LSM6DSV32X Gyroscope Artifacts

[JHenn.1]

We use the accelerometer and gyroscope channels of an LSM6DSV32X for measuring movement parameters of a mechanical system. During some tests we observed artifacts in the gyroscope channel that we cannot explain.

Configuration:

The LSM6DSV32X is connected via SPI and run in Continuous FIFO mode. We have a FIFO Watermark configured and read the entire FIFO whenever its interrupt is detected.

The sensor channels are configured to:

Accelerometer: 3840Hz, +/-32 g

Gyroscope: 3840Hz, +/- 2000 dps

Mechanical Setup:

In order to examine the problem we build a simple mechanical setup and recorded some data with it:

We attached the LSM6DSV32X to a metal bar that pivots around an axle attached to a solid wall. During the recording we lifted the unrestrained end of the metal bar up and dropped it, causing it to hit the wall. After the impact of hitting the wall, the metal bar bounces back and forth a few times.

We repeated this several times over multiple recordings.

This is what the first recording looks like:

Both Signal 2 and Signal 3 look as expected with gyroscope amplitudes of around 400 dps. Here is a close up of Signal 2:

And a close up of Signal 3 (different time scaling in the plot, but very similar to Signal 2:(

However Signal 1 looks completely different. Here the gyroscope amplitudes are close to 2000 dps and the expected patterns of impacts and bouncing back are not visible.

To be sure we did more recordings and while most of the signals looked good, there were several more signals that looked like the example above.

Here is another example:

Question:

For the most part the accelerometer signals look fine, so the problem seem to be in the gyroscope channel only.

During the impacts the accelerometer is almost always clipping (>32G). For our application this is not an issue, however we are wondering if this might cause issues for the gyroscope.

--> Is it possible that we sometimes get artifacts in the gyroscope channel because the accelerometer is clipping?

--> What else could be the problem and cause these artifacts?

Best regards,

Johannes

[PGump.1]

Hi,

I have not looked at the LSM6DSV32X specifically, however, these devices typically use a DSP based on Sampling Theorem for their functionality, and therefore, governed by those limitations.

In your case, Impact analysis can be distorted by low Sample Rates, and DSP’s handling of Overflow arithmetic through filters...

If the front-end of the LSM6DSV32X can be configured for different sample rates or filtering - try there.

I hope this helps.

Kind regards
Pedro

[JHenn.1]

We tried different sampling rates and found that the artifacts occur with all sampling rates.

It is very plausible that there might be effects like mechanical crosstalk between accelerometer and gyroscope or in general overload problems in situations with high accelerations. However I was unable to find any clear information about this in the datasheet.

I was hoping to get a definitive answer here:

--> Is there a maximum acceleration for the gyroscope to work properly?

(The datasheet only lists 20.000g absolute maximum acceleration) 

[PGump.1]

---

@JHenn.1 wrote:

We tried different sampling rates and found that the artifacts occur with all sampling rates.

Unfortunately, the information you provided here doesn't give any sample rates so, I cannot be of help...

---

It is very plausible that there might be effects like mechanical crosstalk between accelerometer and gyroscope or in general overload problems in situations with high accelerations.

---

In case you are not aware, gyroscopes are based on detecting changes in a magnetic field. Acceleration can also affect a magnetic field.

I was hoping to get a definitive answer here:

---

There possibly is. Depending how deep you want to look...

--> Is there a maximum acceleration for the gyroscope to work properly?

 Mechanical characteristics portion of the datasheet has the details you are looking for.

Kind regards
Pedro

[JHenn.1]

Thank you for your reply. 

You asked for sampling rates:

"Unfortunately, the information you provided here doesn't give any sample rates so, I cannot be of help..." 

I had posted the sampling rates in my original post:

"The sensor channels are configured to:

Accelerometer: 3840Hz, +/-32 g

Gyroscope: 3840Hz, +/- 2000 dps"

In addition to that we tested with 1920Hz and 960Hz.

About maximum acceleration for the gyroscope:

You wrote:

"Mechanical characteristics portion of the datasheet has the details you are looking for." 

This is what the datasheet provides:

 

I cant find a specification for a maximum acceleration for the gyroscope. 

[JHenn.1]

--

[PGump.1]

Hi,

I’m sorry, I made some poor assumptions of your understanding of this subject.

The gyroscope data you want is here -

The sampling rates would need to increase, not lower, to lessen the distortion.

Do you know the rise time of the g-force?

Do you understand the characteristics of Impulse Response in Sampling Systems?

Do you understand the characteristics of Saturation in Digital Filters?

Are you sure that the SPI is achieving the sampling rate?

Are you testing for FIFO overrun?

I hope this helps.

Kind regards
Pedro

[JHenn.1]

---

@PGump.1 wrote:

The gyroscope data you want is here -

---

This is the angular velocity measurement range of the gyroscope. In our case we set it to +/- 2000 dps.

As I wrote in my original post, the signals we were measuring have a magnitude of about +/- 400 dps, so they are well within the measurement range.

I will try to clarify this:

The signals we are measuring have moderate angular velocities (~400 dps) but, for short periods of time, very high linear accelerations of much more than 32 g. When measuring these signals we expect the accelerometer to go into clipping, however the gyroscope should still yield useful data.

But as you can see in my first post, when the accelerometer goes into clipping the gyroscope sometimes also produces incorrect data.

So my question was really: Apparently the gyropscope is affected by high linear accelerations. Is there any information in the datasheet about the maximum linear acceleration under which the gyroscope still functions properly? 

[JHenn.1]

---

Do you know the rise time of the g-force?

---

Not yet, we ordered a +/- 200g accelerometer to hopefully measure this.

Do you understand the characteristics of Impulse Response in Sampling Systems?

Do you understand the characteristics of Saturation in Digital Filters?

---

I think I understand most of it, but again, the signals measured by the gyroscope are rather small. So if this was a saturation problem, the accelerometer would have to cause the saturation which would then have to influence the gyroscope data. I am not saying this is impossible, but its hard to find the cause of the issue when there is so little information available.

Are you sure that the SPI is achieving the sampling rate?

Are you testing for FIFO overrun?

---

Yes

[PGump.1]

Hi,

Not yet, we ordered a +/- 200g accelerometer to hopefully measure this.

For my interest, what part have you ordered?

An accelerometer and gyroscope are two separate functions, and these types of MEMS devices are highly integrated. For both of these to be totally accurate, they need to share the same X, Y and Z origins. I don't know how this achieved, however, it implies that either -

• they are separated with the loss of accuracy or,
• they share the same resources at the cost of crosstalk or,
• both...

I have also noted that many MEMS devices don't have the same accuracy across all three axis'.

You could try to attenuate the G-force, keeping it below 32, for your tests.

You could also try rotating the MEMS to determine if the axis' are the same...

I hope this helps.

Kind regards
Peter