1 / 9

Solutions for MDT readout Staturation

Solutions for MDT readout Staturation. Tiesheng Dai Univ. of Michigan ATLAS Muon Week, Feb. 17-21, 2014. MDT Readout Saturation. In EI, readout saturation tube rate ~68KHz, L1 ~70KHz AMT hit buffer overflow increased dramatically seen from online monitors

lona
Download Presentation

Solutions for MDT readout Staturation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Solutions for MDT readout Staturation Tiesheng Dai Univ. of Michigan ATLAS Muon Week, Feb. 17-21, 2014

  2. MDT Readout Saturation • In EI, readout saturation tube rate ~68KHz, L1 ~70KHz • AMT hit buffer overflow increased dramatically seen from online monitors • Produced hit buffer overflow at Lab., where AMT hit buffer overflow occurs after its readout buffer overflow, and need long time to clear situation. • Linked Lab. observation with online, most likely that observed MDT readout saturation also associated with AMT readout buffer overflow

  3. AMT Buffers To CSM 450ns per 32 bits When AMT readout FIFO got nearly full, AMT event builder need much longer time to build an event  L1 trigger buffered in trigger FIFO  hit buffer has to store much more date which results in hit buffer overflow and readout saturation Depth 64 Depth 8 256 depth hit

  4. Solution • Reduce event data size by shorted readout window to ~1.3µs without risk on lost real muon data: gain ~40%  tube rate ~91KHz at L1 ~70KHz • Clear AMT buffers when its readout buffer nearly full with two assumptions: a) AMT readout buffer does not occur two often; b) buffer clear does not hurt too much on real muon data. The test at Lab. by MPI group and AMT simulation by chip designer (Yasuo), it will almost eliminate the AMT hit buffer overflow. But this must be validated by cosmic to see no unexpected effect and collision beam with high luminosity to ensure tolerable read muon hit lost. Expected real data rate will be below red line and above black curve, i.e., fraction of hit lost to be measured with collision data

  5. Shorten Readout Window • Done by configure AMT chip via MDT configuration DB, no foreseen risk by possible impact/works for MDT DQA, GNAM(?), new T0s: • Tune (25ns step) all chamber T0 in region of 50 to 100ns via • L1 delay in CSM : only deal with ~1200 CSMs but maximize MDT L1 latency • (Default) AMT : minimize MDT L1 latency but have to deal with more than 200K AMT chips • All MDT chamber T0s on hand, except new installed EE side A (use corresponding side C EE T0), BOE and BME • Shorten AMT readout window via its configuration • Plan to put as a MDT configuration  easy to select for use • Need to collaborate with Muon calibration group to see how to deal with the configuration switching

  6. Enable Buffer Clear Done by configure AMT chip via MDT configuration DB, expecting of real muon hit lost, inefficiencies to be measured with beam. Nearly full reject mode (current used): (enable_rofull_reject =1. enable_l1full_reject = 1 and/or enable_trfull_reject = 1 ) In this mode the trigger matching will be blocked if the read-out FIFO becomes full and either the L1 buffer is nearly full (256 - 64 = 191 words occupied ) or the trigger FIFO is nearly full (contains 4 triggers). If this occurs event data will be rejected to prevent the L1 buffer and the trigger FIFO to overflow. The event header and event trailer data will never be rejected as this would mean the loss of event synchronization in the DAQ system. Any event which have lost data in this way will be marked with an error flag. Reject data (intend to use): (enable_rofull_reject =1, enable_l1full_reject = 0, enable_trfull_reject = 0 ): As soon as the read-out FIFO full, event data (not event headers and trailers) will be rejected. Any loss of data will be signaled with an error flag.

  7. AMT Chip Simulation from Its Manual

  8. Backup Solution : Pair Mode • Done by configure AMT chip via MDT configuration DB • Data volume is reduced by factor 2 (1 hit = 1 32 bit word, Edge mode 1 hit = 2 32 bit words) • AMT readout window could be further shorted to ~1µs • Gain is more than factor 2 • Need special operation mode to be agreed with ATLAS, detail see Tiesheng’s talk: https://indico.cern.ch/event/301203/contribution/8/material/slides/0.pptx • Preferred ATLAS TTC Operation (one option) • Disable L1 trigger 250µs • Broadcast GlobalReset (EBCRST) • Broadcast ECReset • Broadcast BCReset • Wait for (ATLAS L1 latency + 2µs) then enable L1 trigger Most likely pair mode have to be used in case luminosity above 2X1034

  9. Expected Multiple MDT Configurations Default (current in use) Edge 1.3us Edge 1.3us BufferClear Pair 1.3us Pair 1.0us Pair 1.0us BufferClear It will be good to do MDT configuration selection from MDT control at P1

More Related