[time-nuts] HP5370B anomaly

[Magnus Danielson]

Hi!

On 12/27/2010 06:35 AM, Bruce Griffiths wrote:
>> Considering that:
>>
>> 1) I've measured the same signals on similar or higher performance
>> counters and not seen it with those units.
>>
>> 2) The frequency difference is so low that I do not experience a
>> phase-wrap for the 2476 s long measurement run.
>>
>> 3) The phase slope is much lower than the amplitude of the
>> oscillation, so the linearity error can be ruled out completely.
> Why?
> What about a periodic effect due to crosstalk at the interpolator mixer
> FFs inputs?

The noise seems independent of the slope of the signal, but do depend on 
how I set up the input signals. The three last plots are interesting...

http://rubidium.dyndns.org/~magnus/time/hp/hp5370/HP5370B-inputwander.png

SEP plot:

1 Hz to EXT ARM from TADD-2
5 MHz to the START channel
5 MHz to the STOP channel
http://rubidium.dyndns.org/~magnus/time/hp/hp5370/HP5370B-4.tim

COM plot:
Same as above, but changed the setup from SEP to COM.
http://rubidium.dyndns.org/~magnus/time/hp/hp5370/HP5370B-5.tim

COM no STOP plot:
Same as above, but with the STOP channel disconnected.
http://rubidium.dyndns.org/~magnus/time/hp/hp5370/HP5370B-6.tim

So, I *do* think that there is some disturbance being injected into the 
signal-path, but I do not think it is normal non-linearity which would 
depend on rate of phase. This disturbance is however shifting with what 
signals I have hooked up, even if they are not used.

Looking up on details, the LED detection on A22 is clocked by a 10 Hz 
coming from the A11 board where a RC-based oscillator generates it, so 
it's frequency is separate from any other. And notice how my 10 Hz 
sampling experiment does not give much difference than the 1 Hz sampling 
rate.

So, I suspect the noise I see is the beat frequency between the 10 Hz 
oscillator and my 1 Hz or 10 Hz signal.

The use of ECL FF for LED sampling isn't quite the best choice... sharp 
edges...perfect for injecting noise.

>>
>> 4) The amplitude is stable and does not change with the slope.
>>
>> 5) The unit does not have the 5 MHz noise as I disabled that "feature".
>>
>> I do howver realize that the START and STOP channels blinks at a
>> fairly high rate. These LEDs are driven with signals coming from the
>> A22 board, where all the critical timing also passes by... will try to
>> figure out a way to handle it, but right now it severely degrades the
>> performance of my HP5370B.
>>
>> I think I discovered the effect at least, with a suspect mechanism for
>> it.
> You suspect the LED drive signal?

Or some form related signal. I suspect the LED detection...

>>
>> It would be nice if others would see if they could achieve the same
>> effect.
>>
>> Cheers,
>> Magnus
>>
>>
>
> What is the amplitude of the effect?
> How good a source pair is required to see it?

Some 200 ps or so in amplitude...

It degrades the performance of the HP5370B to the level of my HP5372A. 
Even when setting up the less useful case it falls down a little.

See how the same sources behaves when tossed to a DTS-2070C.

Cheers,
Magnus