8kHz Noise with LIS25BA Accelerometer

[CupOfTea]

I see a constant, very strong peak at 8kHz in all signals measured with the LIS25BA accelerometer without any obvious source. The peak is very sharp, so there is probably a coherent signal coupled into the system, and it's not the sensor resonance or anything like that. It is strongest in the z-axis, but also present in the other axes.

The sensor is mounted on a custom mini board and connected to it's periphery via a shielded cable. The output sampling frequency is WCLK=24kHz of the sensor and the master / bit clock MCLK=BCLK=12.288MHz . I encountered the problem with all mini-boards / sensors I tested.

I'm a bit confused now, because of the digital nature of the transmission, I think the noise has to be coupled into the signal on the chip itself, either mechanically or electrically.

I'm not sure, which countermeasures I can't take. Does anyone know, what might cause this problem or how to counteract it? Is this maybe a known problem? Any help appreciated! If necessary, I can provide further information / material.

[Svan .8]

Signal bandwith is smaller than 2400Hz why would you look for frequency like 8000Hz? I good filter is a fast solution. But we faced simmilar problems, you get this kind of peaks if you don't sample with exact ODR, ODR/2 or ODR/4. Keep searching for Aliasing - Wikipedia 😉

[CupOfTea]

Thanks for your quick reply!

Unfortunately, an additional filter strong enough to filter out this distortion would create other problems in our application.

What do you mean by "exact ODR"? The WCLK is exactly 24kHz, and thus should be the ODR? This ODR is specified in the datasheet as a valid one, thus, I would expect it to work. Do you mean, that the noise always occurs, when not WCLK=8kHz but WCLK=16 or 24kHz is chosen? Or do you mean the problem occurs if there is some frequency jitter? I'm familiar with aliasing, but I'm not quite sure how it does apply in this situation. I would have expected that the LIS25 just does an internal lowpass filtering to remove content beyond 8kHz?

[Eleon BORLINI]

Hi @CupOfTea​ ,

Since the MEMS bandwidth is 2400Hz, I don't believe it is a mechanical effect.

Is the 8kHz peak you are seeing sharp? Can you share the raw data (or the FFT data)?

To check it is an electrical issue, can you also try to change the WCLK from 24kHz to e.g. 8kHz (so that you are excluding the 8kHz in the ODR bandwidth?

-Eleon

[CupOfTea]

Hi @Eleon BORLINI​,

thanks for your reply!

The peak is very sharp, with the right window and resolution it is really only at 8 kHz and 50 dB high. I cannot share the raw data at this point.

Unfortunately I cannot change the word clock due to system limitations.

Could the problem be, that I use ~12MHz for the bit as well as the master clock? From my interpretation of the data sheet, this is not the standard, but should work.

[Eleon BORLINI]

Hi @CupOfTea​ ,

You are right, using ~12MHz for the bit as well as the master clock is not the standard configuration.

It should work, but can you try also one of the other suggested clock configurations?

-Eleon

[CupOfTea]

I will look into it, but unfortunately, I'm not sure, how I can test this, due to hardware limitations.

[CupOfTea]

@Eleon BORLINI​ small update, I tried using a ~12MHz MCLK and a ~3MHz BCLK, but unfortunately, the noise was still present with this configuration. Unfortunately, I still cannot check the suggested WCLK configurations at the moment.