[time-nuts] OCXO startup transient

[Magnus Danielson]

Hal Murray wrote:
> A couple of graphs in case anybody is looking for examples...
> 
> This is a 3 MHz OCXO from Telco surplus.
> 
> The can says EROS-800-M45, 94-12.  I haven't found a data sheet but it's 
> supposed to be equivalent to Isotemp 127-10.
> 
> It's got a mechanical adjustment, no electrical adjustment.  (I haven't 
> touched the adjustment.)
> 
> The can is 2.5 x 2.75 x 1.75 inches.  I'd call it medium quality.  Nothing 
> fancy, but probably better than most of the small ovens.  When I was playing 
> a while ago, I could see changes when I grabbed the can with my fingers.
> 
> It was powered off for at least 6 months before I started this run.
> 
> 
> I'm collecting samples every 5 minutes.  I discarded the first 3 samples 
> after power on.  The first was way low.  (It's so far off that the rest of 
> any graph gets compressed  into a flat line.)  The next two overshot.
> 
> When I run out of other things to do, I'll make another run with samples 
> closer together for the first few hours.  But first I want to watch the drift 
> some more.
> 
> So far, the drift has generally been in one direction.  That's when averaged 
> over several days.  There is enough noise on the hour scale that I wouldn't 
> want to try to predict anything to get better holdover.

Your holdover properties should be evaluated after running for say 30 
days. A frequency compensation get you a first degree correction which 
together with TDEV measures give you an indication of aggregated 
holdover properties.

> To my eye, there is a bend in the drift curve at 4 days.  I can't think of 
> any environmental change that happened then.  I might have bumped it.

Considering the wobble on day 8, I'd say the day 4 shift is within what 
environmental shifts do to you, where as on day 4 the curvature is still 
steep so the added effects fools the eye.

Cheers,
Magnus