Cosmics dt reco fix #7044

ptraczyk · 2015-01-05T16:44:29Z

Meantimer reconstruction is enabled in the cosmic reco sequences. A small tweak was also done in the MT algo - previously the algorithm interpreted segments with time within 20ns of nominal as "in time" and fit segment position and direction assuming time offset ==0. Now this is configurable and switched off for cosmics - for cosmics the 3 parameter fit with time floating is performed always.

Validation results:
http://cmsdoc.cern.ch/~ptraczyk/cosmicMuons.pdf
http://cmsdoc.cern.ch/~ptraczyk/cosmicMuons1Leg.pdf
http://cmsdoc.cern.ch/~ptraczyk/globalCosmicMuons.pdf
http://cmsdoc.cern.ch/~ptraczyk/globalCosmicMuons1Leg.pdf
(efficiency and resolution improve)

cmsbuild · 2015-01-05T16:44:50Z

A new Pull Request was created by @ptraczyk (Piotr Traczyk) for CMSSW_7_4_X.

Cosmics dt reco fix

It involves the following packages:

RecoLocalMuon/Configuration
RecoLocalMuon/DTSegment

@cmsbuild, @nclopezo, @StoyanStoynev, @slava77 can you please review it and eventually sign? Thanks.
@bellan this is something you requested to watch as well.
You can sign-off by replying to this message having '+1' in the first line of your reply.
You can reject by replying to this message having '-1' in the first line of your reply.
If you are a L2 or a release manager you can ask for tests by saying 'please test' in the first line of a comment.

cmsbuild · 2015-01-06T16:24:27Z

+1
Tested at: 0508216
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-7044/1748/summary.html

cmsbuild · 2015-01-06T17:09:23Z

Comparison is ready
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-7044/1748/summary.html

slava77 · 2015-01-08T04:50:10Z

RecoLocalMuon/DTSegment/python/DTMeantimerPatternReco2DAlgo_LinearDriftFromDB_CosmicData_cfi.py

@@ -7,16 +7,17 @@
 # The linear DriftFromDB algos is used.


technically, these are not module initializer files
but none of it has been enforced by the config file model

https://twiki.cern.ch/twiki/bin/view/CMSPublic/SWGuideAboutPythonConfigFile#Including_Standard_Module_defini

meaning, _cff is a more appropriate suffix

this would apply to nearly all of the config files here, right? We could clean this up at some point, though this would require some care as these files are being then referenced in a few other places ..

right.
It's something to put on the list.
Doesn't have to come with this PR.

slava77 · 2015-01-08T05:24:29Z

Hi Piotr

Was this presented in the muon POG meeting already?

slava77 · 2015-01-08T05:56:04Z

RecoLocalMuon/DTSegment/src/DTSegmentUpdator.cc

@@ -122,16 +125,16 @@ void DTSegmentUpdator::fit(DTRecSegment4D* seg, bool allow3par)  const {
    if(seg->hasZed()) {

      // fit in-time Phi segments with the 2par fit and out-of-time segments with the 3par fit
-      if (fabs(seg->phiSegment()->t0())<40.) {
-        fit(seg->phiSegment(),allow3par,0);


this part "allow3par, block3par" had block3par change from 1 in #3724 (~original version) to 0 in #4018, to now back to 1 here.
Is this just empirical (possibly riding a random variation wave) or is there reason behind each time this being an improvement?

The previous change was an error. I spotted this and reverted to the original version. "block3par" blocks the 3 parameter fit, which is supposed to happen if the segment is "in time". then in line 382 the 2-par fit is executed and it replaces the result of the 3-par fit.
The cutoff it now configurable via "intime_cut" so if we set it to -1 "block3par" has no effect - this is what is done in the cosmic sequences now.

ptraczyk · 2015-01-08T13:58:34Z

The intention was to have intime_cut = -1 uncommented from the very
beginning - that was the reason for the parameter to be added in the first
place, before this cut was hardcoded. I believe all the tests that I linked
earlier were done with it uncommented and the fact that it was not present
in the original PR was just an error on my side, made when I was preparing
the code for the pull request (my previous working branch was based on 72).

Right now I'm trying to double-check this (the comparison plots)

On Thu, Jan 8, 2015 at 2:50 PM, Slava Krutelyov notifications@github.com
wrote:

Hi Piotr,

I'll see what comes out from our RECO comparisons (locally run)
and if the diff is confirmed to be small we can proceed.

Please comment on now uncommented intime_cut = -1 in cosmics,
which came with the last commit.

On 1/8/15 7:34 AM, Piotr Traczyk wrote:

hi Slava,
the block3par change has very little effect, I can try to make some
plots to quentify it, but anyhow the befavior before this PR is
logically incorrect - out-of-time segments are fit with 2-par fit and
in-time segments with 3-par fit.
For collision muons the difference is very slight, for things like
cosmics its more visible - if the muons are in time, allowing time to
float (3-par fit) or fixing it to zero (2-par fit) is almost the same
(the latter should have a tiny advantage for example it a segment is
in-time, but the fit thinks it's off by a few ns).

—
Reply to this email directly or view it on GitHub
#7044 (comment).

