I2S - How to sync the start of another DMA with the begin() of the I2S (get access of the DMA pointer of the AudioBufferManager of I2S)

[BertheGrandPied]

Hello,

I try to get a sync buffer with I2S to generate TTL signals (with PIO) synchronise with I2S generation.
I managed to get it done by using another AudioBufferManager sending datas to my PIO program through DMA of this object.

But I got a last problem, I'm not able to perfectly synchronied their start together.
There is a little delay when doing successive call to begin() function of I2S (calling begin function of its AudioBufferManager) and the begin function of my AudioBufferManager for my TTL.

One solution seems to get access to the the DMA pointers of both AudioBufferManager. That could allow to use the option MULTI_CHAN_TRIGGER of RP2040 DMA system.
But this pointers are private on the AudioBufferManager object and it's seems impossible to get access to them.

Is there the possibility to get access to them that I didn't get ?
If not, could be a useful feature added to the object in future or this is a bad idea ? (and that would need to allow access of I2S's AudioBufferManager, it's private too)

Thanks for any help,

LD

[earlephilhower]

I think you need to synchronize the PIO and not the DMA to do what you want. The DMA runs basically async to the PIO, filling the PIO FIFO when it can. If you need something to flip the exact CPU clock the I2S starts, I think you need to write a custom PIO assembly file with an additional sideset pin. Tour main code could set it to 0 and the PIO code, as soon as it emits the 1st I2S bit could set it to 1 at the exact same instant.

As far as exposing the underlying ABM for I2S, I don't think that's a good idea in general because you can break things in strange and unusual ways . For very advanced users who want to program their own DMA controls they can either hack their own installation or rewrite their own I2S and ABM to provide that very specific functionality using the code here as a base. For all other general users, they just want to output sound in whatever format and speed...

[BertheGrandPied]

Thanks a lot for this clarifications and ideas. The problem make more sense now.

I will try to write a custom PIO, seems a really good exercise to learn more about it !

[BertheGrandPied]

A quick update to mark this as answered (the problem was the one you pointed out).

I was able to make a perfect sync with your advice and get a really better understanding of PIO programs.
The final solution I find doesn't use additionnal sideset, I used WAIT commands (on LR clock pin) in my PIO driving my TTL and that did the trick !

Many thanks

LD