Analysis Lessons Learned (d2n)

From Hall A Wiki
Jump to: navigation, search

What follows is a (partial) listing of analysis lessons learned from Transversity, which may be a useful guide for d2n. Much of this is culled from Transversity analysis meeting presentations, accessible through the Transversity ELOG (from experiment webpage). I've given the student's name and the date of the relevant presentation, in italics, for each point. The Hall A Analysis Workshop 2009 is also a useful resource.

Beam Characteristics

  • Calibrate phase difference between raster currents, BPM positions. (Vince, 2009.01.29)
  • Transversity took an ARC measurement on Nov. 17. We rely on the same measurement. (Vince, 2009.01.29)
  • Raster x/y cables found to have been swapped (Jin at Analysis Workshop)
  • Should be able to see RF timing if we have fixed our timewalk properly (Jin, 2009.05.22)
  • What kind of analysis (if any) do we want to do on our HAPPEX runs? Or were these just for charge feedback and not for, say, later quantification of charge asymmetries?
  • Instabilities in BCM u3/d3 (Jin, 2009.06.23, 2009.06.25)
  • Does L1A problem affect helicity readout? (Jin, 2009.10.26)


  • Extended target correction to account for <math>x_{tg} \neq 0</math> (Jin at Analysis Workshop)
    • THaExtTargCor class in analyzer
    • Review of mathematics (Jin, 2009.09.07)
  • Cell measurement summary document posted by Yi in ELOG, 2009.07.24
  • Notes on nominal material thicknesses, radiation lengths posted in ELOG 2009.08.17, 2009.08.19, 2009.08.20. These will have to be double-checked.
  • Energy loss of beam in target must be accounted for; new class, THaBeamElossE06010 (Xin, 2009.09.02)
  • Bar supporting carbon foils bends due to gravity; HALOG entry (Jin, 2009.09.25)
  • Absolute target spin calibration can be done via EPR (3He spin along holding field -> weaker total field (Jin, 2009.09.25)
  • Transversity target diagrams posted in ELOG (Jin, 2009.10.09)
  • Correct for fluctuations in oven temperature (Xin, 2010.01.12)


  • Basic goal: Convert from focal plane coordinates to target coordinates (Miha, Analysis Workshop)
  • Several approaches (Miha, Analysis Workshop):
    • Circular arc approximation: THaOpticsAnalytical, THaVertexTime
    • Effective-midplane approximation: THaOpticsAGen
    • Transport matrix formalism: THaOpticsHRS
  • Momentum reconstruction can get a rough double-check from dE/E ADC signals for PID, knowledge of scintillator physics (Miha, Analysis Workshop)
  • BB Optics strategy discussed in (Xin, 2009.02.03), (Xin, 2009.03.26), (Xin, 2009.05.22)
  • LHRS Optics strategy discussed in (Jin, 2009.09.25)
    • New optimization tool, available at Jin's site
    • Acceptance adjustable by changing variable R; 6 sets of 2d graphical cuts
  • Optics are calibrated with low-momentum data. Can we trust this calibration in a high-momentum regime? (Xin, 2009.03.06)
    • Study of high-momentum validity of optics (Xin, 2009.09.08)
  • Acceptance cuts should be the same for positive and negative polarity (Xin, 2009.04.16)
  • Corrections should be continuous throughout the acceptance (Xin, 2009.03.26)
  • Momentum calibrations must take into account energy loss in material, including e.g. air and target walls (Xin, 2009.05.22)


General LHRS

  • VDC T0 is calibration on which all others rest (Chiranjib, 2009.03.06)
  • L1A sometimes double-pulses and both register as hits (Kalyan, 2009.08.06)
    • Either the first hit or the second one might be the "true" hit
    • Mostly affects ROC3
    • Differentiation: A good event fires one channel, a bad event may fire arbitrary number of channels (Kalyan, 2009.08.13)
    • Bad events can be removed from analysis by flagging them as bad based on channel firings

