[Primexd] [EXTERNAL] Re: Question

Thu May 14 18:30:33 EDT 2020

Hi all,

Let me elaborate a bit on this matter. The pi0 calibration of a cell measures the gain at the average energy of photons from pi0s hitting this cell. This energy can be easily measured. The non-linearity correction applied later should not change the gain at this particular energy. Selecting a sub-sample of the p0s one should be able to change the average energy in this particular cell and measure the pi0 mass for the new average energy.

If the observed non-linearity is caused by the threshold effect, the correction rather belongs to the cluster than to the central cell. One can probably evaluate the function. Let us consider a very simple case: there is only one adjacent cell and it receives an energy fraction 0:E0*a, flatly distributed. The threshold is E_t. Then, on average the cluster will loose an energy of E_t/2*E_t/(E0*a), or a fraction of the full energy of E_t/2*(E_t/(a*E0))**2. The measurable energy will be E0*(1-b/E0**2), where b is a factor depending on the threshold, on the transverse shower profile etc. In our case it will be a correction to the gain measured at a certain energy E_g. So, the observed energy will be
E0*(1-b/E0**2)/(1-b/E_g**2). One can use different functions f(E0), keeping in mind that after the pi0 calibration, for every channel the observed energy will be E0*f(E0)/f(E_g).

If I were doing this, I would first do the pi0 calibration ignoring the non-linearity, then I would try to find a generic function with a variable parameter (or two) for the non-linearity correction, applied to the whole clusters, and fit the parameters selecting sub-samples of pi0s for each central cell. Hopefully, the parameters will be more or less the same. One may repeat the pi0 calibration with a cluster correction: E=Sum(A_i*g_i)*f(Sum(A_i*g_i)), where f may depend on the particular central channel. It may produce more narrow pi0 peaks. Hopefully, no correction will be needed afterwords. Or, one may do as Igal - do pi0 calibration once and with no correction, and apply a correction f(E)/f(E_g) for the clusters for the later analysis. In the first case the correction will be applied to the clusters always. In the second - only after the pi0 calibration.

Eugene

________________________________
From: Primexd <primexd-bounces at jlab.org> on behalf of Shepherd, Matthew <mashephe at indiana.edu>
Sent: Thursday, May 14, 2020 1:26 PM
To: Igal Jaegle <ijaegle at jlab.org>
Cc: primexd at jlab.org <primexd at jlab.org>; Ashot Gasparian <agaspari at ncat.edu>
Subject: [Primexd] [EXTERNAL] Re: Question

Hi all,

I would argue that you should take an approach that is driven by the need to solve problems.  I think this is what Sasha was advocating for.  As you know manpower is very limited.

Regarding this specific discussion, we attempt to use the method Ilya sketches, but as Igal points out, the energy scales and variations are larger.  Certainly by putting a fixed threshold you can simulate the cut better, but one needs to be sure the distribution of the variable being cut is also properly simulated.  We've attempted to do that, but then stopped short of putting the cut in the argument that it is bad to throw away information.

Sometimes you must throw away information to ensure reproducibility in simulation -- this is the essence of fiducial cuts.

We can debate back and forth the pros and cons of all approaches, but it is kind of an academic discussion.  If I were in your position, I would try to look for biggest problems or systematic sensitivities in the simulation.  As Sasha said, this can be done by looking at variations of extracted Compton under a variety of different choices in the analysis.  Then when a signficant effect is observed try to drill down all the way to the root cause, maybe as far as things like Ashot was suggesting to try to control/calibrate gains of PMTs based on LED system.

I think if you don't pursue a pragmatic strategy like this, there is risk of consuming significant manpower to create an alternative simulation approach that, while completely justified and may be similar to that used in other experiments, does not result in a significant reduction of systematic error.  And then you'll have to spend time fixing the real systematics limitations when they arise.

That's just my two cents -- which may not even be worth two cents!

Cheers and I enjoyed the opportunity to talk about this at length yesterday afternoon.

