[time-nuts] Re: Does filtering a TCXO Vtune to reduce low tau ADEV with max 50% make sense?

[Erik Kaashoek]

Hi All,

What I did:
0: Stopped all SPI traffic and display updating to avoid supply load 
changes.
1: Used a demodulating spectrum analyzer to listen to the FM at the 
200th overtone (2 GHz), only white noise (e.g. dominant high frequencies)
2: Use a modulation analyzer to analyzed the FM modulation of the 100th 
overtone (1 Ghz), no visible structure in the demodulated signal and the 
frequency deviation was below 1Hz (is consistent with the measured ADEV)
3: Used a 200 MHz LO into a mixer, mixed with the 20th overtone 
(200MHz), output of the mixer into a very good audio input (1 Hz to 192 
kHz, 24 bit), tuned the LO to zero beat and did an FFT of mixer output 
displaying from 1 Hz to 96 kHz (due to the 20th overtone in reality 
starting at 0.005 Hz), no visible structure in the FFT.
4: Also looked at the output of the mixer with a scope to double check 
for content outside the audio range. The audio looked like noise but not 
really white noise.

When the DAC is set to zero the "white" noise level is a bit lower 
compared to a normal DAC value but due to the 0.1Hz low pass filter 
between the  DAC and the TCXO Vtune the "white" noise can not come from 
the DAC so probably generated inside the TCXO.

As soon as the display update was enabled, clearly visible frequency 
swings at the exact rhythm of the display updates where observed. These 
small swings could be heard as ticking noise in the FM demodulation and 
where visible in audio FFT and on the scope observing the audio.
Stopping or starting the display updates did not have a measurable 
impact on the ADEV.

The TCXO tuning range is about 50 Hz over 2 Volt, the lowest observable 
DAC step leading to a frequency change is about 1E-10 or 0.001 Hz
The DAC output was measured using a 6.5 digit voltmeter and there was no 
observable instability. The DAC output did contain high frequency (above 
100 kHz) leakage observable with a scope, probably coming from the TCXO 
through the shared supply or from the ground lead of the probe acting as 
an antenna for the 10MHz. Due to the ground lead it was not possible to 
reliable measure the absence of the 10MHz after the 0.1 Hz low pass filter.
The TCXO supply sensitivity is 0.5ppm for +/-5% changed. The DAC and 
TCXO used a dedicated 3.3 V regulator and the DAC has an internal band 
gap reference for further stabilization. The input and output of the 3.3 
V regulator where decoupled with 1000 uF low ESR capacitors. No change 
was observed with or without the capacitors.
The 0.1 Hz low pass filter between the DAC and the TCXO reduced the 
noise level just visible when the DAC was set not non zero.

I may have reached the limit of the used components. Which is OK as the 
performance and cost are both within the intended spec.

Erik.

On 20-5-2022 19:40, Bob kb8tq wrote:
> Hi
>
> The typical answer to this issue is to first try to clean up the supply to the
> DAC. If that does not fix the issue, you likely need a better DAC. Unless the
> TCXO has a crazy large tune range, rational parts should be able to do the job.
> If the TCXO has a crazy large range then maybe a different TCXO is the answer.
>
> You can measure this or that to confirm / deny that the problem exists here or
> there in the circuit. An audio analyzer that gets down to a bit under 1 Hz should
> be able to tell you what’s what. Various sound card based approaches likely
> would be “good enough” ( yes, ground loops will be fun to take care of ….)
>
> Bob
>
>> On May 20, 2022, at 9:26 AM, Erik Kaashoek via time-nuts <time-nuts at lists.febo.com> wrote:
>>
>> During testing a TCXO using a correct Vtune for 10MHz output, the ADEV at tau=1 second  was measured at 1.08E-10 (see attached plot[1], trace DAC=normal)
>> When the DAC controlling the Vtune was forced to zero volt output the ADEV at tau=1 s reduced to 5.43E-11 (see plot, trace DAC=0)
>> This clearly hinted at some noise present on the DAC output.
>> Adding substantial filtering between the DAC and the Vtune input of the TCXO and setting to 10MHz resulted in an ADEV at tau=1 s of 7.28E-11 (see plot, trace DAC filtered)
>> All measurements where done with the controller for the Vtune disabled.
>> The plot contains error bars.
>>
>> The extra filtering required for the reduction creates a lag of about 10 seconds in controlling the TCXO and the controller without the filtering was nicely able to correct fairly quickly random walks so I'm a bit worried about the problems this 10 seconds extra lag may create in tuning the controller loop.
>>
>> Is a reduction of the ADEV of a 10MHz reference output with 25% (or at most 50%) for low tau in practice of any relevance if this may cause an increase in ADEV for larger tau as the controller may have more difficulty to correctly adjust Vtune to correct random walks?
>>
>> [1] Plot: http://athome.kaashoek.com/time-nuts/DAC_Filtering.png<DAC_Filtering.png>_______________________________________________
>> time-nuts mailing list -- time-nuts at lists.febo.com
>> To unsubscribe send an email to time-nuts-leave at lists.febo.com