[ntp:questions] Installing more stable oscillator?

Jason Rabel jason at extremeoverclocking.com
Sat Jul 14 03:55:53 UTC 2007


> I guess my unasked question was, "How can one build a better quality, 
> more stable time source?"

Build it as a stand-alone piece of hardware. Like I said before, PC's have
many caveats that prevent a simple oscillator swap from being as useful as
you would want it to be. The combination of all the hardware and software
adds a bunch of variables that just can't be controlled as fine as one would
like. I'm not saying that a swap wouldn't help, it just wouldn't be as
useful as alternative methods using that same oscillator.

You can get PCI cards with TCXO and maybe even OCXO oscillators on them. But
unless you find one cheap or have a specific need for the internal design,
it's probably better to find something slightly older and bigger on eBay. ;)

> I could use NTP to sync my computer's cheap oscillator to an internet 
> source (what I currently do), or to a radio clock (GPS, CDMA, WWVB, etc. 
> -- I'd like to do this, but budget and lack of knowledge is currently 
> preventing this.)...that's not too hard. But how would one build a more 
> stable source of time? If the external source is interrupted, even with 
> NTP adjusting for the system's drift, it will still drift further and 
> further away from the actual time in relatively short order. I'd like to 
> have a system here that can avoid that, mostly "because I can", not for 
> any particular reason.

This is where a lot of companies make their money. Spectracom has a lot of
WWVB receivers that discipline crystals or rubidium sources. Symmetricom has
similar based on GPS & CDMA. When the signal is lost or incomplete the
hardware simply runs off the oscillator until a proper signal can be
regained.

Likewise some GPS timing receivers have TRAIM, which allow them to stop
sending a PPS pulse if it determines that it would not be valid.

These dedicated boxes are specifically designed to be highly accurate
frequency and/or time sources. 

> Surely many of the stratum 1 servers (say, time.nist.gov) that get and 
> distribute time directly from atomic clocks aren't just off-the-shelf 
> servers with cheap, unstabilized system clocks, right? I know that many 
> of the public stratum 1 servers deployed by individuals and 
> organizations get their time from GPS, and are probably ordinary 
> computers, but I have this (again, perhaps incorrect) assumption that 
> the servers that supply the time /to/ the GPS system are not ordinary 
> computers.

Read NIST Special Publication 250-59. It details the Internet Time Service &
ACTS (dial-up) service. They are more or less "off-the-shelf", but have been
modified (both hardware & software) some to suit their specific timing
application. The key thing to note is that a 1PPS signal is fed to the
machines for the time-keeping part.

With large organizations, they will either use existing servers and sync
over the internet. Maybe some will do their own GPS / server combo. But most
will buy dedicated "network time servers" from companies such as
Symmetricom, Zyfer, Arbiter, Endrun, etc...

> Perhaps I misunderstand, but are you saying that it's possible to 
> replace the computer's crystal oscillator with an ovenized quartz 
> oscillator? Or are you saying that one can use the ovenized quartz 
> oscillator (or rubidium, cesium, hydrogen maser, etc.) as an external 
> oscillator, and it would provide PPS input to the computer running NTP?

First, there is a difference between a crystal and an oscillator. An
oscillator contains a crystal, along with bits of electronics in the
package. Most PC motherboards what you see as the little metal twinkie
looking thing is simply a crystal. The rest of the circuitry is on the
motherboard.

But the better route to go anyhow would be to use the external oscillator to
provide the PPS input to the computer.

> Interesting. If one were to have such a setup, and GPS service was 
> unavailable for whatever reason[1], the rubidium oscillator would 
> continue to provide PPS input to NTP, and while the rubidium source 
> would not be ticking at exactly the same speed as UTC, the drift would 
> be extremely minimal, even with extended unavailability?

Drift depends on the source used. A rubidium standard you could probably
flywheel for a few months and your PC would be non-the-wiser. An ovenized
crystal on the other hand can really vary, but it will drift quite a bit
more than the Rubidium. However if your GPS outages are only occurring for a
few hours each day then it wouldn't really matter. Many cell sites have GPS
antennas and Rubidium standards, not for time, but for frequency. If the
frequency between the cell sites drift too much your call will be dropped as
you switch towers.

GPSDO products are also very common. The HP Z3801A is probably the best know
of these products as it was a surplus item from telecom upgrades back in the
late 90's I believe. There are newer GPSDO products, made by the same
companies mentioned before. I have one made by a company called Brandywine.
It never looses lock but when it does it goes into 'holdover' mode and
simply uses the oscillator until the GPS signal returns.

> I apologize for the rather simple questions, but over the last few years 
> I've gone from being a minor time enthusiast to wanting to get more 
> involved. Syncing from internet servers is certainly adequate for my 
> needs here, but I'm looking at doing bigger and better things. 
> Unfortunately, doing so requires lots of rather silly questions. :/

No need to apologize, asking questions is how you learn. ;) I'm in the same
boat as you, time is very interesting to me and I'm still pretty green
myself. Hide the pocketbook as it can get expensive! If I only stopped at
that first GPS that I bought and left it hanging out the window.




More information about the questions mailing list