[time-nuts] Simple solution for disciplining OCXO with 1 PPS

[Magnus Danielson]

Bernd,

On 05/22/2016 07:41 PM, Bernd Neubig wrote:
> Hello Fellow time-nuts,
> I am looking for a simple solution to discipline my 10 MHz reference OCXO in my private lab with an 1 PPS signal from a separate GPS receiver.
> I am curious if there is a solution possible without programming a microcontroller, as I am an old-fashioned "analogue" guy ;)
> I am well aware, that such a solution would have a lot of disadvantages, as it cannot effectively compensate for short-term variations. However I would be happy if such a KISS solution could achieve  a stability (STS) of better 1E-10 over an hour. I know this is a damned long integration time for an analogue integrator...
> If that sounds too weird, I am open to receive advises for a microcontroller based solution.

OK, I assume you can tolerate to work with digital gates.
Square your sine up, if needed.
Divide down in steps of 10 to 1 Hz.
Resynchronize with original 10 MHz using a DFF.
Use a SR-flip-flop to compare the two PPSes.
Do a PI-filter with an op-amp, two resistors and a capacitor (one 
resistor and cap in the negative feedback part, one resistor from the 
source.

That in principle is all what you need.
You can use a standard RC lowpass filter to improve the noise filtering, 
but keep the bandwidth of it high enough not to interfere with the 
damping of the loop.

Far from optimal, but you shuold be able to get it to lock.

Dividers might be not be needed, and removing those will help to make 
divider state less of a track-in issue, thus maintaining lock quicker. 
However, there is a benefit in avoiding at least some of the cycle 
ambiguity, so at least the first divide by 10 should be maintained.

I'm sure this can be optimized.

>
> Thanks a lot for your comments to come.
> BTW: you need not to teach me about basics of short-term stability. I just want to evaluate the limits of a possible analogue solution. For sure, that's not real disciplining, but more like a long-tau integration PLL

Well, since the loop bandwidth will be larger than optimum, more of the 
GPS noise will creep into the output and not really give your 
measurement a good chance. Long-term will work, but short-term not so 
much. Longer tau/lower bandwidth will filter away more of the GPS noise.

Cheers,
Magnus