10MHz GPSDO by "BH3SAP"

[KungFuJosh]

You definitely have a different firmware, I don't recall that setting existing in mine.

[flash2b]

The PPb / PWM menu starts after powering up with a connected antenna after PPS on the  ATGM 336H-5N-31 starts blinking in Red (1Hz). This means the unit receives GPS/Beidu information from the satellites. This may take some time and the antenna position must be ok to receive signal but for my unit it starts very quickly.

If no satellites found, the unit stays on the "10.00Mhz / GPSDO" main screen.

The PWM adjusts the 1 unit PWM = 0,000076 V adjustment voltage to the OCXO. Saving needs to be done so that the next time the unit starts it goes more quickly to 10MHz output.

[KungFuJosh]

I saw the PWM screen. I never paid attention to it after startup, and never saw a PPB value. Oh well, too late now.

[flash2b]

I see you're a big fan of lab instruments, -I have been using them for 30 years 

Well you did not spend much on you FNIRSI 1014D for the PPS measurement which is also a "toy" compared to a real oscilloscope, while the topic starter had a Siglent SDS2504X HD.

[flash2b]

Today I had some time left to examine the start-up behavior of this GPSDO. For this i warmed up (about 4h) my Raca-Dana 1998 (Option 04E) and played around with TestController. It took longer than expected, since I am not a TestController expert, to produce the graph that is attached below but it was a success.



The GPSDO was already calibrated and config saved. It shows that the output stabilizes after about a minute, before that the PPS is already blinking (unit is open) so satellites are found, the GPS modules and antenna work very good with this unit.

[KungFuJosh]

TestController is great, but there's definitely a learning curve, especially when making a chart. Look at my settings for chart screenshot, specifically at the SDM3065X Vdc values for some hints at getting clearer info & better details.

[flash2b]

Yesterday, after a very long measurement session where a friend of mine brought a Miller GPSDO, we hooked up all our counters to the Miller GPSDO on the Ext Ref input and the BH3SAP GPSDO to the counter input with the following result:
 

The difference was 0.001Hz, so not bad  at all. Apart from the quirky controls this GPSDO is actually very good value for money and accurate as well.

[montecri]

Wondering if somebody got the firmware code for it. Can think of several improvements that can be made.

[flash2b]

montecri, do you also have this GPSDO?

If yes, do you have the same version and what is your accuracy compared to other standards. What improvements are you looking for.

[montecri]

Yes, got the exact same model, purple PCB.

Unfortunately, I don't have a standard against which to compare it yet. I used it to calibrate a CTI OXCO using an oscilloscope to put both channels in sync by adjusting the CTI trimpot.

I'd like a richer interface, with lock status, number of satellites in view, uptime, etc.

[807]

I noticed the gravitational effect on the xtal oscillator while playing with this unit.

Using the Leo Bodnar Mini GPSDO as an external reference of my frequency counter & measuring the frequency of this unit, it varies slowly between +2mHz & -2mHz when sitting on a level surface. When the unit is tilted upwards so that the display is uppermost, the frequency increases by around 30mHz. When tilted so that the display is downwards, the frequecy decreases by a similar amount.

It can be seen more noticably when using a 2 channel scope. When left in any orientation it gradually recovers over 2 to 3 minutes.

I have the purple PCB with date code 240705.

The display could have been made much more user friendly.

[dankar]

I bought one of these recently and ended up taking it apart, doing some reverse engineering, and writing a new firmware for it: https://github.com/dankar/gpsdo-fw

I'm not sure if my calibration against the GPS PPS is correct, but it seems to work. I started to reverse engineer the original firmware but that ended up taking too much time.

More information on how the firmware works is in the README.md. Please feel free to try it out. Suggestions and contributions welcome.

There are currently no build/flash instructions, but it's a cmake project using arm-none-eabi-gcc. I used stm32cubeMX to generate a lot of the init code, the project file is "gpsdo.ioc". The Blue Pill dev board inside has SWD connector which can be used to flash it.



Displays shows, from top left to bottom right: PPS indicator, number of satellites, current measured PPB error and UTC time.

EDIT: Oh, and this was quickly thrown together. The code can definitely use more love.

[flash2b]

Hi dankar, thank for you effort for new firmware. Is it possible to add a "bin-file" on your GitHub file so that we can prepare an empty bluepill for easy programming ?

[dankar]

Sure: https://github.com/dankar/gpsdo-fw/releases/tag/v0.0.1

"release.zip" contains both an elf and a bin. This is your reminder to backup the current firmware before flashing this one. I have the original one saved, but uploading it somewhere is probably not okay.

EDIT: Btw, if you plan on replacing the bluepill, you need to know that there is a wire under it, attaching to one of the legs of the xtal. On mine I can lift the bluepill from the headers without cutting the wire, but only barely.

[montecri]

For sure following and testing this project. Thank you!

@flash2b Can you work your magic with the frequency counters on this new firmware?

[montecri]

Just flashed your 0.0.2 version. Testing...
Just flashed your 0.0.3 version. Testing...

