IIS3DWB static and dynamic noise question

[FCook.1]

Hello ST team, I’m using the IIS3DWB (STEVAL-MKI208V1K) together with an Arduino Portenta H7. No filter added through the composite group of digital filters, only the default LPF1. Also, 2g scale resolution.

When placing the sensor on a surface isolated from vibrations (i.e. reading the static DC response), single axis mode, I’m registering a noise of about +-20 mg around the gravity constant offset. Graph in image1.

Similarly, when measuring a 674 Hz signal from a tone generator, single axis mode, I see a noise +-20 mg added to the signal. Graph in image2.

In the datasheet, a noise of 60 microg / sqrt(Hz) is indicated for single axis mode, so I think I’m seeing a bigger noise than expected (around 2 mg for the 674 Hz signal using that datasheet value).

Am I misunderstanding the noise definition of the datasheet in relation to the noise I’m seeing in my experiments? What could be the reason for that noise? Can it be reduced at source without additional filters or averaging?

I’m using the sensor in bypass mode (not FIFO buffer, for reasons I could explain later in the discussion). VDD, VDDIO and GND of the sensor connected to the digital IO pins of the board. I tried both with data-ready interrupt enabled or disabled, noise is the same.

Thanks in advance for your kind help,

[FCook.1]

Hi, I've tried using the sensor HPF, ODR/800.

While I can check it is taking place (automatically eliminating the gravity offset, no need to manually adjust for the offset), the +-20 mg jitter is still present.

What could be the reason for that noise? Can it be reduced without averaging? I'm afraid averaging could reduce the fidelity of the acceleration signal.

Regards,

[Eleon BORLINI · Accepted solution]

Hi @FCook.1​ ,

please note first that the Linear acceleration zero-g level offset accuracy parameter (datasheet, p.5) is ±60mg, so the ±20mg value you see is inside this (but of course, not good since it is not constant).

Note also that the IIS3DWB ODR is 26.6kHz: at which frequency are you acquiring data? Could you use a value equal or higher than this?

Try also removing the HPF, it might help the mid frequency accuracy.

-Eleon