Meeting Agenda:
1) roundtable: updates/progress?
-> redux EEMC geometry, decoder, tests - all
* fyi: latest geometry/validation
http://drupal.star.bnl.gov/STAR/blog/jwebb/2009/sep/21/validation-eemc-mc-geometry
* fyi:SMD orientation reconciliation
http://www.star.bnl.gov/HyperNews-star/get/emc2/3318.html
* how to get on with MC tests?
- additional info/work needed (e.g., next)
* decoder issues
http://www.star.bnl.gov/HyperNews-star/protected/get/starsimu/391.html
* tests/docs needed
-> other status/update
2) AOB
-----------------------------------------------------------------------------
Minutes
Q = question
A = answer
C = comment
Try to identify speaker as I am able.
Filtering
Will Q: News on filtering?
Michael A: Waiting for Victor - not until after collaboration meeting.
EEMC Geometry
Will Q: Any updates since yesterday?
Jason A: Looked at some other layers and no obvious problems.
Will Q: Is number of volumes stable?
Jason A: Think so. Yes.
Will Q: What is polystyrene on both sides of tower megatiles?
Hal A: Was there in original geometry.
Jim, etc. C: Might be trying to mockup other material in wrapper. Aluminum, Kevlar, Tyvek faces
forward. Front plastic 0.15 mm back 0.55 mm.
Jason A: Delay until virtual Monte Carlo? Change would require 2 additional volumes because now
tile is stuck in a large volume of polystyrene.
Jim C: Delaying sounds fine.
Scott Q: Does everyone agree SMD correct?
Will A: Yes.
Jim A: Agree.
Will C: Please double check.
Will Q: How do we move on with testing?
Jason A: Will send email to Victor. Same problem with having to wait until after collaboration
meeting.
Ilya C: Might be able to fix decoder. Email sent to starsimu.
Jason C: I have not looked over code in great detail.
Will C: Maybe ask TOF folks.
Jason C: I think simulations group has control of this.
Jim C: Frank worked on TOF.
Hal C: Also HFT.
Will Q: What tests do we want?
Will A: Have to look at everything.
Michael C: You can do tests now with Geant.
Ilya C: Would have to write new code.
Jason C: I have code to get sampling fraction straight from Geant.
Will C: We care about SMD. Not worth extra work.
AOB
Will Q: Alice what about gamma+jet code?
Alice A: Code from 3 people. Will look at next week.
Minutes from meeting 13 October 2009
Q=Question
A=Answer
C=Comment
-----------------------------------------------------------------------------------------------------------------------------------------
Updates/Progress
Misc.:
Alice and Will C: Someone with karma/more familar with gamma maker should look into TMinuit library issue here: http://www.star.bnl.gov/HyperNews-star/protected/get/phana/388.html
Will C: No Michael,updates on filtering?
Alice Q: Ilya, were you able to run with full geometry yet?
Ilya A: No, adding "east endcap" back in only allows for EEMC only running.
EEMC Geometry Tests
Old:
Discussed yesterday at EEMC meeting
New:
Will and Hal C: Need to label "front" and "back" differently
Ilya C: Exclusion is done by not generating material. Materials added one at a time.
Will Q: What is there? Is it air? What?
Will Q: We compare upper left and lower right?
Ilya A: Yes
Hal Q: UVS refers to particle hitting first U, then V, then S from interaction point?
Ilya A: Yes
Hal C: Seems the wrong way. U plane in front - less hits?
Will C: A megatile layer is directly in front of shower max.
Ilya C: Shower is not so developed so less hits in first plane.
Will C: What bothers me is that 12 and 2 are not the same - same U/V cadence. Or is this because you just don't generate material and we don't know what is taking its place? Many agree.
Ilya C: v-Plane first sectors seem to agree better.
Will C: Should maybe try without LOW_EM option.
Scott C: For pions shower in front layer is narrower than one in back. j-all plot looks fine. This is what we really care about.
Ilya, Scott C: In geometry cvs there is no spacer, sectors 11 and 12 were the same? Or was there air in the spacer?
Hal A: In cvs air was in spacer location.
