Star-fst-l mailing list

[Daniel Brandenburg <dbrandenburg.ufl AT gmail.com>]

Hi Jason, all,

Thank you for some clarification. But the results in these slides do not support that conclusion for the Si detector.

The tracking with sTGC only should have ~100% efficiency at low multiplicity since the sTGCs have 100% intrinsic efficiency and good resolution (and even better granularity in these studies than now) for the requirement of 4 or more hits on track. It is not on the slides but we get the same efficiency in our recent studies (using this tracker) with that implicit requirement so i am confident that cut is what is used in this study. We also know that with this requirement the Si plays a very small roll in the tracking (many tracks get only 1 or zero Si hits, if you require all Si efficiency drops significantly). You can actually see this directly from the later efficiency study using ftsref7/8 which has only sTGCs - and the efficiency is  nearly identical to what you show in earlier slides with Si+sTGCs - this shows that Si platys almost no role.  If I saw these plots with the requirement that all stations were used in tracking I would believe the conclusion about R-segmentation.  To me, this means that these studies are insensitive to the Si radial segmentation since the sTGC hits play the dominant role in tracking (when requiring only 4 or more hits) - especially since the primary vertex to the sTGC hits is the longer lever arm and the primary vertex here had a  small uncertainty. 

So, if this is the study that supports that R segmentation does not matter much then I am definitely not convinced. 

Thanks,

Daniel

On Wed, Jun 26, 2019 at 4:11 PM Jason C. Webb <jwebb AT rcf.rhic.bnl.gov> wrote:

Hi Zhenyu,

I really don't recall the full context of the conversation with Daniel.  But let me clarify my
understanding of the two big bullet points:  segmentation and stereo angles...  and then I'll
address the issue w/r to the global track plots which you mentioned...


1) Radial segmentation:  going back to some the earliest studies:

Silicon only (see ~page 8) --
https://drupal.star.bnl.gov/STAR/system/files/FTS%20Tracking%20Plots%203_0.pdf

Silicon + sTGC (see ~ page 87) --
https://docs.google.com/presentation/d/1skB2Uhg8keu2BX1Eg5ijP7vjo0tNE2x3IzFtb6bsVGU/edit?usp=sharing

Increased radial segmentation showed modest improvement in efficiency at the most
extreme track densities.  But there was not really a make-or-break issue with radial segmentation.
In the Si+sTGC studies, you can look at efficiency vs centrality with different radial vs phi
segmentation ~page 103.

Bottom line: I think the radial segmentation of ~8 will work fine (but more segmentation can
improve matters somewhat in more central collisions).


2) Rastering:  this was always discussed in terms of the Si-only tracker

Silicon see pages 10 and 14
https://drupal.star.bnl.gov/STAR/system/files/FTS%20Tracking%20Plots%203_0.pdf

Shows a class of failed fits.  Believed to be due to small curvature tracks passing
through the same phi strip, leading to a degenerate (straight line) fit.

Rastering was introduced (slide 21), by placing the center of the disks off the beamline by 1mm.
This solved the issue of the failed fits by effectively introducing a stereo angle between
each Si disk.

Bottom line: stereo angles between Si planes are needed in order to reconstruct small curvature tracks.  This
can be introduced via rastering, but could also be built into the detector design.


3) "Jason, I was also told by Daniel that the global track efficiency for those plots is
Cheers,
  wrong. Can you please tell me what is wrong with them?"

"Those plots" refer to slide 42 of the proposal...
https://drupal.star.bnl.gov/STAR/system/files/ForwardUpgrade.v20.pdf

There was nothing wrong with them per se, but they were produced on a very short time scale and there were
open questions at the time which were not fully addressed.  "Those plots" originated on slide 23 here--
https://drupal.star.bnl.gov/STAR/system/files/FTS%20Tracking%20Plots%203_0.pdf

The open question was why (at high pT) we saw global track efficiencies > 100%.  The reason was that the
global tracks were contaminated by combinatorics.  So when I say the global track plots were wrong, I mean
that they had a source of contamination which wasn't properly accounted for.  Primaries, on the other hand,
had the additional requirement that they match the MC vertex.  This seemed to help suppress the conbinatorics
in the study.

Bottom line here... in those plots on page 42 of the proposal, trust the primary tracks as more reliable than
the globals.


Hope this helps,
Jason





On 6/26/19 10:29 AM, Zhenyu Ye wrote:
> Hi Daniel,
>
> I think I was confused with your statements "The study from Jason was later changed though - he is one of the S&C people that made the point to me that more r segmentation was needed. I talked with him/Gene about this last 1.5 week. Also I know that the global track efficiency on those plots is wrong, also from talking with Jason." "OK i talked with Gene about the hit rasterization in R just 1 or 2 ago”.
>
> I guess I know better now what was said to you after reading your email below but still confused to some extent. So is it correct that Gene (and Jason?) pointed out to you the importance of r rasterization, and Jason pointed out to you the need of more r segmentations?
>
> Thanks,
> Zhenyu
>
>> On Jun 26, 2019, at 8:46 AM, Daniel Brandenburg <dbrandenburg.ufl AT gmail.com> wrote:
>>
>> Hi all,
>>
>> maybe just a little clarification:
>>
>> Gene, in our meeting a week ago about the forward tracking we discussed difficulty with the silicon. Specifically I talked about the resolution in R and you guys mentioned that rasterizing the hits was study and found to be important.
>>
>> Jason, we talked about the global efficiency some time ago now. We did several studies of the lambda decays with the older fast sim (with different resolution etc.) and with the newer ones for Si and sTGCs. We found that the lambda daughters (true global tracks, not just primary tracks without the primary vertex included in fit) had a very low reconstruction efficiency ( maybe 10%, I dont remember exactly) - we talked about this and how the tracker does the track following etc.
>> So my comment is not so much that those plots are wrong but that it is definitely not the performance we see with the current detector resolutions. I hope that is clear.
>>
>> Since i was not active with the forward tracking before last year I understand that a lot was done before I was active. I want to make sure we fully understand this issue before we get further along.
>>
>> Best,
>> Daniel
>>
>>
>> On Wed, Jun 26, 2019 at 9:16 AM Zhenyu Ye <yezhenyu2003 AT gmail.com> wrote:
>> Hi Jason, Gene,
>>
>> I was told by Daniel Brandenburg that you made the point to him very recently that more radial segmentations are absolutely needed for Forward Silicon Tracker. This would contradict to the statement in page 41 of the proposal https://drupal.star.bnl.gov/STAR/system/files/ForwardUpgrade.v20.pdf
>>
>> “In addition, we increased the radial segmentation from 8 to 16, 32, and 64 and found no significant improvement on the momentum resolution or tracking efficiency.”
>>
>> Can you please help me understand why you believe more radial segmentations are needed for FST?
>>
>> Jason, I was also told by Daniel that the global track efficiency for those plots is wrong. Can you please tell me what is wrong with them?
>>
>> Thanks,
>> Zhenyu
>

--
.--- .-- . -... -... .--.-. -... -. .-.. .-.-.- --. --- ...-