LSM6DSV16X pitch/roll signal anomaly

[JoanV]

Hi,

We are using a LSM6DSV16X unit to track pitch and roll in our product based on the quaternions reported by the unit's SFLP algorithm (High Accuracy ODR 120Hz)

During the operation and in what seem to be arbitrary times, the pitch and roll signals for such unit present some unexpected overshoot or oscillation (see below) before returning to the expected value.

The graph above shows the IMU pitch signal (purple) with the reading of an absolute encoder (blue) and both signals should be fairly similar. The roll signal below, although not compared to any other signal, should maintain the same oscilatory behavior that is seen before and after the anomaly (vertical axis in º). At the moment of the anomaly, no special movement happened that could have caused such readings.

We have other LSM6DSV16X units in the same usecase and we do not see such behavior as of now.

We were doubting whether this could be a calibration problem of the IMU? Is there any calibration procedure you recommend for these IMU units?

Could this maybe be somehow related to the limitation in the quaternion components variables stored in half precision floats? https://community.st.com/t5/mems-sensors/lsm6dsv-half-precision-error-at-180-degrees/td-p/628771

Thank you in advance,

Joan

[Smithson]

You may have an input from another machine or human in the environment.  If it does not repeat I would suggest that it is the likely cause.  If it repeats then look for external signal first.  

In the Smithsonian it is a railway tunnel, in London it is a railway subway and in Italy it was a 35 Hz motor.  

On a desk it can be a hard drive rotation.  

Just a few thoughts, but I would blame the instrument last.  

[JoanV]

Hi Smithson,

Thank you for your help!

We will keep in an eye for external signals that could cause this disturbance on the IMU signals.

In any case, we are interested to know if there is any calibration procedure recommended for the accelerometer and gyroscope of the LSM6DSV16X IMU. 

Thank you in advance!

Joan

[Smithson]

Based on the response you are getting so far, I would hazard no.  

I am using the same device in the Sensor TileBox Pro, I cannot think of a physics test for calibration that would work. 

Nothing could hold it still enough, now statistically maybe, but it would be hard and time expensive.  

Kappler provides an interesting discussion on the problem.  

[Smithson]

Joan:

These are the acceleration and FFT readings from the same accelerometer in a batch over two days of 19000

they are clearly in error as the fourth to 16h columns are acceleration data.  

I will have to find the files. 

9387	1	56:29.6	15.93125	15.49546	14.35354	-14.6654	-15.7858	-15.6482	0.285111	0.314508	0.349435	0.596007	0.597489	0.629491	1.31191	274.012	0.715587	306.7781	1.31191	302.3718	44.36641	19.90793	42.33891	18.57206	41.62332	18.71853	22.72689
9388	1	56:29.6	15.63601	14.86594	14.49604	-14.8464	-12.5767	-15.6165	0.320757	0.331966	0.302736	0.605239	0.595011	0.581627	1.31191	350.1252	1.31191	342.0719	1.31191	292.9388	35.77936	19.96408	37.80686	22.43483	35.06377	23.55651	22.72689

[JoanV]

Hi Smithson,

Thanks again for your answers. 

We are mostly interested in whether there was a calibration routine that ST recommended for such IMU model that could be done relatively easy (a flag that could be checked for calibration completeness, position the IMU in different orientations to allow for better self calibration, ...)

We've conducted more tests and could see that external mechanical disturbance do create pitch and roll anomalies in the IMU. The anomaly seems to come from the quaternion components coming from the SFLP algorithm. However, we've tested other units in the same conditions and do not show the same behavior, being their signals consistent with the expected values.

We've also run the inbuilt selftest for the IMU that is showing signal anomalies and it passed successfully. Nonetheless, since the IMU showing anomalies seems to be a particular case within the set of IMUs tested, we are inclined to think it is an issue with that particular unit.

We will continue to investigate though.

If you or anyone else might have other calibration suggestions or has observed similar behavior, we would like to hear it:)

Joan

[Smithson]

The Sensor Tile Box Pro I have is returning incorrect temperature, so I understand the pain. 

I have three more devices in London, but it is a long way away.  

Good hunting.  

[Jodo]

Same happened to me. Doing custom made drone with stm32f405 ,+ this gyro.

After 24 minutes of work some oscillations appears randomly

 

[Smithson]

If I was in a lab and saw that signal I would go look for who has turned something on.  

We can pick up the electrical signal to 300 Hz range.  

If you have the signal do a FFT and look for the frequency and then check your equipment.  

[Jodo]

This bug is reproducible each time in the field. Flying 24 minutes and this happens. Logger shows random spikes during ~3-6 ms. And then it goes to normal stage. But drone during this 3-6 simply enable safe mode and falling down. Looks like something wrong with internal counters.