[time-nuts] Re: Simple simulation model for an OCXO?

[Greg Maxwell]

On Mon, May 2, 2022 at 10:01 PM Bob kb8tq <kb8tq at n1k.org> wrote:
> By far the best approach is to use actual data. Grab a pair of OCXO’s and
> compare them. A single mixer setup is one easy ( = cheap ) way to do it. You
> will get the sum of the two devices, but for simulation purposes, it will be *much*
> closer to reality than anything you can brew up with a formula.

But a programmatic way of generating plausible OCXO noise lets you
generally millions of times more data, over a much larger spectrum of
plausible operating conditions-- so that you can test the stability of
an algorithm over a wider collection of conditions.

It's not a complete replacement for using real data-- you should do
that too.  But it can be a much more comprehensive test.