LIS3DH noise while access with I2C

[leca]

Hi,

It seems that accessing the i2c bus add noise to acceleration values read by the LIS3DH. The graphic represent 270 X,Y,Z 10bit samples read at 1344KHz. The FIFO is enable and a watermark level interrupt of 28 is set. At each interrupt, I'm doing a bulk reading of the output registers (28*6 bytes). We can see a cyclic glitch especially on Z axis in yellow that match the bulk reading period. 

Before using the FIFO, I was doing multiple single reading but the result was very noisy.

Is somebody had this problem? Is it a known issue? Is there a way to solve it?

I also include the layout of the PCB of the chip.

[Issamos]

Hello @leca 

I suggest you to refer to the AN3308.

This will help you to configure a filter that can reduce the noise. 

Best regards.

II

[leca]

Hi.

I don't want to used a filter since I want to do a fourier analysis on the signal. What I'd like to know is if it's normal to introduce noise to values when there's activity on the i2c bus for this device. In my point of view, it's not acceptable. So, is it because there's a problem with my pcb layout? Should I use SPI bus? 

Or this is normal behavior in which case, I will find another device.

Thanks.

[Issamos]

I don't think is normal behaviour. But to make sure, try to use the filter and look if the issue persist or not. If we still have problem that can be a PCB problem or an I2C probleme, the best way now is to test using spi and if the problem persist then you have to use another component because that means this is his normal behaviour.

Best regards.

II

[Bob Paddock]

Is that black line just above C3 silk or a break in the ground plane?
If it is the latter is there a large loop area making C3 and C44 ineffective?

The via in the pads of the discretes can cause thermal imbalances that lead to tumbestoning during assembly.for some CMs.