Matt

On May 14, 2020, at 1:05 PM, Igal Jaegle <ijaegle at jlab.org<mailto:ijaegle at jlab.org>> wrote:

Thank you Ilya, for the clarification as the threshold in FCAL is 25 times higher with a 5 MeV RMS from channel to channel, should we chose the same approach i.e. set the threshold at the lowest possible values or try to uniform it for most channels? What do you think?

tks ig.
________________________________
From: Ilya Larin <ilarin at jlab.org<mailto:ilarin at jlab.org>>
Sent: Thursday, May 14, 2020 12:56 PM
To: Igal Jaegle <ijaegle at jlab.org<mailto:ijaegle at jlab.org>>; Ashot Gasparian <agaspari at ncat.edu<mailto:agaspari at ncat.edu>>; primexd at jlab.org<mailto:primexd at jlab.org> <primexd at jlab.org<mailto:primexd at jlab.org>>; Gan, Liping <ganl at uncw.edu<mailto:ganl at uncw.edu>>
Subject: RE: Question

that was read from the table,
it was about 3-5 counts and a count was approximately 1 MeV.
Ilya

________________________________
От: Igal Jaegle <ijaegle at jlab.org<mailto:ijaegle at jlab.org>>
Отправлено: 14 мая 2020 г. 12:54
Кому: Ilya Larin <ilarin at jlab.org<mailto:ilarin at jlab.org>>; Ashot Gasparian <agaspari at ncat.edu<mailto:agaspari at ncat.edu>>; primexd at jlab.org<mailto:primexd at jlab.org> <primexd at jlab.org<mailto:primexd at jlab.org>>; Gan, Liping <ganl at uncw.edu<mailto:ganl at uncw.edu>>
Тема: Re: Question

Thank you, Ilya, What was this threshold in MeV and by how much it was varying from channel to channel?

tks ig.
________________________________
From: Ilya Larin <ilarin at jlab.org<mailto:ilarin at jlab.org>>
Sent: Thursday, May 14, 2020 12:50 PM
To: Igal Jaegle <ijaegle at jlab.org<mailto:ijaegle at jlab.org>>; Ashot Gasparian <agaspari at ncat.edu<mailto:agaspari at ncat.edu>>; primexd at jlab.org<mailto:primexd at jlab.org> <primexd at jlab.org<mailto:primexd at jlab.org>>; Gan, Liping <ganl at uncw.edu<mailto:ganl at uncw.edu>>
Subject: RE: Question

Hi Igal,
in the PrimEx simulations, we just add electronics noise first,
then cut the channel signal at our sparcification threshold:
whatever is below - set to zero.

 Ilya

________________________________
От: Igal Jaegle <ijaegle at jlab.org<mailto:ijaegle at jlab.org>>
Отправлено: 14 мая 2020 г. 12:46
Кому: Ashot Gasparian <agaspari at ncat.edu<mailto:agaspari at ncat.edu>>; primexd at jlab.org<mailto:primexd at jlab.org> <primexd at jlab.org<mailto:primexd at jlab.org>>; Ilya Larin <ilarin at jlab.org<mailto:ilarin at jlab.org>>; Gan, Liping <ganl at uncw.edu<mailto:ganl at uncw.edu>>
Тема: Question

Liping, Ilya, Ashot,

Currently, in the MC simulation, the energy threshold for a detector to be included in a cluster is calculated from the data gains ie it is varying from detector to detector and is run-dependent. To minimize the systematics should we leave the situation as it is or think of a different approach. I am thinking (naively) that it might be better to set this threshold to value at least 20 to 30% to what it is actually and possibly to set the threshold in a way that is the same for most detectors, and all runs both in data and simulation. What are your inputs in this or what did you do in previous PRIMEX measurements?

tks ig.
_______________________________________________
Primexd mailing list
Primexd at jlab.org<mailto:Primexd at jlab.org>
https://mailman.jlab.org/mailman/listinfo/primexd

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/primexd/attachments/20200514/9d7874ff/attachment-0002.html>