[evlatests] Y27 switched power signal, timing

[Walter Brisken]

Hi EVLA tests,

We are struggling to find VLBI fringes to Y27 in experiment TY023.  This 
experiment is thought to be nearly identical to the previous TY022 which 
did show fringes.  The most obvious question to ask is if the timing good. 
Vivek and the analysts have exhausted timing searches via correlation with 
high spectral resolution to +- 60 microseconds.

Examination of the timing of the switched power signals offer a test of 
relative timing within the VLA between 10s of microseconds and the 
switching period (0.1 seconds).  Ken Sowinski tells me that the switched 
power signal is synchronized with the 1 PPS signal.  On the VLBA such 
tests have proved useful to find timing problems local to the digital 
electronics.

I have folded (pulsar style) the total power of 20 seconds of data 
recorded on the Mark5 unit into 4000 time bins.  This allows for searching 
of timing offsets between 25 microsecods and 100 milliseconds.  Attached 
is a plot showing _very_ good time alignment.  In red is the raw folded 
data with a 25 microsecond time resolution.  The transition from the high 
power state to the lower power state occurs exactly where desired.  If 
there is a timing error preventing fringes it must be either global at the 
VLA site (equally affecting the analog systems and WIDAR, including the 
VLBI formatting), or it must be very near an integer multiple of the 
switching period (0.1 second).


Unrelated to the search for hints in a possible timing offset problem the 
shape of the switched power signal comes out.  The actual waveform 
differs from a pure square wave.  The green trace on the plot is a binned 
version of the red trace.  Both show that the first 30 ms after each 
transition there is a transient response.  Probably of no consequence, 
though this differs from the nearly perfect square waves seen at the VLBA.


A final observation is that the total power is low.  The power axis is 
calibrated so that optimal 2-bit quantization would lead to a power level 
of 1.0; a power of ~0.85 is observed.  This low power is observed in both 
channels investigated.  The effect of this would be an unnecessary 
sensitivity loss of about 1%.


Where to go from here? Since this timing test only demonstrates internal 
timing consistency within the VLA we still cannot exclude side-wide timing 
errors greater than 60 microseconds.  Are there any ways to independently 
measure the VLA time to 10s of microsecond precision (e.g., comparison 
with GPS)?


 	-Walter
-------------- next part --------------
A non-text attachment was scrubbed...
Name: y27.switched_power.pdf
Type: application/pdf
Size: 29945 bytes
Desc: 
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20121215/fb715626/attachment.pdf>