VEGAS Current Issues - Updated 21 April 2014
- See Also CICADAGreenBankSpectrometerAstronomicalObservations
- Common to all modes
- Completed Items
- Who: Richard, Jason, Hong Importance: MEDIUM Tune ADCs
- Who: Anish, Ray. Importance: MEDIUM. Normalize spectra and number of integrations per scan
- Who: Anish? Priority: Medium Mode 19 issue [Also mode 24]
- Who: Amanda, ?. Importance: Medium Bank F fails to balance
- : Who: Joe Brandt. Importance: High SWPER of 0.5ms results in integration time of 1ms.
- Who: Paul, Ray. Importance: High. IF Manager not aware of all it's connections
- Who: Richard, Ray. Importance: High Problems switching between modes
- Who: Amanda, Richard Priority: Medium System reacts to setting Doppler track_freq and then setting a specific frequency for auto peak fails with an LO error.
- Who: Amanda, Richard Priority: Medium System first scan fails with banks failing.
- Who: Paul. Importance: High. vegasdm should improve plot labels
- Who: Richard, Ray, Bob. Importance: LOW. FITS file is not written to disk until scan is completely finished.
- Who: Richard, Ray. Importance: LOW. It would be nice for vegasdm to show whether a bank is actually being used in an observation.
- Who: Anish, Ray. Importance: High. vegasdm should display results of requantised samples after digital filtering for modes 4 onwards.
- Who: Richard, Mark. Importance: Medium Spurious IF Manager Errors
- Who: Paul & Ray. Importance: High. VEGAS Manager/VEGAS_IF.xml/Coordinator - 950MHz filter/missing spectral lines.
- Who: Anish, Joe Priority: High Current minimum exposure time is two hardware exposure times.
- Who: Amanda, Melinda, Ray. Importance: High. Bank A not included in all configurations.
- Who: Ray. Importance: Medium. When connection to ROACH is lost, manager doesn't automatically re-connect..
- Who: Richard, Bob. Importance: LOW. GBTIDL displays the wrong bandwidth
- Who: Richard, Ray. Importance: LOW. Make "rc_system_status" a globally-defined alias
- Who: Richard, Ray. Importance: HIGH. Sub-beam nod doesn't seem to work
- Who: Richard, Wolfgang. Importance: LOW. lustre should be mounted on the observing machines
- Who: Richard, Melinda, Glenn, Ray, Joe, Anish. Importance: LOW. Balancing
- Who: Bob, Amanda. Importance: LOW. bandwidth not displayed on GBTIDL plotter for mode 3.
- Who: Richard, Ray, Joe Importance: LOW Actual switch period in not written into ACTSTATE table
- Who Richard, Ray, Joe Importance: MEDIUM Manager should only provide data for requested sub-bands
- Who: Richard, Ray, Joe Importance: MEDIUM The first integration in a scan looks bogus
- Who: Richard, Ray. Importance: MEDIUM. Miscellaneous hardware issues.
- Who: Galen, Richard, Anish. Importance: MEDIUM. Gain adjustment for Galen's module.
- Who: Galen, Richard. Importance: MEDIUM. Mode 3. The `signal to noise ratio' (ie the ratio of `signal' from the CM output to the noise floor of Galen's module) at the output of Galen's module need to be checked with 950 MHz filter selection.
- Who: Richard, Mark. Importance: LOW GFM real-time display
- Who: Richard, Ray. Importance: LOW. Use latest BOF files
- Who: Richard, Ray. Importance: LOW. vegasdm - tries to connect to all banks, even though only BankA and BankB are active
- Mode Specific
Common to all modes
: Who: ??. Importance: Medium measpwr displays show repeatable structure.
Observed by RMP while doing tests for Jim Braatz. All samplers show:
On timescales of tens of seconds, measpwr drops from -20dB to -31dB for one sample. Happens asynchronously between banks, so I don't think it can be due to phase changes or blanking.
The total power is systematically shifting up and down. Looks like a square wave, period ~30 secs, amplitude ~0.3dB.
Status 21 Apr 2014:
Status 28 April 2014:Problem could not be replicated - others are asked to watch for this in subsequent tests - does it stop or just drop in level?
Status 05 May 2014: Continue monitoring during testing
Status 12 May 2014: Continue monitoring
Status 19 May 2014 - June 10 2014: Continue monitoring
Who: TBD. Importance: Medium High. Write a python script to upload ADC coefficients.
