Communication between two STM32 devices via SPI interfaces (master/slave)

[hoppe]

Posted on September 16, 2014 at 17:38

I'm concerned with the communication between two stm32 devices via 4-wire SPI interface, no DMA, Keil RTX.

The first device (SPI master) is a base board with stm32f427, the secind device (SPI slave) is a display with stm32f207 (both our own hardware). The second device receives the voltage supply via the first device. The problem is that the slave (display) starts up a little bit earlier (approx. 4--5 ms) than the master. This causes that the SPI interface of the slave is initialized before than the SPI interface of the master. The master sends data correctly, but these are not recognized by the slave (rx interrupt handler). If I install a small delay before initializing the slaves, then everything is working properly. Does anyone have any idea what exactly is the cause? #spi #stm32

[Tesla DeLorean]

Posted on September 16, 2014 at 20:52

Does anyone have any idea what exactly is the cause?

One might guess the clock is glitching, and that you enable the SPI regardless of the CS state? You'd typically put a logic analyzer on the signals, and describe those, and the data actually received at the slave, if you wanted to understand what was going on.

[hoppe]

Posted on September 17, 2014 at 10:57

Hello, clive1,

thank you for your response. The MOSI signal looks good on the oscilloscope, but the slave receives allways by 1 bit or 2 bits shifted signal. For example, 0x2D or 0x16 instead of 0x5B.

Also

I see

in the debugger

that the

busy

flag

is always

set

.

[rasko]

Posted on November 04, 2014 at 10:37

Hello Omsker

Your problem could be that when setting up the SPI master device the CLK pin change state a few times, presenting fake clocks to the slave. This will always shift the bits, and this is why the busy flag is always set. One solution could be to initialize the slave SPI interface after the SPI master is set up. You could synchronize the chips using the CS pin.