DSPIC33: SPI AND DMA

[dman232323]

Hi,

I would like to read data from flash using SPI and DMA.

I understand that i will need 2 dma channels, 1 to read and 1 to write while I read (dummy).

I can't seem to make it work.

[Ttelmah]

My own setup for doing SPI receive on SPI2, is:

setup_dma(0, DMA_TRIGGER_SPI2RX, DMA_BYTE | DMA_RELOAD_ADDRESS);

Then the DMA start is:

dma_start(0, DMA_SOURCE_ADDR_UNCHANGED | DMA_INC_DEST_ADDR | DMA_REPEATED, DMA_ADC_RX, getenv("SFR:SPI2BUFL"), 256);

I trigger by sending the setup sequence using DMA1. The chip's reply is
then what starts the receive.

[dman232323]

[Ttelmah]

The key thing you are missing is the difference between DMA_CONTINUOUS
and DMA_ONE_SHOT.

[dman232323]

OK, could you please elaborate?
because i think I tried some variation with continues,
but as I see it,
there is a point in my program, in which i would like to get data from the FLASH. It's a spesific moment, let's say it's a user pressing a button.
So, for me, it sounds like a one-shot...

[Ttelmah]

No.
The transfer stops after a block either way, but with 'continuous' the
address counter goes back to the start of the buffer. With one shot it
doesn't. If you don't select this, unless you do another 'setup', it goes
on up the memory.
It goes 'continuous', if you select ping_pong & continuous. Otherwise
it still stops.

Try this:
dma_start(1, DMA_CONTINUOUS|DMA_FORCE_NOW|DMA_PERIF_ADDR, dma1_buffer,256);

I must admit I've never tried sending the setup without DMA. I send the
setup as a defined block with the target address using the other DMA
channel, and the completion of this triggers the receive.

[dman232323]

Regarding the last part of your message, i also tried sending the commands as part of the dma0

int8 dma0_buffer[256] = {RD_OP, ADDR_2,ADDR_1,ADDR_0,DUMMY};

I looked at the signals going out on the MOSI line, they look ok.

regarding the main part of your message, i tried it, i get 0's back.

so my line of code now looks like this:
output_low(FLASH_CS);
dma_start(1, DMA_CONTINUES|DMA_FORCE_NOW|DMA_PERIF_ADDR, &dma1_buffer[0],256);
dma_start(2, DMA_ONE_SHOT|DMA_FORCE_NOW, &dma0_buffer[0],256);
output_high(FLASH_CS);

tried also continues with dma0 did not make a difference

so if MISO lines are OK
There is a problem with the recieve DMA
because when i use standalone SPI it all works fine, MISO MOSI , data is valid.


on a side note, when using continies, i noticed the lines of CLK and MOSI are always active, the dont stop...

Thanks for the help

[Ttelmah]

OK.
The problem you have is that nothing is starting the clocking.
Think about it without clocks, the bytes can't read, and nothing is
saying to the SPI 'start clocking'.
I started the transaction by sending a null byte. This then makes the
receive buffer load, and the SPI starts receiving, and the DMA reads
this.

[dman232323]

[Ttelmah]

No, you are asking it to receive. You are reading the RX buffer, not
writing. That is the point. To send, you have to write, then the read
will occur.

[dman232323]

I will not be at the lab for the next couple of days, but i can't let go

So you are suggesting that on the dma channel i am writing, i will keep the buffer to a size of 5 which is the size of the instruction i am required to send in order to read.
And then execute, when i get a flag that it has ended, the start of the read dma which is let's say 256 or whatever amount of bytes i want to read?

[dman232323]

So... I am gonna need a little more help here.
I am now using DMA0 just to send the data in order to enable the read: OPCODE ADDRESSX3 and DUMMY bytes required by datasheet => 5 bytes.
EDIT:
I seem to be able to do one read of 256 bytes from the flash, but i cannot re-generate the operation, to read multiple pages (chunks).

Here is my current code:

[Ttelmah]

The reading device can start the transfer. You don't need to use the write
device:

[dman232323]

Thanks!
Got it to work.
its functioning well now.

Thanks Ttelmah for the help