fifoClear is break

[Lucasmm016]

When configuring the LSM6DS3 accelerometer sample rate to 6660 Hz and the FIFO sample rate to 6600 Hz, then attempting to use the fifoClear() function, the code gets stuck.

This probably happens because the fifoClear() function does not actually reset the FIFO. Instead, it tries to drain it by reading word-by-word through repeated calls to fifoRead(), while new data continues to arrive. Basically, fifoRead() cannot drain the FIFO faster than new samples are being written into it, so it never reaches FIFO_EMPTY.

The correct solution may be to avoid clearing the FIFO by draining it while in continuous mode. One idea would be to switch the FIFO to bypass mode to reset it, then enable continuous mode again afterward.

The board used for testing is the MG24 Sense.

Example to reproduce the issue:

myIMU.settings.accelEnabled = 1;
myIMU.settings.accelRange = 16;
myIMU.settings.accelSampleRate = 6660; // <--- 6660hz
myIMU.settings.accelBandWidth = 400;
myIMU.settings.accelFifoEnabled = 1;
myIMU.settings.accelFifoDecimation = 1;

myIMU.settings.gyroEnabled = 0;
myIMU.settings.gyroFifoEnabled = 0;
myIMU.settings.gyroFifoDecimation = 0;

myIMU.settings.timestampEnabled = 0;
myIMU.settings.timestampFifoEnabled = 0;
myIMU.settings.timestampResolution = 0;

myIMU.settings.fifoThreshold = 240;
myIMU.settings.fifoSampleRate = 6600; // <--- 6600hz
myIMU.settings.fifoModeWord = 6;

myIMU.settings.commMode = 1;

if (myIMU.begin() != 0) {
    Serial.println("Device error");
    while (1);
  }

myIMU.fifoBegin();
myIMU.fifoClear(); // <--- BREAK!

[github-actions]

This issue is stale because it has been open 14 days with no activity. Remove stale label or comment or this will be closed in 7 days.

[Lucasmm016]

up

[cumin777]

Hi @Lucasmm016, thanks for the detailed report.

fifoClear() clears the FIFO by draining it in a loop. In the continuous FIFO mode that fifoBegin() configures, the sensor keeps writing new samples into the FIFO as fast as it samples. At accel 6660 Hz / FIFO 6600 Hz, samples arrive faster than fifoRead() can drain them, so FIFO_EMPTY never asserts and the loop never exits,which is exactly the hang you saw.

I've opened #37 with a fix that clears the FIFO the datasheet way.

Could you try the branch in that PR with your MG24 Sense and the 6660/6600 Hz config from the issue, and confirm whether fifoClear() no longer hangs? Thanks!

[baorepo]

Merged #37 after static review and CI verification. The fix changes fifoClear() to flush the FIFO by temporarily switching FIFO_CTRL5.MODE to BYPASS and then restoring the previous FIFO mode/ODR, avoiding the high-ODR drain loop hang reported here. Closing this as completed.