Timer-driven Single word Memory-to-Memory DMA transfer, can it be done in HAL??

Christopher Pappas · ‎2017-02-22

Posted on February 22, 2017 at 17:04

Hello everyone again!

I have a special case scenario here that might be impossible to implement.

I wantto transfer (using HAL) from Flash memory to SRAM memory a single word at a time via DMA, using a timer to control the rate at which the transfer happens (using TRGO_Update), but amhaving no luck triggering the transfer at all.

Cananyone offerany suggestions, ideas, or insight on how to accomplish thisoperation in HAL?

Thanks!!

Christopher

#timer #dma #transfer #mem-to-mem

Note: this post was migrated and contained many threaded conversations, some content may be missing.

waclawek.jan · ‎2017-02-22

Posted on February 22, 2017 at 19:37

Which device? What have you done so far?

JW

Christopher Pappas · ‎2017-02-22

Posted on February 22, 2017 at 20:19

Hi Jan,

It is the STM32F767ZI, and the project is being built on the NUCLEO-F767ZI board.

I modified the example 'DMAFlashtoRam'.

The timer (TIM9) is working, and the DMA is transferring from a FLASH buffer of constant words to a single word location in SRAM. I just can't seem to get the timer to control the rate of the DMA xfer (i.e. - it is always running in full speed rather than the rate of the TIM9 Update frequency).

Thanks for taking the time to assist me!

Christopher

waclawek.jan · ‎2017-02-22

Posted on February 23, 2017 at 01:15

Christopher,

I don't Cube/CubeMX so can't help with that
only DMA2 is capable of memory-to-memory transfers in the 'F7xx
each DMA Stream/Channel has a fixed specified trigger source, see DMA2 request mapping table in RM0410 - you have selected DMA2, Stream0 Channel0 - that's triggered by ADC1, so it's good e.g. to transfer the result of ADC conversion to memory
TIM9 is not capable to trigger DMA transfers. See TIMx_DIER register in the timers chapter(s). That's where you need to select/enable, which event within the timer will cause a respective DMA trigger
besides TIMx_DIER, go back to the DMA2 request mapping table - as you can see, only TIM1 and TIM8 can be used as triggers for DMA2
don't use a memory-to-memory mode of DMA. This sounds counterintuitive, but that is a mode which is not triggered from outside, but simply triggers itself permanently to yield as fast transfers as possible. That's what you've observed. Use either of the peripheral-to-memory of memory-to-peripheral modes, it does not matter which one. The direction then determines which address will go to which address register, but it's likely Cube will hide this for you (again, I don't Cube).
isn't there an example for timer-triggered-memory-to-whatever (most probably -to-GPIO) DMA?

JW

Christopher Pappas · ‎2017-02-22

Posted on February 23, 2017 at 04:22

Thanks Jan!! Everything you suggested makes good sense. I am using TIM2, M2P, and DMA1. The request mapping table Channel 3 Stream 1 is triggered by TIM2_UP, so I chose that. I set TIM2_DIER to trigger on Update.

Still not working yet (I can't get the DMA xfer to begin at all now), but on the right track.

Jan Waclawek · ‎2017-02-23

Posted on February 23, 2017 at 09:00

I am using TIM2, M2P, and DMA1.

You can't use DMA1 for memory-to-memory transfers. As I said above, use DMA2 and TIM1 or TIM8.

JW

Christopher Pappas · ‎2017-02-23

Posted on February 23, 2017 at 14:45

I am using Memory-to-Peripheral now (M2P), as I said before!

All the pieces of the puzzle are there in the code now, I just have to figure out Cube/HAL is stopping it from working.

waclawek.jan · ‎2017-02-23

Posted on February 23, 2017 at 14:55

Just do what I told you, don't use DMA1. I'll explain it later.

JW

Christopher Pappas · ‎2017-02-24

Posted on February 24, 2017 at 21:45

There is a TEIF1 Stream x transfer error interrupt happening on the first word transfer.

I will leave it as is for now, unsolved. Maybe I'll come back to debugging it at a later date.

waclawek.jan · ‎2017-02-24

Posted on February 24, 2017 at 21:52

Without downloading those whopping 6.5MB - if you are using Direct Mode (i.e. FIFO off) and it's a FIFO error, then just ignore it (leave it disabled in the interrupt masks). You can find the explanation somewhere on this forum, and it's harmless.

JW