Budget GPSDO - complete

[MIS42N]

Tell me more about PICDIV.
Thanks
enut11

Uses assembler program for a pic12f675 or similar to create pulses from the OCXO 10MHz input. The processor runs without interrupts so relies on accurate loops to produce the desired output.

There are a few different 1PPS programs. The simple ones produce the 1PPS with no synchronisation so are not aligned with an external time source. More complex programs allow alignment to within ±200ns of another source with suitable external input. Due to the instruction time being a quarter of the clock, it would be difficult to adjust that last 200ns.

I don't know it is worth implementing. You want 1PPS, take it from the GPS module. As long as there is a reasonable sky view, they will continue to output 1PPS. The PIC in the GPSDO keeps stats on lost signal, rarely happens.

However, if 100% reliable (but not necessarily 100% accurate) 1PPS is required, for example to run a wall clock, then maybe PICDIV has a place. The OCXO is always running but with an extended loss of input there will be a CPU reboot and there may be a loss or gain of cycles before discipline is reestablished.

The GPSDO program could be written to give extended holdover (if input is lost then restored, the OCXO frequency is changed to compensate for any lost/gained cycles then discipline reestablished). However, that seemed like more complexity that a simple GPSDO was expected to deliver. Currently if signal is lost for around a minute, the CPU is rebooted and runs with the Vco established during calibration (unless the user interface is used to update the default Vco) until the signal returns.

http://www.leapsecond.com/pic/picdiv.htm

[radio_ham]

@MIS42N  Hi, perhaps somewhat late to this party, however I have been lurking and following with considerable interest. I recently built two of your excellent GPSDO designs using your GITHUB gerber files you posted and they work well. The short term corrections are better than 0.04ppb and -0.04ppb over maybe 3 to 4 hours and longer term corrections over 24 hours tend to occasionally jump to around plus or minus 0.08pbb so at worst around 0.16ppb drift. For the most part the stability is well within 1Hz at frequencies higher than 10GHz. The number of satellites I have in view with good signals will vary between 11 and 15.

My setup for testing is presently a dual output linear DC PSU set at 8v to drive PIC16F1455 and the Oscillator and the other output set to 12v for the remaining needs. I am using NEO M8N sat receiver module. The active antenna is a puck patch antenna mounted high up in my conservatory with approximately a 240 degree view of the sky. The sat rx module is plugged directly into the GPSDO board. The GPSDO is also in the conservatory so there is a temperature variation, particularly when the sun shines through but the correction appears to cope with that extremely well. In these extreme temperatures I find the 256 second correction adjustment works fine.

Some observations from my experience on the build for others contemplating this project are these. I have used male PCB headers for J1, J3, J4, J5 and J8. For J6 and J7 I have used female PCB headers. J6 does not matter so much as this is a two pin header and can be used as a link if so desired but J7 needs to be a socket should anyone wish to plug the satellite board directly into the GPSDO PCB. If you do this make sure that U8, the UA9637 is not soldered on the board or not plugged into an IC socket if used, as this loads the 1pps from the GPS receiver board. I have used low profile PCB sockets for U8 and also for the PIC chip to no obvious detriment. Do I need to improve this? Likely not for my needs, but having got the bug and because I can I have ordered a few VECTRON C4550A1-0213 10MHZ OCXOs from Ali-veryfast and a LEA M8T Huewai board to see if there will be any difference. I had five PCBs made by a well known Chinese PCB Manufacturer for $5.00 plus approx £20.00 UK postage so I shall make two more boards up and eventually compare the difference. Do I look at implimenting my own RTK system? Probably not as this will be entering the realms of diminishing returns!

@MIS42N, congrats on an excellent GPSDO design and a heartfelt thanks for sharing this. I appreciate the time and effort you have put in.

[MIS42N]

Do I need to improve this? Likely not for my needs, but having got the bug and because I can I have ordered a few VECTRON C4550A1-0213 10MHZ OCXOs from Ali-veryfast and a LEA M8T Huewai board to see if there will be any difference. I had five PCBs made by a well known Chinese PCB Manufacturer for $5.00 plus approx £20.00 UK postage so I shall make two more boards up and eventually compare the difference. Do I look at implimenting my own RTK system? Probably not as this will be entering the realms of diminishing returns!

@MIS42N, congrats on an excellent GPSDO design and a heartfelt thanks for sharing this. I appreciate the time and effort you have put in.

Thank you for the kind words.

I wonder if it is a design suited to further refinement as you intend. The original design parameters were (1) to produce something that had respectable performance for minimal cost (2) to self calibrate so there were no annoying parameters that needed fiddling (3) to be self checking (hence the long averaging periods). This did involve some compromises, the main one being the 1PPS detection which is only to the nearest 25ns. Perversely, this has proven to be adequate when using low cost GPS modules that jitter the 1PPS then applying averages. If there were no jitter, it wouldn't work.

You have addressed most of the simple improvements with the well position antenna and the quality power supply and you've noted that environment plays a part too. Varying temperatures do have effects that are hard to pinpoint, so having a stable temperature helps.

However, you are now bitten by the bug, and it will be interesting to see how the design stands up. I hope you share your results.