LSM6DSV80X FIFO quaternion output always shows X=0, but raw register read is correct

[AhilanAshwin]

I’m trying to read the 3-component “game rotation” quaternion out of the LSM6DSV80X’s FIFO, but every time I pop a FIFO record I get three half-precision floats [x,y,z] where x is always zero and y,z look plausible. When I read the same fusion engine registers directly (not via FIFO), I get a full four-component quaternion [w,x,y,z] where x is nonzero—as expected. Ive connected the STEVAL-MKI247A to my ESP 32 via SPI with the following configuration 

// Define SPI pins
#define CS_PIN 5    // Chip Select pin
#define SCK_PIN 18  // SPI Clock pin
#define INT_PIN 25  // Interrupt pin
#define MOSI_PIN 23 // Master Out Slave In pin
#define MISO_PIN 19 // Master In Slave Out pin

What ive done: 

• Enabled the SF-LP game rotation vector and timestamp in the embedded-function bank, then

• Configured FIFO streaming mode and set watermark = 1, and

• In loop() I parse the data from the FIFO queue and immediately compare it with the raw register read. 

1. How i enabled SF-LP FIFO batching:

fifo_sflp.game_rotation = 1;  
fifo_sflp.gravity       = 0;  
fifo_sflp.gbias         = 0;
lsm6dsv80x_fifo_sflp_batch_set(&dev_ctx, fifo_sflp);
lsm6dsv80x_fifo_mode_set(&dev_ctx, LSM6DSV80X_STREAM_MODE);

 
2. FIFO-read + tag check:

lsm6dsv80x_fifo_out_raw_get(&dev_ctx, &f_data);
if (f_data.tag == LSM6DSV80X_SFLP_GAME_ROTATION_VECTOR_TAG) {
  uint16_t hx = (f_data.data[1]<<8) | f_data.data[0];
  uint16_t hy = (f_data.data[3]<<8) | f_data.data[2];
  uint16_t hz = (f_data.data[5]<<8) | f_data.data[4];
  float x = npy_half_to_float(hx);
  float y = npy_half_to_float(hy);
  float z = npy_half_to_float(hz);
  Serial.printf("FIFO FLOATS : x=%.4f y=%.4f z=%.4f\n", x, y, z);
}

3. Read raw registers for comparison

uint16_t raw_q[4];
lsm6dsv80x_sflp_quaternion_raw_get(&dev_ctx, raw_q);
float w = lsm6dsv80x_from_quaternion_lsb_to_float(raw_q[0]);
float x = lsm6dsv80x_from_quaternion_lsb_to_float(raw_q[1]);
float y = lsm6dsv80x_from_quaternion_lsb_to_float(raw_q[2]);
float z = lsm6dsv80x_from_quaternion_lsb_to_float(raw_q[3]);
Serial.printf("RAW  FLOATS : w=%.4f x=%.4f y=%.4f z=%.4f\n", w, x, y, z);

4. Output:

FIFO FLOATS : x=0.0000 y=-0.1245 z=-0.9414  
RAW  FLOATS : w=0.2949 x=-0.1245 y=-0.9414 z=-0.0973

Any example on how to retrieve the sensor fusion output from via the fifo would be most appreciated!