[time-nuts] Frequency standards for different tau in Allen Dev measurement

[Bob kb8tq]

Hi

Backing up a bit:

If you are going to measure something like a crystal oscillator against a PPS, you need
to be careful of phase slip. The oscillator may be off frequency by many ns/s. Working out
what happens as it crosses a 100 ns boundary is not as easy as one might think. In addition,
counters tend to get a bit confused (in TI mode) when two signals happen at exactly the
same time. That confusion leads to data errors.

The “normal” answer is to divide the device output down to a frequency that is low enough
that it is unlikely to “slip” over the course of the run. 100 Hz is what JPL seems to like to
use. Just like the PPS, the divided out also needs to be a square wave. Slow sine wave
edges create problems with counters.

=====

One good thing about a very short period on one input: The time base of the counter really
drops out. If all you are measuring is 10 ps out of 100 ns, an error of < 20 ppm in the counter
is “adequate” ( 100 ppm resolution / 5 = 20). Even running on a bare crystal, the counter
probably would hit this sort of number. With an OCXO in the counter, it very much does 
not have any impact on the readings. 

=====

Like all test gear, the 5370 hits its specs with proper care and feeding. There is a check 
procedure to see how it’s doing. There are lots of tweaks and steps involved in aligning 
it to hit the proper specs if it turns out to need some help. It is *very* common to find 
that a 5370 needs some help after a few years ….

Lots of fun !!!!

Bob

> On Mar 1, 2020, at 12:19 AM, Taka Kamiya via time-nuts <time-nuts at lists.febo.com> wrote:
> 
> I took several Adev measurements.  I would love to get senior time-nuts' opinion. 
> 
> Retrace seems to be good, but results overall seems to be off.  As far as I know, they all should be 1 order of magnitude better. (hence line 5 experiment)
> 
> Setup:
> They are all done with HP5370 feeding TimeLab every one second into Channel A.  It comes from GPSTM (modified).  External reference to HP5370 is Efratom PRFS-102.  Output is every 1 second.  Trigger point was carefully set.  That's common on every measurement.
> 1)  Dark Blue lineThis is a sort of "reference" line.  DUT is HP105B 5MHz output.  TimeLab is set to FREQUENCY mode.
> 2)  Pink lineThis is HP105B 5MHz output in INTERVAL measurement.  Channel A is connected to GPSTM (T-bolt) 1 pps line
> 3)  Green line  
> This is PRS-10 #1 10MHz output in INTERVAL measurement.  Channel A is connected to GPSTM (T-bolt) 1 pps line
> 
> 4)  Red lineThis is PRS-10 #2 10MHz output in INTERVAL measurement.  Channel A is connected to GPSTM (T-bolt) 1 pps line
> 
> 5)  Light Blue lineSame as #4 above except 1 pps is coming from FT/4040A through TAPR 1 pps converter unit
> Questions:
> I realize there are two "references" here.  One is a reference going into HP5370 external reference line.  Another is what's connected to channel A. (1 pps)  Right now, I have independent supplies.  I'm thinking I can use one 10MHz reference, split the line, one goes to ext ref, another goes to TAPR converter then goes to channel A.  My fear is, everything will now be in LOCK-STEP that I'd be manufacturing a result.  Is this allowed??
> Blue vs pink:  everything is the same except for measuring method.  Blue line is frequency.  Pink line is T.I.  Seems blue line is getting a better result, except it has 60Hz interference.  Does this mean, even with this counter (which is not known for accuracy/resolution in frequency mode)  frequency method is superior?  Or is this averaging going on that the result is artificially better?  (like HP53132A)?
> 
> Another weirdness is, TI reading has very few digits.  XX.XX nS is the most I'd get. The counter's minimum resolution is 20pS, so the LSD is correct.  The period of 10MHz is 100nS.  So, MSD, too, makes sense.  But this is limiting dynamic range.  Would that mean input signal from DUT is too high?  Or, inversely, 1 pps into channel A is too low?  What can I do?  (other than spending $100K)  
> 
> For this measurement, I didn't give FT/4040A enough warm up time.  It has been on for about 2 hours.  I usually give much more warm up time.  But at the same token, I do not want to keep this on, nor wait 2 days before starting measurements.
> (I am accepting donation of HP5065A)
> 
> I used the best cable I have....  per advice of Bob.
> --------------------------------------- 
> (Mr.) Taka Kamiya
> KB4EMF / ex JF2DKG
> 
> 
> 
> <IMG_3968.jpg><adevTest.png>_______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.