Armbian Stretch 4.19.13 (Orange Pi Lite) - SPIDev0.0 Generates Extra SCK Cycle

Felix · January 23, 2019

Hi,

I moved from Jessie to Stretch. On Jessie i looked under a scope and generated SCK (Orange Pi Lite is Master) is clean. However, after i switch to Stretch, my SPI Slave doesn't work anymore and further investigation reviews that it has similar problem as previous builds that extra clock cycle is generated before data is transfered. The behaviour is exactly similar to following image posted by another forumer but for earlier builds. I am not sure whether any of you experience the same problem. Thanks

my armbianEnv.txt

verbosity=1

logo=disabled

console=both

disp_mode=1920x1080p60

overlay_prefix=sun8i-h3

rootdev=UUID=25d9957f-6c04-4354-aa09-6b9c6f162257

rootfstype=ext4

overlays=spi-add-cs1 spi-spidev

param_spidev_spi_bus=0

param_spidev_spi_cs=0

param_spidev_max_freq=100000000

usbstoragequirks=0x2537:0x1066:u,0x2537:0x1068:u

Felix · January 24, 2019

There is no deny that SCK has extrac burst before data is transferred. Further investigation confirms that it happens right before the edge of CE0 (in my case active low). If you use the kernel to control CE0, then you are safe. However, if you use a manual way to control CE0 (like me), then you will have problem due to slave may mistakenly interprete the burst as a clock and read in a wrong info. I am not sure whether i consider luckily, the kernel generated CE0 is active low (by default?). If you need an active-high CE0, may be someone can share some thought on how to achieve this. Cheers

martinayotte · January 24, 2019

53 minutes ago, Felix said:

If you need an active-high CE0, may be someone can share some thought on how to achieve this.

Did you tried "spi-cs-high" ?

From https://www.kernel.org/doc/Documentation/devicetree/bindings/spi/spi-bus.txt

Quote

- spi-cs-high - Empty property indicating device requires chip select active high.

zador.blood.stained · January 24, 2019

There is an old upstream mainline kernel issue that never was fixed: https://groups.google.com/forum/#!topic/linux-sunxi/Ar9A_OYzk1Q

And we can't pull random patches without proper long term testing because they may break more things than they fix.

Felix · January 27, 2019

On 1/25/2019 at 12:28 AM, martinayotte said:

Did you tried "spi-cs-high" ?

From https://www.kernel.org/doc/Documentation/devicetree/bindings/spi/spi-bus.txt

I have not tried it though because it is not documented in /boot/overlays/READMExxx for SPI