PUPPI Observing in Folding (Coherent) Mode

Basic Info

• The main PUPPI computer at Arecibo is called 'puppimaster' and the only user account currently is called 'gpu'.
• When logging in to puppimaster as gpu, you will already have all relevant PATH/etc set up (ie, no sourcing guppi.bash is required).
• The gpu cluster nodes are only accessible from puppimaster, and are called gpu01, gpu02, ..., gpu09. As of March 2, 2012, gpu04 is offline for repairs.
• The monitor programs guppi_status and guppi_gpu_status are available on puppimaster and work exactly the same as at Green Bank. These are both good to have open during an observing session.

Basic How-To

• Log in to puppimaster (or use existing vnc session on :1 and :2)
• On a different machine (e.g. fusion00), start Cima (Cima runs very slowly on puppimaster), select Stable version, then Pulsar mode.
• Copy a PUPPI-friendly Cima conf file for your chosen receiver from /share/obs4/usr/pulsar/p2624 to your project directory.
• Make sure winking cal is turned off in Main Cima Menu:Pulsar Observing.
• Main Cima Menu:Load Configuration, select the .conf file you copied over, Load and apply.
• Main Cima Menu:Power Control, Adjust All.
• On puppimaster, in a shell type guppi to get to the guppi command prompt, then execute these commands in sequence: mode(‘c512’) to set the number of channels (512 in this example), reset(‘800’) to set the bandwidth in MHz, arm() to start hardware. The IF/LO Setup section below describes the parameters of several available configurations. The Setting up the PUPPI HW section gives more details about the commands.
• On puppimaster, in a normal shell:guppi_adc_hist. Two histograms will be displayed, one for each polarization; they should be Gaussian and closely overlapping.
  • Sample output from guppi_adc_hist:
    ADC Power level info:
PolA (FPGA3): Mean=-1.030 RMS=21.897 Min=-96 Max=82
Add 0.8 dB attenuation (for target RMS 20.0)
PolB (FPGA1): Mean=-0.947 RMS=20.178 Min=-82 Max=86
Add 0.1 dB attenuation (for target RMS 20.0)
  • To check current attenuation values, on puppimaster: puppi_atten. The attenuation values can be changed in steps of 0.5 dB within the range 0 - 15 dB.
  • If necessary, set new attenuation values with
    puppi_atten -A [dB] -B [dB]
  • If PUPPI attenuation values are already at 0.0, but guppi_adc_hist recommends decreasing the attenuation: Main Cima Menu: Power Control: Use the arrow buttons in the line labeled "Attenuation(dB)" to reduce the attenuation for both polarizations by the recommended amount.
• Main Cima Menu:Cima Utilities; in "Send to VX-works" field type setcal lcorcal, then hit Enter. Confirmation of the command is displayed in the Cima log window. Erase it and then in the same field: cal25 on (hit Enter) to turn the cal on manually.
• On puppimaster, in normal shell: puppi_observe -b L -c [full path to .par file]. This takes a 90s cal scan at L-band. Par files are located in /home/gpu/tzpar. The Running an observation section describes available options for the puppi_observe script.
• On puppimaster, in normal shell: guppi_gpu_status (to monitor all GPUs), guppi_status (general backend status).
• On puppimaster, log into one of the gpu nodes to check cal data: e.g. ssh gpu05. In this case, the data is in /data/gpu/partial/gpu05.
• On gpu node, pav -X [puppi .fits cal file] to look at the cal data.
• When the cal scan is done, Main Cima Menu:Utilities; in "Send to VX-works" field type cal25 off (hit Enter) to turn the cal off manually.
• On puppimaster, in shell: puppi_observe -b L -T [obs. time in s] [full path to .par file] (example for L-wide observation). Data will be in the same directory as the cal files.
• On gpu node, in the data directory: pav -SFT [puppi .fits file] (shows the folded pulse profile).
• pav -GTdp [puppi .fits file] (shows frequency subband vs. pulse phase).
• pav -YFp [puppi .fits file] (shows subintegration vs. pulse phase).
• To end an observation at any time, on puppimaster: puppi_stop.

IF/LO setup

Several PUPPI-friendly CIMA config files are available in the p2624 (NANOGrav) project area. These are:

• generic_327_newcima.conf - The old ASP 327 MHz setup works for PUPPI as-is. PUPPI center freq=327 MHz, BW=+100 MHz.
• generic_430_newcima.conf - The old 430 MHz config also works. PUPPI center freq=430 MHz, BW=+100 MHz.
• puppi_1410_newcima.conf - L-Wide setup for PUPPI+ASP+WAPPs. The center freqs for the various backends are:
  • PUPPI = 1380 MHz, BW=-800 MHz
  • ASP = 1410 MHz, band_dir=+1 (Note changed band_dir vs old L-Wide config)
  • WAPPs = 1700, 1195, 1500 MHz, BW=50 MHz
• puppi_2030_newcima.conf - S-Low config putting PUPPI in the bottom half of the receiver band with the 1700-1880, 2040-2400 MHz filters in place. Center frequency/bandwidth info:
  • PUPPI = 2030 MHz, BW=-800 MHz
  • ASP = 2350 MHz, band_dir=-1 (Note changed band_dir vs old S-low config)
  • WAPPs = 2085, 2205, 1850 MHz, BW=100 MHz
• puppi_2800_newcima.conf - S-Low config putting PUPPI in the upper half of the receiver band with the 2600-3100 MHz filter in place. Center frequency/bandwidth info:
  • PUPPI = 2800 MHz, BW=-800 MHz
  • ASP = 2650 MHz, band_dir=+1 (Note changed freq vs old S-low config)
  • WAPPs = 2950, 2750, 2850 MHz, BW=100 MHz

