Yet another DIY GPSDO - yes, another one

[Johnny B Good]

@TomD

 Just a wild guess (I'm still a newbie when using the Arduino IDE) but could it be as simple as an incorrect reference to the nano in the C runtime Library statement?

[TomD]

Team,
problem solved. The reason is not clear but after delete and re-installation of the STM32 package (this time 2.2.0) the compilation and load run well. Now I will test and implement the other options. Many thanks

Again Andrew great job!

[TomD]

Thank you John,
I am a basic Arduino user and the issue was solve by reload of the STM32 package.

Thank you for your quick reply

[TomD]

First test running well 

Next steps:
- implementing the options
- testing
- installation in a nice alu-case
 

[Scrachi]

Hi All,

The PCB is here, but..... not perfect... The footprint I've made for the OCXO was reverted, the connector for the INA219 too and finally I've also reverted the pins GND and VCC on the 74HC14. I need to check if my eyes are not reverted too.... After a surgical operation with some wires all is working as expected.

Attached I've uploaded the schematic and the PCB's Gerber files (with all above issue fixed, but I've not manufactured theses PCB for testing purpose).
I've made 2 versions of the PCB :
- 1st one for 3D printed case, very similar with the version I've already manufactured, I will design a 3D case for printing soon.
- 2nd one to fit on a 100mm large extruded case, all components need to be directly soldered on the pcb except the lcd and the outputs will be connected using u.fl connectors.


EDIT 11-16-2022 : Schematic updated, to correct 74HC14 GND and VCC pins

Hope it will be useful


Mick   

[MIS42N]

The PCB is here, but..... not perfect... The footprint I've made for the OCXO was reverted, the connector for the INA219 too and finally I've also reverted the pins GND and VCC on the 74HC14. I need to check if my eyes are not reverted too.... After a surgical operation with some wires all is working as expected.

My first PCB I did the same. KiCad thinks you are looking from the top, you think you are looking from the bottom. Newby error (lesson learned).

Your 10MHz output would stress if connected to a 50 ohm load, and may not survive a short. You put 220 ohm in series for the 1pps output, maybe do the same for the 10MHz.

Your results may be rubbish for a few days. I've found it can take a week for the OCXO to settle down.

Good work. Interested to see the results.

[Scrachi]

The PCB is here, but..... not perfect... The footprint I've made for the OCXO was reverted, the connector for the INA219 too and finally I've also reverted the pins GND and VCC on the 74HC14. I need to check if my eyes are not reverted too.... After a surgical operation with some wires all is working as expected.

My first PCB I did the same. KiCad thinks you are looking from the top, you think you are looking from the bottom. Newby error (lesson learned).

Your 10MHz output would stress if connected to a 50 ohm load, and may not survive a short. You put 220 ohm in series for the 1pps output, maybe do the same for the 10MHz.

Your results may be rubbish for a few days. I've found it can take a week for the OCXO to settle down.

Good work. Interested to see the results.

Hi,

Well noted for the 10MHz protection, I will do some test and update the files on the above post.

Thanks.

Mick

[TomD]

Hi ,
some remarks (just for verification purpose):
- The 10MHz rectangle output is not buffered (only the 10MHz sinus one) in the KidCAD schematic, as the 74HC14 have 6 unites I would use two of them
- I am a fan of alu-enclosure (especially for HF frequencies) I would suggest to have 100mm (or 99mm) on one side of the PCB and around 5mm gap to the electronic parts. That will allow to slide the PCB in the EU-design alu-enclosure
  without additional mounting-screws. 3d print cases are not a solution for me because of the HF frequencies in my HAM radio shack.
- I would (optionally) have the 1pps/10Mhz pins also as a SMA connector output. On Ali-ex... I bought BNC-Connector to SMA connector and 100mm coaxial-cable in between. This ensure better HF radio handling and easy mounting.

just as remarks not as a critic...

One question: When I power on the OLED show long time ... 'calibrating' and then the DAC value is either very low....4k or very high 56k therefore wrong 10Mhz frequencies about 15HZ too high or low. Then the unit move the frequency in the right direction
but that will take a week to be on target level. Once the DAC was at 35k and the 10MHz was nearly correct. Any suggestion how to optimize the calibration, or deselect and take a start value of 35k? And how?

Thanks in advance


 

[Scrachi]

Hi,
About the 220 ohm resistor to protect the 10MHz output, I've tried to connect one but I've a strange behavior, the output is not anymore stable, I've about 25Hz swiping around the 10MHz with the resistor but nothing without. Someone have an idea about the issue?

