Jones matrices for beamforming

The purpose of this feature is to clarify the process of polarization
calibration in the phased-array beamformers for pulsar observations
(especially high-accuracy timing).  At present, there is a rather
vague requirement for Jones matrices to be provided by SDP to CSP for
use in beam-forming.

Starting point is the memo "Real-time Polarization Calibration in the
Phased-array Beamformers" (van Straten 2015), attached to this ticket as memo.pdf.  Specific questions
(probably with different answers for MID and LOW):

1. Requirements for calibration measurements during the observation:
are Jones matrix elements supposed to come from pre-existing
calibrations (boresight and/or voltage beams) or is it anticipated
that calibration observations will be simultaneous or interleaved with
pulsar observations?

2. How many beams? For example, Van Straten (2015) suggests one high-precision
timing beam, on boresight, with correction before summation, for MID.

3. Calibration before or after summation, or hybrid?

4. Required update rates.  Correction for instrumental response on
scales smaller than the bandwidth: this requires matrix convolution -
how many coefficients?

5. Inclusion of RFI masking: (how) is this done?

6. Assume that whoever calculates the Jones matrices does not need to include
parallactic angle rotation (I think that this must be the case, as leakages should be fixed in the dish.station frame: confirm (otherwise there are implications for the update rate).

Some work has been done by Grant Hampson and CSP, specifically to assess the data rate for LOW (attached as LMC_datarates.pptx).