sPHENIX MAPS tracker discussion

["Huang, Jin" <jhuang AT bnl.gov>]

Thanks for the immediate response! Few follow up:

 

 

          -> All indications we see from the commissioning data show the MVTX detector operated properly. Youngil says it is the important message we can send to the community. What do you think of it?

 

I see. OK.

 

         -> We made a big modification following your comments. There was a big change in the poster contents, and we overlooked your points in a hurry.

 

Thanks for the update. This is a complex topic to explain and hard to fit into the space on the poster. If you have a public talk or paper describing this method, it could be helpful to provide a link (or QR code).

 

Also good to practice on responding to a few possible questions before the poster session, e.g.

• How layer information is propagated to module and sensors?
• Where input track information goes into the NN?
• What was the training-validation procedure?

 

 

        -> The original plan was to describe the alignmnet procedure, but we had a few issues and what I wanted was, 1. detector perfomred decent and 2. we can start the alignment finally.

 

OK. For those purposes, I would suggest write on the poster along the line that these residuals are from the first look of the data. While further studies are on-going,  detector performance was decent, and we can start the alignment procedure. I agree with you that this is an important message.

 

I would suggest removing the last paragraph describing the details on the residuals and keep it for verbal discussion in person. The explanation on the shifts seems speculative. Without a study to prove or disapprove them (e.g. with simulation), the current paragraph might draw questions and disagreement from a reasonable reader.

 

Cheers

 

Jin

 

 

 

______________________________

 

Jin HUANG

 

Physicist, Ph.D.

Brookhaven National Laboratory

Physics Department, Bldg 510 C

Upton, NY 11973-5000

 

Office: 631-344-5898

Cell:   757-604-9946

______________________________

 

From: Jaehyun Kim <xhaxha AT yonsei.ac.kr>
Sent: Wednesday, August 30, 2023 9:16 AM
To: Huang, Jin <jhuang AT bnl.gov>
Cc: Liu, Ming <ming AT bnl.gov>; Anthony Frawley <afrawley AT fsu.edu>; Youngil Kwon (ykwon AT yonsei.ac.kr) <ykwon AT yonsei.ac.kr>; SPHENIX MVTX <sphenix-maps-l AT lists.bnl.gov>
Subject: Re: [EXTERNAL] Re: ML alignment poster?

 

Dear Jin,

 

Here I embed responses to your comments.

 

Yours,

Jaehyun

 

2023년 8월 30일 (수) 오후 7:26, Huang, Jin <jhuang AT bnl.gov>님이 작성:

Dear Jaehyun,

 

Very impressive work on the initial analysis and initial alignment study for MVTX!

 

Please find my comments below, which focuses on the poster:

 

• In general, please remember to change all “sPHENIX Internal” plots to sPHENIX preliminary after their approval.

          -> Sure, I will keep that in mind.

• Abstract: “we verify the successful operation of the MVTX detector in the commissioning stage” is very broad suggest change to
  -> We analyzed the collision data from the commissioning stage

          -> All indications we see from the commissioning data show the MVTX detector operated properly. Youngil says it is the important message we can send to the community. What do you think of it?

• Cluster Characteristics:

• Figure 1 is hard to tell the difference and message.

• Instead of the right figure, suggest use Fig4 in the analysis note which is quite telling for the background.

                                 -> Okay. We included the plot as recommended. 

• “Approximately 50 of the readout frames have small numbers, specifically 12”

• Is this expected? Given collision rate and frame rate, the percentage of frames without collisions should be calculable.

                                  -> It is an observation based on data. Most of the remaining 50% is also background events. 1/50 will contain ZDC coincidence.

• If this is expected, please just comment on the poster that “the MVTX uses streaming readout, recording all hits grouped in time frames of XX us. Only the time frames those contain collision events were used in this analysis.“

                                   -> We clarified how the detector was read out.

• Alignment with AI ML Approach

• This section is very unclear. For example

• What is s1 s2, f(), s1, s2
• What is the activation function in NN?
• How layer information is propagated to module and sensors?
• Where input track information goes into the NN?
• What was the training-validation procedure?

                                   -> We made a big modification following your comments. There was a big change in the poster contents, and we overlooked your points in a hurry.

It is probably impossible to explain all these given the space available. So a link to external documents on the algorithm that explain it in detail would be useful.

• “Decent amount of high purity particle trajectory samples will trigger the  application” -> “(give order of magnitude) high purity particle trajectory samples is required for alignment procedure”

                                   -> The more, the better. With less statistics, we can group sensors together in a bigger unit.

• Summary and Outlook

• \Delta_{XY} is a strange notation. Given tis definition you probably meant for \Delta_{r\phi}, i.e. the other component of \Delta_{r} is not yet included

                                   -> The effect by the magnetic field is supposed to appear in the XY plane. And hence we introduced a simple nomenclature. We can write the _expression_ r \Delta(phi) as you mentioned, but we still need to introduce two coordinates.

• A large net shift in \Delta_{XY} is very strange.

• For a chi2 based straight line fit, if there is a net shift, the fitter would just shift every track by that amount and achieve a better chi2.
• Misalignment and field bending also introduce shift towards opposite directions cross layers, rather than single positive directional shift as observed in the plot

                                    -> We didn't do any alignment but the generous crude trajectory finding. The offset of 650um, or half millimeter or so, will mean the detector alignment at the hardware level is decent. 