Hi ,
some remarks (just for verification purpose):
- The 10MHz rectangle output is not buffered (only the 10MHz sinus one) in the KidCAD schematic, as the 74HC14 have 6 unites I would use two of them
- I am a fan of alu-enclosure (especially for HF frequencies) I would suggest to have 100mm (or 99mm) on one side of the PCB and around 5mm gap to the electronic parts. That will allow to slide the PCB in the EU-design alu-enclosure
  without additional mounting-screws. 3d print cases are not a solution for me because of the HF frequencies in my HAM radio shack.
- I would (optionally) have the 1pps/10Mhz pins also as a SMA connector output. On Ali-ex... I bought BNC-Connector to SMA connector and 100mm coaxial-cable in between. This ensure better HF radio handling and easy mounting.

just as remarks not as a critic...

One question: When I power on the OLED show long time ... 'calibrating' and then the DAC value is either very low....4k or very high 56k therefore wrong 10Mhz frequencies about 15HZ too high or low. Then the unit move the frequency in the right direction
but that will take a week to be on target level. Once the DAC was at 35k and the 10MHz was nearly correct. Any suggestion how to optimize the calibration, or deselect and take a start value of 35k? And how?

Thanks in advance

The alu enclosure, the PCB gerber named "GERBER_ExtrudedCase.zip" is adapted for an extruded alu enclosure, it's 100mm large and made to fit into the case without screws.
For the 3D printed case version, I usually cover the interior of my electronic cases with copper tape grounded to the PCB.


About PWM dac value, I guess you can set it manually to 35k using the serial terminal with the command "SR 35000".


Thanks
Mick

[AndrewBCN]

Hi ,
some remarks (just for verification purpose):
- The 10MHz rectangle output is not buffered (only the 10MHz sinus one) in the KidCAD schematic, as the 74HC14 have 6 unites I would use two of them
- I am a fan of alu-enclosure (especially for HF frequencies) I would suggest to have 100mm (or 99mm) on one side of the PCB and around 5mm gap to the electronic parts. That will allow to slide the PCB in the EU-design alu-enclosure
  without additional mounting-screws. 3d print cases are not a solution for me because of the HF frequencies in my HAM radio shack.
- I would (optionally) have the 1pps/10Mhz pins also as a SMA connector output. On Ali-ex... I bought BNC-Connector to SMA connector and 100mm coaxial-cable in between. This ensure better HF radio handling and easy mounting.

just as remarks not as a critic...

One question: When I power on the OLED show long time ... 'calibrating' and then the DAC value is either very low....4k or very high 56k therefore wrong 10Mhz frequencies about 15HZ too high or low. Then the unit move the frequency in the right direction
but that will take a week to be on target level. Once the DAC was at 35k and the 10MHz was nearly correct. Any suggestion how to optimize the calibration, or deselect and take a start value of 35k? And how?

Thanks in advance

Hi TomD,
First, well done!
And yes, you can disable the calibration, or increase the OCXO warm up time, or optimize the calibration in various ways, and even set the DAC initial value close to the optimum for your particular OCXO, all these options are available in the source code before compilation, and you also have complete control to set the DAC value after the GPSDO has powered up and configured itself.
Check the source code, it's well commented exactly so people can experiment with it.

I am sorry that I have been so busy at work and that should continue until mid-november.  plus COVID-19 twice...

[nealix]

@Scrachi  (Mick):

Are you able or willing to share the KiCad footprint and symbol for the STM32 BlackPill?
I have been watching training videos for Kicad and have it installed.
I wish to try designing my first PCB for this also.   But AndrewBCN got very busy with
his job before I could receive the actual footprint and schematic symbol for the BlackPill.
I suspect by watching some more videos on youtube, I could learn to make the footprint,
but if you already have it and are willing to share, that would save reinventing the wheel.

Cheers,

Neal

[Scrachi]

Hi Neal,

Yes sure, I've got a STM32 footprint bundle including the black Pill from here : https://github.com/piit79/Kicad-STM32


I've modified the schematic and the PCBs adding a 220ohm protecting resistor in series with the 10MHz output as recommended by MIS42N and I've finished the assembly of the 3D printed box, I will upload all this stuff later.


Mick

[Scrachi]

Hi,

I've finished the last version of the PCB including the 220ohm protection resistor in series with the 10Mhz output. I've also finished the 3D prited case and the assembly of everything.

Find attached the latest version of :
- The Shematic
- The PCB Gerber files for the 3D printed case
- The PCB Gerber files for the 100mm aluminum extruded
- The PDF of the front and back panel for the 3D printed case (Note it need to be printed with a scale of 105%)
- The 3D case STLs can be downloaded here : https://www.thingiverse.com/thing:5582069

The inside of the case is covered by copper tape connected to the GND.

[TomD]

Mick, great job looks good and it works!! 

