This is about a study of the daisychain of three synthesizers in the PDEV spectrometers.

These are run from a single 10MHz reference which is daisychained through all three.

The chain is as follows (at the start of this investigation):

Source: Ref distro amplifier in rack 2-9 in the radar equipment room.
1st PDEV synthesizer in chain:  "A/D clock"
2nd PDEV synthesizer in chain:  "L.O. High"
3rd PDEV synthesizer in chain:  "L.O. Low"
Phase glitch measurement system: from Ref Out on "L.O. Low" synthesizer

There has been some history of unexplained error messages on the last two units in the
chain, complaining about loss of reference lock.

The data discussed herein was taken with Dana's phase comparison rig, comprising:

> A 10MHz quadrature demodulator comprising MCL coaxial splitters & mixers
> A dual differential to single-ended low noise baseband amplifier
> A COTS DAQ system, sampling both I & Q at 100 Hz
> Software written by Dana to analyze the results

When setting up this apparatus in a new situation, I make a test record in which
I temporarily interrupt the signal input to measure DC offsets in the two mixers.
I then record these values for use in subsequent measurements with the given setup.

This system measures the phase as ATAN2(Q, I), then optionally (and usually)
unwraps the phase.  This approach avoids the need for some kind of mechanical
phase shifter (hard to realize at 10 MHz) in one of the signal paths to get a
90 deg phase difference between the two inputs as required by a single mixer
used as a phase detector.  Also this approach allows tracking of phase drifts
over many full cycles, as opposed to only about one radian with a single phase
detector. 


03 Oct 2013:

The first dataset was started late yesterday afternoon and was allowed to continue
until early this morning, mostly as a check that the equipment setup was working
correctly.  The raw data file was named 'trial-0.dat'.

The first stage of processing was done with 'daq4phdf.exe' to select the time range
and processing bandwidth under consideration and to generate two different 2nd stage
files as follows:

a) trial-0A.bin (0.1 sec/SA; length = 344251):  somewhat narrow band
b) trial-0B.bin (0.01 sec/SA; length = 5606381):  much wider bandwidth, taken to
   better show details of the big glitch.

These files were then viewed using 'cansplzz.exe' for final interpretation.  

'trial-0A' revealed that the glitch was not properly resolved at this bandwidth,
but did yield low noise results for general phase wandering.  It did show a
possible "aftershock" of the glitch, although is is not certain.

'trial-0B' showed the glitch itself as well resolved; indeed there appeared to be
two slightly separated components.  While the S/N was not as good here, the same
general phase drift could be seen, including the possible "aftershock" following
the big glitch.

Interestingly, there were no errors generated by any of the PDEV synthesizers 
during the collection of this record.

I will now commence further recording in a series of files up to perhaps a couple
of days each, in the hopes of catching one of the events that throw errors in the
synthesizers.  (new recording started ~09:40 Thurs 03 Oct 2013)


05 Oct 2013:

The file that was started at ~09:40, 03 Oct, was terminated at 08:21, 05 Oct.
It is named 'event_1.dat', and was captured because sometime during 04 Oct
the two trailing sig gens recorded "ext ref lock failure" error(s).  The
event(s) probably occurred between mid-late morning and mid-afternoon, possibly
during a thunderstorm around lunchtime on Friday.

Analysis of the data shows several events, of generally similar character:
> An initial severe phase disturbance, followed by
> An interval of about 1 second with a fairly steady frequency offset of about
  0.16 Hz, followed by
> Recovery of phase lock to the original phase (with or without a cycle slip).

These are best seen in 'event_1b.bin' using 'consplzz.exe'.

event_1a.bin:  0.01 sec/SA, 1263444 Sa  (too narrowband)
event_1b.bin:  0.01 sec/SA, 1432077 Sa  (wideband, excellent view)  

Started a new data collection at 08:29  05 Oct 2013 (Sat morning).