VEGAS Pulsar Project Meeting: 2015 October 12th 2:00 - 3:00pm ET

Room / Connection Details
  • GB-137 / Soc 280
  • 192.33.117.12##7144
  • 434-817-6443

Housekeeping Details

  • Main wiki page
  • gbsapp Mailing List
  • Location and regular date and time of meeting - agreed weekly on Mondays; GB-137 / Soc-280; 2-3pm ET
  • ETK codes:
    • Electronics: 432510.GB3015
    • Software: 893252.GB3015
    • Science: 432540.GB3015

Resource Allocations
BLOSS, MARTIN ALAN 0.05
BRANDT, JOSEPH JOHN 0.10
CREAGER, RAY 0.65
DEMOREST, PAUL 0.05
RICHMOND-DECKER, JUSTIN 0.20
FORD, JOHN M 0.05
MELLO, MELINDA JANE 0.05
MCCULLOUGH, RANDY L 0.25
RAY, JASON 0.55
PRESTAGE, RICHARD 0.05
LYNCH, Ryan 0.25
VAN TONDER, VEREESE 0.10

Present

  • Randy, Vereese, Jason, Joe, Justin, Ryan, Scott, Ray

Agenda

BOF file status updates

  • c1500x4096 mode
    • Had 15 timing errors in build from Friday, with worst that was just over 0.6 ns
      • Not wholly satisfied yet but will go ahead with some tests on srbs
      • All errors were in PFB blocks
    • 14 were removed by fixing the PFB block, but now there are 16 in the FFT block
      • Will work on eliminating errors in FFT block later today
      • Looking for ways to eliminate layers of logic, will attempt similar strategy used to eliminate errors from PFB block
    • Will plan on using srbs for tests on Thursday of this week
  • Further test results of BOFs with unexpected bandpass and/or signal
    • All changes that need to be made for saturated bandpass due to improper scales have been noted and will be pushed through during next builds
    • All the coherent modes need a small edit to correct nchan calculation
    • Many fixes from the c1500x4096 build will be used as basis for other coherent modes
    • Still tracking down bandpass issues for i0800x0064, 0128, and 8192
    • Looked at design file for i0800x8192 mode and the logic for unwrapping packets looks OK. Still looking for cause of duplicated signal. Need to investigate further.
  • Correction of known errors
    • All files have small errors that require a rebuild, and these will all be done under the latest library version
  • Randy noticed that the software registers for the port numbers are 32-bit but only 16 bits are used. Would it cause any problems to redfine using 16 bits?
    • Joe and Ray don't think so
    • katcp can handle 16 bits and it should be transparent to the software
    • If this doesn't work we won't see any packets

Software

  • Justin will fill in developer sections of MR and send it to Melinda this week
    • Ryan will send to Richard once Justin gives his OK
  • Paul and Scott wrote a script to query GBT status SQL database and fill in shared memory
    • It looks like this is already running as these values show up correctly in vegas_status
    • Manager is currently setting some parameters
    • Scott asks if PSRFITS files will need to be filled or if they will be written in a process-ready format like GUPPI?
      • The plan is to emulate GUPPI
    • Ryan will send Justin a list of what parameters are missing from PSRFITS headers to see if Manager is trying to write them
  • Switching signal generation
    • BOF files do contain switching signal blocks
    • Ray thinks the switching signal is part of the backend classes
      • There are two for VPM modes
      • Backend sends switching signal generator to ROACH, I/O handled by ROACH
      • Ray and Joe will look at this first and contact Ryan when needed
    • Randy and Jason can test BOF files on srbs, and also test Manager
  • Matrix HPC
    • Coming along
    • New Pivotal Tracker stories have been made
    • Looking at guppi_daq for threads and modes that it uses
    • Justin will be joining forces with Ray on the pulsar end
      • Working on spectral line fits writing just to familiarize himself with Matrix code
  • TBIN rounding issue
    • Thought this may be the cause for the drift in the 1500 MHz BOFs
    • Increase precision in shared memory
    • Ryan and Justin will sit down to look into this
    • Justin suggests using a string instead of a float to store this
    • Need to check precision of type being used

GBT testing

  • Joe implemented transpose on GPU instead of CPU
    • Ryan tested this with VEGAS hardware for all available 800 MHz and 1500 MHz coherent modes and saw no packet loss
  • Joe found bug in guppi_daq

Wideband AP (Randy)

  • Finished circuit board layout and all the Gerber files
  • Randy talked with John Sheldon and will get quote for circuit boards and likely place order Wednesday

Hardware/software discussion with Dave (All)

  • Berkley crew had a meeting with BL board and they stressed the importance of taking data and making it public in January
    • Would like to jump start some hardware testing to stay on this timeiline
    • Plan on shipping a new server to start spearheading hardware installation and layout
    • Server should arrive today along with SPF+ mezzanine cards
    • Waiting on new switch, but will likely connect directly between the mezzanine card and the 4U computer so that we can test data flow
    • The amount of data is not critical
    • Would like to hook into one of the ROACH's on VEGAS this week
      • Use Banks C -- H in case work needs to be done on Bank B
    • Only need 4 links from the ROACH since there are only 4 ports on the computer
      • May be able to use existing SFP cards since only one port is populated on Banks B -- H
      • Can use only 3 ports if necessary
      • Will still plan to install other boards as there is no need to delay
    • Not yet sure what will be run on new computer, but interest in using PSRFITS format since it is self-contained
      • Find the right guppi_daq version on github
  • Dave will sit down with Ryan to learn how to operate VPM observing
    • Right now IF path and scan starts can be handled using Astrid, VPM manager parameters set using Ryan's script
    • Basically ready to run with 3-steps
    • GB plans to have config tool integration ASAP which will allow for full integration with existing software
  • General data management plan
    • During BL observing time data will just be recorded to disk
      • Will likely use disks that are local to BL machines
      • guppi_daq may write files
      • Our guppi_daq will not be getting packets so don't need to worry about where data is being written
    • After data are recorded and other projects are running, data will be processed and analyzed
    • "Interesting" data will be stored long-term, other data will be overwritten
    • May be interested in multi-casting to allow packets to flow to different machines
  • Initital system will be a SERNIDIP6 running guppi_daq; Joe wonders how Manager will run?
    • Will need a Manager running on the BL machine, but Ray wonders if it would be easier to run dealer/player
      • What is the long term plan? Best to do that instead of working on an ad hoc fix?
      • White board session to map stuff out 9:00 AM in the library

Date / Time of Next Meeting

  • October 26

-- RyanLynch - 2015-10-12
Topic revision: r2 - 2015-10-12, RyanLynch
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback