Add support for 2-pin stepper driver

[GrbavaCigla]

I’ve implemented a simple PIO-based driver for 2-pin (STEP/DIR) stepper motors for my own project, and I noticed that this library currently only provides a 4-pin stepper driver. Would adding support for 2-pin stepper drivers be of interest to this project? I’d be happy to open a pull request if it aligns with the project’s goals.

Additionally, I was considering adding acceleration support, but I’m not sure how feasible that is using pure PIO assembly. Any guidance or preferences on that would also be appreciated.

Also, I have a question regarding @CBJamo pull request:

embassy/embassy-rp/src/pio_programs/stepper.rs

Line 75 in b8afe4f

let clock_divider = calculate_pio_clock_divider(freq * 136);

The target value is multiplied by 136. I assume this corresponds to the number of instructions in a loop, but I wasn’t able to reconcile that exact number when counting. Could you clarify how 136 is derived?

[CBJamo]

I didn't write the stepper pio program or the setup code, I only moved it into embassy-rp instead of being in examples. It looks like @kalkyl did the stepper pio example I pulled from.

Additional pio programs are welcome. I moved them into embassy-rp to make them more reusable and to focus effort into making them all bug free and featureful. I think baking acceleration into the pio program would be difficult, but it'd be an interesting thing to try. If not, baking it into the higher level rust interface is still very useful. Take a look at the ws2812 module as an example of the higher level code doing some of the work, while the pio handles the timing sensitive parts.

[kalkyl]

Hi! Yeah i wrote that one, sorry for not being more clear in my calculation there... It was a long time ago but from the top of my mind without doing the math again: freq = "number of full 4-phase sequences per second" and i guess i counted 34 cycles per step in the pio asm, times 4 for a full 4-phase sequence = 136 cycles. +1 cycle every 8 steps, but close enough for rock n roll

[GrbavaCigla]

Thanks for the input! I’ll draft a pull request once I’ve finished implementing acceleration and deceleration.

[GrbavaCigla]

I’ve considered multiple approaches and I’m not sure which is best.

My plan is to use a PIO program that generates pulses with a loop-based delay, where both the delay and pulse count (N) are fed via DMA. During acceleration/deceleration I’d stream delays with N = 1, and during the cruise phase I’d use larger N with a constant delay.

My concern is memory usage, since acceleration requires many small updates.

Another approach I’ve considered is using periodic IRQs to adjust the frequency or push newly calculated delay values instead of precomputing everything and sending it via DMA.

[GrbavaCigla]

I’ve implemented a solution in my project with what I think are reasonable trade-offs. I’ll finish it soon and then port it back to the embassy library.