S1, S2m

  • Check ADC pedestal stability over time (Chiranjib, 2009.03.06)
  • Timing calibration from s2m (Jin, Analysis Workshop, citing Chiranjib)
    • Two-bar-hit events, with tight ADC cut: rough offset alignment, get below 1 ns
    • With tight ADC cut and RF structure, use fits to extract matrix elements for flight path matrix (up to 2nd order corrections in x_{pf}, th_{pf})
    • Using RF structure, get fine bar offsets and time walk corrections
  • Timewalk: separate left side from right, since right side defines the time (Chiranjib, 2009.04.16)
  • Pathlength: Check dependence of individual focal plane variables (Chiranjib, 2009.04.16)
    • Theta is largest correction, so do it first, then X.
    • Has to be corrected for RF structure (Chiranjib, 2009.06.13)

Gas Cerenkov

  • Check stability of SPEs over run period (Chiranjib, 2009.03.06)
  • Pedestals may change over time/configuration (Xin, 2009.08.26)


  • Gas problems can affect timing spectra (Xin, 2009.08.26)
  • L1A problems also affect t0 calibrations (Chiranjib, 2009.09.03)


General BigBite

  • New BigBite class for Transversity experiment: THaBigBiteE06010 (Jin, 2009.03.04)
  • GEANT3 simulation of BigBite can couple to analyzer (Xin, Analysis Workshop)
  • Positive optics may need small correction at bottom of magnet (region was outside Transversity acceptance) (Xin, 2009.04.16)
  • 0.5% momentum increase (normalization factor) needed for d2n as compared to Transversity calibration (Xin, 2009.04.16)
  • Contamination from photons; we don't have coincidence data to check this as Transversity did (Xin, 2009.09.25; Jin, 2009.10.01)
    • Can use events with shower/preshower energy, but no track (Kalyan, 2009.11.20)

BigBite Shower/Preshower

  • Shower sum #15 never reaches T1 trigger (was this fixed during d2n?) (Jin, 2009.03.04)
    • Can tell from lack of self-timing edge
  • We want BB timing to be precise enough to identify beam bunches (Jin, 2009.04.09)
  • Timing calibration done bar-by-bar (in comparison with neighbors) (Jin, 2009.04.09)
  • Check ADC pedestal stability over time (Chiranjib, 2009.03.06)
  • Radiation damage over time manifested as 5-15% gain drop (Chiranjib, Analysis Workshop)
    • Calibrate gain separately for several run periods
    • E/p calibration correction factor based on total accumulated beam charge (Kalyan, 2009.05.22)
    • Does not degrade uniformly, region to region (large-angle degrades degrade less) (Kalyan, 2009.07.16)
    • Details of correction in (Kalyan, 2009.09.25)
  • Two "bad blocks" in shower with low efficiency (in ADC signals) (Kalyan, 2009.09.25)
  • Shower calibration minimizes chi^2 (Kalyan, 2009.04.09)
    • Do uniformity checks on energy resolution across acceptance
  • New shower software
  • Trigger hole documentation in second talk posted under (Xin, 2009.09.17)
    • Xin mentions pulser tests to gauge size of effects, to be undertaken at end of d2n -- was this done?

BigBite MWDC

  • Things to watch out for: HV changes, gas-related problems, wire-chamber trips (cut them out, like beam trips) (Xin, 2009.02.03)
  • Xin found one cable mapping problem during a low-luminosity run -- do we have this fixed? (Xin, Analysis Workshop)
    • Xin had set up the database for our positive data and had incorporated those changes.
  • Wires not always centered; cable lengths not always matched (Xin, Analysis Workshop)
  • At "high" beam current (14 uA), we get lots of false tracks (Xin, Analysis Workshop)
    • Fix: additional cuts (optics, track quality, track-calorimeter match, PID, yield stability)
  • MWDC sensitive to timewalk. T1, T6 times are very different. (Xin, 2009.03.06)
  • MWDC calibration procedure (Xin, Analysis Workshop; details and plots in Xin, 2009.02.03)
    • Chamber position, geometry, and cable mapping
    • Overall drift time calibration ($t_0$)
    • Conversion: Drift time $\rightarrow$ drift distance
    • Overall chamber position
    • t0 calibration: wire by wire
    • Position calibration: wire by wire
    • Iterate the wire-by-wire calibrations...

