LIS3DHH accuracy and stability

[Kirill Rzhav]

Hello! I am R&D inclinometer based on LIS3DHH. Firstly I started my works with LIS2HH12 as MEMS accelerometer due to his best accuracy and resolution. I wanted to get accuracy about 1-2 angular minutes. All was good when I turned on almost all internal filters and using software moving average. Even with 7 sec response it was ok for my solution. But after few experiments I discovered problem with temperature stability. I decided use internal temperature sensor to compensate temperature coefficient and started temperature experiment (first with termal camera, but after without, because was a lot noise from camera). And I can't find out temperature coefficient for that. Next figure shows angle for one axis (red), temperature from internal sensor (blue) and reference signal from another special inclinometer. So, angle changed unpredictably on different days.

Next figure shows one angle (green) and temperature (blue). Here some unpredictable overshoot.

In this case I cant solve the problem with temperature and discover new LIS3DHH. I made a new PCB layout for it. There only 2 internal filters and with using moving average. The best result I've got only 5 angular minutes accuracy. Then I start temperature experiment for LIS3DHH and it was better than LIS2HH12.

Can you help me, how I can solve temperature stability problem in LIS2HH12 or accuracy in LIS3DHH. Maybe better to chose another accelerometer for inclinometer solution? Is it possible to have stability angle with 1-2 angular minutes accuracy with MEMS technology?

[Miroslav BATEK]

OK, if I put aside the jump.

You changed the temperature from 45°C to 5°C and the tilt angle in X axis changed -90 to -40 degrees. This is impossible, obviously you have something wrong. See below the specification of the LIS3DHH.

[alessio31183]

Sorry Miroslav I was speaking about that sudden jump around the 5°C.

I was not speaking about the temperature movement. That's normal, I know and it's what I compensate in firmware during the temperature test in the climatic chamber.

Of course each accelerometer has a different temperature behavior and it's compensate one by one.

[Miroslav BATEK]

It is not normal to have 50 angle degrees change (from -90 to -40) change if you change the temperature from 45°C to 5°C!

[alessio31183]

No sorry the data that you see is the raw accelerometer data without the tilt computation.

It was only to check if the jump coming from my tilt computation in firmware.

Attached you see the same picture but with the tilt value

[alessio31183]

Miroslav, sorry, have you got the point?

[Miroslav BATEK]

Yes, I got your point. Let me discuss with sensor specialists.

[alessio31183]

Hi Miroslav, do you have some news?

In the meantime I give yout another example on how data changes dramatically and suddenly without any motivation

 

Moreover, this behavior happens going down only with the temperature and NOT going up

Thanks in advance

[alessio31183]

Miroslav I've found the problem.

THE RESIN PROCESS ON THE ACCELEROMETER MAKES THIS STRANGE BEHAVIOR.

UNBELIEVABLE

Usually I did the process for all other accelerometer that I used in the past

Probably this time the resin was to hard and with a different thermal coefficient so it stressed the accelerometer position makes this sort of offset.

I don't want what to say. Sorry to everyone for the time that I made you lose.

Now I will do some other tests in order to be sure that this is the right solution for my application.

[Miroslav BATEK]

OK, @Community member​ thank you for the information.

[alessio31183]

Hi Miroslav,

I'm just looking carefully to the Recommended power-up sequence explained in the datasheet.

Basically it says that the accelerometer need to connect the 3.3V remotely with a couple of capacitor close to the component.

10uF for the 3v3 (pin8) and 100nF (pin7) in order to make the Twait.

Do you think that this is can involve in some bad behavior if not observed?

Thanks in advance