GPS 1 pulse-per-second signal

I would check out how paging companies (remember them?) used GPS to simulcast their transmitters in a given area. The timing resolutions were much, much tighter than what you're looking for. I want to say there's a 10MHz carrier you can lock to, but I just don't recall. It was a long time ago....

Glenayre was a major player in this technology so I would start there. Depeneding on how many you needed, I'll bet you can get this gear very cheaply in the used market. -mpm

This is very true; the output pulse is different. Also, if a module is not specified as a timing source, there can be the significant jerks of phase on the 1 PPS signal.

It depends on the accuracy. If you need a precise synchronization, you should use modules which are specified as the timing sources.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Use an ovenized local reference and lots of averaging...

Certain of Rockwell's Jupiter chipsets output a 10kHz reference signal--that allows much faster acquisition. Googling frequency reference + jupiter + gps gives some good examples.

HTH, James Arthur

Richard Henry hath wroth:

Yep. The time between pulses is guaranteed to be exactly 1 second. However, the time of any particular pulse is NOT guaranteed to be synchronized with any particular time or event. If you take two GPS receivers, and compare the 1 pps outputs, they could easily be at quite different times. However, each succeeding pulse will be exactly

Sure. Get a receiver that is suitable for running a clock, not an interval timer. The complexity and cost will depend on how accurately you're trying to sychronize. Lacking numbers, I can't offer any real solutions.

Maybe something like this will work:

"The GPSClock 200 has an RS-232 output that provides NMEA time codes and a PPS output signal. About a half-second before, it outputs the time of the next PPS pulse in either GPRMC or GPZDA format. Within one microsecond of the beginning of the UTC second, it brings the PPS output high for about 500 ms."

If you can live with some atmospheric and propagation drift, the cheapest solution is a WWVB 60KHz receiver and clock.

The problem is that WWVB only works in North America.

Another interesting oddity about GPS is that the position error reported by two otherwise identical units will be dependent on which satellites have been acquired. This problem comes up a lot in differential GPS time & distance measurements.

What sort of things are you trying to do? How accurate do they need to be synchronized?

The GPS units I've worked with all specify that the leading edge of the pulse will be very close to the second tick of UTC.

If you need (or could use) a PC to do whatever you want to do, I'd plug the GPS unit into a PC running ntpd.

If you only need time within a few 10s or 100s of ms, you may not need the GPS unit.

Lots of info in either comp.protocols.time.ntp or

The Garmin USB 18 LVC is popular because it's the lowest cost unit with a PPS signal. Some assembly required.

GPS timing receivers, such as Trimble Resolution-T or disciplined oscillators such as Trimble mini-Thunderbolt or Jackson Labs Fury will output a 1PPS whose leading edge is within +/- 15 nsec of UTC at any location on the earth (once fully locked). The pulse widths vary from one vendor to another, but the leading edges are the only significant data in the 1PPS outputs. Different vendor receivers have better or worse jitter specs, but the fact that all are aligned to UTC (or GPS master time which is an integral number of seconds offset from UTC) makes them suitable for aligning clocks that are scattered around the world.

These timing receivers all make use of their GPS position in the world to account for propagation delays, and provide a stable pulse that is very accurate in frequency as well as time-aligned to a universal standard. Accuracy is lower if your clocks are moving.

Steve

Thanks for the responses and links.

What I have learned so far from this thread, the links referred to, and a little studying on my own (please correct any errors I make here):

There is an instant every second where UTC time transitions from one second to the next. The point of 1 PPS signal is to allow determining that exact instant so that devices placed remotely from each other can synchronize their operation without direct contact. An example use of such a system would be synchronizing a radio transmitter with a receiver.

If a device knows its exact postion on the earth it only needs a good signal from one satellite to determine the second transition instant. If the device does not know its position, or knows that it is moving, it needs data from 4 satellites to determine the position and thus the accurate timing.

Even though there is an indicator of the exact instant in the satellite's transmitted data, the signal cannot simply be received, decoded and used as is. The receiver must account for the estimated rf space and atmospheric propogation delays, a local estimate of antenna and cabling delays, snd some electronic circuitry delay. It can then adjust the received signal back to UTC second transition time, and compare that calculation to an estimate derived from an internal clock, adjusting the internal estimate accordingly. Hopefully the internal estimate will be good enough that the adjustment will never be more than one cycle of the reference clock. The estimate is what is used to generate the 1 PPS UTC-synch ronized pulse.

Watch the jitter on the pulse. Some GPS implementations do it in software - with a few microseconds of jitter. Some receivers generate the pulse in hardware and can achieve nanosecond jitter. The later ones are normally special designs for time keeping.

It very much depends on your application. All GPS PPS implementations are accurate long term.

To synchronize things, you need to watch the *specified* edge of the PPS signal (usually the 0-1 transition), which usually marks the moment which you'll find in the NMEA messages of the next second.

In other words, the timestamps in the NMEA will give you time accurate to the second and the edge of the PPS signal takes it to fractions of a second.

Kind regards,

Iwo