[evlatests] L302 and phase spins.

[vdhawan at aoc.nrao.edu]

		PHASE SPINS on L302

Data of October 27, UT 05:30 to 08:30. X-band.

1. Continuum data at 0.4s integration
	Short scan ~2min on 3C48, long scan ~15min 3C84, the source pair
	repeated 4 times plus part of a 5th. 

	Phase spins at:

	ant 14, IF A, C. First 4 long scans spin, 5th does not - 
			 the L302 has recovered. 

			Short scans do not spin. 
			B, D IFs never spin.

	ant 16, IFs A, B, C, D spin in the first 4 long scans,
				not the 5th.

			Short scans do not spin.


2. In between each ~20min continuum cycle was a ~20min cycle, 
	same pattern, line mode, 3s integrations, same 2 sources,
	5 cycles plus part of a 6th.

	Phase spins at 

	ant 14, IF A, C; on 6th long scan only

	ant 16, IF A, B, C, D. 4th and 6th long scans.

The details above show that the phase spins are absent in the line
data but present in the continuum data interspersed in time.

In addition to the spins, there are seemingly random offsets in phase 
when an antenna returns to a source, on most IFs & most scans.

The rate of phase change (fringe frequency or fringe rate)
was comparable to the natural rate from changing geometry,
so I thought it may be due to timing errors or mis-application 
of the geometry model. 

So, I

a. Attempted to extract L302 DDS frequencies from the monitor logs.
   These are not accurate to the (sub-Hz) level required for
   this particular goose chase.

b. Commands are not echoed to the log, so no luck checking what
	the L302 thought it was doing.

c. Finally, I took the geometric delay (TAU_geo) column in the 
   AIPS CL table, for antenna/source. Differentiated that to get 
   TAU_dot, the rate of change, in sec/sec. The 'natural' fringe rate 
   from Earth rotation for an antenna on a given source is FR = 
   (Tau_dot * F), F=8.4GHz. 

   The FR on a given antenna, (when it spins at all) matches
   (FR_1 - FR_2) the difference in FR for the 2 sources,

	i.e.,

   The spin is due to the L302 not updating the model for the 
   (long) scan after it is done with the short scan. The rates
   on the short scans are so slowly changing that a missed update
   just shows up as a phase offset with hardly any rate, when that
   source is re-visited 20 min later.  

   The antenna is correctly pointed at the desired source - it is the 
   fringe tracking that is not happening for that source.

   If my verbal handwaving is unconvincing, I have plots to wave - 
   come by or drop an email.

   Some questions:

   How exactly does the L302 gets confused? (Ken S. suggested the FPGA
   that gets the updates might be confused); 

   Why does it recover without reboot?

   Why does the line data appear normal while the continuum data is hit by
   L302 not updating? 

   Why is this not seen in baselines runs where many sources are
   observed in short order? (perhaps a recent breakage?)

   One suggestion is to try a test where a few sources are cycled over, like
   the test of Oct 27.

Vivek.