LPF2 on IIS3DWB and anti-alias

[vix]

I use IIS3DWB with no LPF2 enabled, and I understand the Frequency response shown in Figure 13 at page 20 of the datasheet.

I know that I can enable low-pass LPF2 filter, but from the same graph I don't understand if this filter can be enabled as a hardware anti-aliasing filter, or not.

I mean, after I enable ODR/4 (as an example), is it safe decimating ODR by 4, or not?

This decimation is safe if LPF2 is a hadrtware filter that "cuts away" every frequency above (ODR/2)/4.

From the graph it doesn't seem so, but i't not 100% clear IMHO

Thanks in advance for the clarification.

[Federica Bossi]

Hi @vix ,

If I understand correctly, you want to set an ODR, enable the LPF2 and then undersampling without folding, is it right? If yes, you can do this. But note that when the datasheet reports ODR/4 it is the ODR you set divided by 4 and not (ODR/2)/4. So to implement what you want, you need a step lower than ODR/8, so the ODR/10 or ODR/20.

If this helps you, please mark my answer as "Best Answer" by clicking on the "Accept as Solution" button, this can be helpful for Community users to find this solution faster 🙂

[vix]

Hello Federica,

as far as I understand I cannot change ODR in IIS3DWB.

I don't see any way to do this in the datasheet.

Can you find a way?

So, ODR is fixed to 26.667 kHz (nominal).

And so my question is: when I enable digital filter with ODR/4 (5.6kHz cutoff), is it safe downsampling (i.e., decimation only) to ODR/4 without any software decimating filetr (as an example halfband)?

My tests on a shaker suggest that this is not true, and a software decimator filter is necessary.

Can you explain a little bit more?

[Federica Bossi]

Hi @vix ,

It's not true for the Nyquist theorem. You need to downsample to ODR/10 and not ODR/4.

I confirm that the ODR is fixed at 26.6kHz.

If my reply answered your question, please click on Accept as Solution at the bottom of this post. This will help other users with the same issue to find the answer faster 🙂

[vix]

Hi Federica,

I don't understand: ODR/10 is 2.66kHz, and the Nyquist theorem requires that the signal doesn't have any frequency above 1.33kHz.

And so a cutoff 5.6kHz is too high.

Is there something I don't understand properly?

[vix · Accepted solution]

After my investigation, I can say that buit-in low-pass LPF2 filter cannot be used as a hardware anti-aliasing filter.

If a downsampling is required, a suitable software low-pass filter must be applied before decimation (to avoid aliasing).