[dankar]

Yeah, I did some changes and is now up to v0.0.3. Doesn't mean it's better though Second "page" accessible via the encoder now shows more digits of the measured PPB.

Some thoughts:

* The VCO is controlled with a simple PWM DAC. In case you want to run the USB peripheral to get a virtual serial port with comms/NMEA you would need to run the timers at 10MHz. We get a bit more resolution when measuring the frequency when running at 70MHz (assuming the PLL maintains a good lock). But running at 10MHz and using 16 bits of the PWM, we only get 150Hz out. This might be a bit low for the DAC circuitry, which I haven't investigated closely (it's a couple of low pass filters, but I haven't checked the values of the components). At 70MHz and 16 bits we get 1kHz. It would be possible to run USB and get PWM at 1kHz with 13 bits precision instead. Does this matter? I haven't done the math to check that.

* It might be a good idea to switch to some averaging for the adjustments once it's close to stable.

* The 128 seconds averaging to show ppb on the display is perhaps a bit overkill, but from some quick math it's needed to get down to mHz range. On the other hand, it's currently just an indication and is not used for the adjustments.

* It might be interesting to generate the output PPS from the MCU instead of from the GPS module. The GPS module can skip pulses and have some phase jitter, but the MCU output would be rock solid since it's clocked by the OCXO. One could also add a buffer amplifier/schmitt trigger to fix the impedance issue. But then again, at that point it might be more reasonable to build one of the many open source GPSDOs.

* If my measurements are correct, it seems to be sensitive to ambient temperature changes. Insulation around the OCXO might be a decent idea.

[montecri]

I don't have yet the equipment needed to test the accuracy of the device. For now, can contribute with functionality tests and suggestions only.

Mine has been running solid on 0.0.3 for the past 3 hours. Slow to reach the desired precision, but that is more or less given on a long-running device like this one. Some time to reach the working temperature and settle on the desired PPB is kind of expected.

Suggestion: How about making the PPS screen indicator a "spinning bar"? Like, alternating in sequence the characters | / - \  |  / - \ to create the illusion of a bar spinning?

[dankar]

I don't have yet the equipment needed to test the accuracy of the device. For now, can contribute with functionality tests and suggestions only.

Mine has been running solid on 0.0.3 for the past 3 hours. Slow to reach the desired precision, but that is more or less given on a long-running device like this one. Some time to reach the working temperature and settle on the desired PPB is kind of expected.

Suggestion: How about making the PPS screen indicator a "spinning bar"? Like, alternating in sequence the characters | / - \  |  / - \ to create the illusion of a bar spinning?

It shouldn't be that slow. If my memory serves, the original FW reaches it much faster, and there's a lot that could be improved in that department.

Suggestions on the UI are very welcome! It's hard to squeeze in meaningful information on 8x2.

[dankar]

I've now also added access to the GPS module via two pins on the bluepill. The module (ATGM336H version) can be configured via GnssToolkit and it seems like you can configure it for stationary usage. Not sure exactly how to do it though, and I don't think it would matter that much for this device.

EDIT: Firmware should now also work if you have a ublox-module, which it seems like someone had.

EDIT2: README is updated with current state of the firmware. I don't have any other ideas on what to implement for the moment, unless anyone tests the firmware and finds that the calibration isn't working, or something along those lines.

[Jdw1953]

I also have newer purple board (240705) and agree the menu is not very robust. Anyone figure out what is intended by adjusting CONTRAST? The number changes with rotation of knob but nothing I can see is changing with it.

[flash2b]

Even newer than my purple board (240617) so getting popular ?

Contrast on my GPSDO seems also have little effect turning the knob. Since there is no additional info on the firmware on the internet available we can only guess.

[montecri]

Guys, check out dankar's firmware v0.0.12 for this GPSDO. For me, it's much better than stock (to which I'm never going back).

[Gertjan]

I've now also added access to the GPS module via two pins on the bluepill. The module (ATGM336H version) can be configured via GnssToolkit and it seems like you can configure it for stationary usage.....

Hi dankar,
Do you think it is possible to get NMEA data out of this GPSDO?
It would be a big improvement to be able to monitor reception (number of satellites, signal strength) etcetera. That would make one more confident in the quality and reliability of it's output signals...

[dankar]

I've now also added access to the GPS module via two pins on the bluepill. The module (ATGM336H version) can be configured via GnssToolkit and it seems like you can configure it for stationary usage.....

Hi dankar,
Do you think it is possible to get NMEA data out of this GPSDO?
It would be a big improvement to be able to monitor reception (number of satellites, signal strength) etcetera. That would make one more confident in the quality and reliability of it's output signals...

That has been done, but you need to solder wires onto two of the pins of the MCU board (PA2 (TX) and PA3 (RX), 9600 baud). Unfortunately there is no easy way to use the USB port on the board for that. My firmware also shows the number of satellites on the display. But be warned, I don't think anyone has tested the accuracy of my calibration code yet.