If you want an "existence proof" for a simple, mostly analog, a few digital counters, no software or microprocessor GPSDO, look at the "Miller" GPSDO. He designed it for his own use, then put it into production because of demand for a simple, cheap GPSDO. It has been characterized, and works well for the simple circuit.
The schematics are published, since it started as a hobby project. http://www.jrmiller.demon.co.uk/projects/ministd/frqstd0.htm http://www.jrmiller.demon.co.uk/projects/ministd/frqstd.htm You can buy his, or build your own with his design as a starting point. He does use a GPS with 10 kHz output to simplify some of the timing and integration issues Bob referred to. --- Graham == On Mon, May 23, 2016 at 8:21 PM, Bob Camp <kb...@n1k.org> wrote: > Hi > > Ok, so how would you do a pure analog GPSDO? > > The GPS receiver and that side of it are what they are. I’ll assume that > you have a 1 pps out of a module. > > Your OCXO needs to get to 1 Hz via dividers. You can do that with digital > dividers or with a chain of regenerative > dividers. One is a bit more analog, the other may be “ok” under the “don’t > go to crazy” ground rule. > > You now have a PPS that is off somewhere relative to the GPS. A push > button will get them into rough alignment. > Your OCXO is quite likely a bit high or low. A multi turn pot on the EFC > will let you get it within 1x10^-9 without a > lot of crazy work. A reasonable counter tied to a reference will let you > do this. > > Net result: The pps signals are roughly aligned and drifting < 1 ns / s. > Considering the delta between them is > bopping around by 10 ns, that’s quite good. > > Run a very normal bipolar charge pump off of the delta between the two pps > signals. Fire a sample and hold when > the transition is over. You now have a (maybe) +/- 60V signal that > corresponds to the phase error. Since you are using > film capacitors, the 60V comes along for free. Taking it to the maximum is > just a way to save money on caps. > > Next up, do a fairly simple 20 second time constant R/C filter. That will > take out a lot of the hopping around and make > the rest of the system a bit easier to quiet down. You now have a somewhat > linear +/- 60V signal that tells you how > far off phase the setup is. After the RC you have a high input impedance / > low drift buffer amplifier. Yes that’s a little > tricky. > > Next you need a P and an I term. Both need to be variable as the system > calms down. A rotary switch will do fine for > this. Relays might also do the job. The P is a bank of resistors, each one > to scale the buffered R/C to your control amp. > The I goes off to a similar set of resistors driving an integrator. Net > time constant there will be in the 200 to 2,000 second range. > That’s were the ovenized caps come in. You also need a really good amp as > part of the integrator to buffer out the signal. > > The nice thing about doing it this way is that you can *see* it all > happening. There is a nice *clunk* noise as the filter > steps off. Each number in the filter has a (likely large value) resistor > that sets it up. To change the filter characteristics, > you swap out resistors or twiddle pots. > > If you do the math, even with 60 V on the system, you probably don’t want > anything over 1 meg ohm involved. At 2K seconds > that gets you to a pretty big film capacitor bank. Even the 20 second > lowpass isn’t exactly small by the standards of fancy > capacitors. > > There are a few interesting tidbits like wire wound / high value / low > temp co resistors that would help things a bit. Swapping > those in and out as you change filter settings experimentally could get a > bit crazy. > > The net result should be a good starting point for a GPSDO. You still > would need to spend all of the time working out values > and matching it up to your OCXO. The need for a good local reference and > good measurement gear while doing this still is > a limit, just like the pure digital approach. > > Bob > > > > > On May 23, 2016, at 12:46 PM, Nick Sayer via time-nuts < > time-nuts@febo.com> wrote: > > > > > >> > >>> If that sounds too weird, I am open to receive advises for a > microcontroller based solution. > >> > >> If you want to go that way, probably the simplest solution would be to > >> take one of Nick Sayers boards, pull out the GPS receiver and feed the > >> PPS input from your GPS receiver. > > > > It’d be kind of an awkward fit. For the OCXO/TCXO, you’d need to pull > the oscillator as well as the GPS (I believe you said you had an oscillator > already), and your EFC would be 1.65 volts wide centered on 1.65 volts. > That’s unlikely to be absolutely correct for your oscillator. You could > change around the Vref for the DAC, but at that point I’d consider > redesigning the board for your purposes instead. > > > > That said, I think it’d be easy to adapt the circuit and code for a more > arbitrary setup. And I believe my system is good down to the ADEV 10E-11 > level at tau 1s or so. I don’t know how much better it can do, as I’ve > simply not tried to go below that (and I likely couldn’t properly measure > the results anyway). > > > > There’s also the FE-5680 board, but it has an RS-232 level shifter in > place of the DAC. On the other hand, it does have a very nice 2A @ 15V > power supply, which likely is very close to what you’d need for a really > good OCXO. A mash-up of that with the DAC put back in might be closer. But > either way, you’re designing a new board, I think. > > _______________________________________________ > > time-nuts mailing list -- time-nuts@febo.com > > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > > and follow the instructions there. > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > _______________________________________________ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.