STM32H723 .. 735 RM0468 System Architecture table wrong

[LCE]

... or something I don't understand?

Looking at RM0468, 2.1 System Architecture "Table 2. Bus-master-to-bus-slave interconnect" shows that DMA1 and DMA2 have access to basically everything except TCM, including APB4 peripherals and SRAM4.

Is there some hidden feature to let DMA1 / DMA2 actually access "D3 domain"? If yes, how?

Am I reading that wrong, or is it another documentation blunder?

It would be so helpful to use DMA1/2 with SAI4...

[LCE]

Okay, at least DMA1/2 can access SRAM4!

Now only the "APB4 peripherals" entry is wrong?

[Tesla DeLorean]

Seem to recall APB4 / GPIO / RAM4 has it's own DMA unit, BDMA on H7's ?

[waclawek.jan]

Not all 'H7 created equal.

Should be complete connectivity from D2 to D3 in this sub-family:

JW

[LCE]

@Tesla DeLorean yes, BDMA can handle D3 internal stuff. BDMA is a little different from DMA1/2, mainly working without FIFO, but I can probably live with that

@waclawek.jan it looks per DS/RM as if DMA1/2 could access the D3 peripherals, but they cannot because the DMAMUX for DMA1/2 has no connection to the D3 peripherals. That's a little misleading, at least.

At least DMA1/2 can access SRAM4 in D3, that might help.

[waclawek.jan]

> DMAMUX for DMA1/2 has no connection to the D3 peripherals

That does not mean that DMA1/2 can't *access* D3 peripherals. DMA *requests* (aka triggers) and DMA *transfers* are not bound to each other, even if that's the most common usage (there may be a logical bound, though, which may be not that obvious - e.g. a request).

For example, you can use TIM4 to trigger DMA1 to transfer from memory into SPI6 DR, thus initiating an entirely correct SPI Tx.

Maybe the SAI4->TIM5_ETR interconnection can be abused for what you intend to do? I don't use this chip and yes, it's apparently a beast.

JW

[LCE]

> Maybe the SAI4->TIM5_ETR interconnection can be abused for what you intend to do?

That sounds interesting, I'll check...