Home » Blogs » perev's blog

HFT hit selection

Updated on Tue, 2015-04-07 11:32. Originally created by perev on 2015-03-02 22:30.

The last puzzle in Sti was HFT hits selections. We requested at least 4 HFT hits on track
but sometimes tracks was accepted with even 1 Hft hit. I spend a lot of time to understand,
with no success. Then I rewrite the code with the same logic but more carefully and clean.

In addition I slightly changed the logic, namely:
Old: Fit Kalman outside to in; From the upper node attempt to fit to outside; Refit.
Now:Fit Kalman outside to in;  Refit; From the upper node attempt to fit to outside; Refit.
The difference, after refit, the accuracy of upper node is better and probability to find the proper hit
in the way to outside is better.
Thi case (outside fit) is very rare but it is better to make it good

In result, problem with hits was solved. But Sti is working now slightly different.

Definitions:
Globals:                        tracks with >   10 hits
Good globals : global    tracks with >= 15 hits
GoodPrimaries: primary tracks with >= 15 hits
Good hits(gits) : hits from good tracks

Stastics to test was ruther small, only 22 events.
It should be checked by our standard QA


Improvements:

  1. Correct selection of Hft hits;
  2. Globals  +3.1%
  3. GoodGlobals +3.3%
  4. Using of TPC hits  +2.7%
  5. Using of HFT hits increased (+10%)
  6. Tracks with HFT hits increased (+10%)
  7. DcaXY & DcaZ improved (8%)
Degradations:
  1. Number of primaries decreased (-2.3%)
  2. Number of GOOD primaries decreased(-0.6%)
  3. Efficiency slightly less 40%>39%
Efficiency defined as TracksWithHftHit/TracksThruHft*100
May be decreasing of primaries is  related to better accuracy ???
In average looks like better


The related code is  committed now. But production was not affected..


More detailed numbers and pictures below.

 
  Current Fixed Comments
 
lobals        1470  1515  +3.1%
Good globals        1055  1090  +3.3%
 Primaries          567  554  -2.3%
 Good Primaries          470  467  -0.6%
 Used Tpc hits       46.8%  48.0%  +2.7%%
 Used Ist hits        30.%  33.7%  +11%%
 Used Pxl hits        3.8%   4.2%  +10.5%%
 Good track Tpc hits       26.6  26.5  - 0.4%%
 Good track Ist hits       0.2  0.23  +13%
 Good track Pxl hits       0.42  0.46  +9.5%
 tracks thru HFT      12086  12918  +832
 DcaY      0.67  0.51  +7%
 DcaZ      0.55  0.50  +9.1%
Efficiensy      40%  39%  -2.5%% 
Current version
===============

nFiles =1 nEvents = 22 <nEvents>=22
nGlob =      32344 <nGlob> =  1470 nGlobMax= 3364
nGgoo =      23211 <nGgoo> =  1055 nGgooMax= 2207 Ggoo/Gtot=0.718
nPrim =      12479 <nPrim> =   567 nPrimMax= 1231 Ptot/Gtot=0.386
nPgoo =      10342 <nPgoo> =   470 nPgooMax= 1003 Pgoo/Ggoo=0.446

Tpc Hits =   1500405 used    702194(46.80%) perTrack 21.710
Pxl Hits =    264215 used      9948( 3.77%) perTrack 0.308
Ist Hits =     16377 used      4900(29.92%) perTrack 0.151

Tpc Gits =   1500405 used    617958(41.19%) perTrack 26.623
Pxl Gits =    264215 used      9686( 3.67%) perTrack 0.417
Ist Gits =     16377 used      4775(29.16%) perTrack 0.206
Fixed version
=============
nFiles =1 nEvents = 22 <nEvents>=22
nGlob =      33332 <nGlob> =  1515 nGlobMax= 3649
nGgoo =      23987 <nGgoo> =  1090 nGgooMax= 2376 Ggoo/Gtot=0.720
nPrim =      12208 <nPrim> =   554 nPrimMax= 1216 Ptot/Gtot=0.366
nPgoo =      10289 <nPgoo> =   467 nPgooMax= 1046 Pgoo/Ggoo=0.429

Tpc Hits =   1500405 used    719448(47.95%) perTrack 21.584
Pxl Hits =    264215 used     11209( 4.24%) perTrack 0.336
Ist Hits =     16377 used      5519(33.70%) perTrack 0.166

Tpc Gits =   1500405 used    634718(42.30%) perTrack 26.461
Pxl Gits =    264215 used     10961( 4.15%) perTrack 0.457
Ist Gits =     16377 used      5398(32.96%) perTrack 0.225
====================================================================


Histogram of CURRENT Sti tracks                   
xxx - tracks missing HFT
000 - tracks thru hft with 0 hft hits
1234- Pxl1=1 Pxl2=2 Ist=3 Sst=4
efficiency HftTracks(hits>0)/HftTracks(all)*100 = 40%

Histogram of Fixed Sti tracks
  DcaY current version  DcaY FIXED version
 
DcaZ current version DcaZ FIXED version
b
»