Jim C: The scale difference for u and v plots makes it hard to interpret. A 1-D plot would be helpful.
Hal C: Can Ilya send me geometry file with how you did these cases?
All: Energy plots looks good
Plans:
Will Q: Is there separation in towers?
Hal A: "wiggle" Option is there, not implemented. But isolation gap in eta is there. Not sure about phi.
Will Q: Do we need more documentation? Jason added lots of documentation in the form of comments. What about variables - i.e. LOW_EM option variables?
Will Q: What about TWIST/EGGS?
Hal A: EGGS is similar to Geant3. Worry about hadrons not doing a good job - what about FLUKA (sp?)? Seem to remember lots of parameters that dictate how particles are "tracked."
Jim C: Most backgrounds for gamma/jet are EM anyway.
Everyone: We seem stuck until full geometry is working.
Hal C: Someone should contact Jerome directly and see what we can do.
Alice C: Victor just replied to you.
Minutes 20 October 2009
Q=Question
A=Answer
C=Comment
-----------------------------------------------------------------------------
Sampling Fraction with New Geometry
Ilya C: Rechecked many things- including labels, re-ran cvs version.
Hal Q: Didn't you do this earlier? Where some parts where brighter?
(Link:
http://drupal.star.bnl.gov/STAR/pwg/spin/ilya-selyuzhenkov/photon-jets/2009/10-oct/20091012-jaso
n-eemc-geometry-position-correlations)
Ilya A: Was EEMC only and number of hits.
Scott Q: Is z-scale the same on all?
Ilya A: Yes
Jim Q: Which geometry year?
Ilya A: 2006. Double checked all 2006 LOW_EM option.
Jim Q: Can you run with 2009 geometry?
Ilya A: Yes, how would it help?
Jim A: Something to check with a change.
Ilya C: Strange SMD is consistent, but not towers.
Jim Q: Why is 2nd preshower so different from 1st?
Ilya A: y-axis scale is different (see 10^-3) in left plot
Scott Q: Why any difference at all in preshower? Wasn't all difference put in the SMD layer?
Jason A: Some minor changes: changes to steel structure, changed depth of polystyrene in
scintillator
Will Q: Any change in thickness of polystyrene on either side of tiles?
Jason A: No
Jason C: GPrint TMED. Print tracking parameters at beginning of run and at end of run to see if
LOW_EM option is actually getting picked up. Also check endcap materials - something strange
outside may be affecting something inside. Will send in email.
Jim Q: Bright spots on borders and dark centers of tiles?
Ilya A: yes
Scott Q: What is threshold in "towers above threshold" plot?
Ilya A: 3 sigma
Jim Q: How to print out all volumes?
Jason A: Is a command, maybe not useful (TMI).
Jim C: Take stand-alone geometry and change front plate to lead. Then we would know if material
in front increases sampling fraction or not. (Should not increase) Also like to see color
wheels for preshower.
Scott C: Try to figure out at what depth we see the effect.
Jason C: With my current code this is hard to do in Geant.
Hal C: 5 sections: 2 preshower, posthower and 2 towers. 2 tower sections are added together. In
EEmcUtil area.
Will C: Not shown, but postshower isn't crazy looking.
Hal C: When you calculate energy for sampling fraction how many towers do you use?
Ilya A: This is all towers above threshold. Figure 5 is just the highest tower.
Jim Q: No chance full geometry has defined in it an active volume that shouldn't be there?
Jason A: No.
Jason C: Figure 5 seems to be telling us something opposite figure 2. Maybe our intuition about
fig 2 is off.
Hal C: Yes, but difference between eemc stand-alone and full is not the same in figure 5 either.
Scott c: Maybe we need to look at MPV and not the mean
Jason Q: How do you determine thrown position? NO cuts on SMD response?
Ilya A: Geant, no.
Hal Q: Are you losing event from TPC ribs?
Ilya A: Off-scale.
Scott c: Between thrown energy of 20 to 30 make a sampling fraction distribution so we can see
the shape.
Hal Q: Did you change the tower geometry at all?
Jason A: Just added a feature to allow wiggle, but not implemented.
Jim Q: Gap in towers?
