[Martin Lorton] acquired a GPS-disciplined oscillator. He wasn’t quite sure what to do with it, so he did a little research and experimentation. If you have about two hours to spare, you can watch his videos where he shares his results (see below).
The unit he mainly looks at is a Symmetricom TrueTime XL-DC, and even on eBay it ran over $500. However, [Martin] also looks at a smaller unit that is much more affordable.
So what do you use something like this for? The idea is simple. A high-quality crystal (or rubidium) oscillator is disciplined to agree with the timing signals from GPS or another type of navigation satellite. Because the navigation satellites are kept very precise, the crystal oscillator can be accurate down to the nanosecond level.
In the military and commercial world, the timing from such an oscillator can synchronize multiple receivers which is necessary for applications such as passive RADAR. GPS timing by itself has excellent long-term stability, but since the unit of timing is one pulse per second — combined with things like multipath and other propagation effects — the short-term stability isn’t always great. A crystal oscillator, however, has great short-term stability. They tend to drift with temperature (mitigated by ovens) and simple aging of the crystal and other components. By combining both, you can obtain excellent short-term performance that holds over long periods, too.
Depending on how it is set up, such an oscillator can be accurate to within a few parts per trillion very shortly after turn on. Even low-end devices will be able to operate in the parts per billion range. Of course, you are going to need an antenna that can see the sky, too.
We covered hacking these out of cellphone sites, before. We’ve also seen GPS synchronizing PC clocks if that interests you.
The link for this blog page is broken, as is the link from the blog itself.
Seems to work for me. Which link are you having trouble with?
Now its better.
GPSDOs are about finding the best of both worlds.
GPS can give timing results on the order of the tens of nanoseconds, but it’s short term stability is rather poor because of the effects of the atmosphere and ionosphere that very over the short term.
By contrast, an oven controlled crystal oscillator (OCXO) has excellent short-term stability, but over longer sampling periods will tend to wander.
The art of making a GPSDO is selecting the best time constant for your oscillator. By keeping a looser rein over better oscillators, you can keep the worst of the GPS jitter out of the output, and instead relegate GPS to the role of countering the oscillator’s tendency to slowly wander.
Rubidium oscillators (at least the ex-telecom ones you get on eBay) are actually Rubidium disciplined OCXOs (the FEI ones are actually tunable synthesizers whose reference is the Rubidium physics package). The FEI ones actually suffer on the low end. So they’re ok for lab reference, but too noisy for microwave reference usage or SDR. Rubidium offers the same long term stability as GPS. I’ve had excellent results using GPS to calibrate these oscillators. Once calibrated they are quite trustworthy in offline (that is, no GPS) scenarios. I just got done using such a setup as a reference for a frequency counter at Maker Faire Bay Area, where GPS just doesn’t work because of the heavy amount of radio noise.
What is the purpose of these? I’m all for precision and accuracy, but I can’t think of a use case.
BTS’es (cell phone towers) need that kind of accuracy to cram in so much traffic. Such oscillators could also be used for calibrating radio equipment or for precise physics experiments.
Also a ton of setups like TV studios (keeping cameras and audio in perfect sync avoiding buffers), heck, having very accurate clocks can help with just general networking. I’ve been tempted to make a few tiny (GPS tamed) OCXO for mobile video production avoiding the need to tie cameras together on the sync line.
Almost the first project I did on an FPGA dev board was a basic frequency counter.
There are techniques for increasing resolution – but you also need to work on accuracy. For calibration, you need a reference, and GPS is “the duck’s guts” as some locals would say. [ie. best there is]
The level of accuracy required for a clock is actually sort of astonishing. 10 parts per million is around 30 seconds a month. Any worse than that, and most people would likely regard it as noticeably inaccurate.
When you buy crystals, you can get them with a stability specification of 10 ppm, but what you must understand is that that doesn’t mean that they’re within 10 ppm of the frequency specified. The actual frequency will be determined by the capacitive loading of the system in which it’s installed. What that specification means is that if you buy a thousand of them and put them in the same circuit, they’ll all be within 10 ppm of each other.
This all came as quite a rude shock to me early in the development of the Crazy Clock.
Getting back to your question: what’s the purpose? I use a really, really good GPSDO as a lab frequency reference for my frequency counter, which I use to determine the “batch” drift of a manufactured run of Crazy Clock controllers. I also use it to calibrate an individual board (down to ~0.5 ppm) to order as a value-add for buyers. I have a particularly good GPSDO also because I use it to characterize the performance of GPSDOs that I sell on Tindie.
But fundamentally, GPSDOs illustrate the difference between accuracy and stability. Improving stability is what a crystal will get you, but accuracy is what you need for timekeeping, and by itself a crystal can’t provide that.
These are a fun project to make.. Build, with heavy modifications, a Silicon Chip chip modded to use a USSR era(pre aged) double oven crystal. And a second hand trimble timing GPS module from old GSM towers.
Now it sits on my desk and syncs everything else. Plus tells me the time.
I’m curious about the implementation, but not enough to watch a couple of hours of video. Are the two time sources combined with a Kalman filter?