How accurate is a GPS 1pps signal?

[Dave Boechler]

I find variation in specs on this..  Is it like rubidium clock accuracy?
Best information I can find is this, which is quite old, but seems to be in the 20ns range even back then.
http://www2.iee.or.jp/ver2/honbu/14-magazine/log/2005/2005_08c_08.pdf

Could you use is as a cheap standard for home equipment calibration and other accurate pulse circuit design?

Dave it sounds like a good project/video lesson...

[Jay_Diddy_B]

The GPS 1pps signal is very accurate. Each GPS satelite has four caesium or rubidium atomic clocks.
The jitter comes from the changes in propagation and the receiver specifications.

HP made GPS disciplined OCXO frequency standards. These were used in cell phone base stations.
Google HP Z3801A frequency standard.

These can be treated as primary frequency standards and do not need to be calibrated.

There are several other designs which basically are a PLL using the OCXO as the oscillator and pulling the frequency to lock to the GPS 1pps signal.

Google GPSDO

John

[Dave Boechler]

Thanks for the quick response!

DIY version: http://www.diyphysics.com/2012/02/19/d-i-y-gps-disciplined-10-mhz-frequency-standard-gps-based-universal-time-clock/

Sounds like a fun project..

[ejeffrey]

The long term stability of the 1pps is perfect(*).  There are short-term sources of drift such as atmospheric fluctuations and jitter in the trigger circuit.  That is why it is typical to use a very stable oscillator such as an OCXO or rubidium clock and lock that to the GPS signal with a very slow feedback loop.  GPS receivers optimized for timekeeping will perform better in this fashion.  It is also possible to retroactively adjust measurements to reduce systematic errors based on atmospheric observations.

(*)  The GPS satellite signals are based off of onboard atomic clocks which are in turn synchronized to ground based atomic clocks that are the primary references for timekeeping.  They represent the passage of time at the earth's geoid.  There will be a small but measurable frequency offset if you are at significant altitude due to general relativity.

[joelby]

One thing to note is that not every module puts out a perfect 1 pps!

I used one that had such an output, but found that it was not very stable at all. Digging through the documentation revealed that the pin wasn't intended to be used at all.

Here's a blog post I wrote on the topic: http://tumblr.tristesse.org/post/6687033745/gps-synchronisation

Infinite persistence view of the pulse:

[gxti]

Depends entirely on the receiver. Timing-oriented receivers like Trimble Resolution T (or Resolution SMT), or Motorola Oncore GT+ or UT+, have on the order of 15ns jitter. If the datasheet doesn't say anything then assume microseconds at best and milliseconds at worst. The receivers I mentioned are readily available on ebay for not a lot of money if you want something to play with.

[Bored@Work]

Could you use is as a cheap standard for home equipment calibration and other accurate pulse circuit design?

GPS is one of the best time sources you can get, and many people use it for that.  The best source of information for DIY is a mailing list (and its archive) out there called time-nuts. http://leapsecond.com/time-nuts.htm For commercial solutions (if they sell to you), look at companies like Symmetricom http://www.symmetricom.com/

[jeroen74]

http://n1.taur.dk/simplexdo/

This guy's site (without a reasonable index page :/, use Google for it) has all kinds of interesting timing reference stuff.

[kkp]

This guy's site (without a reasonable index page :/, use Google for it) has all kinds of interesting timing reference stuff.

True, I really should make an index page. I mostly throw stuff up and send links to the people I want to see it.

Cheap nav-type gps'es aren't really good timing wise - but then again, one person's 'awful' may be another person's 'awesome'. We all start somewhere. I did PPS measurements on two common receivers, the garmin GPS-18x, and the S1513R which is a Skytraq Venus 6 series.
http://n1.taur.dk/gps18x/
http://n1.taur.dk/s1513r/

Together with a cheap OCXO, you get, from a 'normal' user's perspective, a rather awesome _traceable_ reference.

My advice: If building a GPSDO has your interest, build one, quick. Then, if you are dissatisfied with it, build another. Don't plan endlessly for a better one you never get around to building.

This one I never got around to finish:
http://n1.taur.dk/gpsdo2a.pdf
It is basically 'lab notes' on a GPSDO, the most interesting part being the super-cheap interpolator to get nanosecond resolution from a 10MHz clock. Some of the PICs have a roughly-a-nanosecond interpolator on-chip, but I didn't have one of those.

/Kasper Pedersen

[SeanB]

I just read the serial output to do a sanity check of the nettime setting. Normally agrees well.