LIS2DW12 Unexpected interrupt event

[MIR.1]

Hi,

I am using LIS2DW12 accelerometer for tilt angle detection.

I expect an interrupt event when the tilt angle is crossing the threshold and is working fine.

But an interrupt event is unnecessarily occurring during physical shocks/knocks which is not expected.

I have read all the interrupt sources(all_int_src, sixd_src, wake_up_src, tap_src, status_dup).

I have given a physical shock/knock but no changes in any of the above mentioned registers.

But still interrupt is triggering for shocks /knocks.

Can you please suggest a solution for this ?

[Eleon BORLINI]

Hi @MIR.1​ ,

You could first try with increasing the FullScale of the device, from +-2g to +-4 or +-8g, so that it is less sensible to the small spikes.

Then, in order to avoid any DC effects, you can configure the high pass filter (see the application note AN5038 p. 12 for it).

On the other hand, you could also try with cutting the high frequency disturbances, setting the low pass filter on the acquisition chain as described in the same app note p.8.

Let me know if you can make any progress in this way.

-Eleon

[MIR.1]

Hi Eleon,

I am using the full scale setting as +-8g now.

I have tried to introduce high pass filtering also.

But when we add HP filtering, it will not trigger the interrupt for normal wake up threshold detection.

Still the shock or knock will be detected even after HP filtering.

Then from ( application note AN5038 p. 21,third paragraph), I planned to increase the wake up duration by reducing the value of ODR from 200 Hz to 12.5 hz.

Then the shock/knock detection is reduced very much but it occurs very rarely.

Can you suggest a method for avoiding shock/knock detection even if it occurs rarely?

[Eleon BORLINI]

Hi @MIR.1​ ,

good that you are progressing in your application!

To address the issue of spikes (i.e. high frequency components in the signal that -even if they are true data- might interfere with the desired signal), you could now set a low pass filter, in particular trying to decrease the Low-pass filter 1 bandwidth, as shown in the same app note AN5038 p.10.

The frequency windows defined by the combination of the high pass filter and the low pass filter should be centered around the first harmonic of the signal of your interest, and the band should be able to filter out the unwanted data.

-Eleon

[MIR.1]

Hi Eleon,

Thank you very much for your response.

I have already tried this.

Shock/knock detection is reduced but not completely avoided.

Please let me know is it possible to avoid the complete avoidance of unwanted detections( for shocks /knocks).

Regards,

Prajin.I.R

[Eleon BORLINI]

Hi Prajin,

Are you receiving false positives also after having increased the wake-up duration and after having added a LP filter? Which filter cut-off frequency did you set?

By the way, you might think to set a "filter" at software level: you could characterize ex-ante the pattern of the mechanical spikes that generate false positive, and try to ignore them in port-processing... don't know if this option is feasible, but characterizing the application and it's weakness is a mandatory step for checking its robustness.

-Eleon