AVR basics: SPI on the ATMEGA – Part 2

In Part 1 we got the SPI bus set up on an AVR ATMEGA328P microcontroller. Now let’s start using it.

Settings pins

Before we get going, we need to set up the pins for the SPI bus on the AVR (which we’re using in master mode). I’m using the ATMEGA328P here, so I’m going to define some macros to make things clearer. You can adapt these to your needs if using a different microcontroller.

#define SPI_SS_GPIO PB2



#define SPI_SCK_GPIO PB5

Now let’s use these to set up the pins.

SPI_MOSI_DDR |= (1 << SPI_MOSI_GPIO);   // MOSI as output
SPI_SS_DDR |= (1 << SPI_SS_GPIO);       // SS as output
SPI_SCK_DDR |= (1 << SPI_SCK_GPIO);     // SCK as output
// MISO should be configured automatically as input as that's default state
// on GPIOs, but if you want to be emphatic
SPI_MISO_DDR &= ~(1 << SPI_MISO_GPIO);  // MISO as input

SPI_SS_PORT |= (1 << SPI_SS_GPIO);      // take SS high to deselect slave
SPI_MISO_PORT |= (1 << SPI_MISO_GPIO);  // set pullup on MISO

To simplify matters later, we might also want to set up a couple of macros to control the SS line.


Now we’re set up, let’s look at the register that does all the hard work.

SPDR – SPI Data Register

This is the third of the key SPI registers and it works like magic. Just write a byte to this register and the AVR will fire up SPI and send the whole byte of data down the wires without any further work on your part. And thanks to that dance between shift registers that I mentioned in the first part, when it’s done doing that, that same SPDR now contains the incoming data. So an exchange of data between master and slave consists of the master doing this:

  • Take the SS line low to enable the slave device.
  • Write a byte to SPDR.
  • Wait for the exchange to complete.
  • Take the SS line high again.
  • Read what’s in SPDR.

That’s it. How do we know if the exchange is complete? Simple – the SPIF bit in the SPSR register gets set. So all you need do is wait for that. In the example below we simply hang around in a while loop. That’s blocking and you might find it wasteful of clock cycles if your microcontroller is hard-pressed. If so, you’ll want to look into enabling SPI interrupts that will get triggered when SPIF gets set. You can then get on with other things while the SPI data exchange takes place. But that’s a bit advanced for this post.

So here’s a bit of code that sends a byte of data over SPI and gets one back.

uint8_t value = 0xFA; // randomly chosen value for demo purposes
SPDR = value;                  // initiates transfer
while( !(SPSR &amp; (1 << SPIF))); // wait for SPIF bit to be set

And that’s it! SPDR now contains whatever was sent back by the slave – most likely nothing of any interest.

And that’s the bit about SPI that can catch you out. The byte that the master just sent to the slave may have been a command of some kind. To get the slave’s response, just send another byte of any arbitrary value and the slave will send you its response.

In fact, you might have to send and receive multiple bytes. I mentioned in Part 1 that I’d messed around with a serial RAM chip. Here’s how you go about writing a value to a memory location on that chip:

  1. Take the SS line low.
  2. Send the write command (0x02). Ignore what comes back from the chip.
  3. Send the high byte of the 16-bit memory address. Ignore what comes back from the chip.
  4. Send the low byte of the 16-bit memory address. Ignore what comes back from the chip.
  5. Send the one-byte value to be stored at that location. Ignore what comes back from the chip.
  6. Take SS line high.

So in this case, all the traffic is one-way and all the data coming back from the chip (actually just zeroes) is thrown away. Now let’s read a value from a memory location. This is what you do:

  1. Take the SS line low.
  2. Send the read command (0x03). Ignore what comes back from the chip.
  3. Send the high byte of the 16-bit memory address. Ignore what comes back from the chip.
  4. Send the low byte of the 16-bit memory address. Ignore what comes back from the chip.
  5. Send any one-byte value you like, which is ignored by the chip. However, the chip sends back the value held at the memory location which it has already stored in its shift register as a result of the above operations.
  6. Take SS line high.

The value held at the memory location is now in SPDR.

And here’s what that looks like on an oscilloscope.

That might seem like a lot of work to get one byte of data from the memory. But the 23LCV512 is typical of many SPI devices in that, once the initial contact is made, continually prodding it will return more information.

So, for example, if you wanted to get 32 bytes of data, you’d use the same process as described above, just sending the start address, but at step 5, instead of sending one byte’s worth of junk data – in order to prompt the slave into sending data back – you’d send 32 of them. But not all at once. After each byte, you’d need to retrieve what’s been set in SPDR and store it somewhere because that’s about to get overwritten after you send the next junk byte.

And that’s it for basic SPI use.

Leave a Reply

Your email address will not be published. Required fields are marked *