[time-nuts] Reality check - understanding MV89A aging figures

[Bob kb8tq]

Hi

OCXO aging is not “linear” in the fashion you are trying to interpret it. Put another way,
daily aging does not equal yearly aging / 365. Hourly aging does not equal daily aging / 24. 
The drivers are not that simple or that well behaved. If you find a spec that calls out a lot
of details, there may be a daily rate “after X days on power”. There also might be a monthly
and yearly number. 

A typical OCXO will have an daily aging rate that decrease day to day for (most likely)
the first 30 to 90 days after power on. What it does past that is very dependent on a 
bunch of process variables, both in the crystal *and* the circuit. Eventually the daily
aging rate will “bounce” back and forth from positive to negative. The net effect will 
be quite low, even if the daily numbers still are very noticeable. 

All this *assumes* that temperature / voltage / load / whatever are kept constant enough
to not impact things. On a 10811 “whatever” includes both pressure and humidity. On
a MV89 one would hope those are less of an issue.

There are a number of classical formulas for modeling crystal aging. The gotcah is that 
they work well for bad crystals ( = ones that age a lot ). They tend to be less useful 
for low aging parts. The formula *assumes* there is one dominant driver, in a good 
crystal, that’s not the case. 

Fun !!

Bob

> On Sep 3, 2020, at 5:14 PM, Matthias Welwarsky <time-nuts at welwarsky.de> wrote:
> 
> Hi,
> 
> since I'm using the MV89A timebase of my counter for all measurements, this 
> should provide a possibility to also evaluate the stability of that timebase, 
> shouldn't it?
> 
> When I measure a GPSDO, while it is locked the long term stability will be as 
> good as the GPS clock. Therefore, all drift I'm seeing in the measurements 
> will be from the counter timebase.
> 
> Now, TimeLab tells me the drift is around 3E-5 Hz per hour. Pulling up the 
> well-known datasheet for the MV89A, depending on the OCXO class the aging is 
> between +/- 3e-8 and +/- 5e-9  per year. Since no unit is given on the data 
> sheet I'm assuming it is relative accuracy, right?
> 
> The datasheet doesn't give daily aging figures but states that the typical 
> correspondence between yearly and daily is a factor of 1e-2.
> 
> Therefore, is the following calculation correct:
> 
> 5e-9 / year => 5e-11 / day => around 2e-12 / hour, relative accuracy
> 
> 2e-12 / hour * 10 MHz = 2e-5 Hz / hour
> 
> If that's correct, my timebase would be still in spec, not best in class but 
> quite OK, at least as far as aging is concerned.
> 
> BR,
> Matthias
> 
> 
> 
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.