NRF9160 zephyr SPI - CS stay low too long

[Mateusz Gancarczyk]

Hi,
I'm working on nRF9160 with Zephyr (SDK 1.7.1) and one SPI device. I analyze CS signal with correlation to SPI SCK. Is that normal that CS signal stays low much longer than SCK signal?

For example:
SPI SCK = 2 MHz:

CS goes low ~8us before SCK and goes high ~21 us after the end, while communication takes 12us.

SPI SCK = 8 MHz:

CS goes low ~7us before SCK and goes high ~7us after SCK, while communication takes 3us.

When I working with time sensitive SPI devices - these delays matter a lot. Is there a chance to reduce them?

&spi0 {
	compatible = "nordic,nrf-spim";
	status = "okay";
	sck-pin = <21>;
	mosi-pin = <22>;
	miso-pin = <23>;
	cs-gpios = <&gpio0 24 GPIO_ACTIVE_LOW>;
};

Of course - I set CS gpio delay to 0.

 

[Sigurd]

Hi,

I see similar results here.

Zephyr SPI API:

nrfx driver(based on this example https://devzone.nordicsemi.com/f/nordic-q-a/70105/how-to-use-spi-interface-in-a-zephyr-sample-example/288587#288587 ) 

When the CS pin is set high/low is handled in the SPI driver(SW). There could be ways to optimize the driver SW here, but if you need minimal/no latency here, and the SPI sensor support that, I would look into using (D)PPI to toggle the GPIO. https://devzone.nordicsemi.com/f/nordic-q-a/63466/control-of-ss-pin-of-spi-with-ppi

[Saleh]

Hi

It seems the problem is coming from Zephyr side, a similar question was asked before and results show that this annoying delay is the normal behavior of the current Zephyr's version.

https://devzone.nordicsemi.com/f/nordic-q-a/81966/slow-spi-performance-with-zephyr

[Mateusz Gancarczyk]

I figure out this problem - looks epic :) :

I can't share a code because it's commercial but:
it use 2 gppi channels. One connect SPIM Start event to gpiote toggle pin task and second connect SPIM end event and the same gpiote toggle pin task.