Jason A: Isolation groove is there in eta.
Hal C: Figures in addition to s.f. for 1X1 tower - 2x2, etc.
Ilya C: 3x3 similar to figure 2, 2x1 different from both.
Will Q: What about Jan's comment about kONLY="MANY"
Jason A: Yes I use it, you must when volumes overlap. GEANT says don't because computationally
expensive
Minutes 27 October 2009
Q=Question
A=Answer
C=Comment
---------------------------------------------------------------------------------------------
(1)Filtering
Will Q: Status of filtering?
Michael A: Victor is still working on filtering. Problem with TPC in 2009.
Michael C: I cannot test the code because I don't have access to it. Victor is working on 2009
geometry fixes.
Jim C: I thought Victor had to approve some code.
All: Much confusion about what the situation is.
There are 2 filters - one for pythia and one for bfc. Victor is providing the hook to put
your filter in bfc. That is what we are waiting for.
(2)EEMC Geometry
Jason C: LOW_EM option does not get executed in stand-alone
Will C: Some of the effects we've seen are probably turning on and off LOW_EM and not
stand-alone versus full geometry
Alice Q: So you aren't bothered by difference in sampling fraction anymore?
Ilya A: No we just understand some things.
Ilya C: There might be other differences between stand-alone and full.
Will C: Maybe we should move away from stand-alone geometry
Jason C: I changed EM cuts in preshower layers so they were consistent with ones in tower stack.
Ilya Q: Changes in materials before preshower layer?
Jason A: No changes in materials
Ilya C: More low energy in preshower.
Jason C: This is different cuts being applied in the preshower. Cuts are lowered with respect
to cvs. Maybe did not finish this task.
Jim C: Not sure that "all" really looks like the rest added together.
Will C: Biggest difference appears to be in 1X1. Energy shared among towers differently.
Scott and others C: It doesn't make sense that red goes up and black goes down when look at
total energy vs. thrown energy. Why do changes only show up when we put stuff in front of the
detector? Even if you say black is right and red is wrong, it's still puzzling.
Ilya Q: DetP geom y2006g has no effect in standalone?
Jason A: Yes
Many: So standalone will always be the same no matter what year and options.
Will C: Comparing to standalone not worthwhile. Things to look at: LOW_EM option, tile configuration.
Jason C: Postshower looks fine. This means postshower hasn't gotten moved outside poletip. Also check FILL ESEC - try setting each deltaz to 0. That puts scintillators in same position as cvs code.
Jim C: We need to compare to data.
Will Q: What option does standalone pick for endcap?
Jason A: The default - which is full endcap.
Scott Q: Any changes that would effect outside of endcap?
Jason A: None that I can think of. Should be more than 2mm of clearance in front of endcap - i.e. overall length of endcap changed.
Jason will send email to list about cuts and might have time to document other changes.
3 November 2009
Very brief meeting, everyone still working on understanding geometry.
(1)Filtering
Michael c:Victor asked Michael for refinements to code. Michael is working on it.
(2)Geometry
Will C: Seems to be lead. For now not focusing on standalone but on full STAR geometry but with new layers turned off. Need to keep this in mind. Why does this change + LOW_EM option produce such a big change?
Ilya and Alice C: Trying some experimental coding on ELED block just to rule things out.
Ilya C: Changes in sampling fraction with modified ELED block and LOW_EM on or off seems significant.
Scott C: I think there is a problem with the way the block is defined.
Will C: We need to settle on some numbers - particle cuts, Low_em, sampling fraction
Minutes 17 November 2009
Q=Question
A=Answer
C=Comment
(1)Geometry
Jason C:y2006dev geometry is defined. QA needs be done before a production release is made.
After this is closed making significant revisions will be more difficult.
Ilya Q: What is ETA for 2006?
Jason A: In auto-build now, will be available tomorrow.
Ilya C: Now pure lead is consistent with lead alloy. Distributions in all layers look
consistent as well. Is change with LOW_EM option reasonable?
Hal C: What about extra time for LOW_EM?
Michael A: It's a factor of 6 for very high energy photons, but less than time to run bfc.