I searching for some support:
Please see the pic below. It seams as it would working well but I have some doubts.
1. The Frequency is the base /default freq. and without the =.xx value.
2. The PWM value is stable at 32400 (I set this value as it was the number that brings my prototype to 10MHz). 
3. See the in..it before the voltage measurement

I transferred the elements from the breadboard (see earlier posted pic) to the development board (final version). Just the 74HC driver was added an the 100nF caps  and a proper power supply.
I check over 2 days all parts several times ... measured the signals out (1pps/10MHz sinus and 10Mhz rec), Vctl=1,67V and sats=12 and GPS signal is fixed and green led on GPS blinking.
Also SW 0.5j and 0.6c tested.
Always the same ... the program loop on 'Please wait Calibrating...'. The only way is to skip calibrating in the program and than it works like on the picture. I guess without GPS disciplined..

Any hint/idea/tip? Everything is welcome

Thank you in advance

[nealix]

Hi Tom:

Personally, I have seen the firmware fail to exit the "Calibrating...." routine if the 1 PPS signal is not
getting into the STM32 CPU on the right pin, or not getting there at all.
Do you have a scope, so you could probe the GPIO pin right at the STM32 module and see if the
1 PPS pulse is showing up?    Also double check that the pin# defined in the code for 1 PPS input
is the same physical pin# actually wired up on your STM32 module.  (I made that mistake once :-)  )

Another mistake that had the same result (failing to exit calibration), is when I connected the 10MHz
output of the crystal oven to the wrong breadboard point (wrong pin on STM32 CPU).   

Again, not sure this is YOUR specific problem, but it is a good place to start for diagnosing.
Try to double-check that the 10MHZ and the 1 PPS signals are both getting into the proper pin
on the STM32 module using a scope or some other test instrument.  And check that the software
code is set to use those same pin numbers for those signals.

Neal

[TomD]

Hi Neal,
thank you, I took your points and measure with a scope the signals direct on the STM32 board. PA15....10MHz check, PB10....1pps check, PB9... PWM out check, PB1...Vctlin check.
Is following line defining PB10?
FreqMeasTim->setMode(3, TIMER_INPUT_CAPTURE_RISING, PB10);
I never found any other PB10 definition so it should be the only one and correct.... Version 006c line 2743

I will now install and test the Bluetooth module to see the serial monitor data under production mode....
Question: I use a DC-DC down converter to get 12V (13,9V) to 6,8V and two LM1084 to have proper 5V and 3.3V. That's a bit over the top but should create proper power.
Is the STM32 sensitive on RF or ripple on the powerline?

[MIS42N]

Question: I use a DC-DC down converter to get 12V (13,9V) to 6,8V and two LM1084 to have proper 5V and 3.3V. That's a bit over the top but should create proper power.

My design also uses the same scheme - nominal 12V->buck converter->7V->two LDOs->5V. The supply is split so the OSC5A2B02 and its control voltage has its own LDO to isolate it from power glitches caused by other bits. It was a worthwhile improvement, so don't think it is 'over the top'. The oscillator now doesn't glitch when equipment under test is attached, previously it was necessary to wait a minute for the system to be stable.

[TomD]

Hi MIS42N, thanks for the confirmation.
I will test with double 5V supply , one for the OCXO and one for the STM32.
Another point, I scoped the pin PA15 (10MHZ) on the STM32 coming from the resistor 180ohm and OCXO out. The OXCO is a CTI OSC5A2B02.
I measure 5V rectangle voltage (refer to the pic). Is the STM32 max voltage not 3,3V for the ADC. Did I overload the STM ADC?
As the eff voltage is 2,5V (5v/2) and just the raising signal is used ==>no problem? Right or wrong?

[MIS42N]

Hi MIS42N, thanks for the confirmation.
I will test with double 5V supply , one for the OCXO and one for the STM32.
Another point, I scoped the pin PA15 (10MHZ) on the STM32 coming from the resistor 180ohm and OCXO out. The OXCO is a CTI OSC5A2B02.
I measure 5V rectangle voltage (refer to the pic). Is the STM32 max voltage not 3,3V for the ADC. Did I overload the STM ADC?
As the eff voltage is 2,5V (5v/2) and just the raising signal is used ==>no problem? Right or wrong?

Sorry, can't help you there. My system is very different. Although it uses the same OCXO (the CTI OSC5A2B02) it uses a PIC16F1455 processor with a 74HC04 to buffer the 10MHz. It all runs with 5V (except the signals from the GPS are 3.3V. The PIC is happy to work with that using TTL level inputs). Originally that was the whole system, but there were problems every time the system was attached to something else. Which is why I went to 12V and a split supply (and optocoupler for the serial TX/RX). If you are interested, it's written up here: https://www.eevblog.com/forum/projects/budget-gpsdo-a-work-in-progress/. The current circuit is attached to Reply #53.

