[time-nuts] Caveats on Allan Deviation with ultra stable oscillators

[Magnus Danielson]

Corby,

On 05/29/2014 06:15 PM, cdelect at juno.com wrote:
> Recently I have been testing some Quartz oscillators with
> exceptional stability. (low parts in 10-13th)
>
> I've noticed something that jumped out at these performance levels!
> Normally I test oscillators on a DMTD system with either
> an ultra stable Quartz or a Hydrogen Maser reference.
>
> On these latest oscillators at the longer Tau (>100sec.)
> the Quartz versus Quartz data shows much better performance
> than the Quartz versus Maser!
>
> What is happening (I think) in this case is that both Quartz units
> have exceptionally low and similar ageing.
> This similar ageing rate masks the true stability performance at longer
> Tau.
> The Maser data (with no effective ageing in this case)
> reveals the true stability at the longer Tau!

Do you have plots to share?

What you describe sounds like you do not remove the drift from your 
data, which means that you run into the drift limitation. The linear 
drift shows up as a ADEV(tau) = D*tau/sqrt(2) curve.

Have you tried to use the Hadamard deviation instead? Hadamard removes 
first degree frequency drift.

Cheers,
Magnus