LSM6DSL Gyro 800Hz Noise

JWoeb.1 · ‎2021-01-25

Hello,

When configuring the LSM6DSL ( STEVAL-MKI178V1 ) to sample Data at 6664 HZ ( without any digital Filters active, FS= +- 16g, +- 250deg/s ) there seems to be an consistent 1deg/s @ 800Hz noise on some gyro axis. The same signal but stronger ( > 10deg/s @ 800Hz ) is also apparent on a proprietary board.

What causes this noise ( it seems to come from the LSM6DSL itself ) and why is it stronger in some modules than in others?

Doesn't this signal violate the specified noise density of 4mdps/sqrt(Hz)?

STEVAL-MKI178V1:

proprietary Board:

The setup to measure the noise was a LSM6DSL Eval Board ( STEVAL-MKI178V1 ) supplied with 3.3V via a lab supply and connected to a STM32 MCU. The MCU was connected to the LSM6DSL via SPI with a Clock Speed of 6MHz. The LSM6DSL was configured to create an Irq once new data is available ( no FIFO ) at which point the MCU read the Gyro, Accelerometer and Timestamp values from the IMU.

In case of the proprietary board the Samplerate was set to 3332 Hz.

Thanks

Eleon BORLINI · ‎2021-01-27

Hi @JWoeb.1 ,

Thank you for the detailed picture you shared.

It's quite strange that this noise is platform-dependent, if it is related to the sensor...

Is this noise always peaked at 800Hz or is it varying with the ODR? Could you try to set ODR < 800Hz (for example 200Hz) and check if you see again the noise?

I'm wondering also if this noise could be related to some calculation error: do the two acquisition boards share the same firmware/software?

-Eleon

JWoeb.1 · ‎2021-01-27

Hi,

thanks for answering. If set the ODR to a sample rate below the noise signal at 800 Hz it is filtered away by the decimation filter inside the sensor in case of the STEVAL-MKI178V1 board. But because it is ten times stronger on the proprietary board it is not completely filtered away and it is still visible. It is only 0.14 deg/s @ 70Hz which is not a huge problem anymore but we would like to be able to use higher ODRs.

The two acquisition boards do not share the same software. In case of the STEVAL-MKI178V1 we connected it to some STM32 MCU. The MCU was connected to the LSM6DSL via SPI with a Clock Speed of 6MHz. The LSM6DSL was configured to create an Irq once new data is available ( no FIFO ) at which point the MCU read the Gyro, Accelerometer and Timestamp values from the IMU.

In case of the proprietary board the software does more or less the same but I dont think it is caused by the way the data is read from the LSM6DSL. We already activated 25us timestamping in the LSM6DSL to see if we missed or mixed up some samples but the timestamps seem to be correct.

JWoeb.1 · ‎2021-01-27

Our main worry is that whatever causes this noise, which we maybe could filter away in post processing as it is only at a certain frequency, could also cause some other inaccuracies / noise because currently we are trying to use these sensors to measure displacement and deflection of precision measurement equipment mounted in vehicles and we currently don't get to the accuracies we would need.