I think this part introduce more question on the analysis than benefit and I would recommend NOT to show the residual distribution until it is understood

                                     -> The original plan was to describe the alignmnet procedure, but we had a few issues and what I wanted was, 1. detector perfomred decent and 2. we can start the alignment finally.

Cheers

 

Jin

 

 

______________________________

 

Jin HUANG

 

Physicist, Ph.D.

Brookhaven National Laboratory

Physics Department, Bldg 510 C

Upton, NY 11973-5000

 

Office: 631-344-5898

Cell:   757-604-9946

______________________________

 

From: sPHENIX-MAPS-l <sphenix-maps-l-bounces AT lists.bnl.gov> On Behalf Of Ming Liu via sPHENIX-MAPS-l
Sent: Monday, August 28, 2023 11:23 PM
To: Jaehyun Kim <xhaxha AT yonsei.ac.kr>; Anthony Frawley <afrawley AT fsu.edu>
Cc: Youngil Kwon (ykwon AT yonsei.ac.kr) <ykwon AT yonsei.ac.kr>; SPHENIX MVTX <sphenix-maps-l AT lists.bnl.gov>
Subject: Re: [Sphenix-maps-l] [EXTERNAL] Re: ML alignment poster?

 

Dear Jaehyun,

 

Thank you for the nice poster. I have a few minor comments/suggestions to improve it with the latest updates:

 

1. Abstract:

You need to update the abstract as the current one was written very early this year before the MVTX commissioning.   

Mention that sPHENIX beam commissioning in 2023 was carried out with Au+Au collisions.  

 

2. Cluster Characterization:  -> suggestions:

Usually, a highly energetic charged particle generates a signal upon hitting the MVTX sensor layer, resulting in the activation of a

cluster of neighboring pixels. This behavior is consistent with the findings in the test beam data. Nevertheless, during the initial

stages of commissioning, notably large clusters were detected, which were ascribed to unanticipated high background related

to the beam.

 

-----

Dr. Ming Xiong Liu P-3, MS H846 Physics Division   Office: 505.667.7125 Mobile: 505.412.7396 Los Alamos National Laboratory

 

 

From: Jaehyun Kim <xhaxha AT yonsei.ac.kr>
Date: Friday, August 25, 2023 at 10:19 PM
To: Anthony Frawley <afrawley AT fsu.edu>
Cc: Ming Liu <mliu AT lanl.gov>, Youngil Kwon <ykwon AT yonsei.ac.kr>
Subject: Re: [EXTERNAL] Re: ML alignment poster?

 

Dear Tony,

 

I uploaded the poster and note draft to the invenio.

 

1. Poster

https://sphenix-invenio.sdcc.bnl.gov/me/requests/892191b1-846f-4cf7-9941-0de47fa17dac

 

2. Analysis note

https://sphenix-invenio.sdcc.bnl.gov/me/requests/7567e0f1-ae45-4da1-9ad1-d1703f244830

 

Your comments will be appreciated.

 

Yours,

Jaehyun

 

2023년 8월 24일 (목) 오후 11:07, Anthony Frawley <afrawley AT fsu.edu>님이 작성:

Dear Jaehyun,

 

Thank you for the draft. I will look at it and send comments.

 

Best regards

Tony

 

---

From: Jaehyun Kim <xhaxha AT yonsei.ac.kr>
Sent: Thursday, August 24, 2023 5:40 AM
To: Anthony Frawley <afrawley AT fsu.edu>
Cc: Liu, Ming Xiong <mliu AT lanl.gov>; Youngil Kwon (ykwon AT yonsei.ac.kr) <ykwon AT yonsei.ac.kr>
Subject: Re: [EXTERNAL] Re: ML alignment poster?

 

Dear Tony,

 

Here I attach the version of the poster for QM being worked out currently.

We just made a limited progress to initiate very crude alignment.

Also, we have approval issues to finalize the poster. So, we are pressing hard on the analysis note.

 

The attached poster can change significantly depending on the approval process.

We will do our best to meet the high standard of the collaboration, but it's quite challenging.

 

Your comments will always be welcome.

 

Yours,

Jaehyun

 

2023년 8월 24일 (목) 오전 3:23, Liu, Ming Xiong <mliu AT lanl.gov>님이 작성:

Youngil and Jaehyun,

 

Do you have a draft  poster for people to review?

 

Cheers

Ming

 

Get Outlook for iOS

---

From: mxliu <mxliu AT rcf.rhic.bnl.gov>
Sent: Wednesday, August 23, 2023 11:21:04 AM
To: Anthony Frawley <afrawley AT fsu.edu>
Cc: Liu, Ming Xiong <mliu AT lanl.gov>
Subject: [EXTERNAL] Re: ML alignment poster?

 

Hi Tony,

Let me ping Jaehyun and Youngil about it.
We have a few event displays for preliminary approval.
Ming

On 2023-08-23 07:18, Anthony Frawley wrote:
> Hi Ming,
>
>  Is there a draft of the AI - ML alignment that we could look at?
>
>  Thanks
>  Tony