Vyacheslav (Slava) Krutelyov
TAMU: Physics Dept Texas A&M MS4242, College Station, TX 77843-4242
CERN: 42-R-027
AIM/Skype: siava16 googleTalk: slava77@gmail.com
(630) 291-5128 Cell (US) +41 76 275 7116 Cell (CERN)

—
Reply to this email directly or view it on GitHub
#7044 (comment).

slava77 · 2015-01-08T22:01:04Z

Are there plans to update the cosmic data reconstruction (cosmic scenario in the matrix)?

slava77 · 2015-01-09T05:50:09Z

+1

for #7044 e159def

based on tests in CMSSW_7_4_X_2015-01-07-1000 (test area sign485)

Some observations

cosmic reco in cosmic scenario is unchanged
most changes (in pp configuration), as expected start from dt segments and then propagate downstream. changes in the cosmic muon sequences are larger.
the "regular" muon reco is visibly affected only in chi2 of the DT segments, the effects at higher level are much smaller

1TeV muon sample

10 GeV muon gun

time distribution for cosmics reco change quite a bit (as I understand, as expected)

in ttbar with PU35@25ns (wf 25202) the regionalCosmicCkfTrackCandidates has some fluctuations affecting mostly CPU cost: this module takes about 5 sec to run on events with muons (almost as slow as all tracking ... not related to this PR, just mostly a reminder that some review and fixing is pending there

muon pt10 also shows visibly better residuals

Again, for pt=10 GeV muons: There seems to be more hits on the DYT muon fits (good, but not quite obvious)

slava77 · 2015-01-09T05:50:32Z

please test
(to get jenkins auto-merge)

cmsbuild · 2015-01-09T05:55:16Z

The tests are being triggered in jenkins.

cmsbuild · 2015-01-09T05:55:19Z

This pull request is fully signed and it will be integrated in one of the next CMSSW_7_4_X IBs unless changes or unless it breaks tests.

cmsbuild · 2015-01-09T07:47:05Z

+1
Tested at: e159def
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-7044/1773/summary.html

cmsbuild · 2015-01-09T07:56:03Z

This pull request is fully signed and it will be integrated in one of the next CMSSW_7_4_X IBs unless changes (tests are also fine). This pull request will be automatically merged.

Cosmics dt reco fix

ptraczyk added 2 commits January 5, 2015 16:31

enable MT reco in cosmic sequence, small fix in the algorithm

10dee6d

clean up and remove NoDrift local reconstruction sequences

0508216

cmsbuild added this to the Next CMSSW_7_4_X milestone Jan 5, 2015

cmsbuild added comparison-pending pending-signatures reconstruction-pending tests-pending labels Jan 5, 2015

cmsbuild added comparison-pending pending-signatures reconstruction-pending tests-approved and removed comparison-pending pending-signatures reconstruction-pending tests-pending labels Jan 6, 2015

cmsbuild added comparison-available pending-signatures reconstruction-pending tests-approved and removed comparison-pending pending-signatures reconstruction-pending tests-approved labels Jan 6, 2015

slava77 reviewed Jan 8, 2015
View reviewed changes

cmsbuild removed comparison-pending pending-signatures reconstruction-pending tests-pending labels Jan 9, 2015

cmsbuild added comparison-pending fully-signed reconstruction-approved tests-pending labels Jan 9, 2015

cmsbuild added comparison-pending fully-signed reconstruction-approved tests-approved and removed comparison-pending fully-signed reconstruction-approved tests-pending labels Jan 9, 2015

cmsbuild added a commit that referenced this pull request Jan 9, 2015

Merge pull request #7044 from ptraczyk/CosmicsDTRecoFix

a0c9b34

Cosmics dt reco fix

cmsbuild merged commit a0c9b34 into cms-sw:CMSSW_7_4_X Jan 9, 2015

ptraczyk mentioned this pull request Jan 15, 2015

DT cosmic reco switch to meantimer #7177

Merged

ptraczyk mentioned this pull request Feb 11, 2015

enable Meantimer reco in cosmic sequence - backport to 73X #7665

Merged

abbiendi mentioned this pull request Mar 11, 2015

Cosmic muon validation #8199

Merged

abbiendi mentioned this pull request May 26, 2015

complete the recoMuon validation workflow for cosmic muons #9279

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Cosmics dt reco fix #7044

Cosmics dt reco fix #7044

ptraczyk commented Jan 5, 2015

cmsbuild commented Jan 5, 2015

cmsbuild commented Jan 6, 2015

cmsbuild commented Jan 6, 2015

slava77 Jan 8, 2015

slava77 Jan 8, 2015

ptraczyk Jan 8, 2015

slava77 Jan 8, 2015

slava77 commented Jan 8, 2015

slava77 Jan 8, 2015

ptraczyk Jan 8, 2015

ptraczyk commented Jan 8, 2015

slava77 commented Jan 8, 2015

slava77 commented Jan 9, 2015

slava77 commented Jan 9, 2015

cmsbuild commented Jan 9, 2015

cmsbuild commented Jan 9, 2015

cmsbuild commented Jan 9, 2015

cmsbuild commented Jan 9, 2015

Cosmics dt reco fix #7044

Cosmics dt reco fix #7044

Conversation

ptraczyk commented Jan 5, 2015

cmsbuild commented Jan 5, 2015

cmsbuild commented Jan 6, 2015

cmsbuild commented Jan 6, 2015

slava77 Jan 8, 2015

Choose a reason for hiding this comment

slava77 Jan 8, 2015

Choose a reason for hiding this comment

ptraczyk Jan 8, 2015

Choose a reason for hiding this comment

slava77 Jan 8, 2015

Choose a reason for hiding this comment

slava77 commented Jan 8, 2015

slava77 Jan 8, 2015

Choose a reason for hiding this comment

ptraczyk Jan 8, 2015

Choose a reason for hiding this comment

ptraczyk commented Jan 8, 2015

slava77 commented Jan 8, 2015

slava77 commented Jan 9, 2015

slava77 commented Jan 9, 2015

cmsbuild commented Jan 9, 2015

cmsbuild commented Jan 9, 2015

cmsbuild commented Jan 9, 2015

cmsbuild commented Jan 9, 2015