Combined several related entries
Status 02 June 2014:
Status 10 June 2014: see
June 10 Notes
Who: Joe. Importance: High. Mixed mode integration time calculation.
Status 14 march 2014: Added to list.
Status 19 march 2014:In progress. Joe implemented the naive option and is working on the more involved solution.
Status 24 march 2014:Still under way
Status 31 March 2014: Slaved switching (AKA mixed modes operation) has been tested on a single test node (SRBS), but will need testing on the 'real' VEGAS HW. Wednesday is probably too soon, anytime afterward works for me. There is some intersection of this with the blanking discussions, so it may be desirable to hold-off testing until both are fully defined.
Status 07 Apr 2014: Fix in place, expect for 16K mode (need Berkeley involvement). Waiting before releasing.
Status 14 Apr 2014:
Status 28 April 2014:Developed, but unreleased/tested
Status 05 May 2014: No Activity
Status 12 May 2014: should be tested this week
Status 19 May 2014: Released but may be other issues that will be tested 5/20-24
Status 27 May 2014: Tested using mode 19 observations on May 15, 2014. Preliminary results where that the mode 19 data behaved as expected. Next step is to test other modes to see that they are okay.
Status 02 June 2014: Test week of 6/9
Status 17 June 2014: Go ahead and test while single mode issues are still being resolved
Who: Ray, Paul. Importance: Low. Refactor code to use ROACH interface provided in YGOR library.
Status 07 Apr 2014: - 10 Jun 2014: No Activity
Who: Richard, Anish, Ray, Melinda. Importance: MEDIUM. Possible to choose configurations which do not produce any data.
- Current status (10-Feb-2014): Ray to add a warning in the Manager if the configuration will produce no data.
- In e.g. Mode 2 it is possible to select a switch period of 10ms, which produces a phase length of 5ms. But the minimum integration time is 8ms, and the result appears to be no data. At a minimum, the Manager should flag an error. Do we want config tool to fail to validate?
- Current Status (18-Feb-2014): Assign these to alternative SDD member (Paul) as a VEGAS diversification effort.
- Current status (24-Feb-2014): [Still active]
Status 10 March 2014: Melinda to make this an illegal status in the config_tool; Paul to implement warning in Manager
LOW.
Status 17 march 2014: No report
Status 24 march 2014:Testing on March 25
Status 31 March 2014: Work is waiting for update to mode table (minimum blacking time)
Status 7 April 2014: RP to follow up with Anish.
Status 14 Apr 2014: Joe to ponder
Status 21 Apr 2014: Joe not available to update - Ray to join pondering then update Melinda
Status 28 April 2014: No Report
Status 05 May 2014: MR Ready for Melinda to review.
Status 12 May 2014: Same
Status 19 May 2014: Same
Status 27 May 2014: Discussing unlikely occurrences that will complete MR
Status 02 June 2014: MR Complete
Status 10 June 2014: MR in process
Who: Amanda. Importance: LOW. Investigate the ripple seen in mode 1 bandshape.
- Current Status : (10-Feb-2014) will be investigated at Q-band as part of support scientist "training"
- Current Status (18-Feb-2014): Reviewing the tests that should be used to characterize and correct
- Current status (24-Feb-2014): [Amanda is looking at it as testing and training allows]
Status 10 March 2014: Ongoing
Status 17 march 2014: Jason and Amanda to investigate 18-March maintenance
Status 24 march 2014: Amanda is still looking at data
Status 7 April 2014: Will look into it this week.
Status 14 Apr 2014: A new bof file should correct the likely cause. The new file has been installed and then repeat the previous test. Scripts should be available for re-use.
-
- A quick test of Mode 2 & 3 should be added if time permits to confirm that there is no common interaction
Status 28 April 2014: No Report
Status 05 May 2014: May be related to setting of sg_period. Ray and Glenn to agree on requirements; Ray to modify manager and then Amanda and Jason (or Randy) to retest.
Status: 12 May 2014 - will be investigated this week
Status 19 May 2014: Paul is fixing via MR
Status 26 May 2014:
MR: this part of the MR is progressing; python and C++ code for dealing with this has been developed and tested. Just need to integrate it into the manager.
Status 02 June 2014:
Status 10 June 2014: see
June 10 Notes
Who: Mark. Priority: Low VEGAS dictionary in Astrid is case sensitive.
