Difference between revisions of "CountingDaq 220218"

From Moller Wiki
Jump to: navigation, search
(→‎Agenda)
(→‎Minutes)
Line 22: Line 22:
  
 
==  Minutes ==
 
==  Minutes ==
−
Recording on Zoom:   
+
Recording on Zoom: [https://jlab-org.zoomgov.com/rec/share/nTeNXwpFUZfSplLR3R5743bYLqb30eSlJdW_tom_HY8miZaeQQ3NPlUdNKyhNwq8.AT6jEMq4gXd-ZMe1?startTime=1645203309000]  
  
−
Participants:
+
Participants: B. Moffit, P. King, D. Armstrong, C. Ghosh, H. Liu, M. Gericke, D. McNulty
 +
 
 +
*  Discussing data rates
 +
**  The links from the MPDs to the VTP are 1Gbps links, so ~5kHz with 6 samples per event should be fine
 +
**  The VTP optical link is 10Gbps Ethernet, so we could handle ~40% occupancy at 5kHz in the GEMs, if we had one VTP
 +
**  When we do zero suppression, the passed data has all six time samples for the channels that are above threshold.  The zero suppression is done in the VTP
 +
**  Chandan comments that his simulation using the beam generator in an open sector at 100nA has a rate of about 10MHz (Mollers and elastics); my estimates were based on a Moller rate of 30MHz, so occupancies might be lower than I estimated.
 +
**  David points out that most of our running would not be at the very highest luminosity.  Also we can prescale our triggers down to keep the data rate under control.
 +
***  We should not need to use the multiple event builders like SBS is going to need to do in the next running period.
 +
**  David also comments that SBS is getting higher occupancies than they were expecting; this is coming from xrays making single events in individual layers.  David has a large beam-generator simulation file from Chandan that should include xrays, so he is planning to try to look at this in simulation soon
 +
**  We'll need to prescale our triggers; as Chandan had said above, the rate per sector is ~10MHz at 100nA, and the maximum event rate we can take is ~5kHz.  So we'll need prescales of order 2000.
 +
**  Taking about the FADC window
 +
***  I had estimated data rates based on 38 4-ns samples (152ns).  Bryan says that that due to data packaging/transfer issues we might be as efficient using a 256ns window (64 samples).
 +
****  It's best to use 2^m samples in the FADC to take advantage of the bandwidth.
 +
***  With raw mode at 5kHz (and using 38 samples per event), I'd gotten ~850Mbps for the crate with 14 FADC modules; so we should plan on using the VTP readout for the FADCs
 +
***  Also note that we could take most of the events just in pulse parameter mode, and only get the full raw mode periodically.
 +
**  Hanjie asks if the network from the hall to CHA was upgraded to 10Gbps for SBS; Bryan says yes, and they have been able to saturate that link as a test.
 +
***  Probably we can get both the GEM readout and the FADC readout within one 10Gbps link
 +
*  Last meeting Michael had stated the ring five signals had a 50ns FWHM, but he corrects that they are more like 20-25ns FWHM.  But this should still be fine with the 4ns sampling in the FADCs.

Revision as of 01:12, 4 March 2022

Back to Main Page >> Data Acquisition Meetings >> Counting mode DAQ meetings

previous meeting << >> following meeting

Meeting information

 Meeting time:  12:00pm Eastern
   
 Zoom connection information
 Join ZoomGov Meeting
 https://jlab-org.zoomgov.com/j/1603800887?pwd=aXVmSnN6aWZWN3RBdXB4RUhGckZJdz09
 
 Meeting ID: 160 380 0887
 Passcode: 903612
 US Tool free phone: (833) 568 8864
 Find your local number: https://jlab-org.zoomgov.com/u/ab8341cjcG

Agenda

  • Activity updates
  • General discussion

Minutes

Recording on Zoom: [1]

Participants: B. Moffit, P. King, D. Armstrong, C. Ghosh, H. Liu, M. Gericke, D. McNulty

  • Discussing data rates
    • The links from the MPDs to the VTP are 1Gbps links, so ~5kHz with 6 samples per event should be fine
    • The VTP optical link is 10Gbps Ethernet, so we could handle ~40% occupancy at 5kHz in the GEMs, if we had one VTP
    • When we do zero suppression, the passed data has all six time samples for the channels that are above threshold. The zero suppression is done in the VTP
    • Chandan comments that his simulation using the beam generator in an open sector at 100nA has a rate of about 10MHz (Mollers and elastics); my estimates were based on a Moller rate of 30MHz, so occupancies might be lower than I estimated.
    • David points out that most of our running would not be at the very highest luminosity. Also we can prescale our triggers down to keep the data rate under control.
      • We should not need to use the multiple event builders like SBS is going to need to do in the next running period.
    • David also comments that SBS is getting higher occupancies than they were expecting; this is coming from xrays making single events in individual layers. David has a large beam-generator simulation file from Chandan that should include xrays, so he is planning to try to look at this in simulation soon
    • We'll need to prescale our triggers; as Chandan had said above, the rate per sector is ~10MHz at 100nA, and the maximum event rate we can take is ~5kHz. So we'll need prescales of order 2000.
    • Taking about the FADC window
      • I had estimated data rates based on 38 4-ns samples (152ns). Bryan says that that due to data packaging/transfer issues we might be as efficient using a 256ns window (64 samples).
        • It's best to use 2^m samples in the FADC to take advantage of the bandwidth.
      • With raw mode at 5kHz (and using 38 samples per event), I'd gotten ~850Mbps for the crate with 14 FADC modules; so we should plan on using the VTP readout for the FADCs
      • Also note that we could take most of the events just in pulse parameter mode, and only get the full raw mode periodically.
    • Hanjie asks if the network from the hall to CHA was upgraded to 10Gbps for SBS; Bryan says yes, and they have been able to saturate that link as a test.
      • Probably we can get both the GEM readout and the FADC readout within one 10Gbps link
  • Last meeting Michael had stated the ring five signals had a 50ns FWHM, but he corrects that they are more like 20-25ns FWHM. But this should still be fine with the 4ns sampling in the FADCs.
Retrieved from "https://moller.jlab.org/wiki/index.php?title=CountingDaq_220218&oldid=2894"