I am currently (April, 2017) working with a GPS Disciplined Oscillator (GPSDO), which is an add-in option for the FlexRadio 6500 transceiver. This device is mainly intended as a precise frequency standard. It is not clear how easily one can access precise timing information. Can it be part of an NPTD server? We will see.
Older Project (Garmin 18-X)
I am using a Garmin 18-X GPS, which is an inexpensive bare-bones hockey puck with bare wire leads ($64 at Amazon).
Some links I found helpful:
Using xgps on Ubuntu, I can see about 8 satellites with the receiver a few feet away from my west-facing window. It doesn't seem to require a great fix to provide the 1 pps output.
- The most common application would as a computer time reference. You set up the USB interface as described in the links. That allows you to align your computer's internal clock to UTC to "high accuracy". The accuracy is limited by your computer OS and how well it can track the external pulse, It would be a nice test to use the OS to generate an output 1 pps tick according to its idea of time and compare with GPS. I would expect milliseconds of jitter, at least, because common OS's are not real time systems. Despite the jitter, the long-term time stability should be excellent, and your computer can become an NTP server at "Stratum 1".
- Another application I am interested in is as a way to time-tag received HF signals. See HF Time of Arrival. In this work, we are interested in using a fairly narrow 1 pps pulse as a broadband "tick" that can be added to a received HF signal for time-tagging purposes. The Garmin unit provides a 100 msec pulse that rises precisely on the 1 second UTC tick. How precisely is that? Garmin says ±1 µsec, certainly good enough for time-tagging the audio data.
It turns out (no surprise) that the USB port of my computer is a rich source of RF hash! Using the USB port as an unfiltered source of +5 volts to power my interface is not a great idea. I can hear lots of "music" on my ham receiver that is related to the various activity cycles in the computer. Entertaining and interesting if you're into computer security perhaps, but not good for sensitive RF reception.
Perhaps a separate 5 V supply will be good enough, or perhaps I will have to abandon the USB data connection, too. I don't really need to be at Stratum 1. Ordinary NTP-based Internet time sync is good enough for nearly all purposes.
I came across an Analog Devices USB Digital Isolator chip. Info here. It provides a fully isolated connection that should allow you to use USB connections in noise sensitive systems without getting the computer hash problem. I'd like to try it, but it's a SOIC_W package, which is a little awkward for my prototyping technology...