[time-nuts] Re: What phase variations to expect in a DMTD due to temperature fluctuations?

[Erik Kaashoek]

Hi Attilla

I checked the source of the temperature drift with some cold air. The 
ADC is not sensitive, the mixers are.

The side channel based  phase stabilization is working well. Kurt 
Paulson did some 24 hours tests
With some degrees temperature difference over a 24 hour period the ADEV 
floor hits 2e-17 at 10000 s and the ADEV follows 1/Tau till 1e-16 from 
1e-13 @ 1s
Here is the resulting ADEV floor:
http://athome.kaashoek.com/time-nuts/DMTD/ADEV_floor.png
Blue is some degrees temperature variation, pink with very stable 
temperature
And here is the phase deviation over the same period:
http://athome.kaashoek.com/time-nuts/DMTD/ADEV_floor_phase.png
Much more testing using real use cases is needed before I dare to rely 
on these results.

We have started a separate email group to discuss the DMTD to avoid too 
much traffic on the time-nuts mailing list, some more testers are 
expected to join soon.
If you are interested , let me know and I will send you an invite.
Erik.

On 22-11-2022 11:37, Attila Kinali via time-nuts wrote:
> Moi Erik,
>
> On Wed, 16 Nov 2022 16:57:59 +0100
> Erik Kaashoek via time-nuts <time-nuts at lists.febo.com> wrote:
>
>> As often with your advice, I'm not smart enough to understand.
>> The digital down mixing to zero Hz is done with I/Q mixers where sin/cos
>> of the internal LO is multiplied with the input signals and then average
>> over some samples to get I and Q of the signal vector as the output
>> frequency is zero Hz.
> Exactly. Going to zero Hz means that the LO frequency is the same as the
> RF frequency. Which means the (complex) output gives you a vector that points
> into the direction of the phase difference between the RF and LO signal.
>
>> I would expect I would need a separate set of I/Q mixers for the side
>> channel and by summing over many samples the noise would be reduced
>> sufficiently to get a relevant I and Q signal even if the side channel
>> has 40dB lower amplitude.
> Yes. You are looking for changes due to temperature fluctuation, which are slow.
> A time constant in the  milli-second range is sufficient.
>
>> I may have used the word FFT to describe the
>> operation of the I/Q mixers as computationally they look a lot like a
>> single bucket FFT
> Yes, there are many ways how to implement an FFT in hardware. One of them
> is using narrow band filters made out of (damped) oscillator structures.
> And that is, in its working principle, quite close to what the down-mixing
> approach does.
>
> On Wed, 16 Nov 2022 18:28:33 +0100
> Erik Kaashoek via time-nuts <time-nuts at lists.febo.com> wrote:
>
>> I went ahead and did a fairly simple implementation of the side channel.
>> Both channels show, apart from the noise, the same phase drift in the
>> order of some ps.
> So you got a result that is close to what we expected to see.
>
>> This seems to confirm the drift originates in the mixers.
> Be careful with this conclusion. The drift can originate anywhere
> where the signals travel different paths. It could be equally well the
> clock of your two ADCs that's the culprit, because they too take different
> paths after the splitter. Figuring out which part is the dominant source
> of delay takes a lot of care and experimental skill.
>
>
>> Now I need a 3 channel DSS to test this. Maybe I can use the
>> 10MHz ref out as the 3rd channel. And I have to find a way to inject the
>> side channel without creating additional leakage between the two main
>> channels.
> Be also careful with signals that are correlated to your clock or LO signals.
> These can show higher or lower noise than the actually measured signals.
>
> The injection needs to be done in a way that isolates the two paths. You can
> either use some isolation amplifier that are closely matched, or start with
> a strong signal and add some 30dB damping in the two paths after the splitter.
>
> 				Attila Kinali
>