- Current status (24-Feb-2014): [Likely a bigger issue than VEGAS - add validation for vegas.vfreq dictionary ]
Status 10 March 2014 - 02 Jun 2014: No activity
Status 31 March 2014: New item
Status 7 April 2014: No activity.
Status 14 Apr 2014: Richard to raise with Sci Staff and revisit priority along with config tool behavior
Status 21 Apr 2014: Richard in discussions with Melinda
Status 28 April 2014:
Status 05 May 2014: MR drafted and circulated for review. In progress
Status 12 May 2014: Same
Status 19 May 2014: Same
Status 10 June 2014: MR in process
Completed Items
Who: Richard, Jason, Hong Importance: MEDIUM Tune ADCs
- Current Status: (10-Feb-2014): should be performed this week.
- Current Status (18-Feb-2014): Jason will contact Hong and if availability is a problem, Dan to locate an alternative resources
- Current status (24-Feb-2014): [Started, but issues stopped progress last week]
Status 10 March 2014: Ongoing
Status 17 march 2014: Hong and Jason need to communicate, Hong needs certain hardware access and will show Jason how to do the setup. Substantial improvement has been seen with the original sample.Jason to contact Hong
Status 24 march 2014: Hong has finished and provided a tarball and procedure. Randy,Jason, John to meet and send questions to Hong. Then do the rest of the ADCs and determine how to apply fixes.
Status 31 March 2014: Amanda will look at data this week
Status 7 April 2014: Jason to obtain values/"run the tuning" during non-maintenance.
Status 14 Apr 2014: Perform 15-April
Status 21 Apr 2014:redo
Status 05 May 2014: Complete
Who: Anish, Ray. Importance: MEDIUM. Normalize spectra and number of integrations per scan
- Current Status (10-Feb-2014): Software development group to decide whether to leave this in the sdfits filler, or move it to the manager
- Anish to describe to Ray what is required.
- This should take care of the following items noted:
- Normalization of spectra with exposure time.
- Number of integration in a scan should be equal to that requested by the user. Number of integration in a scan for data from different banks should be same if they are configured in the same way. (Ray and HPC software person) - Done
- Current Status (18-Feb-2014): Assign these to alternative SDD member (Paul) as a VEGAS diversification effort. Investigate if this is a good idea.
- Current status (24-Feb-2014): [No activity]
Status 10 March 2014: Paul is implementing this. We should make it permanent or not implement it (as opposed to having it happen sometimes); SDD to decide which way to proceed.
Status 17 march 2014: No report
Status 24 march 2014: Code is in place for release after benchmarking tests
Status 7 April 2014: On-going. Trying an alternate implementation, will test and then if it works sufficiently then we'll release.
Status 14 Apr 2014: Test this week
Status 21 Apr 2014: base problem fixed - Paul will add Nan parameter
Status 28 April 2014: No Report
Status 05 May 2014: Complete and released
Who: Anish? Priority: Medium Mode 19 issue [Also mode 24]
Status 24 March - one or two spectra are missing. Needs additional testing, hopefully during the Tuesday maintenance period (14:00 - 16:30). # filter taps indicate a different number of blanking clock ticks than currently implemented per spec. New spec needs to be experimentally derived.
Status 31 March 2014: Blanking time discussions. I've been working on a page trying to summarize the discussions thus far, and lay out what needs to be updated.
Status April 7 2014: Joe was written wiki. To distribute this week.
Status 14 Apr 2014: Work has started and unit tests will be used to initially verify the fix - needs a system test to determine how to test on the system (repeat of tsys tests?).
Status 21 Apr 2014: completed - see mixed mode discussions
Who: Amanda, ?. Importance: Medium Bank F fails to balance
See email from Amanda.
Status 10 March 2014: Amanda to follow up with Steve and Galen. John will give them a heads-up. Work-around is to set converter rack attenuators by hand, but data still looks suspicious.
Status 17 march 2014: Discussed with Steve and "moved" line to new bank and problem moved too.
Status 24 march 2014:Galen looking at converter rack
Status 31 March 2014: Two channels need attenuation changes, but nothing was determined to be out of tolerance. Steve will be doing KFPA tests and will see if any bands fail to balance
Status 14 Apr 2014: To be investigated 15-April
Status 21 Apr 2014: Fixed
: Who: Joe Brandt. Importance: High SWPER of 0.5ms results in integration time of 1ms.
