SPIM3 tx data corruption, byte carried over from previous operation.

[Anthony Ambuehl]

NCS 2.7

NRF52840

SPIM3 at 32 MHz.  

I have CONFIG_NRF52_ANOMALY_198_WORKAROUND=y and have confirmed the workaround is being activated by setting breakpoints in my code.

I am chasing down a NOR flash page program failure.  I found the failure occurred after a byte from the previous single-byte SPIM operation was driven at the start of the next SPIM operation.

The expected sequence from spi_nor driver:   WREN (0x06),   Page Program 0xA26900 (0x02, 0xA2, 0x69, 0x00, data)

The sequence observed on SPI bus: WREN (0x06),  Page Program 0xA26900 (0x06, 0x02, 0xA2, 0x69, data)

The page program sequence included an extra 0x06 at the beginning and lost a byte of the address from before the data started.

Although this is an nrf52840 and I have seen multiple posts asserting anomaly 109 is not present.. maybe it actually is present? 

[Einar Thorsrud]

Hi,

There are some unfortunate issues with SPIM3. In this case, I wonder if it would help to use double-buffering and ensure that the buffers are on separate RAM sections, as shown in this post?

[Anthony Ambuehl]

I gave up and switch to the older/slower version of the SPIM peripheral.  If I have time and client is willing to pay I'll spend time trying to figure it out.