Pulsar Observing Documentation
This page describes the capabilities for pulsar observing with the VLA. It also provides information about how to set up pulsar observations, including what modifications need to be made to the observing scripts and VCI files that are produced by the OPT.
Phased-array Observing (YUPPI)
Currently the most well-tested pulsar observing style is based on phased-array processing. Originally developed for VLBI, in this mode the WIDAR correlator outputs a coherent (voltage) sum of the signals from one or more antennas, after applying phase/delay calibration solutions that are determined in real time. For VLBI observations, the data are sent to a Mark5 recorder and saved for later offline correlation. For pulsar observations, the data are instead directed to the Correlator Backend (CBE) compute cluster at the VLA site. The "Y" Ultimate Pulsar Processing Instrument (YUPPI) is a set of software that runs on the CBE and can process this data stream in real time. Additional details of the software can be found at the EVLAPulsarSoftware
wiki page. The allowed processing modes and resulting data products are very similar to those of standard single-dish pulsar observing; readers are assumed to have at least basic familiarity with this type of observation.
Independent of processing mode, the limit on the total bandwidth than can be handled is primarily set by the networking constraints involved in getting the data into the CBE cluster:
- subbands -- Each CBE node can receive at most two subbands. As there are 32 total nodes, this means at most 64 subbands can be processed.
- bits -- Either 2, 4, or 8 bits can be selected for the summed voltage data stream.
- bandwidth -- The total data rate per subband can be at most 1 Gb/s. This leads to the following constraints on per-subband and total bandwidths:
- 32 MHz/subband, 8 bits = 2 GHz max total BW
- 64 MHz/subband, 4 bits = 4 GHz max total BW
- 128 MHz/subband, 2 bits = 8 GHz max total BW (but note that processing 8 GHz is currently untested)
- Subbands narrower than 32 MHz are also allowed.
- resolution -- In all processing modes, the subband width sets an upper limit on frequency resolution (ie, combining subbands to produce wider channels is not possible). See the processing-specific sections below for additional constraints on time/frequency resolution.
An appropriate correlator configuration (VCI file) can be constructed as follows:
- The "array phasing" option must be selected for all desired subbands, this can be done in the OPT/RCT. Note, I have an existing set of pulsar configurations, these should be made available.
- The resulting VCI file will initially have the data routed to the Mark5, it must be changed to instead send the data to the CBE. The
pulsarize_vci program can automatically convert these files appropriately, with the number of bits selected based on the subband BW.
In order to observe in these modes, observing scripts require the following:
Power level setting:
- phasing scans -- Appropriate phase calibration ("determine autophase") scans must be included, and the resulting phase and delay calibration must be applied to the pulsar scans, this can be done in the OPT. With large numbers of subbands, there can be up to ~100 seconds latency in solving for the calibration solutions. At least 12 10-second phasing scans are recommended for each phase calibration.
- pulsar configuration -- Pulsar processing is enabled by including an appropriate scan intent (all begin with
OBSERVE_PULSAR). Mode-specific options are set using additional intents; see details below. This currently can not be done in the OPT, these intents must be added to the script by hand.
- switched power -- The 10 Hz switched power calibration signal can interfere with pulsar signals, and it should be turned off for pulsar scans. This can be done in the observing script by making a copy of the
LoIfSetup object and calling its
setNoiseTubeOff() method. It may be useful to pair each pulsar scan with a short calibration scan where the 10 Hz is on; best practices for VLA pulsar calibration are still under development.
By default, phased array data is passed through an automatic gain control (AGC) loop that levels the power appropriately for the number of bits in use. This feature has been observed to cause large power fluctuations and dropouts that are extremely detrimental to pulsar observing. It should be disabled following an initial setup period to allow the power levels to come into a good range. This currently must be done manually during the observation
/home/asg/bin/cmib agcGainOff; /home/asg/bin/cmib agcWindowOff
In fold mode, each subband is split into smaller frequency channels, coherent dedispersion is optionally applied, and the data are detected and folded modulo the current predicted pulsar period into some number of phase bins. Fold mode is selected by including
as a scan intent. The relevant limits on each parameter, and intent name within the observing script is:
- frequency resolution -- The coarsest possible resolution is 1 channel per subband. The narrowest is effectively arbitrary, but will at some point be limited by FFT length; if more than 1024 channels per subband are desired, additional testing is recommended. The number of channels per subband is specified via the
- bins -- Any number of pulse phase bins up to 16384 can be used. The number of bins is specified via the
- integration time -- Fold dump times as short as 1 second have been tested. The maximum dump time is in principle set only by the scan length. The dump time is specified via the
PsrFoldIntTime intent, in seconds.
- polarizations -- Either 1, 2, or 4 polarization products can be selected (1 implies summed polarizations). This is specified via the
- pulsar ephemeris -- A valid ephemeris ("par file") for folding must be supplied. Currently, only TEMPO-compatible par files are allowed, although TEMPO2 support could be enabled if there is demand. The full path to the par file is set via the
TempoFileName intent; this must exist on a filesystem that is accessible from the CBE nodes.
TempoFileName is set to the special value "CAL" the data will instead be folded at a constant (topocentric) period. The folding frequency is specified by the
PsrCalFreq intent, in Hz. If unspecified, the frequency will default to 10 Hz.
- coherent dedispersion -- By default, coherent dedispersion will be applied using the dispersion measure (DM) specified in the par file. It will be disabled for "CAL" folding. The maximum DM allowed is not currently well characterized; it will depend on the total bandwidth, channel bandwidth, and observing frequency.
The data are recorded in fold-mode PSRFITS format, using PSRCHIVE's standard 16-bit data encoding. For typical observing parameters, the resulting total data rate is generally small (~few MB/s). The output data rate can be calculated (in bytes per second) as 2 * Nsubband
. If parameters are chosen so that this rate exceeds ~50 MB/s, the configuration should be tested in advance.
Folded (aka binned) visibilities.
Although not specifically a pulsar mode, for completeness we note that it is possible for WIDAR to produce visibilities with integration times as short as 5 ms (this is also known as "FRB mode"). This mode may also be useful for observations of slow pulsars. Currently, a well-tested configuration exists for 5 ms integrations, 256 MHz total bandwidth (2 subbands), 256 channels, and 2 polarization products. Variations on this that produce similar total output data rate may be possible, but will require additional testing prior to use.