[ntp:questions] Re: NTP algorithms to predict clock skew and/or jitter?

Ulrich Windl Ulrich.Windl at RZ.Uni-Regensburg.DE
Wed Aug 10 10:45:46 UTC 2005


"skillzero at gmail.com" <skillzero at gmail.com> writes:

> Are there NTP algorithms to predict clock skew given previous NTP clock
> updates? My clock runs about 5.8ms/min slow and I'd like to
> programmatically predict this so time drifts much less between NTP
> updates. I have time syncing to about 100 microseconds now, but I'd
> like to get it down to about 22 microseconds (1 audio sample at 44,100
> Hz).

I'm afraid most audio hardware uses its own clocks and timer interrupts. Only
a few allow feeding an external clock.

> 
> Also, are there NTP algorithms to reduce the effects of jitter? I'm
> seeing about 50 microseconds of jitter and I was hoping to reduce the
> effects of it.
> 
> For some background, I'm syncing time from a single server to 1-8
> clients on a private network. The clients are embedded devices with
> limited capabilities (e.g. no floating point, limited memory, etc.) so
> I'm not able to use the full NTP 4 code. I've written my own
> implementation of SNTP 4 and a local clock code based on processor
> cycle counts to maintain a nanosecond counter with 60 slew
> adjustments/sec.




More information about the questions mailing list