Correlator Setup

The correlator configuration depends entirely on the science goal. For continuum science, choose the widest bandwidth per subband with coarse spectral resolution; typically the NRAO default settings (possibly retuned to alternative center frequencies) are sufficient. For spectral line, choose the subband bandwidth and the spectral resolution that best fit the scientific objectives of your project. For pulsar observations, spectral resolution and pulse phase (bin) resolution should generally be chosen to maximize bandwidth while still resolving the pulse; your project may have other specific requirements depending on the goals. Detailed information on the correlator is available in the WIDAR section of the VLA OSS.

Issues that should be kept in mind are:

General and Spectral Line Considerations:

The widest (128 MHz) subbands in a baseband do not overlap. Additionally, a few channels may need to be flagged at either subband edge because of the higher noise due to filter roll-off. If the science goal requires a homogeneous sensitivity sampling over multiple 128 MHz subbands, we recommend tuning the second baseband at a frequency that is offset by a fraction of a subband width with respect to the first baseband. This, however, removes the possibility to place the second baseband freely in the receiver band to do other science. Also, anywhere from 8 MHz to up to 30 MHz at the edges of the basebands may be noisier, so you should not rely on a spectral line that would be close to a baseband edge. For more details, see the Subband 0 subsection of the Spectral Line section in the Guide to Observing with the VLA.

As described below, each subband must be calibrated independently. Therefore, in the case of a narrow subband bandwidth, choose flux density and complex gain calibrators that are strong enough for adequate S/N at this subband bandwidth. For more information about successful calibration, see the Exposure and Overhead in this guide, and the Calibration section in the Guide to Observing with the VLA.

It may be necessary to Hanning smooth your data in order to get rid of Gibbs ringing (for the theory behind this phenomenon see Gibbs phenomenon). Lower frequency bands (X and below) are prone to strong Radio Frequency Interference (RFI); flagging the RFI could be close to impossible unless you first Hanning smooth your data. This necessity should be taken into account when choosing the spectral resolution of your proposed observations, since the effective resolution will be lower than the original (pre-Hanning smoothed), even though the number of channels will stay the same. Note that the frequency resolution (FWHM) of un-tapered spectra is 1.2×Δν (where $Δ ν$ is the channel spacing) and the resolution of Hanning-tapered spectra is 2.0×Δν .

For spectral line observations, given an expected line width, it is a good idea to select a spectral resolution that will allow for at least 4–5 channels across your line, or twice that many when Hanning smoothing. There are a number of tools available online to identify molecular line rest frequencies such as the Lovas Catalog and Splatalogue. The frequency range covered by ~10 to 50 GHz (X to Q-band) contains a large number of diagnostically interesting atomic and molecular transitions. For continuum data only, it is wise to check whether the chosen frequency range contains potentially strong spectral lines.
Any correlator configuration should stay within the allowed data rate limits for GO and SRO observing. This might become an issue with complex correlator setups using recirculation and/or short integration times. RCT-proposing will report the data rate for a specific correlator setup.
Spectral line correlator configurations are typically set up with the Resource Catalog Tool for proposing (RCT-proposing) (see next section) in the proposal stage (PST) and with the Resource Catalog Tool (RCT-observing) in the OPT for the actual observations.

Pulsar Observations:

There are two different pulsar observing modes available at the VLA: phased-array (also called "YUPPI" mode) and phase-binned imaging mode.

In the phased-array mode, the individual antenna data streams are summed coherently in real-time to produce high time resolution data on a single sky pixel. Coherent de-dispersion can optionally be applied, and the data can be written out either as high-time-resolution spectra, or folded modulo a known pulsar timing ephemeris. See the Pulsar Observing section of the OSS for more details.
In phase-binned imaging, the antenna cross-correlations (visibilities) are averaged into different pulse phase ranges using a known pulsar timing ephemeris. This gives data for the entire field of view, but generally with much worse time resolution than is available in phased-array mode.

For phase-binned pulsar observations, there are constraints and trade-offs between the total number of subbands and the bin width. There is also an output data rate limit that will affect the maximum number of bins allowed for a given dump time and number of channels. These issues are described in more detail in the Pulsar Observing section of the OSS, under the "Gated or binned visibilities" heading. The Resource section of the PST will prompt for the relevant information and compute the data rate for pulsar binning modes.
In general for pulsar binning observations, the bin width (via number of bins) should be chosen to be comparable to or smaller than the pulse width. The spectral resolution (via subband bandwidth) should ideally be chosen so that dispersive smearing within a channel is comparable to or smaller than the bin width. Dispersive smearing can be computed as Δt_DM = 8.3μs x Δν(MHz) x DM(pc cm^-3) / ν(GHz)³.
Dump time for pulsar binning should generally follow standard guidelines for interferometric imaging (see for example the Field of View OSS section). Binning observations require the target source to have a known timing ephemeris, in TEMPO-compatible format. The ephemeris does not need to include absolute phase information, but it should be accurate enough so that the pulse phase drift (due to any apparent pulse period error) within a dump time is much smaller than the bin width.