# [ntp:questions] Setting the maximum rate of change

Spoon devnull at localhost.com
Fri Mar 23 15:33:21 UTC 2007

```David Woolley wrote:

> Greg Dowd wrote:
>
>> If I understand correctly, the question regards slew rate limits.
>
> That's not my understanding.  My understanding is that he want to
> constrain the first derivative of the frequency correction.

Correct. I want to know if it's possible to constrain the acceleration.

> That is
> the second derivative of the offset.  Slew rate is used to describe
> the frequency correction, particular when it is being used to reduce
> an offset error.
>
> What's probably implicit is that, in the specification he is trying
> to implement, this is the derivative including the first derivative of
> the crystal frequency, so the 25mHz/27MHz/s (about 0.001 ppm/s)

NB: it's 75 mHz/s / 27 MHz i.e. 10 ppm/h

> might
> be totally taken up by variations in the crystal frequency.  Actually,
> given that he wants this condition met after one minute, but the machine
> probably won't reach thermal equilibrium for more like 15 minutes, I think
> it will almost certainly be violated by the crystal frequency on its own.

Assume the systems have been running for 24 hours.

As far as I understand, clock drift rate (what I call acceleration) is a
function of temperature, voltage, and what else? Assume temperature is
kept almost constant. I can't say anything about voltage. Since we are
using ordinary power supplies, it could probably be slightly improved.

> In that case, the only way of achieving conformance is to NOT constrain
> the first derivative of frequency correction, so that there is some
> chance of the frequency being compensated fast enough to compensate for
> the crystal drift.

Would the oscillator drift rate be that bad in the situation described
above?

Regards.

```