Minimum RP2040 clock speed

[bruelltuete]

Is there a minimum clock speed the RP2040 must have for the driver to work? If I run it at 48 MHz then the test pattern check at https://github.com/georgerobotics/cyw43-driver/blob/main/src/cyw43_ll.c#L1468 fails. At 96 MHz it's fine.

This was a bit surprising because the RP2040 drives the clock signal, so I was expecting it to be mostly indifferent to what speed it runs at.

I wanted to use cyw43_arch_gpio_put() from pico_cyw43_arch and hit this test pattern fail.

[bruelltuete]

Alternatively, what do the different values coming out of SPI_READ_TEST_REGISTER mean?

[peterharperuk]

I don't think we've tested this. If you slow the clock down I think you will slow down the pio state machines which will slow down spi. I would have hoped spi would continue to work when you slow it down. I guess it doesn't?

The spi test register should only return the value it expects. If it sees something else it's not talking to the chip properly.

[kilograham]

Try changing define here to one of the other PIO programs, e.g. spi_gap010_sample1

https://github.com/raspberrypi/pico-sdk/blob/2e6142b15b8a75c1227dd3edbe839193b2bf9041/src/rp2_common/cyw43_driver/cyw43_bus_pio_spi.c#L34

The "turn-around" time for the SPI seems to be somewhat dependent on clock speed for no apparently good reason.

[dpgeorge]

On tests with an STM32F7xx as the host MCU, I did find that there was a minimum host CPU frequency for which the CYW43 would work. It was something around 48MHz. I never investigated the reason for this.

[rbryson74]

Just adding my experience here. I had the same issue on a PicoW. After setting the clock with set_sys_clock_48mhz() the test pattern check would fail. Setting the SPI_PROGRAM_NAME to spi_gap0_sample1 resolved the issue.

[peterharperuk]

I've now tested the driver running from xosc at 12MHz, you can speed up the pio to compensate. The pio lives in the pico sdk so this shouldn't be a driver issue.