Status 14 Apr 2014: Needs to be addressed in Manager. Spec is correct, ("fftrates in S/W wiki). Joe to investigate.
Status 28 April 2014:Released and tested - incorporated in Friday's test data
Who: Paul, Ray. Importance: High. IF Manager not aware of all it's connections
This is fall out from the incomplete fix to the 'completed' issue: 'Spurious IF Manager Errors'.
Status 04 April 2014: Added to list. Fix in progress.
Status 07 Apr 2014: Fix in place. Needs testing.
Status 14 Apr 2014: Done and released
Status 21 Apr 2014:
Who: Richard, Ray. Importance: High Problems switching between modes
Tried to switch between mode 20 and mode 1 in Bank A, and got the error message:
Status 10 March 2014: In progress.
Status 17 march 2014: This problem has yet to be resolved. Spent Friday creating a manager capable of running without hardware so that I could test issues like this without contending for limited resources. In the meantime VEGAS should be turned off then back on again via the coordinator, to make sure the coordinator and banks to be used are in sync.
Status 24 march 2014:Ok as long as you work through coordinator, needs to accommodate exceptions
Status 07 April 2014: Released fix. Waiting to see if anyone reports issue re-appearing.
Status 14 Apr 2014: fixed as long as you work through coordinator
Who: Amanda, Richard Priority: Medium System reacts to setting Doppler track_freq and then setting a specific frequency for auto peak fails with an LO error.
Status 17 March 2014:New
Status 24 march 2014: Ready to test
Status 31 March 2014: Config tool test failed, plan to retest this week.
Status 7 April 2014: Ready to test during maintenance this week.
Status 14 Apr 2014: Fixed and released
Who: Amanda, Richard Priority: Medium System first scan fails with banks failing.
Status 17 March 2014:New
Status 24 march 2014: No activity
Status 7 April 2014: We think this is fixed.
Status 14 Apr 2014: Done
Who: Paul. Importance: High. vegasdm should improve plot labels
From Richard: "label the red and green lines. Ideally, you could label them by feed number and polarization; if that is not possible, just VEGAS sampler number (i.e. J1 through J16, as labelled by the IF Manager CLEO application)"
Status 07 Apr 2014: started. proceed if not difficult to do. vegasdm to be released tomorrow during maintenance tomorrow.
Status 14 Apr 2014: done and released
Who: Richard, Ray, Bob. Importance: LOW. FITS file is not written to disk until scan is completely finished.
So, gfm and sdfits cannot start looking at the data until the scan is over. It would be nice to periodically (once every 30 secs or so) flush the file to disk, so the analysis programs could get started.
- Current Status (18-Feb-2014): Experiment with small part of data being written so that data processing can begin. Watch for file system performance degradation. Assign these to alternative SDD member (Paul) as a VEGAS diversification effort.
- Current status (24-Feb-2014): [Still active]
Status 10 March 2014: Paul will proceed to implement this.
Status 17 march 2014: No report
Status 24 march 2014: In progress
Status 31 March 2014: Dependent on benchmarking tests
Status 7 April 2014: By design to write as fast as possible to lustre file system. No further work.
Who: Richard, Ray. Importance: LOW. It would be nice for vegasdm to show whether a bank is actually being used in an observation.
Status 10 March 2014: Ray will work on this when he gets around to it.
Status 17 march 2014: Cleo release on March 16 may perform this function. Richard to validate.
Status 24 march 2014:On the list to code
Status 31 March 2014: Plan to release Wednesday - also in Cleo
Status 7 April 2014: Complete/released.
Who: Anish, Ray. Importance: High. vegasdm should display results of requantised samples after digital filtering for modes 4 onwards.
Status 10 march 2014: Ray will tackle this after the switching modes bug is addressed.
Status 17 march 2014:no update
Status 24 march 2014:Paul is working issus
Status 31 March 2014: Paul has completed his work, but Ray may want to re-factor. Go ahead and release 2-April.
Who: Richard, Mark. Importance: Medium Spurious IF Manager Errors
Configured to just use Bank A (so VEGAS inputs J1 and J2). IF Manager produces the error message:
- VEGAS: J16 has no IF signal and connections only as far as ConverterModule16:J9
Repeated for all of J3 through J16. But we don't care about these. This spurious error message is confusing, and a potential cause for concern.
Status 10 March 2014: Paul to investigate.
Status 17 march 2014:Fixed?
