[time-nuts] Fast frequency counting question

Bruce Griffiths bruce.griffiths at xtra.co.nz
Sun May 4 21:13:09 EDT 2008

John Miles wrote:
>> Probably impractical as the sampling noise is so high that it will
>> require  tens of thousands of measurements to get the noise down to the
>> required level and meanwhile the temperature etc will have changed
>> significantly.
> Heh, that's a good point, you can't just sit there and trigger the thing
> 10,000 times per second for 24 hours...
>> Also the relatively low ADC resolution and nonlinearities
>> will tend to limit the effective resolution in determining the zero
>> crossing locations.
> I stand by my assertion, though, that all you're going to get with high
> sampling precision is a lot of well-characterized random noise determined
> almost solely by outside factors.  It's an interesting question, but one
> that can't be answered without knowing more about the symptom that started
> the whole thread.
> -- john, KE5FX
If the oscillator AV were to follow a 1E-12/Tau law into the microsecond 
region then at 10us the AV will be around 1E-7.

A measurement system noise sampling jitter of a little less than a 
picosecond should suffice to characterise this.

It is unlikely that a GPS TCXO has that low a noise it may be perhaps 
10x this.
Thus measurements of the startup transient characteristics to the 1ppb 
level in 10us windows are perhaps somewhat futile.

The 16Hz variation in a ~ 200kHz signal is equivalent to noise of about 
4E-7 at 40MHz with a 10us gate.

A substantial part of this may well be due to the TCXO and local 
oscillator phase noise with some contribution due to amplified mixer noise.

More detail on the end users measurement techniques are in order to see 
exactly how and what they are measuring is required.
If in fact they want to characterise the zero crossing jitter more 
accurately during startup subpicosecond (but not 10 femtosecond) system 
noise can probably be justified.
This is easy to achieve with a high resolution (14-16 bit) ADC directly 
sampling the 40MHz signal.

100x smaller measurement system noise is perhaps only justified if one 
were observing the startup transients of a much higher Q oscillator 
however these would last much longer so in practice perhaps it isnt 


More information about the time-nuts mailing list