[time-nuts] Help me make some sense of adev measurements of SR620's own clock

[Dr. David Kirkby (Kirkby Microwave Ltd)]

After sorting out some GPIB issues, I finally got to be able to make
some measurements on my Stanford Research SR620 time-interval counter.
I thought it sensible to first try to determine the performance of the
counter, which is using its own high stability clock (option 001). So
no external reference, such as one derived from GPS, is used.

I took the 10 MHz reference output from the SR620 via a cable about
0.5 m to the A input of the counter, which also had a BNC T on the
counter. From the A input to the B input is a cable about 3.6 m long
(longer than I would have liked with hindsight). I then measured the
time-interval every second for 55488 seconds - it is actually still
collecting data. The data file is here.

http://www.kirkbymicrowave.co.uk/time-stuff/ref-out-to-A-3.6m-cable-to-B-rev4.zip

The format should be pretty self-explanatory. Note the counter sample
size is 1000, so it takes 1 sec. Note A is an high impedance input,
and B 50 Ohms, which seems a logical choice if tapping off a bit from
a 50 Ohm cable for the A input.

# Data collected with ./tic version 0.01
# GPIB address 17 on host 'buzzard'
# Data collection started at: 23:2:55 GMT on 24/01/2015 (day/month/year format)
# Instrument settings are as follows:
# Sample size: 1E3
# Trigger level (external): -0.21 V
# Trigger level (A): -0.01 V
# Trigger level (B): -0.01 V
# Coupling (A): AC
# Coupling (B): AC
# Termination (external): 1000000 Ohm
# Termination (A): 1000000 Ohm
# Termination (B): 50 Ohm
# Mode = Time
# Column 1 is the time from the SR620 in seconds
# Column 2 is a hash(#) character, used to denote a comment
# Column 3 is the delay in seconds since data was first collected
1.7540E-8 # 0.000000 s
1.7538E-8 # 0.988158 s
1.7538E-8 # 1.976571 s
1.7538E-8 # 2.964327 s

The times recorded, about 17.5 ns, are consistent with what one would
expect with a cable about the length I have.

I then used Tom's "adev1"

http://www.leapsecond.com/tools/adev1.htm
http://www.leapsecond.com/tools/adev1.c

to analyze the data.

drkirkby at buzzard:/tmp$ adev1 1 < ref-out-to-A-3.6m-cable-to-B.txt

** Sampling period: 1 s
** Phase data scale factor: 1.000e+00
** Total phase samples: 56165
** Normal and Overlapping Allan deviation:

       1 tau, 1.0066e-12 adev(n=56163),     1.0066e-12 oadev(n=56163)
       2 tau, 5.2611e-13 adev(n=28081),     5.2820e-13 oadev(n=56161)
       5 tau, 2.4461e-13 adev(n=11231),     2.4397e-13 oadev(n=56155)
      10 tau, 1.5235e-13 adev(n=5615),      1.5188e-13 oadev(n=56145)
      20 tau, 9.8477e-14 adev(n=2807),      9.9323e-14 oadev(n=56125)
      50 tau, 5.7764e-14 adev(n=1122),      5.9520e-14 oadev(n=56065)
     100 tau, 4.1609e-14 adev(n=560),       4.2643e-14 oadev(n=55965)
     200 tau, 2.7712e-14 adev(n=279),       2.8362e-14 oadev(n=55765)
     500 tau, 8.1848e-15 adev(n=111),       9.3519e-15 oadev(n=55165)
    1000 tau, 4.9553e-15 adev(n=55),        5.2360e-15 oadev(n=54165)
    2000 tau, 3.3500e-15 adev(n=27),        3.0661e-15 oadev(n=52165)
    5000 tau, 1.8873e-15 adev(n=10),        1.4325e-15 oadev(n=46165)
   10000 tau, 8.6819e-16 adev(n=4),         8.6732e-16 oadev(n=36165)
   20000 tau, 1.4849e-15 adev(n=1),         6.3165e-16 oadev(n=16165)

I'm puzzled about, is how to interpret this, and if interpretation is
correct, my counter might not be in spec.

I thought from reading Wikipedia and

en.wikipedia.org/wiki/Allan_variance

that at 1 tau, the Allen Deviation represents the RMS deviation
between two observations 1 second apart.  So that is 1.0066 ps.

The SR620 counter's display has resolution of 1 ps, and supposedly a
25 ps rms single short resolution. Would I be right in assuming that
after 1 second (1000 samples), I would expect to see an adev of
25e-12/sqrt(1000) = 8e-13, suggesting my counter is not achieving the
25 ps rms resolution, but rather sqrt(1000)*1.0066e-12=31.8 ps? I've
run Autocal on this counter, and put the oscillator on frequency with
one of the calbytes, but have not done any other adjustments. Needless
to say its an eBay purchase, and I doubt has been near a cal lab in
years.

Again based on a 25 ps rms resolution, would I expect after 500
seconds (50,000 counts), to see an adev of 25e-12/sqrt(50000)=1.118 x
10^-13, rather than the 8.1848e-15 the data shows?

Also, why would the adev rise at 20000 tau, when this is only
measuring the time between its own reference, and a version delayed by
about 17.5 ns due to a few metres of cable? But maybe there's not
really enough data at 20000 seconds.

Do most people save time information as I have done there, or phase
information? I'm guessing the two are easily related, but I'm
wondering what will work with most peoples software. What I like about
Tom's is it compiles easily on my Unix box, without me having to use
Windows. But I note some of Tom's software wants phase, and the other
time.

I'm still collecting data, so will at some point upload a larger file
to http://www.kirkbymicrowave.co.uk/time-stuff/ with more data. I'll
certainly keep this going for 24 hours, unless I am doing something
stupid and should abort now.

Dave