Status 24 march 2014:Move to 'complete'
Who: Paul & Ray. Importance: High. VEGAS Manager/VEGAS_IF.xml/Coordinator - 950MHz filter/missing spectral lines.
Status 11 march 2014: Added to list.
Status 17 march 2014:Fixed. Last data looks good. Some discussion of exception processing precedence
Status 24 march 2014:Move to 'completed'
Who: Anish, Joe Priority: High Current minimum exposure time is two hardware exposure times.
Need this to be one exposure time for pulsar observations. Joe to investigate.
Status 17 March 2014: fixed
Who: Amanda, Melinda, Ray. Importance: High. Bank A not included in all configurations.
- If KFPA is configured for beam 1, banks E-H are used. Bank A is not configured so there's no switching signal. This will also affect two beam receivers, although according to Ron the affected beam will probably be beam 2 in those cases.
- Current status (24-Feb-2014): [Amanda will consult with Melinda]
Status 10 March 2014: Ongoing
Status 17 March 2014: Fixed
Who: Ray. Importance: Medium. When connection to ROACH is lost, manager doesn't automatically re-connect..
Status 11 march 2014: Ray will tackle this.
Status 17 march 2014: Fixed
Who: Richard, Bob. Importance: LOW. GBTIDL displays the wrong bandwidth
See scan 3 in /home/sdfits/TGBT13B_502_33/TGBT13B_502_33.raw.vegas. Should be 850MHz, but displays 1.000 GHz.
- Current status (24-Feb-2014): [Amanda will update documentation]
Who: Richard, Ray. Importance: LOW. Make "rc_system_status" a globally-defined alias
In my home directory, I have a file rc_vegas_status which I copied from Ray, to display the shared memory. We should put this somewhere (like in /home/gbt/etc or somewhere), with an alias defined for it, so everyone can run it without needing to copy it from Ray's directory?
- Current status (24-Feb-2014): [Interim fix is in place, but long term fix is to implement the DIBAS solution in VEGAS]
Who: Richard, Ray. Importance: HIGH. Sub-beam nod doesn't seem to work
Running wnodsnod from
TGBT13A_501 consistently failed with the error:
Aborting: deadline missed arm time is in the past
- Current status (24-Feb-2014): [Not exclusive to nodding - Data has been taken, Frayer to review data - Fixed]
Who: Richard, Wolfgang. Importance: LOW. lustre should be mounted on the observing machines
- Current status (24-Feb-2014): [Done]
Who: Richard, Melinda, Glenn, Ray, Joe, Anish. Importance: LOW. Balancing
- Current Status : (10-Feb-2014): Need to check Melinda's VEGAS option has been added to balancing
- Potential setting of FFTSHIFT and equivalent L1/LBW1 etc parameters may be added later as an expert-user mode requirement is identified
- What is the correct balance, -20dBm or -25dBm ?
- Glenn Jones confirms -20dBm is correct (possibly even a little on the low side).
- Anish This need to be checked along with the gain adjustment of Galen's module. I suspect the current adjustment is not ok for low-freq (1.4 and below) galactic plane obs.
- In testing of 13th August with the telescope looking at zenith, the VEGAS balancing seemed to come out around - 13dBm, with no error messages.
- Balance command needs to be updated for L1/LBW1, L8/LBW1 and L8/LBW8 modes. The power level should be adjusted so that the signals are `properly' re-quantization after the hardware filtering. The voltage snapshot after the filtering has to be readout and the power estimated from these values should be adjusted to a specified value by changing the digital gain. This can be a task manager can handle and balance command can evoke this task after balancing the adcpower when the above said modes are used.
- Current Status (18-Feb-2014): Solved by agreeing not to do it unless an expert user requires it
Who: Bob, Amanda. Importance: LOW. bandwidth not displayed on GBTIDL plotter for mode 3.
- Current Status : (10-Feb-2014) Amanda to provide Bob some data.
- Current Status (18-Feb-2014): Believed fixed by Bob G.
Who: Richard, Ray, Joe Importance: LOW Actual switch period in not written into ACTSTATE table
- Current Status : (10-Feb-2014): Ray / Joe to implement
- Current Status (18-Feb-2014): Fixed
Who Richard, Ray, Joe Importance: MEDIUM Manager should only provide data for requested sub-bands
- Current Status: (10-Feb-2014): fix ready to be released
- Current Status (18-Feb-2014): fix released, to be tested Thursday/Friday 20/21 February.
