Given the correlator modes discussed above and listed in Appendix A, especially 4 IF mode, there is a wide range of possible observing strategies. The following configuration is popular: tune one IF pair, AC, to one central frequency (or velocity), and the other IF pair, BD, to a frequency which is offset. This offset can have any value as long as the second frequency falls within the same observing band (e.g., L-band) as the first and is not farther apart than 500 MHz (which is set by the 4.5-5 GHz range into which most signals are initially mixed). This arrangement is useful when measuring, for example, two of the lines of OH in both polarizations, right hand circularly polarized for IFs A and B, and left hand for C and D.2.6
An alternative use of the ``4 mode'' is to choose the central frequencies of each IF pair in such a way that the two bands are set adjacent to each other, partially overlapping to cover the edge channels. In that case one can gain nearly a factor of 2 greater bandwidth while preserving the frequency resolution. When observing in this mode it is important to do a careful bandpass calibration (cf. 3.2) to ensure a seamless match when combining the data from both IF pairs. Observers of recombination and weak absorption lines should avoid this strategy because of the uncertainties in generating such a seamless match. Similarly, if a very accurate continuum subtraction is an issue, this option should not be used.
It is also possible to specify a wide bandwidth for IFs A and C (e.g.,
6.25 MHz) to cover the redshifts of several galaxies in a cluster
while IFs B and D use a higher frequency resolution over a narrower
bandwidth (e.g., 1.56 MHz) to focus on one cluster member. Although it
is even possible to mix different bandwidths within one IF pair this
is not recommended. The main reason is that the
fringe stopping is done for the frequencies to which IFs A and D are
tuned. If the frequencies at IFs A and C (or B and D) are not equal
there is a residual fringe frequency, 
-
, which is
. For example, when observing IF A with a
1.56 MHz bandwidth and IF C with a 3.125 MHz bandwidth, the frequency
used in calculating the fringe rate is in error by 0.758 MHz. This
effect can cause decorrelation within an integration period when
dealing with long baselines and/or large differences in bandwidth
which cannot be corrected for. At the least it causes phase winding in
one of the IFs involved. AIPS task CLCOR contains an option which
calculates the phase corrections for IFs B and C after which the
standard complex gain calibration can be applied.
There is a second, more subtle, effect of which observers contemplating this mode should be aware. This was reported by Mehringer (1992, VLA Test Memo No. 161). The VLA beams for the two polarizations have a slightly different pointing center, an effect known as beam squint. Observations made with different bandwidths for each IF have thus different pointing centers. This can lead to errors when attempting to use line-free channels taken from an IF which has a broad bandwidth (for example IF C which measures left hand polarization) from a more narrow band (IF A, which records right hand polarization).