[Halld-offline] event-number-dependent calibration

[David Lawrence]

Hi Mark et al.,

    One major constraint we have is in the JANA framework itself. It was 
designed with a brun (for "begin run") callback that automatically gets 
called when the run number changes. It was done this way on the 
assumption that the run number would be the primary index of the 
calibration system. If we did have the ability in the database to index 
on a finer level than run number (presumably event number), then the 
framework would need to be modified to detect that condition in the data 
stream so the appropriate callbacks could be called.

    I appreciate your point though. If we are starting from scratch on 
the calibration/conditions DB design and it's fairly easy to design in 
this feature, we should probably do it. The fact that the framework 
doesn't really support it now doesn't mean we shouldn't include it as a 
risk mitigator. We (mainly Dimitri) just need to think carefully about 
how to include it as a more or less dormant, out of the way feature as 
you suggest.

    In regards to Curtis' question on how long a single run should be, 
IMHO it should be on the order of an hour. This is driven by the human 
time scale which comes into play not only in run stopping/starting, but 
in logistics of offline analysis. It is also a reasonable measure for 
the size of the chunk of data you're willing to throw away if a problem 
is detected in it. The cost in $ of acquiring the data tends to be a 
linear function of time, not  of terrabytes. Just my $0.02.

Regards,
-David

Curtis A. Meyer wrote:
> In Crystal Barrel, I think that our calibration constants were tagged by
> valid time periods. This of course meant checking the validity more often
> than by run, but did allow them to change more often. I am pretty sure 
> that
> we never had them change that often though.
>
> The real question comes down to how long is a single run? If they are 
> an hour
> or less, then it might not be worth the trouble to build a system to 
> handle this.
>
>      curtis
>
>
> On 6/2/10 3:06 PM, Mark M. Ito wrote:
>> David,
>>
>> I had a thought: I remember in G7 they had some kind of problem with the
>> tagger(?) where they had to apply a different correction for different
>> sections of a single run. This had to be done with some ad-hoc system of
>> course. We might think about this since we are starting from scratch: is
>> there some way to apply calibrations constants that vary within a run. I
>> think you would want to have this as a non-default feature, something
>> that only exists if needed.
>>
>> Just wanted to add that to the feature wish list. Something to think
>> about at least.
>>
>>     -- Mark
>>
>> _______________________________________________
>> Halld-offline mailing list
>> Halld-offline at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/halld-offline
>>
>>    
>
>

-- 

------------------------------------------------------------------------
 David Lawrence Ph.D.
 Staff Scientist                 Office: (757)269-5567   [[[  [   [ [       
 Jefferson Lab                   Pager:  (757)584-5567   [  [ [ [ [ [   
 http://www.jlab.org/~davidl     davidl at jlab.org         [[[  [[ [[ [[[
------------------------------------------------------------------------