[AndrewBCN]

Hi Neal,
thank you, I took your points and measure with a scope the signals direct on the STM32 board. PA15....10MHz check, PB10....1pps check, PB9... PWM out check, PB1...Vctlin check.
Is following line defining PB10?
FreqMeasTim->setMode(3, TIMER_INPUT_CAPTURE_RISING, PB10);
I never found any other PB10 definition so it should be the only one and correct.... Version 006c line 2743

Yes, PB10 is the 1PPS input (output from the GPS module).

I will now install and test the Bluetooth module to see the serial monitor data under production mode....
Question: I use a DC-DC down converter to get 12V (13,9V) to 6,8V and two LM1084 to have proper 5V and 3.3V. That's a bit over the top but should create proper power.
Is the STM32 sensitive on RF or ripple on the powerline?

Like any digital part, the STM32 is relatively immune to noise and ripple on its power supply inputs.

[AndrewBCN]

Hi MIS42N, thanks for the confirmation.
I will test with double 5V supply , one for the OCXO and one for the STM32.

As MIS42N explained, it can't hurt to separate the 5V power supplies for the OCXO and the rest of the circuit.

Another point, I scoped the pin PA15 (10MHZ) on the STM32 coming from the resistor 180ohm and OCXO out. The OXCO is a CTI OSC5A2B02.
I measure 5V rectangle voltage (refer to the pic). Is the STM32 max voltage not 3,3V for the ADC. Did I overload the STM ADC?
As the eff voltage is 2,5V (5v/2) and just the raising signal is used ==>no problem? Right or wrong?

The STM32F411CEU6 has +5V tolerant digital inputs, so no problem there, as far as I can tell. Of course, one should not feed a +5V square 10MHz signal to the STM32 internal ADC.

[TomD]

Team, problem identified 
I ordered an entire set (GPS/MCU) and use another OCXO to fix the problem (no calibration =>loop/no frequency counting). Build the GPDSO on a breadboard.
I started with 2 resistor 10k (each) to reduce the voltage of the 10MHz and 1pps signal by 50% to the MCU. Same problem... and my scope didn't show a nice signal.
Then I use the original 100 and 180 ohm and the signal improves significant. This also solve the issue. 
So a proper handling of the 10MHZ and 1pps signal to the MCU is key otherwise you will have the same issues as I had.
Now I will solve the issue on the ready installation unit (see above). I guess a better shielding and a coax for the signals will solve the issue as well.

Thank you for your support and the great project.

PS. I love the bluetooth capability as you can control the parameter and adjust if required.

[Johnny B Good]

Team, problem identified 
I ordered an entire set (GPS/MCU) and use another OCXO to fix the problem (no calibration =>loop/no frequency counting). Build the GPDSO on a breadboard.
I started with 2 resistor 10k (each) to reduce the voltage of the 10MHz and 1pps signal by 50% to the MCU. Same problem... and my scope didn't show a nice signal.
Then I use the original 100 and 180 ohm and the signal improves significant. This also solve the issue. 
So a proper handling of the 10MHZ and 1pps signal to the MCU is key otherwise you will have the same issues as I had.
Now I will solve the issue on the ready installation unit (see above). I guess a better shielding and a coax for the signals will solve the issue as well.

Thank you for your support and the great project.

PS. I love the bluetooth capability as you can control the parameter and adjust if required.

 I strongly suspect the trouble you'd experienced lay with insufficient loading on the OCXO's output (less or not true with the buffered square wave output OCXO types). The sine wave OCXOs usually output directly from the oscillator circuit itself which will be designed to operate with a 50 to 75 ohm load (a 100 ohm load proved sufficient in my case to cure the alternating duty cycle effect I'd observed in my own GPSDO which had created a 2ns jitter on the squarer's output). Without that necessary loading, the oscillator amp is most likely to start clipping, introducing a deleterious 'latch up' effect causing a 'pogoing' effect on the sine wave's duty cycle.

 Regarding the PPS output from the gps module, placing a 33 ohm 'stopper resistor' in series right at the output pin helps take the sting out of the odd order harmonics such a fast switching signal can generate (circa 2ns rise and fall times - quite impressively fast imo )

[geggi1]

Would you be able to upload the PCB in scale as PDF or DWG?
I was thinking about etching the PCB at home with the toner acetone method.

Hi,

I've finished the last version of the PCB including the 220ohm protection resistor in series with the 10Mhz output. I've also finished the 3D prited case and the assembly of everything.

Find attached the latest version of :
- The Shematic
- The PCB Gerber files for the 3D printed case
- The PCB Gerber files for the 100mm aluminum extruded
- The PDF of the front and back panel for the 3D printed case (Note it need to be printed with a scale of 105%)
- The 3D case STLs can be downloaded here : https://www.thingiverse.com/thing:5582069

The inside of the case is covered by copper tape connected to the GND.

[Scrachi]

Hi geggi1,

Yes I can try to export it from Kikad if you know how to, please let me know. I guess on the pdf you want only the copper traces?

Mick