[Frost] The current status after 14th iteration of DC calibration

[Franz Klein]

Sungkyun and Volker,
we all appreciate the efforts Sungkyun put into the DC calibration. He 
got good results earlier this month after copying SL5 and SL6 parameters 
and t_max - much better residua than before!!! There was only a small 
run range ~62800, which had comparatively large mean values.
I think we should keep the good values of the 13th iteration.

The other observation appears to be more serious to me: I know that 
there are differences in the simulation and reconstruction results 
between 32-bit and 64-bit versions (gcc4.1) - but in my g13 and g9a 
studies it was of order ~0.1%. I haven't studied the details for gcc4.4 
yet (Redhat6) but found already several problems with processing 
conditions. If you use jlab machines, I will compile and check dc3 as 
soon as the computer center installed tk-devel-8.4 on jlabl1 (I sent 
them a request).

Sorry that we (Eugene and me) did not respond to Sungkyun's mail from 
last Friday ... too much troubles with g14 startup ...

Happy Thanksgiving
Franz

On 11/23/2011 10:45 AM, Volker Crede wrote:
> Hi all,
>
> Sungkyun has been trying to calibrate the drift chambers for g9b for quite a while, but unfortunately we have not seen any significant improvements recently (and it is really time to finish this). He has a lot of experience with this, so we doubt that there is anything wrong with the general procedure. The data themselves also seem to be okay; Sung initially started off with the calibration of those runs with run numbers above 62850 and that went very smoothly, very similar to g9a. We are having difficulties now with the earlier run numbers (that we left out initially) and the fine tuning of the rest.
>
> After a discussion I had with Sung yesterday, our suspicion is that things have become worse when we we started using the 64-bit version of dc3 (some of the calibration software is still based on FORTRAN). A recent calibration improvement of the earlier run numbers involved simply using the constants from the later runs but no actual calibration. In the most recent step (see Sungkyun's email below), he actually applied another real iteration... and things got worse again for no obvious reasons.
>
> Has anybody else seen strange things during the transition from 32 to 64 bits? Franz, is there still a way to use a 32-bit version of dc3 for some tests?
>
> Best wishes
>
>        Volker
>
>
> On Nov 18, 2011, at 5:15 PM, Sungkyun Park wrote:
>
>> Hi Eugene and Franz
>>
>> I have finished another iteration (14th) of DC Calibration.
>> The result plots are updated in the following web :
>>
>>    http://hadron.physics.fsu.edu/~skpark/DC_Cal_g9b/DC_Cal_nov1811.html
>>
>> In this iteration, I only calibrated runs whose means are not located between -50 and 50.
>> After finishing them, means of super-layer 5 and 6 became worst.
>> For example, in the ran range between 62321 and 62400, mean of super-layer 6 is shifted down too much.
>> I do not understand this situation.
>> Basically, I have used the same way as g9a. That process is very simple.
>> Do you have any idea about this situation?
>>
>> Sung
>> Florida State University
>>
>>
>>
>>
>> Florida State University
>> _______________________________________________
>> Frost mailing list
>> Frost at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/frost
>
>

-- 
===============================================================
                   Franz J. Klein, Associate Professor
                   CUA, Department of Physics
                   Washington, DC 20064
   office: Hannan Hall 206          phone: 202-319-6190
   or: Jefferson Lab,CC F-243       phone: 757-269-6672
---------------------------------------------------------------