SPI slave receive and return array using SPI.transfer(data,size)

Projects Programming
1

Hi,

I try to do that :

Master send [1,3,6]
Slave increment items and return [2,4,7]

MASTER :

byte datas[3] = {1,3,6};

SPI.transfer(&datas,3); // or SPI.transfer(datas,3);

Serial.println("Master[0] =>"+String(datas[0]));
Serial.println("Master[1] =>"+String(datas[1]));
Serial.println("Master[2] =>"+String(datas[2]));

SLAVE :

volatile byte buff[3];
byte indx;

void loop()
  if (process)
  {
    Serial.println("Slave[0] => "+String(buff[0]));
    Serial.println("Slave[1] => "+String(buff[1]));
    Serial.println("Slave[2] => "+String(buff[2]));
    process=false;
   }
}

ISR (SPI_STC_vect)
{
   byte c = SPDR; // read byte from SPI Data Register
   buff [indx++] = c;

   SPDR = c+1; // return increment
   
   if (c==6) {
    process =true;
   }
}

Slave receives the right datas :
Slave[0] => 1
Slave[1] => 3
Slave[2] => 6

but master return :
Master[0] => 7 (expected 2, but return last item)
Master[1] => 3 (unchanged, expected 4)
Master[2] => 6 (unchanged, expected 7)

What is the solution ?

Best regards,

Vincent

2

have you tried Nick Gammon's tutorial of SPI communication between arduinos?

does that work OK with your setup?

3

Yes, it works byte by byte, but i try to use SPI.transfer(array, size) function.

4

vblocs:
Yes, it works byte by byte, but i try to use SPI.transfer(array, size) function.

one possible reason why that function is not working here is that it is too quick. if you had a look at Nick's example, in the 'transferAndWait' function there is a small delay to give the slave enough time to process.

you could implement a similar function but takes in an array instead; something like this maybe (UNTESTED!):

void transferAndWait (byte *arr, uint8_t size)
{
  uint8_t i = 0;
  byte a = SPI.transfer (arr[i]);
  
  delayMicroseconds (20);
  
  for (i = 0; i < size - 1; ++i) {
    arr[i] = SPI.transfer (arr[i + 1]);
    delayMicroseconds (20);
  }

  arr[i] = SPI.transfer (0x00);

} // end of transferAndWait

hope that helps

5

Yes, byte by byte works.

SPI.transfer(firstByte);
delayMicroseconds(10);

SPI.transfer(secondByte);
delayMicroseconds(10);
etc...

But, how to use one shot : SPI.transfer(datas,size) and get slave datas ?

I post a very simple example. But in my case, I send command and parameters : [commandType, paramA, paramB]
and slave return values.

6

vblocs:
Yes, byte by byte works.

SPI.transfer(firstByte);
delayMicroseconds(10);

SPI.transfer(secondByte);
delayMicroseconds(10);
etc...

But, how to use one shot : SPI.transfer(datas,size) and get slave datas ?

I post a very simple example. But in my case, I send command and parameters : [commandType, paramA, paramB]
and slave return values.

I just explained to you why the 'standard' function may not work for your case and provided you with a custom function based on nick's example that would do the the same thing (ie take in an array)....

please try the custom function and see if it works for you!

7

Thanks.
I try it...