SPI behaviour is different running the same software on two Nucleo-G474RE boards

[CTapp.1]

I am using SPI 1 on a pair of Nucleo-G474RE boards that are running the same software image. On of the interfaces is working correctly, the other is not.

A logic analyser shows the following:

Working

Not working

From this, it looks as if there is a spurious clock cycle / glitch at the start of the transfer. What could cause this?

[LCE]

Are you sure that the firmware is 100% identical?

And the hardware? External pull-resistors, caps, ... ?

Chip revision?

[AScha.3]

>glitch at the start of the transfer. What could cause this?

some capacitive coupling between cs and clk , maybe long wires close together.

a: reduce port speed cs pin , --> slow.

b: reduce coupling or use screened cable (also for your test-wires , use a screened scope tip and test)

[CTapp.1]

Yes, I have programmed the same image from the same Cube project onto the boards. The boards have nothing else connected to them so that I could eliminate any external influences on the signals.

[CTapp.1]

Both chips are 0x469 Rev X.

[LCE]

So the only difference might be the scope wiring? Maybe there's the problem?

Your screenshots look like a PC scope or logic analyzer, do you have a stand alone scope to verify that?

Wildly guessing... 😉

[TDK]

Make sure all pins are initialized and the SPI peripheral is enabled prior to CS going low. This can be done by sending an initial dummy transfer with CS high.