Who: Richard, Ray, Joe Importance: MEDIUM The first integration in a scan looks bogus
Who: Richard, Ray. Importance: MEDIUM. Miscellaneous hardware issues.
- This should take care of the following items noted.
- Valon freq setting through roach serial port (Ray)
- HPC -- manager start issue; need to reset the coordinator and manager (Ray)
Who: Galen, Richard, Anish. Importance: MEDIUM. Gain adjustment for Galen's module.
- Mode 3. The `signal to noise ratio' (ie the ratio of `signal' from the CM output to the noise floor of Galen's module) at the output of Galen's module need to be checked with 950 MHz filter selection.
- Testing on Aug. 21, 2013 shows S/N ratio of 19 dB for IF Module 10 (from CM1, CM5), 21 dB for IF Module 11 (CM2, CM6). -GW
- For some of the receivers the CM attenuators are close to the maximum attenuation of CM modules after executing astrid balance command. This large attenuation in CM modules prevents the balancing of VEGAS to power level close to -20 dBm as measured by VEGAS manager.
- Should adjust gain of Galen's module so that the `signal' coming from Converter module should be at least 20 dB above the noise floor of Galen's module for all receivers after balancing. This 20 dB `signal to noise' ratio should also be present while observing Galactic plane at frequencies <= 1.4 GHz.
- Gain of the modules is not adjustable, would require new design and fabrication. I suggest adding 10 dB of attenuation (for a a total of 16 dB) between IF Modules and A/D input for optimum Noise versus P1(dBm) tradeoff given levels measured on Aug 21, 2013. -GW
- Now the total attenuation between Galen's IF module and ADC input is 25 dB (which includes the 15dB attenuator added earlier) - Anish
Who: Galen, Richard. Importance: MEDIUM. Mode 3. The `signal to noise ratio' (ie the ratio of `signal' from the CM output to the noise floor of Galen's module) at the output of Galen's module need to be checked with 950 MHz filter selection.
- Checked, see previous item. -GW
Who: Richard, Mark. Importance: LOW GFM real-time display
- Scan (SIMPLE, Track) not handled by the Spectral Line plugin. Should it be? Currently it is not; prefer to focus on real-time streaming as the solution
Who: Richard, Ray. Importance: LOW. Use latest BOF files
- This should be implemented as low priority once the BOF file edits have settled down.
Who: Richard, Ray. Importance: LOW. vegasdm - tries to connect to all banks, even though only BankA and BankB are active
Who:
John, UC-B. Importance:
HIGH. Need to pin our tasks to cores
- Current Status: Complete
- For maximum throughput we need to pin our tasks to cores and pin the interrupt handlers for the network and GPU to certain cores.
- Note also that threads cannot be arbitrarily pinned, so whoever does the assignment needs to experiment with different configurations to find the optimal one.
Who:
Anish, Melinda, Ray. Importance:
HIGH. Complete frequency configuration for VEGAS in Config Tool and Manager.
Current Status : Complete
Melinda updated config tool to configure modes 4 to 9. The new config tool sets the sub_frequencyA and filter_bw correctly for mode 1 to 9 as per the report
http://www.gb.nrao.edu/vegas/report/skyfretobb.pdf
A report describing the programing of switching signal generator for all modes is now available at
http://www.gb.nrao.edu/vegas/report/ssgprogram.pdf
- Anish needs to describe the algorithms in detail to Melinda and Ray.
- This should complete the following items noted:
- CRVAL1 is modified by valon freq and subfreq. Subfreq settings need to be fixed by looking at the need for all modes. (Ray, Melinda, Mark Wagner)
- User specified sky freq is currently tuned to baseband freq of 775 MHz for mode 1, 2, 3. Should this be changed to 750 MHz ? (Melinda)
- VEGAS filter_bw stayed at 1150 - even after Anish re-configured. Needs to be set by config_tool
Who:
Amanda,Ray, Joe?. Importance:
MEDIUM. Buffers that are not full are not written to disk.
- At the end of a scan, any buffers that are not completely full should be written to disk. Joe Brandt may have a fix in place already for DIBAS.
Who:
Richard, Ray. Importance:
HIGH. Modes 2 and 3 are not taking any data in switched observing modes.
Problem was a bug in the blanking calculations.
