nRF9160 with Zephyr - SD card & fatfs errors

[jenga]

Hi,

I'm working on a project using Zephyr and ncs 1.7.1. One of its functions is to stream WAV files from a SD card. It only ever reads data.

The issue I'm having is every so often of playing a sound file on loop (after about 5 mins) Zephyr reports an error: <err> fs: file read error (-5). The errors happen randomly - it does not seem to be an issue with any particular point in the play/loop process. The SD card becomes totally unreadable and any further reads also fail.

I've tried all of the normal approaches - retrying reads, closing and re-opening the file, unmount and remount the drive, recalling the SD card init functions (via disk_access_init()), etc. No approach seems to work (there will either be a hard error or the recovery simply doesnt work) and resetting the entire system is the only reliable recovery.

Is there a method to successfully reinitialize the SD card and fatfs without resetting the whole system? I feel like the drivers and Zephyr should be able to accommodate this functionality.

The SD card is running at 8MHz (the maximum allowed frequency on the 9160)

[code-magic]

Have you tried a lower spi-max-frequency?

I had the same exact error on a 5340 when mounting the FS on an SD card over SPI. While not the same instance for the error, the fix for me was lowering the spi-max-frequency.

Good luck.

[jenga]

Good point. Turning down the frequency instead of being able to deal with errors isn't great, but if it works, it works.

spi-max-frequency in the proj.conf doesn't change anything. I have to go in to sdmmc_spi.c and change SDHC_SPI_MAX_OPER_SPEED to see a different frequency. Sweet.

[jenga]

So changing the SD clock frequency to 4MHz appears to have solved the issue. 

The problem now is the data rate is so low it cant keep up with the sampling rate of the WAV file!

[Hakon]

Hello,

the max SPI frequency for the SDHC driver should be higher than 4Mhz. You should try probing the SPI pins when it breaks down to see what happens.

[jenga]

I totally agree it should work above 4MHz. However I've tested the SD card overnight (SPI CLK at 4MHz, playing a WAV file on loop) and have not encountered the -5 error again. That would indicate it's a signalling issue.

Catching the error on an oscilloscope won't be easy - when the error does happen, it happens randomly after many minutes.

A fellow engineer suggested that perhaps the CS signal causes this issue. Is there an option in Zephyr's SPI driver to hold CS low?

The board I'm using is custom, so I guess I can't rule out being a design/routing issue. Seems unlikely though.