Status of dE/dx Calibration Yuri Fisyak October 17, 2002
Mar 17, 2016
Status of dE/dx Calibration
Yuri FisyakOctober 17, 2002
Outlook
• Why ?• What ?• Where ?• Conclusions
Why ?
• Last calibration was done in March, 2002:– σ(dE/dx)/(dE/dx) = 8.2% for 76 cm track– STAR CDR (p. 4C-33)
• σ(dE/dx)/(dE/dx) = 0.47 N-0.46(Ph)-0.32, P = 1atm• σInner = 14.3%, h = 1.15 cm, N = 12;• σOuter = 7.7%, h = 1.95 cm, N = 32;• σ = 6.8% for 76 cm track in TPC
– H.Bichsel simulation: • σ = 7.3 % for 11 * 1.2 cm + 31 * 2.0 cm
– I am not happy that ~1% is missing
Why (cont.) ?• Comparison new and old tcl• H.Bichsel’s calculations reproduce data only
qualitatively. Is this due to calibration procedure ?• H.Bichsel claims that we have non linearity in dE
measurement. Can we check this ?• Can dE/dx calibration be done in one pass? Can we
move it into fast online ?
What does calibration include?• Applying pad correction obtained from pulser data (Fabrice did this and I will not talk about)• For good clusters (used in fit, no overlaps) • For good global tracks (No. fit points ≥ 30, Track length in TPC > 40 cm)• Z =log[(dE/dx)measured/(dE/dx)predicted for π]
•Fit Z-distribution with Gauss(μ,σ) + pol3 in +/-3σ range. It is supposed that we have ~80% π and μ corresponds to π.
•μ should not depend on
•Time, Pressure
•Sector, row
•Drift distance
•…
What does prediction mean ?
0.45 GeV/c
•Hans Bichsel made presentation about his model during the last collaboration meeting. He is preparing STAR TN on this subject.
•PAI model (Photo Absorption model for Ionization energy loss, V.Grishin V.Ermilova,S.Kotelnikov, NIM A309(1991) 476) gives the same predictions but PAI simulation program (provided by P.Nevski) has problem at low β (it was not tuned for this range). For this reason I will use Bichsel’s calculations.
9.6% @ 76 cm
Resolution before calibration
What does March calibration mean?
• March calibration included:– The same procedure as for Year 1 data– As prediction it was used Sirrf– Calibration was done for all tracks (no restriction on
momentum)– Time dependence : overall gain correction factor each
few hours (1-4)– Sector and pad row correction– Drift distance correction– Result : σ = 9.6 % → σ = 8.2 %
Red – 40% truncation
Green – 30% truncation
Black - fit
8.2% @ 76 cm
9.7%
9.9% =>9.8% after setting convolution flag
New tcl a little bit worse than old one
New calibration• Calibration is based on tracks with 0.4 < p < 0.5 GeV/c
(~MIP for pions: βγ = p/m = 4).• Calibration has been done for new tcl only.• As prediction it was used Bichsel’s calculation with dx
dependence (see next slide) i.e. Z =log[(dE/dx)measured/(dE/dx)predicted for π(βγ,dx)]
• This calibration gives σ = 8.8% (instead of 8.2% obtained in March because it was done only for 0.4<p<0.5GeV/c but resolution is obtained for all momenta).
Calibration done for p in [0.4,0.5] GeV/c with new tcl
σ = 8.8% @76 cm
Check of ADC nonlinearityFor uniquely identified tracks:• σ < 15% and• v = log[(dEdx)/(dE/dx)J],
where J = [e,π,K,p,d]• |v| < 3σ for only J, and • |(dE/dx)J - (dE/dx)k| >5σ, for
J≠KPlot predicted dE versus
measured dEInner
Outer
It is seen clean nonlinearity besides saturation (with offset ~1keV).
It is not clear what is origin for this nonlinearity:
•ADC, it has to be checked with pulser data
•tcl ? due to threshold effects?
Time dependenceGain variation versus time. ~5%
Pressure variation versus time
Oxygen concentration variation versus time
Time dependence (cont.)
• Drift distance correction: e+AdO
where A = 1.75e-6 (1/cm/ppmO2),
provided by A.Lebedev (ALEPH data) d – drift distance (cm), O – concentration of Oxygen (ppm)
Pressure
•Both Drift distance and Pressure correction allow to remove time dependence.
•TPC Gain Monitor cannot be used (at least now) because its results depend on magnetic field.
< 1 %
After pad level correction from pulser calibration it still exists gain variation versus sector and row
Dependence of gain correction versus drift distance (after accounting absorption) for Inner rows has opposite behavior with respect to expected. ADC nonlinearity is overcorrected !?
7.4% @ 76 cm
For both + and -
For positive only
Negative only
Bichsel shapes
Inner
fit by φ(μ+(1+σ)z), where φ(z) is Bichsel shape and z = log(dE/dEmost probable);
Outer
Both inner and outer rows are reasonably well described by Bichsel shape.
Conclusions
• In sense of dE/dx old and new tcl give comparable results.• It is observed significant nonlinearity in dE with respect to H.Bichsel
calculation.• Accounting this nonlinearity allows to improve resolution by ~1%
(8.2%=>7.4%) which is matched with our expectation.• To understand the nonlinearity is necessary to get pulser data with
different signal amplitudes.• If the linearity will be understood than it will be possible to move
dE/dx calibration in fast online.• With new calibration the Bichsel shape of dE/dx distribution
describes data pretty well. Thus we can claim that model calculations gives quantitative agreement with data.