +    }
>>> 
>>> Yuri suggested that these lines of code in
>>> Sti/StiTrackNodeHelper.cxx, modifying some TPC hit errors, might be
>>> responsible for the differences seen in nightly tests with this PR
>>> vs. DEV for pre-iTPC datasets where the alignment should be
>>> unaffected. Processing a few events with and without this bit of
>>> code is quite easy, and comparing the log files obviates that the
>>> bulk of the difference in track counts does disappear when removing
>>> this section of code.
>>> 
>>> However....from looking at the log file, I see that the new
>>> alignment scheme appears to alter TPC distortion corrections for
>>> these old, pre-iTPC datasets without any change to the
>>> reconstruction chain. This is in addition to the modification to TPC
>>> distortion corrections ("OSectorAlign") that Yuri wants to impose
>>> with the "CorrZ" chain option he has proposed for iTPC-era datasets
>>> for which I have expressed an interest in seeing a justification
>>> (through data, not anecdotally).
>>> 
>>> < StMagUtilities::XTWIST        =  -0.38 mrad
>>> < StMagUtilities::YTWIST        =  0.25 mrad
>>> ---
>>>> StMagUtilities::XTWIST        =  0 mrad
>>>> StMagUtilities::YTWIST        =  0 mrad
>>> 
>>> < StMagUtilities::WestClock     =  -0.43 mrad
>>> ---
>>>> StMagUtilities::WestClock     =  0 mrad
>>> 
>>> I do not recall this being discussed before, neither in the context
>>> of pre-iTPC datasets, nor with the new alignment. If the TPC
>>> alignment calibration was truly performed without any correction for
>>> the distortions due to the misalignment of the E and B fields
>>> (represented by the XTWIST & YTWIST numbers), I have significant
>>> concerns about this TPC alignment. Hopefully I am missing something
>>> and this is not the case.
>>> -------------------------
>>> 
>>> I have placed output of the nightly test without this particular
>>> code (but including the rest of this PR) at this location:
>>> 
>>> This PR minus hit error change :
>>> 
>> /star/rcf/test/gitdev/daq_sl302.stica/Mon/year_2016/AuAu200_production_2016_64bit/
>>> 
>>> This is for comparison with other iterations:
>>> 
>>> DEV :
>>> 
>> /star/rcf/test/dev/daq_sl302.stica/Sat/year_2016/AuAu200_production_2016_64bit/
>>> This PR :
>>> 
>> /star/rcf/test/gitdev/daq_sl302.stica/Tue/year_2016/AuAu200_production_2016_64bit/
>>> This PR before the Sept. 11th commits :
>>> 
>> /star/rcf/test/gitdev/daq_sl302.stica/Wed/year_2016/AuAu200_production_2016_64bit/
>>> 
>>> I am making note of the last two in particular because Yuri's study
>>> involved the output from the last one (those jobs were started on
>>> September 11th before the time of Yuri's last commits), even though
>>> data with those commits was available. (Side note, please pay no
>>> heed to the "Tue" or "Mon" in the gitdev paths, as I moved files
>>> there from other days to avoid them being over-written.) I will note
>>> that I didn't see any effective impact of those commits on this
>>> particular test's track counts (i.e. the last two outputs may be
>>> identical).
>>> 
>>> —
>>> Reply to this email directly, view it on GitHub [2], or unsubscribe
>>> [3].
>>> You are receiving this because you were mentioned.Message ID:
>>> <star-bnl/star-sw/pull/702/review/2348366613 AT github.com>
>> 
>> 
>> 
>> Links:
>> ------
>> [1] 
>> https://github.com/star-bnl/star-sw/pull/702#discussion_r1787896945
>> [2] 
>> https://github.com/star-bnl/star-sw/pull/702#pullrequestreview-2348366613
>> [3] 
>> https://github.com/notifications/unsubscribe-auth/AEMNVR3S4SFYAFRVTDX3AF3ZZ24EFAVCNFSM6AAAAABMGT4I5OVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDGNBYGM3DMNRRGM

-- 
Flemming Videbaek
senior scientist, emeritus
videbaek @ bnl.gov
Brookhaven National Lab
Physics Department
Bldg 510D
Upton, NY 11973

phone: 631-344-4106
cell     :  631-681-1596