1 TPG digitization improvements Progress description Rikard Sandström Geneva University Software parallel session 31 Mars -04 CERN

2 Objective 1.Take digitization closer to physical reality 2.Add all information needed for reconstruction 3.Increase speed 4.Create cleaner code

3 1. More realistic digitization Previously –All particles used the same ionization model, including photons... –Digitization put all clusters from one ”hit” (slice of gas) at the same coordinates. –The deposition length used assumed particles parallel to beam line. –Electrons in the ionization clusters were not drifted individually to the readout. –There was a hard cut in number of samplings a signal was allowed to fill.

4 1. More realistic digitization (cont) Now –Only charged particles generate hits, and mu+ are treated explicitly using histogram files. –Clusters are positioned along a track approximated by line segments using the momentum vector of the hit. –The path of each hit last until either the cylinder wall or an adjacent gasslice is hit. –All electrons in every cluster are drifted individually to the GEMs, where they are amplified. –The cut in number of samplings is now adaptive. –Support for dead strips added.

5 1. More realistic digitization (graphic) Hit Cluster Readout Beam dir Electrons drifted from cluster delta e- momentum mu+ e- Charge on strip

6 1. More realistic digitization (Electronics response) 1 The electrons arrive at the strip at different times due to – different vertex position – longitudinal diffusion (drift) 2 The electrons arriving at the strip gives a voltage vs time – U(t) ~ (t/tau) 2 exp(-t/tau)+noise where tau is decay constant (75 ns) 3 Signals belonging to the same sampling have their amplitude summed 4 Digits are checked against threshold and written to output file One strip, U(t) : Samplings (diff 500ns) voltage time threshold

7 2. Interface to Reconstruction –Columns in Sim.out: Slice nr Up/down stream PID Deposited energy [keV] Time of hit [micro sec] Hit position vector Hit momentum vector –Columns in Digits.out: Up/down stream Sampling nr Layer Strip PID Amplitude [V] Drift time [micro sec] Time of hit [micro sec] Cluster position vector Hit momentum vector A number of parameters were added to both Simulation and Digitization:

8 3. Code optimized for speed A helper class, C4DVector, was created to structure the TpgDigits before they were combined and checked against threshold. –This lowered the number of operations from appr 500 M to appr 0.4 M per event. (Hence, what used to take a weekend instead takes 3 min.) –Later improvements reincreased the number of operations by an order of magnitude.

9 4. Cleaner, Object Oriented code TpgHitBank is reads and writes code to output files. TpgEvent creates clusters using hit info and after processes TpgDigits TpgCluster creates TpgDigits (electrons) and drifts them to readout. TpgDigits makes electrons into signal on strips. TpgEvent TpgDigits TpgHitBank TpgCluster If TpgCluster can get a file of its own, TpgDigits should inherit from it instead of from TpgHitBank

10 Snapshot of Digits.out [PID] 10% dead strips 90%He, 10%Methane No cooling No background

11 Snapshot of Digits.out [Volt]

12 Next Steps TpgCluster: –Move TpgCluster to its own file? –Its name is by chance identical to a class in Reconstruction. Pack output file automatically to save diskspace? (1 event = 0.5 Mb) Start using the code in Reconstruction –Simulate both response of MICE TPG and HARP-like TPC With and without RF background HARP gas & He90CO 2 10 Strip spacing (pitch) Field 4T resp 0.7T Tracker length (see Emilio’s talk)