[time-nuts] statistical distribution of initial frequency error in tcxos

[Mark Goldberg]

On Wed, Mar 13, 2019 at 9:02 AM jimlux <jimlux at earthlink.net> wrote:

> If I go out and buy 100 TCXOs with a spec of, say, 50 ppm, what does the
> distribution of the initial frequencies (and, I suppose, the frequencies
> after aging) look like.
>
>
If you are buying 1000, you probably will get the manufacturer to test to
meet your specs. At 100, you will likely get what you get from them. I am
buying VCTCXOs 50 at a time and they have quite a variation. Basically they
meet spec for temperature variation, but many just barely meet spec. They
typically are 1-2 ppm off before adjusting them on frequency and drift 1-2
ppm in the weeks after soldering too. After that, maybe 1 ppm per year.
This is only one example of one manufacturer in a custom frequency for me.
If you buy a standard frequency, you will likely get the benefit of better
manufacturing control they can afford for high volume parts. I am buying
parts specified at +/- .28 ppm over 0 to 70 degrees and they just meet
that, but the frequency is anywhere at any temperature. It is often not a
linear function.

I've also seen hysteresis, frequency jumps, etc, with only about 80% of the
parts acceptable to me. Those effects are not considered or tested by the
manufacturer at this low volume, so I have to do the temperature testing
and cull the unacceptable ones. You will probably have to do the same, age
and test yourself to get acceptable results.

Regards,

Mark