Post Go back to editing

ADIS16136 GYRO_OUT and GYRO_OUT2 combination using SPI and using the IMU Evaluation Program

Hi. I have configured the ADIS16136 using the IMU Evaluation program in the following way:

SMPL_PRD: 000Fh; 2048 SPS

DECI_RATE: 0006h; decimate to 32 SPS

AVG_CNT:    0005h; -3db at arround 15hz (.007 f/fs)

I have a number a number questions, since I don't know how the datasheet agrees with the IMU Evaluation Program, and I haven't used the gyroscope by SPI yet to confirm how it is normally used:

1) Are the numbers I put after the ; correct? Specially I have my doubts about the filter.

2) If I were using this by SPI, I should read GYRO_OUT, then shift it left by 6 bits, then read GYRO_OUT2, and do a bitwise and. That would give me a 22 bit twos-complement integer, and after converting it to signed multiple by 0.018275 / 2^6 to obtain degrees per second. Is this ok?

3) The IMU Evaluation Program, when configured this way and using data capture with "Use Scaled Data" unticked, seems to be giving me full 32 bits numbers, and that I must use 0.018275 / 2^16 as bit weight to obtain degrees per second. What internal arithmetic is it doing with the contents of GYRO_OUT and GYRO_OUT2? Is that bit weight correct to use with unscaled data from the evaluation program? Is it interpolating the real 22 bit number?

Thanks!

EDIT: I'm attaching a sample .csv of data captured not using the use scaled data checkbox.

Datalogtest.csv.zip
  • Hello PeterSate,

    Here are some preliminary answers. I can set up and test this scenario if necessary.

    1) I don't see any issue with your settings, but I don't know what your objectives are.

    2) This approach sounds correct.

    3) Yes, that is correct. The ADIS16136 only communicates in 16-bit registers and the evaluation software does not automatically truncate this result at this time. This is on our list for future consideration, but for now, you can get either 16-bit or 32-bit results.  In your case, I would suggest using un-scaled data and truncate the result down to a 22-bit number through your scaling operation.

    Does that help?

    Best,
    Mark