Jason C: We've talked about being OK with a factor of 2, but not 10. 2<6<10 so who knows.
Michael C: New TPX simulator takes like 10 minutes per event.
Scott C: We really care about backgrounds when it comes to time because we need so many more
statistics and don't think as many high energy particles there.
Will C: Would be a good idea to re-run all tests that were done last week when we thought things
were final last Friday.
Hal and Alice C: Think it's a good idea to look at a small-ish generic QCD sample with the idea
that if we find problems they can be fixed before we are locked in forever. Alice will try to
come up with a proposal.
Michael C: For 2009 geometry, many fixes to the barrel make it not backwards compatible. So
2006 geometry will not be ideal for barrel.
Jason C: I can look into making this backwards compatible. I'm also planning on propagating
EEMC changes all the way back to 2003.
(2)Filtering
Michael C:There has been progress but cluster has been down last few days so a bit stalled.
24 November 2009
Q = Question
A = Answer
C = Comment
(0)Filtering
Michael C:No progress
(1)Code
Alice C: Still not getting smd strips for endcap
Michael, Ilya C: Looks in log file for warnings or errors from StGammaRawEvent
(2)Geometry
Will C: 40% increase of running with LOW_EM option. Seems to improve data - MC agreement,
especially for shower max.
Ilya C: 2009 may appear to agree better than 2006 with 2006 data might be due to cuts on eta
mesons not placed on the Monte Carlo
Jim C: It looks like maybe the LOW_EM option is over-correcting
All C: We need a sample with and without LOW_EM option. Some discussion if you can run with
LOW_EM option and then undo it in analysis, and generally decided not. Still planning on coming
up with a request for a test sample - but many techinical difficulties now with StGammaMaker and
filtering, etc.
(3)Fast Simulator sampling fraction
Will C: Don't think this was measured with the test beam. It wouldn't be in written form
anywhere if we did, not same geometry.
Jim C: If we account for material in front of detector then it will change from year to year.
Hal C: I would be in favor or leaving it at 5%.
All C: The scale factor might then be 1.02. If we want more precise a very detailed study will
be required. This is something the test run can tell us as well.
Jason C: LOW_EM option makes a lot more sense from a physical prespective. The cuts without it
are too high.
8 December 2009
Q=Question
A=Answer
C=Comment
(1)Filtering
Ilya C: Victor sent a note saying things were working to some list, somewhere.
Michael C: We now have a bfc framework. AT the moment it's just the barrel filter.
Will Q: Can you explain tuning more?
Michael A: You have to run enough simulation to show you aren't biasing yourself for whatever
threshold you select.
Things to consider:
--Trigger? => Will A: Not below 5
--Software analysis cuts?
Ilya Q: Is there documentation on how to use this?
Michael A: No, Victor will be doing that.
(2)Alice's post
I didn't take notes while talking. General conclusion is to repeat, double-checking the trigger
simulation. Some debate about if just looking at gamma candidates is the best thing to do.
(3)Hal's post
Hal C: I will continue to run some tests in the background.
Ilya Q: x-axis is generated energy/pT?
Hal A: Yes
(4)LOW_EM option part 100
Will C: Barrel already has 100 kEV limits buried in geometry.
Michael Q: How do you calculate the time?
Ilya A: Time between starting and ending starsim.
Michael C: But big overhead in loading to memory. For barell get 9 sec without LOW_EM and 34
sec with LOW_EM.
Hal Q: Why is this an issue? Aren't we doing physics better this way?
Will A: We need lots of MC - 2006, 2009 200 GeV and 500 GeV.
Jim A: We may be "over-simulating" and wasting CPU, especially in endcap.
Hal C: I think we should put lowest cuts on materials right before the scintillator - this is
not what is in the geometry. We should fix the geometry.
Michael C: From barrel studies seems we need low cuts in lead and plastic.
Hal C: In endcap, then, we need in cladding. For first run, do LOW_EM option and then refine.
Michael C: For barrel, the global LOW_EM cut doesn't add much time.
Matt Q: How many radiation lengths upstream of endcap?
Will/Hal A: 1.5-2.
Matt C: We can't really use different options because they you get different sampling fractions.
