sPHENIX tracking discussion

[Thomas K Hemmick <hemmick AT skipper.physics.sunysb.edu>]

Indeed, this is a coupled question with the event rate (higher than trigger rate).

On Fri, Aug 18, 2017 at 3:28 PM, Spencer Klein <srklein AT lbl.gov> wrote:

Tony, all:

If the TPC gas is slowed down, then the problem of overlapping events from different beam crossings will grow.  This is a more tractable problem than that of too many overlapping clusters from a single ion-ion collision, but it should not be neglected, and it is important to do the optimization globally, and not just for a single event.

Spencer

On 8/18/17 11:34, Anthony Frawley wrote:

Hi All,

Here are some notes on the very useful tracking meeting this morning.

There were two topics of discussion:

1) After the Director's Review, while looking into how to make the TPC simulation more realistic, we realized that the present simulation uses too small values for the shaping and ADC bin times for the TPC readout. Cluster occupancies are much worse than we thought, and are too high in central collisions with the present readout speed and gas drift velocity. We discussed what to do about this.

2) We need (and have long planned) to move to a realistic simulation of the TPC that is capable of predicting the performance of the detector. At present, some of the expected performance parameters are essentially an input to the simulation.

Summary of the discussion:

1) The existing simulation has been modified to correctly represent the effects of occupancy. G4 simulations of 0-4 fm Au+Au collisions (very central) now show cluster occupancy for the innermost and outermost TPC layers of roughly 30% and 15% respectively (these are still slightly optimistic, but correct within 10% or so). This results in track reconstruction efficiency in the 60% range in central Hijing events. The performance in low occupancy events is essentially unchanged, because the TPC Z resolution is not very critical.

The cluster occupancy is due to the r-phi cluster width and Z cluster length at the readout pads. The r-phi cluster width is set by physics. The Z cluster length is set by the shaping time of the amplifier in the SAMPA chip. The binning in Z is set by the ADC clock rate. At a gas drift speed of 6 cm/microsecond, the present shaping time produces a cluster length of 3.1 cm in Z, and the ADC clock rate (9.4 MHz) produces a Z bin width of 0.64 cm.

There are two ways to reduce the cluster length in Z.

a) Reduce the SAMPA chip shaping time by a factor of two (and concurrently increase the ADC clock rate to make the Z bins smaller). This is possible but the former will have some cost (see Tom's slides). This would reduce the cluster length by about a factor of 2 (see my slides).

b) Reduce the gas drift speed. This is Tom's suggestion, and he thinks it may be possible to do it without big effects on ion feedback or performance. This could bring an additional factor of 2 reduction in cluster Z length.

Both of these things together would bring the cluster occupancy back to the levels we had in our (incorrect) simulation, effectively eliminating it as an issue.

Action items:

-----------------

Tony will make a pull request for the updates to the existing simulation code that get realistic occupancy.

The TPC experts will pursue the option of reducing the SAMPA chip shaping time.

The TPC experts will look into slowing down the gas.

2) We had presentations from Carlos and Martin related to implementing a more realistic TPC simulation that would predict performance.

Carlos and Veronica have been working on a modular framework that will separate the various stages of the TPC simulation. Martin is thinking about how to implement in the simulations the results of test bench data for the zigzag readout pads.

After some discussion we came to the following picture, in which we move away from the idea of electron "clouds" that represent only the average behavior and instead try to build fluctuations into the simulation:

a) G4 will step the track through the TPC gas volume, and tell us the energy loss in each step.

b) The energy loss will be converted to electrons by sampling the appropriate statistical distribution.

c) Each electron will be drifted to the face of the GEM stack, with appropriate path distortion and randomization.

d) The position of the drifted electron at the face of the GEM stack will be used to generate charge on the zigzag pads through a response function prepared from test bench data.

e) The charge on each pad will be built up as successive electrons are added.

In this way, the simulation will use only response functions derived from measurements or from detailed simulations, and there will be realistic fluctuations in electron production and transport.

Action items:

-----------------

Carlos will make a pull request for the new framework, which will exist in parallel with the old one until we are ready to switch completely.

Work will start on the production and transport of electrons to the GEM stack.

Martin will work on how to incorporate the GEM stack / readout pad response into the simulation.

If I got anything wrong please correct me.

Cheers

Tony

_______________________________________________
sPHENIX-tracking-l mailing list
sPHENIX-tracking-l AT lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/sphenix-tracking-l

_______________________________________________
sPHENIX-tracking-l mailing list
sPHENIX-tracking-l AT lists.bnl.gov
https://lists.bnl.gov/mailman/listinfo/sphenix-tracking-l