Who:
Anish, Melinda. Importance:
LOW Investigate topo / doppler tracking
- I had set vframe='topo', but it appears that it attempted to doppler track the data (F0=2.165 GHz and Fsky=2.175 GHz). Should I have left dopplertrackfreq blank instead of setting it to 2165 GHz? (F0 is the restfreq set by the user and Fsky is the freq at the reference channel in GBTIDL. For mode 1 the reference channel is 512. The user specified restfreq (if there is no Doppler track) will be tuned by the configtool at a freq 10 MHz (1% of the bandwidth) away from reference channel. This is because we expect large spectral value at channel 512 and spectral line users may not want to tune their line to the channel 512. Anish )
Who:
Bob, Richard. Importance:
MEDIUM. getfs cannot display in Units of Jy
- TGBT13B_502_06 - Scan 10. getfs,10 works fine, but:
GBTIDL -> getfs,10, units="Jy"
% GETFS: Result is all blanked - probably all of the data were flagged
% Program caused arithmetic error: Floating divide by 0
% Program caused arithmetic error: Floating illegal operand
GBTIDL ->
this is because there is not antenna FITS file, so the default elevation is 0.0. Not worth fixing
Who:
Bob, Richard. Importance:
HIGH. Bank B data cannot be displayed in GBTIDL
- TGBT13B_502_06 - fill to /home/scratch/rprestag/vegas/TGBT13B_502_06. Both banks seem to fill ok. When I use filein to select Bank A, everything displays correctly. When I use filein to select Bank B, summary displays the scans correctly (10-15), but gettp produces error: % GETTP: No data found, can not continue.
The problem is that IFNUM=0 for bank A and 1 for bank B. But scan_info
was only returning the number of unique IFNUM values and the code was assuming that the only legal IFNUM values were 0 to (n_ifnums-1) and so there was no way to select ifnum=1 when only bank B data was opened.
I've modified the scan_info structure (which is what is used for these argument sanity checks) to return the list of IFNUM, PLNUM, and FDNUM values actually found in the opened data. Checks are now against those values and the error message when the requested IFNUM , PLNUM or FDNUM is not found now lists the possible values.
I don't want to change the default case if IFNUM is not supplied to use the first one one found instead of IFNUM=0 because I think the former is more likely to be confusing - especially now that IFNUM (as of the upcoming release) is set in the GO fits file - i.e. it's no longer something sdfits or GBTIDL generates.
Who:
Richard, Bob. Importance:
LOW. SDFITS behaviour in case of missing data
- Sometimes VEGAS produces a file with a Header but no data. Filler skips over these with no error messages. Should flag an error.
- Example data in TGBT13B_502_06
Bob has added an appropriate error message.
Mode Specific
Completed Items
Who:
Anish, Ray. Importance:
LOW OBSBW and BASE_BW.
- The VEGAS FITS file has a parameter in the primary header, BASE_BW. It also has a series of key=value pairs copied from shared memory, and this includes OBSBW. In some of the data we took yesterday see e.g. /lustre/gbtdata/TGBT13A_504_26/VEGAS/2013_08_12_19:52:06A.fits these are different BASE_BW is 1400MHz, and OBSBW is 800MHz. Should these be the same? If not, what is the difference between the two values?
Who:
Anish, Ray. Importance:
LOW limit on acc_len.
- Manager should check whether acc_len has gone out of range when an expert user change the valon frequency. The possible value of acc_len's range needs to be specified and put in the manager.
Who:
Anish, Ray, Joe. Importance:
LOW integration time per state
- While using mode 2 with switch period 100 msec and was having 2 states an occasional problem in integration time noticed. Most of the time the value of the integration per state is 42 msec; so including the blank time of 8 msec (previous hwexp for mode 2) the total is 50 msec. Occasionally the switch period of one of the states goes to 50 msec and the other state goes to 34 msec.
Who:
Anish, Ray. Importance:
Low (
Should be fixed during Tier 2 testing)
- Spec_tick for mode 2 and 3 is currently set to about 8 msec. acc_len parameter along with the FFT cycle length in the FPGA decides this period. Spec_tick is also used to latch the switching signals in the FPGA and for computing the period of the switching signal. The parameter acc_len can be reduced to have a blanking period of 2 msec or so for cal switching.
acc_len for mode 2 and 3 is set to 128. This gives the min exposure time of 1.4 and 2 msec for mode 2 and 3 respectively. -- Anish