Setting up the PUPPI HW

Setting up the hardware (loading FPGA designs, changing clock rate, etc) is done via the guppi> prompt. Just run 'guppi' on puppimaster to open it up. At the guppi> prompt, run the following commands to set up or change modes:

• mode('c512') to load 512-channel coherent mode. Use mode('c64') for low-freq receivers. mode() with no arguments will show both the currently selected mode, and all available choices.
• reset('800') to change to 800 MHz BW. Use reset('100') for 100 MHz BW on the low-freq receivers (327, 430). This command only needs to be run when changing the bandwidth.
• init() sets up appropriate values in the FPGAs. Note that the reset() command will automatically call init() so there is no need to run it twice in that case.
• arm() starts data flowing to the computers. Important: Do not run arm() once you are actually recording data! It will reset your observation.
• A typical command sequence for setting up an 800 MHz bandwidth observation is: mode('c512'); reset('800'); arm(). If no receiver changes will happen during a session, this should only needs to be done once at the start of the session.

Once you have done the above steps, you can check that the input power levels are ok by running guppi_adc_hist at the puppimaster command line. Important: Do not run guppi_adc_hist while recording data!

PUPPI has programmable attenuators that can be used to tweak the input power levels via the program puppi_atten. Attenuation for pol A and B can be adjusted separately from 0 to 15.5 dB in 0.5 dB steps. puppi_atten with no arguments will return the current settings. To give new attenuator settings, use -A and/or -B, e.g.:

• puppi_atten -A 13.5 -B 8.0

Running an observation (taking data)

The script puppi_observe will run a coherent fold-mode observation. This will fill standard timing-appropriate observing options into the shared memory, then start a coherent-mode observation (ie, no manual arm() is required). The script takes one argument, which is the par file to use for folding. The full path should be used to specify the par file, eg:

• puppi_observe /home/gpu/tzpar/B1937+21.par

The script also accepts the following options (as well as -h for usage info):

• -c to take a cal mode scan.
• -b to set the receiver band. Default is L-band, other currently allowed values are -b 327, -b 430, and -b S.
• -f To override the default center frequency for a band.
• -T to set the scan length in seconds (default is 1 hour for pulsars and 90 sec for cals).
• -m to override the default observing mode. Allowed values are COHERENT_FOLD, COHERENT_CAL, SEARCH, CAL, or RAW.
• -I to take summed-polarization data. Default is full-Stokes.
• -n To override the default number of channels.
• -p To override the default polarization (e.g. if you want to record only one polarization channel).
• -P To set the project ID.

By default, this script assumes you have set up the PUPPI hardware and IF/LO settings such that:

• L-band: PUPPI center freq (IF=1200 MHz) is 1380 MHz; the bandwidth is 800 MHz.
• 430: PUPPI center freq (IF=750 MHz) is 430 MHz; the bandwidth is 100 MHz.
• 327: PUPPI center freq (IF=750 MHz) is 327 MHz; the bandwidth is 100 MHz.
• S-band: PUPPI center freq (IF=1200 MHz) is 2030 MHz; the bandwidth is 800 MHz.

Finally, this also by default assumes you want to use the following pulsar observing settings (these are currently hard-coded in the script):

• Real-time coherent dedispersion and folding.
• 2048 pulse profile bins.
• 10 sec profile integration time.
• 25 Hz cal signal in cal mode.

The observing mode can be changed to record the channelized voltage data directly to disk by using the -m RAW option to puppi_observe.

To stop an observation before the specified scan length finishes, run puppi_stop.

Plotting/checking data

For coherent mode, the data are recorded on the gpu nodes, same as happens in GUPPI. To look at the files, log into any gpu node and look in the /data/gpu/partial/gpuXX directory. The usual PSRCHIVE plotting commends (eg, pav and psrplot) are available to show the data. For convenience the plot_last script can be run on any gpu node to automatically plot the most recently taken file.

When a scan finishes, the data from all the nodes can be transferred to puppimaster and combined. On puppimaster, go to /data/puppi/Pnnnn (nnnn = project ID), and run the get_data and combine_psrfits commands, for example:

• get_data puppi_55993_B1937+21_0080_0001.fits
• combine_psrfits puppi_55993_B1937+21_0080_0001.fits

And of course replace puppi_55993_B1937+21_0080_0001.fits with your actual data file name.