[time-nuts] Re: 10 MHz TCXO periodically jumping 20 mHz up and down, cause identified

[Bob kb8tq]

Hi

Sounds like you might want to try a different TCXO. Or even 
try an OCXO. Even with a good TCXO, 1x10^-9 accuracy is
a stretch. I’m assuming we’re talking frequency accuracy so 
the usual “one sigma” time accuracy not the overriding number. 

GPS module “output jitter” forces you to a fairly long filter time 
constant to achieve better than that. Hitting the usual “>99% of the
samples at a 10 second gate" sort of frequency accuracy spec
drives you to numbers out past 100 seconds. 

The TCXO (even a good one) in a fairly normal ambient gets hit
by all sorts of small thermal changes. You can try this or try that,
in a real world portable setting … there is “thermal rumble”. The
TCXO will struggle to stay on frequency as it sees that rumble.

Bob

> On Feb 19, 2022, at 3:24 AM, Erik Kaashoek <erik at kaashoek.com> wrote:
> 
> Thanks for all the excellent info
> The TCXO is actually a VC-TCXO at 10MHz intended for use in a cheap GPSDO where the ambition is to have 1e-9 frequency accuracy so the 2e-9 jump was just too high.
> The VC-TCXO had its own low noise voltage regulator but the Vtune was connected to a variable voltage divider between GND and the Vcc of the VC-TCXO  through a low pass filter so even with its own voltage regulator any change in current in the TCXO can change the supply voltage and feedback through the variable voltage divider.
> It was expected that only slow changes in current could happen because of temperature changes as the temperature control is very fast, loop bandwidth well within one second, and these where filtered out  by a low pass filter after the variable voltage divider.
> To test if the jump up and down was caused by a current to Vtune feedback the Vtune was set to Vcc, 1/2 Vcc, 1/4 Vcc and GND.
> The result was interesting:
> Vtune | Max jump
> Vcc    | 6e-9
> 1/2 Vcc | 3e-9
> 1/4 Vcc | 1.5e-9
> GND | no jump
> This suggests the jump is indeed caused by feedback from the current changes through the variable voltage divider into Vtune and there is a digital circuit inside the VC-TCXO with changing current consumption causing the 107.34 seconds periodicity.
> To confirm the feedback was indeed the cause the variable voltage divider was connected to a 3.7 V battery instead of Vcc and indeed, no more frequency jumps!
> The ADEV of this cheap VC-TCXO with Vtune at 1/2 Vcc is not bad:
> 0.1 s | 2e-10
> 1 s | 1.5e-10
> 10 s | 0.9e-10
> 100 s | 2e-10
> With the Vtune at GND the ADEV is even much better so there is still some more investigating to do.
> Some DAC's with internal voltage references have been ordered to test if it  is possible to connect the DAC to the same Vcc (to save cost) and still have a stable but variable Vtune without feedback .
> If this does not work the DAC will need its own voltage regulator.
> Again thanks to all the people that replied, I'm learning a lot.
> Erik.
> 
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com -- To unsubscribe send an email to time-nuts-leave at lists.febo.com
> To unsubscribe, go to and follow the instructions there.