(5)Other
Will C: I think there actually should be less material in TPC now.
Jim C: Agree, seems to be less just looking at it.
15 December 2009
Short meeting with Scott, Will, Jim and Alice in attendance. Mostly touched on the to-do list. Waiting for geometry from Jason and Ilya (both at BNL now). Alice can work on filter next week - not on vacation until Dec 24. News from BNL is no collisions still - maybe this weekend.
Meeting 2 February 2010
Meeting was attended by Alice B, Mike B and Ilya S. We discussed the current status of the filter. A brief summary of action items:
(1)No complaints so far to raising gamma maker thresholds to correspond to 2009 l2 trigger thresholds, which will probably remain the same for some time. These new thresholds are a seed threshold of 4.2 and a cluster threshold of 5.5 Email sent to list about this here. Currently the pythia filter thresholds are set 4 sigma lower than this and bfc filter thresholds are set 2 sigma lower than this. These will changed accordingly. FYI: definition of sigma from EEMC NIM paper is sigma(E)/E = 16%/sqrt(E)+2%. Note: the current gamma maker thresholds are 3.6 and 5.1, from the 2006 trigger configuration.
(2)We discussed current performace of pythia filter. Suspect some increased rejection may come from tighening eta boundaries in the filter along with radius of the cluster cone in eta-phi space. Mike also suggested adding a vertex cut for the particle track vertices. This kind of cut currently exists in the bfc filter. We think these changes need to tested, along with the changes in (1) before we move forward. Ilya will check what the vertex cut in his analysis is.
(3)Small scale tests of the bfc filter with the 2 sigma thresholds are promising and not showing false rejections.
(4)A (too) small scale test of the thresholds in (1) showed no acceptance in the 2-3 GeV partonic pT bin. Will run more events to verify it is excluded. Concerned that we will probably not be able to eliminate the 3-4 GeV bin. Alice would not like to set the gamma maker cluster et threshold above 6. Agreed, and that maybe even that is too high.
(5)We discussed the issue of the gain uncertainty in the filter. The concern is that a filter which is "too tight" will exclude events that we will need for testing the gain uncertainty in a final analysis. There are multiple concerns here :
i. How to actually do this. Jason sent a suggestions that would involve modifying the fundtion StEEmcSlowMaker::setTowerGainSpread(float s) to include a variable MEAN, which is currently hard-coded to 1. The new code would look like:
void StEEmcSlowMaker::setTowerGainSpread(Float_t s, Float_t MEAN)
{
// initialize tower gain factors to 1
for ( Int_t sec=0;sec<kEEmcNumSectors;sec++ )
for ( Int_t sub=0;sub<kEEmcNumSubSectors;sub++ )
for ( Int_t eta=0;eta<kEEmcNumEtas;eta++ )
{
// mTowerGainFact[sec][sub][eta]=1.0;
Float_t f = -1.0E9;
while ( f <= -1. || f >= 1.0 )
f = gRandom->Gaus(0., s);
mTowerGainFact[sec][sub][eta] = MEAN + f;//this was 1.0+f
}
}
this requires a change to the code, but it is small and MEAN could be set by default to 1.0 Then we would re-process the gamma trees with the slow simulator set with MEAN = 1-1sigma uncertainty and MEAN=1+1sigma uncertainty and test the rejection of the filter. Ilya would like to see this code checked into cvs, which would require someone with karma - probably Oleksandr.
ii. What number to use for the uncertainty. It seems that 5% is widely accepted as correct. Please object if you think this is wrong. Please object if you think this method is not sufficient to test the filter with uncertain gains.
Meeting 11 March 2010
Simulation
--------------------
Will: One of last things on list was timestamps. He can try to generate some.
Beam - have positions, but no rms widths. Set to zero.
2-3 bin - don't generate for now
Michael: Ask Gene about timestamps
Alice: I will update page
Task Force and Projection
-------------------------
In general, most discussion focused on estimate for background uncertainty. We decided that using the
sideband argument is OK if we state that this what we are doing and that any additional uncertainty due to
using the sideband under the signal is ignore.
Alice will try to send draft tomorrow now that there is a plot.