Query about CPU Clocking

[darksidelemm]

The RS41 boards have a 24 MHz crystal available ('HSE'), but it appears to need to be explicitly enabled to be used. Looking at the code in here it seems like the internal oscillator ('HSI') is being used: https://github.com/Nevvman18/rs41-nfw/blob/main/fw/arduino-ide_variant-files/variants/STM32L4xx/L412RB(I-T)xP/generic_clock.c#L36

Drift information on the internal oscillator, from the STM32L412 datasheet:

RC oscillator drift could lead to errors in modulation symbol rate timing.

I did some basic testing to see what the Horus binary modulation rate error (a 'ppm' value coming out of the demodulator) looks like when comparing a board running RS41-NFW to one running RS41ng (which definitely does use the external crystal). This was only over a temperature drop of about 20-30 degrees, so it's possible the impacts could be worse at lower or higher temperatures.

These results aren't definitive, but do seem to suggest there is some modulation symbol rate offset when running RS41-NFW. I haven't yet quantified what amount of symbol rate error will start to degrade demodulation, but it's something I'll be looking into.

UPDATE: I ran some simulations, and at weak SNRs (-5 dB, which is about the 50% packet error rate point), the demodulator performance starts to degrade with more than 0.3% (3000 ppm) symbol error rate, and completely breaks at around 0.6-0.7% symbol error rate.

RS41ng - starting temp was about 25 degrees C, end was -5 degrees C:

RS41-NFW - similar range to above:

Interestingly RS41-NFW seems to start out low, then get better at the temperature drops.

Anyway, running from the crystal would probably be a good idea...

[Nevvman18]

Hi Mark,
Noted, thanks for the thorough tests and information. I'll work on it in the next release.

Best,
Franek

[darksidelemm]

I'd definitely recommend that you definitely use the external crystal by default.

There has been a lot of issues with the DFM17 boards, which only have the internal RC oscillator. On these boards the RC oscillator frequency is 'trimmed' using the 1PPS from the GPS receiver, and until the GPS receiver has lock, decoding of packets is very unreliable due to the symbol rate error.

I expect the same issue could be happening on the RS41 if you are only using the internal RC oscillator.