[time-nuts] Choosing an Oscillator...

Tom Van Baak tvb at LeapSecond.com
Sat Jul 28 15:26:24 EDT 2007

> Hi gang,
> I'm still headed towards getting a Stratum 1 NTP server up and going.
> From what I've learned so far, higher quality PPS signals aren't going 
> to do a lot for me if my server has a wiggy clock.  That makes perfect 
> sense.  Now I just have to figure out how to get a good clock into my 
> server a la the soekris clock-block approach.

Do you know yet what level of performance you're looking
for? At some point you may find most of the jitter is due to
software, interrupts, and caches rather than the oscillator
itself. But, as you say, attacking the clock is the first step.

> So what do people recommend for somebody working on a small budget?
> It looks like Efratom Rubidium standards are available on ebay, but the 
> going rate looks like about $200, which is a bit much for me.

There are also power and heat sink issues with an Rb. Note
rubidium or cesium may be total overkill (then again, this is
the time-nuts list, so perhaps nothing is excessive).

You can find 1, 3, 5, 10, 15, 50, 100 MHz oscillators on eBay
for a fraction of the price of a Rb. Many of these OCXO
require only a single 12 or 24VDC supply.

> I also see that HP 10544As are available on ebay for a lot less, 
> probably under $100.  I guess I could put it in a well insulated box and 
> discipline it using my Garmin and a Reflock (I do need to learn how to 
> work with CPLDs...).

The 10544 and 10811 oscillators are already ovenized and
so an insulated box is not really required. Note these very
oscillators are used in hundreds of hp products and none of
them, even the 105B, add any layers of insulation. The split
power requirements for the 10544 and even the 10811 are
more complicated than most OCXO.

Do you need to discipline anything if you're using NTP? I
thought, given a clean GPS-derived 1pps input, NTP had
provisions for frequency offset and drift in the local oscillator.


More information about the time-nuts mailing list