Recirculation vs Baseline Board Pairs

Creating a spectral line resource is similar to creating wide band resources as outlined in the previous section, except for the more advanced specification of the subbands and subband frequency tuning and possible Doppler setting of the frequencies.  Eventually, resources with the requested correlator settings will be pre-filled from information submitted to the PST during the observing time allocation procedure.  However, at this stage none of this has been implemented.

For how and when to use this observing mode, refer to the Spectral Line section of the VLA Observing Guide.

The NRAO Defaults resource catalog contains full polarization dual IF pair spectra-polarimetry resources (the wide band continuum resources). If they appeal to you, you can copy/paste them in a personal catalog just as for the wide band resources above and edit them as needed. Check the spectral line resource properties very carefully as the spectral line resources in the NRAO Defaults have a fixed sky frequency whereas you probably want to use a rest frequency in combination with Doppler setting.  Most likely, however, you will opt to create your own resource, just like creating a wide band resource previously.  Some items that need extra attention are described below, but first a small detour to outline the options for creating large numbers of narrow frequency channels.

Recirculation vs. Baseline Board Stacking

Spectral line observations are typically constrained by the requirement to have the best spectral resolution (i.e., narrowest spectral channel width) combined with the best velocity coverage (i.e., widest observing bandwidth), the latter perhaps also for calibration purposes (as gain calibration is done per subband).

The maximum subband bandwidth is 128 MHz and can be decreased by factors of 2 to 31.25 kHz(*). Total bandwidths wider than 128 MHz are achieved by placing subsequent bandwidths next to each other with the caveat that the few channels next to a subband edge should be considered lost for line work (on either side of the boundary, for continuum work as well). To obtain a contiguous bandwidth which is near-homogeneous in sensitivity and without any subband boundary gaps, another baseband would be placed some MHz offset (e.g., half a subband bandwidth) from the original baseband to enable subband stitching in post-processing. Note that this limits the number of available basebands for other line settings.

A single Baseline Board Pair (BlBP), out of 64 available BlBPs, can handle 256 spectral points divided over the polarization products (polProd). That is, it can deliver 256 spectral channels in single polarization, 128 spectral channels in dual polarization, or 64 spectral channels in full polarization. The channel width (which is slightly less than the spectral resolution) is then simply the subband width divided by the number of spectral channels.

With one BlBP per subband as standard, the selected subband bandwidth is thus limited to 256/polProd channels which may not be narrow enough to achieve the desired spectral resolution. There are two ways to overcome this limitation: Baseline Board Stacking uses more of the limited amount of hardware and Recirculation uses limiting the total bandwidth per subband and CPU cycles. Both have their disadvantages and the choice depends on the science requirements. If either can be used, we recommend using Recirculation.

Baseline Board Stacking uses additional BlBPs to compute extra channels in a subband (of any width up to 128 MHz). Each additional BlBP increases the number of spectral channels by 256/polProd. As there are only 64 BlBPs total, and as every subband uses a minimum of one BlBP, Baseline Board Stacking reduces the number of subbands that can be observed to less than 64, and in the extreme case to a single subband of 128 MHz or less. When most of the 64 BlBPs are being used and all of the subbands are required (instead of some being recognized as less important and thus "desired" versus "required"), the observations will not take place if one or more BlBPs are inoperable. The 3-bit NRAO default setups use 64 BlBPs with the subbands at the baseband edges "desired" to allow continuation of operations when not all BlBPs are available.

The Recirculation option uses software to compute extra channels. For this, CPU cycles are "freed up" by limiting the subband bandwidth fed to the BlBP to less than 128 MHz to obtain more lags (in factors of two), running the data through the board for a second, third, etc., time; hence Recirculation.  As subband bandwidths (and CPU cycles needed to process them) can be decreased by factors of 2, each halving thus allows a doubling of the number of channels in the subband, currently up to a factor of 64. This does not require additional BlBP hardware and thus retains the possibility of using all subbands, albeit at less subband and less total bandwidth. Currently, subbands of 128 MHz must use Baseline Board Stacking to achieve more than 256/polProd channels.

Baseline Board Stacking and Recirculation can be used simultaneously in the same subband (if less than 128 MHz), and configurations with multiple subbands configured with either or both are allowed. Note that here a correlator setup can still request all 64 BlBPs and thus designating some subbands as "desired" is still highly recommended, but Recirculation gives the option to use less than 64 to achieve the number of spectral channels.  However, requesting a large number of channels, whether or not with Baseline Board Stacking and/or Recirculation, yields higher data rates than normal with the default integration times. To remain within the limits set by the observatory, longer integration may be needed which has an impact on time averaging smearing in the larger array configurations and thus on the field of view.

The choice for one or the other, or even for less channels than anticipated, depends on the trade-offs that can be made for the science goals and remain a responsibility of the observer.

(*) When observing at such narrow subband bandwidths it is good to check with the NRAO Science helpdesk. There are other operational constraints, in particular the F-shift in bringing down the baseband, that need to be considered and that may decrease the usable subband bandwidth to much more than a (symmetric) few channels compared to the general case described above.