To document concepts and understanding of what can be done with the
STAR TPC for possible high statistics datasets at RHIC beginning in 2021.

I. DAQ

1) What are the limiting factors for the TPC DAQ rate?

2) To what degree can the TPC operate in a streaming-data mode?

   a) What would be the expected impact on data volumes?

   b) What are the requirements of such a mode for offline reconstruction and embedding?

II. Anodes & Multi-wire chamber

1) In the absence of destructive testing in-situ at STAR, what can we understand from operation of other experiments?

2) What are the possibilities for test chambers?

3) What are the possibilities for, and implications of lowering voltage on the anodes?

III. Gating Grid

1) What are the capabilities/limitations of the existing gating grid (and incoming new driver)?

2) What is the status of alternatives to the existing gating grid, and what are the practical/logistical factors to their implementation in STAR?

IV. Gases

1) What are the advantages and/or disadvantages to using different gases in the STAR TPC?

V. iTPC

1) What are the impacts of (not) having iTPC? 

I. DAQ   

Quoting a section from Tonko's email to the TPC hypernews:

2) The current electronics, as-is, has a number of limits which keeps the data rate around (say...) 2 kHz of triggered events. (But, as I mentioned in my little streaming Proposal, each triggered even contains a number of min-bias events which we might want to use.)   

We have a number of options, IMHO:
  a) ~3 kHz we can get if we reshuffle the FEEs, RDOs in a judicious manner, use the spares to increase the data parallelisation, redo the firmware, get more DAQ PCs etc.  Simpler parts of this I plan to do for FY16 already, BTW.
  b) for 5-6 kHz we could think of a mixed mode where we upgrade the electronics in the inner sector only but then use the old electronics (freed up by the iTPC upgrade) to parallelise the outer sectors even more (e.g. readout the outer sector with 7-8 current-RDOs instead of 4)  Note that by using the limited streaming trick this already represents 25% of streaming (50 μs X 5 kHz) so, personally, I would shoot for this option, in some form.
  c) for more than that we must replace all the electronics and do some sort of real streaming mode. And this means significant money & effort and I don't believe that STAR can do this (private opinion but based upon the large costs & effort of the e.g. ALICE O2 Upgrade). Option b) above at 5 kHz might already be beyond our means.
  d) if we don't have an iTPC Upgrade we could also reach 5-6 kHz by only designing new RDO boards but using the old FEEs. The main bottleneck of the current DAQ1000 system is in the RDO and the optical links and DAQ PCs.

3) At some point (say around 3-4-5 kHz) the data rate into RCF might become an issue and we need to see how to solve it.
 

Tonko's proposal is to achieve high event acquisition rates via trigger rates similar to what exist currently at STAR, but using a large window of acquisiton such that multiple other minbias events occur and are recorded during the window. The window needs to be ~40 μs longer for the TPC than other detectors to be sure that the last event of interest during the window has time for the event to complete its drift. Tonko has proposed something like an 85 μs window of acquisiton for the TPC.

Tonko's intent is that the DAQ reader encapsulates the complications of this scheme for the offline software, presenting individual "events" to the offline software just as now, and parsing the full DAQ events for the individual "events" internally. Ignoring other systems, from a TPC perspective only, this would simplify the transition to this scheme. However, for offline software as a whole, there would need to be significant testing/QA all the way through physics analyses to ensure that subtleties like event skipping & filtering are handled properly for all purposes (e.g. calibrations, physics productions, embedding). Estimates on the effort involved are perhaps ~half an FTE-year for Tonko to implement, plus additional shorter term commitments from numerous other people involved in the testing/QA stages.

A more involved but also more efficient approach would be to do normal tracking over this meta-event and only later "move" the track origins relative to the collision time.
In this way one can assume that the tracking time (aka CPU effort) will approximatively double (since there will be ~2x more tracks in the 860 bunch crossing than in the usual ~400).
However, since we will have a larger number of min-bias events (say 10) this would represent an overall reduction in Offline processing time by ~5X compared to
the simple & quick naive mode above. Albeit at the expense of more effort in the reconstruction software.
 

II. Anodes & Multi-wire chamber

Link to Richard's Talk on TPC Aging (from 2009 TPC Review)

Link to Yuri's Talk on TPC Aging (from 2009 TPC Review)

Here is the best estimate of accumulated charge on the TPC anode wires (from Yuri's message here).
 

In Tonko's proposal, the charge/current on the anode wires at high luminosities scales with the length of the acquisition window (at low luminosities, the triggered event becomes the dominant contributor, while at high luminosities the non-triggered events dominate). 

III. Gating Grid

The electronics of the Gating Grid (GG) drivers is not an issue if we replace the old drivers with something newer. And this is actually in progress and we do have prototypes (designed by Tim Camarda) in the works. The plan is to test them in the real TPC in the next run.  The cost is low (~40-50 $k) for the full replacement so this is not something we should dwell on and we might consider this done.  The only thing that we need is a maximum speed that we should test against since this has ramifications on our choice of power supplies. 5-10 kHz is trivial.

The existing GG was tested for limitations regarding ion backflow (IBF) into the TPC, which would cause additional distortions of the data. The data suggests two operating regions where IBF is not significant:

1. If the GG is open for less than ~150 μs and closed for more than ~300 μs in each cycle, then no ions can fully traverse the region of influence of the GG wires during the brief time that is open (lest they by reined in by the GG wires during the next closing). This is determined solely by the drift patterns around the GG wires.
2. If the GG is open for less than ~1.4 ms and closed for more than ~2.0 ms in each cycle, then all ions produced at the anodes will arrive at the GG during a time when it is closed. This is determined solely by the range of times it takes to drift from the anode to a GG wire.

Historical operation of the TPC has kept it in region 1 (though not strictly, as the closed period is not guaranteed to be that long, the limitation on average GG opening rate near ~2 kHz appears to typically keep the short closures infrequent enough to be negligible). Operating in region 2 may be a concern for high luminosity running where anodes may not be able to handle the high charge/currents (it may, however, be a valid option for the low luminosity BES II running).

It is currently believed that Region 1 could be exploited to run with a significantly longer opening time than the standard TPC drift. That would allow operation as proposed by Tonko with triggering on rare events, and looking for additional minbias events within an extended DAQ window following the trigger.

IV. Gases

(Info from Jim and Alexei)

V. iTPC  

1. The gridleak (GL) fix that is planned for new inner sectors (more details by Yuri here) will not be done; thus at higher luminosities (if higher luminosities are run) distortions will be bigger.
2. Likewise the iTPC will allow us to lower the anode voltages for the same MIP due to modified pad response(larger pad size), reducing GL.
3. w/o iTPC the option we are either stuck in the slow mode (Tonko's option a) or we need to design and manufacture new RDO boards (Tonko's option d)