I wrote some code, the way i usually like by making references to datasheet chapters etc. Everything was going well but the card did not work. I knew it was ok since it worked when plugged to my PC, but it was not responding to my init commands over SPI.

I googled and googled but nothing seemed to work. I tried my no-name logic analyzer but for some reason linux didnt recognise it - it used to work a few years ago but no more i guess.

Eventually i decided to check all the data that is being transmitted to the STM32 chip by changing all the writes into read/writes. Turns out MISO line was high all the time, non stop. This activated some alarm bell in me since it looked like an electrical error.

Eventually i found some random thread about arduino with what seemed to be a similar problem. What i did not mention is that SD cards work with 3.3v (afaik) which is not a big deal since stm32 also works on 3.3v. Sometimes one might want to plug them into a 5v circuit though and to accommodate that, the board i bought contained a voltage level regulator type of chip forgot the name. Essentially converter between 3.3. and 5v signals.

Apparently, in some cheaper no name modules like mine, this type of chip bugs out and keeps on holding the MISO line high despite the fact that SD card has let it go long time ago. This sounds like exactly what is happening to me.

Solution? Glorious:

Bypass the voltage converter. 3.3v should still be considered high state even by 5v circuits, so it should work.

I appreciate the random person who just bumped into a random thread after 2 years of inactivity and just dropped this knowledge. These folks are the backbone of the internet and im glad to have been in his position once or twice myself.

I myself did not feel like soldering that wire, especially since the pad on the socket is super small and these modules are very cheap. Since STM32 works on 3.3v anyways, and atmegas also have low voltage versions nowadays i decided to just buy another module which has no additional ICs on it - no voltage regulators, converters etc, just plain SD card socket and routed connections, maybe a resistor or two.

Funny thing in the end, but also kind of irritating, since it took me long enough to finally sit down and start tinkering with this stuff. Also it might turn out there's a different problem at play and i just bought some more electrojunk for no reason. But then again, i put some more sensors and SPI fun stuff into the basket when i was orderign the new module, so there's always new stuff to look forward to.

Don’t use the Raspberry Pi 2040 to build a DCC decoder.

It’s not really wrong, of course, and it gives you ridiculous amounts of computing power, but to get that you need miles and miles of extra components. Like a dozen of decoupling capacitors, an external oscillator, and of course an external flash memory. You can maybe do that sort of thing in H0, but for real model trains, it’s just not ideal. Pick something like an ATTiny or STM32C0 (I’ve done both for different projects): Cheap, just one or two decoupling capacitors, no external oscillator required and no external flash memory needed either.

The versions of the RP2350 with built in memory may be a better idea, once they fix the GPIO issues anyway, but still, there’s the oscillator and just way too many capacitors.

(Some of these concerns may be different if you want to add sound support, which requires more memory anyway. But I’d argue that sound in model railroads is a mistake anyway. I’ve never seen an N scale locomotive that sounds good, it’s all just an annoying expensive gimmick.)

Also while we’re at it, don’t use a cheap standard bridge rectifier. The steep slopes of the DCC signal mandate a really fast rectifier, typically Schottky diodes. This is particularly important when you have RailCom on your layout. You may need four individual diodes instead of one rectifier, but since you won’t need all the capacitors of the RP2040, you still win out in the end (it’s also typically smaller than the rather large SMD standard rectifiers).

STM32F103C8T6 Development Board STM32 ARM Core Module The STM32 Development is based on the ARM Cortex M3 STM32F103C8T6 Microcontroller from STMicroelectronics. The controller works on 3.3V and can be powered by 5V since it has an internal 3.3V regulator. It has dual crystal oscillators (8MHz and 32KHz) which runs the internal RTC and supports Deep sleep modes for energy constraint battery operated applications.The Board also has two on-board LEDs, one red color LED for power indication and a Green color programmable LED connected to GPIO pin 13. The Board should be toggled between MCU boot mode and Programming board using the two header pins.