[time-nuts] Re: Problems and solution when using a Rigol DDS signal generator for counter testing

[Magnus Danielson]

Hi Erik,

On 2022-08-19 10:32, Erik Kaashoek via time-nuts wrote:
> During the tests of a frequency counter using a Rigol DG812 signal 
> generator locked to an external 10MHz clock it was observed that the 
> output frequency of the DG812 was not stable
> This can clearly be seen in [1], blue trace, where the counter was 
> warming up and the frequency of the frequency variations changed over 
> time.
> To test if the external locking was related to the instability the 
> external lock was disabled which lead to huge instability. [2] Pink trace
> As can be seen in the pink trace the huge variations disappeared when 
> the coax cable with the 10MHz reference clock was disconnected (at 48 
> seconds) and appeared again when the cable was reconnected to the 
> clock input (at 86 seconds), regardless of the Rigol running the whole 
> time on its internal clock.
> This seemed to hint at some interference between the external clock 
> signal and the internal frequency reference happening when running 
> both at internal or external clock.
> To test this an  external clock signal was applied at the frequency 
> (+/- 0.01 Hz ) of the internal clock.
> The instability of the output from the Rigol locked to an external 
> clock close to is internal frequency disappeared completely ([1], red 
> trace) and the remaining frequency variations where within the 
> measurement uncertainty of the used frequency counter.
>
> [1] http://athome.kaashoek.com/time-nuts/Rigol%20locking%202.png
> [2] http://athome.kaashoek.com/time-nuts/Rigol%20pulling.png

A few good experiments.

The details of the lock mechanism can be the culprit here. For instance, 
if it is a phase lock, then dead-band mechanisms can explain this. 
Similarly can a frequency lock experience similar dead-band mechanisms.

One dead-band mechanism can be the resolution of the steering, causing 
one to alternate between nearby values to average near the actual value.

I've run into this before and some designs create such deviations that 
show up in both phase and frequency as you look carefully. For many 
purposes it may suffice just fine, but for others not so much.

If you set offset too close, it can take time before you see the 
adjustment, so test with offsets inbetween to validate that frequency 
gone down.

Yes, I've been hit by similar issues before. Any generator will cause 
phase-deviations so question is just how small deviations.

Cheers,
Magnus