BigBite Tracking

  • Xin wanted to take a coincidence elastic run to evaluate tracking efficiency at high luminosity. (Xin, 2009.02.03)
  • Don't use scintillators; about 10% of Transversity events fail to fire the scintillators (Jin, 2009.03.04)
  • Calorimeter doesn't help in tracking either; Z position has large uncertainties. Instead, match calorimeter to track found without calorimeter. (Xin, 2009.02.03)
  • An error in trigger formation time leads to an error in electron drift time, which leads to an error in computed drift distance (Jin, 2009.03.04)
  • Tracking inefficiencies (esp. from timing) may disproportionately affect some particle types (e.g. pions) (Jin, 2009.03.04)
  • Additions to tracking software (Xin, Analysis Workshop)
    • New class: SpecialWirePlane for scintillator (not used in production?)
    • New method for converting drift time to drift distance: Pol2FitTTD
    • Road::CollectCoordinates(), Road::Fit() give slightly more tracks in exchange for slightly less speed
  • Tracking chi^2/d.o.f. ( brought near 1 with trigger timing calibrations (Xin, 2009.04.16)
  • Efficiency discussion (Xin, 2009.07.09)

BigBite Cerenkov

  • On 2009.07.14, Yi posted a tech note to the Transversity ELOG about PID using the BigBite Cerenkov

BigBite PID

  • Kalyan explored the possibility of using scintillator plane to improve PID; scintillators do show clear difference between pions and electrons (Kalyan, 2009.10.01)


  • Trigger issues described in (Xin, 2009.02.27):
    • "Pulser goes to signal lines of PMTs"
    • Bad filter: 30% reduction in individual ADCs for some preshower signals, small angle side
    • NIM bin problem: Shower amplifier from 1-23
    • Talk of post-d2n studies to understand effects on threshold, acceptance,...
  • Since T3 arrives at the trigger supervisor before T5, it "steals" T5's timing. (Xin, 2009.03.06)
    • Can be corrected as a deadtime effect.
    • Affects mostly early runs (November 2008 and before)
  • Trigger timing can be checked via scintillator TOF (Jin, 2009.03.04)
  • L1A problems caused difficulties with VDCs, other things (Chiranjib, Analysis Workshop)
  • T2 trigger + Cerenkov rejects 20% more pions than T6 (Joe, 2009.03.06)
  • Shower cut insufficient for PID since pions have such a large energy range in shower (Joe, 2009.03.06)
  • Deadtime asymmetry is an additive effect to physics asymmetry and must be accounted for (Jin, 2009.06.18)
  • When the signal edge is outside retiming window, can get distortions by confusing one trigger signal for another. Need to exclude this "sideband." (Xin, 2009.09.03)
  • Small effect from T1/T6 double-counting (Joe, 2009.09.25)
  • EDTPM (T8) rate changed dramatically during Transversity (Xin, 2009.09.25)
  • Error analysis for pre-scaled triggers: Rob Feuerbach's tech note posted on ELOG (Vince, 2009.09.26)
  • Statistical fluctuation in pre-scaled events: (Yi, 2009.11.17)


  • Some gated scalers seemed to be unstable, or disagreed between LHRS and BigBite copies (Chiranjib, Analysis Workshop)
    • Lots of plots supporting this are shown by Min Huang/Xin Qiang, 2009.07.24
    • Fix: Ungated scalers were used in the analysis, and were gated in software during the skimming process.
    • LHRS scalers seemed to be affected by crosstalk from gas Cerenkov (Kalyan, 2009.06.04)
  • Scaler insertion needs to be synchronized to L1A events so that you can cut data and scaler info for the same time period (Jin, 2009.06.18)
    • DL.*, DBB.* variables automatically synched to L1A, but contain limited info
    • Other options: every 100 events (but asynchronous), every 2 s
  • Can cross-check consistency of scalers that should be related to each other. Here are Transversity's cross-check results (Jin, 2009.07.02)
  • Charge/livetime scaler note: Xin, 2009.09.22


