[time-nuts] Re: Tuning a GPSDO loop for optimal disturbance handling

[Erik Kaashoek]

Hi Bob
Good to hear.
As the internal logging does collect the DAC settings and the frequency 
of the TCXO versus a Rb standard I'm lucky to have both. And full scale 
linearity testing of the DAC's hopefully will show any issues there. 
They should be linear within 0.2 step.
I've already build the single mixer+low noise 24 bit microphone input 
sound card setup and locked a 4GHz PLL to a Rb and used the 400th 
overtone of 10MHz from the TCXO so the lowest audio frequencies (20Hz) 
are in reality 0.05Hz, low enough to connect to the 1Hz gate time 
measurement with the counter.
At first impression the phase noise is not terrible, now the calibration 
needs to be done by comparing with a 10MHz source with known (bad) phase 
noise.
The audio analysis SW used can do dB/log Hz.
Will be interesting to share the results here and hear your feedback.
Erik.

On 6-3-2022 18:49, Bob kb8tq wrote:
> Hi
>
> DVM’s show up from a lot of places. Indeed some are nutty when
> shipped from here to there. The shopping process is always going
> to be “delivered price” based.
>
> The reason for looking at the dac output is: The frequency just jumped
> 2x10^-8 … was it from the dac or internal to the TCXO ? The oscillator
> drifted 5x10^-9 it one minute, was it the DAC / Vref? …
>
> Phase Noise:
>
> The “quick / simple” way to do phase noise is with a single mixer setup.
> You run both ports at “max” ( so 7 dbm on a 7 dbm mixer) and lock them
> in quadrature. A fairly simple audio amp based on any of a number of
> op-amps boosts the output to something an audio spectrum analyzer
> or sound card can “see”. You would like an opamp with something in the
> 1 nv/Hz vicinity in terms of noise.
>
> Once you get the setup going, it’s just a matter of calibrating things. There
> are a variety of app notes and papers on that part. It normally involves
> unlocking the loop and measuring the phase slope of the beat note with
> whatever you are using to look at noise. ( The op amp “preamp” normally
> gets switched out for this step).
>
> Not crazy expensive or hard to wire up. I’ve built the circuit a lot of times
> using “dead bug” sort of construction. They all seem to work.
>
> On a typical GPSDO design, you are looking at noise inside about 10 KHz.
> Sure there could be issues anywhere, but the most likely stuff is below the
> max limit on the typical sound card. Low end wise, it would be nice to get
> to 1 Hz. That isn’t going to be easy with the typical sound card.
>
> Yes that glosses over all the “joy” of tracking down ground loops and other
> local noise sources. You are measuring a *very* low level signal. Quiet
> supplies and good grounding are part of this.
>
> Bob
>
>