[time-nuts] SR-620 noise floor question

[Matthias Jelen]

Dear Time-Nuts,

I played around with my SR-620 counter and TimeLab and got 
strange ADEV curves, so I started to investigate a bit.
The SR-620 is equipped with Option 1, the high performance 
Wenzel OCXO - specified with 5E-12 at tau = 1s.

I tried to evaluate the counters noisefloor. I fed 10 MHz 
signal to the start channel and, via a BNC Tee, the same 
signal to the stop channel. A 1 pps pulse does the arming 
once a second.

I tried 3 diffferent combinations - please see the attached 
TimeLab plot:

1. 10 MHz from a Trimble TB to start and stop, reference 
internal (blue line)
2. 10 MHz from the counters REF OUT to start and stop, 
reference internal (pink line)
3. 10 MHz from Trimble TB to REF IN, start and stop, 
reference external (green line)

All clocks should be abt. one magnitude better than the 
noise floor which I would expect from a TIC with 25 ps 
resolution, so I wouldn´t expect any difference between 
those 3 curves, but only the measurement in which the 
internal OCXO was not used shows the expected performance 
with abt. 2E-12 at 1 second. The measurement which used only 
the internal clock was abt. one magnitude worse, and the 
adev curve shows some ripple in the 8 to 40 seconds range.

I had a look at the ref out with a spectrum analyzer, but 
the smallest resolution bandwith I have available at home is 
6 Hz - no spurs noticeable in that region.

Is it reasonable to argue that the oven is out of spec? Did 
anyone on the list encounter a similiar effect with a SR-620 
before? Or am I simply overlooking something?

Thanks a lot for any comments and have a nice weekend,

Matthias
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SR620.png
Type: image/png
Size: 86801 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20150620/3f2cca46/attachment.png>