[SBS_sim] g15 replay results

[Ole Hansen]

Hi Bogdan,

FYI, I re-ran the analysis with Lubomir's exact cuts, and there is no
significant change in the behavior of either ghost track fraction or
tracking efficiency.

One correction: My original numbers used a kinematics window of 20x2
cm^2, not 20x0.2 cm^2. Lubomir had 18x0.2 cm^2.  I have adjusted my
window to this size now. I am also now assuming that the area refers to
the last tracker plane, not the first. Either way, the differences are
small.

For consistency, I also have to reduce the "good reconstructed track"
cut from 0.6x0.6 cm^2 to 0.6x0.2 cm^2 and from the first plane to the
last. This causes slightly smaller tracking efficiency (e.g. 78->75 % at
0% background) since the y-resolution has a small tail beyond 2 mm.

Stay tuned for more results tomorrow (snow day!)

Regards,
Ole

On 12/15/2010 08:06 PM, Ole Hansen wrote:
> On 12/15/2010 07:59 PM, Bogdan Wojtsekhowski wrote:
>> Hi Ole,
>>
>> I would like to check that you have these selection parameters
>> which Lubomir used (attached). 
> My cuts are slightly different (20x0.2 cm^2 instead of 18x0.2 cm^2 and
> 10x10 mrad^2 instead of 30x0.7 mrad^2). These cuts only affect the ghost
> track fraction.  I'll adjust the cuts and recalculate the numbers. I
> suspect they will be comparable, at least I didn't find the angle cut to
> make much difference.
>
>> More see in his report 1/22/2010.
>> I would like also to remind recommendation to made simple
>> event display. You can not miss hits at "0%" background.
>>
> Well, we can miss hits, as you can see, but we clearly should not.  It
> is obvious that there is some kind of bug.
>
> I am presently working on a simple event display. It's the tool of
> choice for a problem like this, no question.
>
> Ole
>
>>> Hi folks,
>>>
>>> attached are two plots with the tracking efficiency and ghost track
>>> fraction of Evaristo's latest g15 replay. The replay was done with
>>> analyzing only the first 6 planes and so should be directly comparable to
>>> last week's results. Interesting points:
>>>
>>> 1) The tracking efficiency with 0% background is only ~78%.  This gives us
>>> a good handle on chasing an obvious bug in the reconstruction.
>>>
>>> 2) I get slightly higher tracking efficiencies than I got with g9-g14 for
>>> all background strengths except 100%.  For example, for 10% I get 72% vs.
>>> 65% before.  This is odd since nothing should have changed.
>>>
>>> 3) The 100% background run seems to have only ~3810 events.  We could use
>>> a little more statistics. Recently, I've been replaying 5000 events for
>>> each setting.
>>>
>>> Results as a function of number of planes will follow shortly.
>>>
>>> Regards,
>>> Ole
>>>
>>> PS: To reproduce these results, please check out the current software as
>>> follows:
>>>
>>> setenv CVS_RSH ssh
>>> cvs -d :ext:cvs.jlab.org:/group/halla/analysis/cvs co TreeSearch-GEM
>>>
>>> PPS: If you'd like to get regular updates on the SBS tracking simulations,
>>> please subscribe to this mailing list. You can do so via a web interface
>>> at JLab (see the signature for the link).
>>>
>>> --------------------------------------------
>>> Parameters (for reference/expert use):
>>>
>>> number replayed = 5000
>>>
>>> search_depth = 12
>>> 3d_ampcorr_maxmiss = 4
>>> 3d_ampcorr_nsigma = 0.18
>>> 3d_chi2_conflevel = 1e-6
>>> 3d_maxmiss = 4
>>> maxmiss  = 3
>>> maxpat   = 1000
>>> maxslope = 0.1
>>> chi2_conflevel = 1e-6
>>> xp.res       = 7e-5
>>> maxclustsiz  = 4
>>> adc.min      = 200
>>> split.frac   = 0.1
>>> maxhits      = 1000
>>> maxsamp      = 3
>>> adc.sigma    = 0.2
>>>
>>>
>>> _______________________________________________
>>> SBS_sim mailing list
>>> SBS_sim at jlab.org
>>> https://mailman.jlab.org/mailman/listinfo/sbs_sim
>>>
> _______________________________________________
> SBS_sim mailing list
> SBS_sim at jlab.org
> https://mailman.jlab.org/mailman/listinfo/sbs_sim