  • Basic cuts: (Jin, 2009.01.29)
    • beam trip
    • trigger type
    • vertex
    • quick PID
  • Corrections to pairs (Jin, 2009.01.29)
    • angular reconstruction
    • polarization (target, beam)
    • dead-time
  • Xin posted a useful note on asymmetry formation, to the ELOG on 2009.07.08
  • Adding "fake spins" and forming asymmetries in witness channels helps understand propagation of false asymmetries (Xiaodong, 2009.07.24)
  • Best to form local asymmetries, within individual runs, before taking them to longer time scales (Xin, 2009.07.24)
    • Some problems (like L1A issues) trend with time, making local asymmetries more valid than longer-term ones
  • Needs a bit of thought to transform raw asymmetries to physics asymmetries ... phase space, bin centering, etc (Xin, 2009.07.24)
  • Compare asymmetries from different triggers, single tracks vs. multi-tracks, etc (Xin, 2009.09.10)
  • False asymmetry in livetime (Xin, 2009.09.13)
  • False asymmetry from differing yields/acceptance in various momentum bins (Xin, 2009.09.17, first talk)
  • To reduce false asymmetries, form asymmetry of many quantities (e.g. mean preshower energy), check for correlations between desired asymmetry and each of these quantity asymmetries (Xin, 2009.09.17, first talk)
  • Cut on track quality (chi^2/ndof) (Joe, 2009.09.25)

Skimmed Root Files

  • Be sure to check replay output for warning messages -- can flag interesting problems (Xin, 2009.06.18)
  • Skimming howto in (Xin, 2009.06.18)
  • Things removed by skimming process (Chiranjib, Analysis Workshop)
    • Beam trips
    • Chamber trips
    • Run periods with high deadtime
    • Run periods with trigger issues
  • Things introduced by skimming process (Chiranjib, Analysis Workshop)
    • An "ss" scaler (see Xin, 2009.08.26)
    • Various kinematic variables computed offline
  • Stages of the skimming process (Chiranjib, Analysis Workshop)
    • Shift scalers. Label beam trips.
    • Shift scalers. Label chamber trips, deadtime/trigger issues, etc.
    • Introduce kinematics variables, R-cut, BigBite acceptance cut.
    • Reduce root files to contain major variables (dramatic decrease in file size).
  • Cross-check results in each skimming pass with those from the previous pass! Details of Transversity's implementation in (Xin, 2009.11.20)

Data Quality Checks

  • Run DB problems
    • Check for missing or doubled end-of-run info in HALOG (Xin, 2009.05.22)
    • Check for overly short runs (Xin, 2009.05.22)
  • Stability check: Find mean/rms of some quantity during run, then plot this value run-by-run to check stability over experiment. (Chiranjib, Analysis Workshop)
    • Gas Cerenkov sum (after calibration, main peak was fitted with a Gaussian)
    • Aerogel A1 sum (after calibration, main peak was fitted with a Gaussian)
    • Pion peak in lead glass counters (after calibration, main peak was fitted with a Gaussian)
    • VDCs for T3, T5 after offset calibrations
    • Left S2m TDCs (right not considered, since self-timed)
    • Scalers: gated and ungated
    • PID (quantities of electrons, pions, etc) (Xin, 2009.05.22)
    • Yield (Xiaodong, 2009.07.24)
    • Reconstructed vertex, momentum (Xin, 2009.08.13)
  • Do variable drifts (e.g. in yield) coincide with different target polarizations? (Xin, 2009.09.25)


  • SIMC code for Transversity (posted on ELOG by Kalyan, 2009.06.24), documented by Kalyan as follows:
    • SIMC code can be found here... /w/work5602/transversity/montecarlo/simc_poltar I suggest we put all the relevant Monte Carlo code in the "montecarlo" folder, so that interested person can copy them.
    • There were some recent changes to fortran compiler on JLab machines, it will affect SIMC code. The details are given by Dave Gaskell here .. /w/work5602/transversity/montecarlo/simc_poltar/
    • You can find how to run the simulation at /w/work5602/transversity/montecarlo/simc_poltar/README
  • GEANT3 code for BigBite (Xin, 2009.11.20)


  • Watch out for difference between abs, fabs in cuts (Joe, 2009.11.20)