Constant z axis offset for -1g

Hello LoremIpsum thanks for posting.

Does this same drift on the bottom sensor repeats sistematically on different boards? 75mg is rated as the max. limit for the offset on the 2g range, thus 99.7% of parts should present in worst case this offset on startup.

Still, there are some other sources of error that can change the offset value as; non-simetrical PCB layout, a thermal gradient due to a power regulator routed nearby, presence of vias under the footprint or handling points further than 3cm are some of the factors that can help inducing tension in the sensor's package and translate in some extra 0g bias drift. On top of that, there is also place for variability from part to part on the sensitivity of the part, or non-mechanical factors as using an external supply instead of the internal LDOs can also lead to offset drift if it is not precise enough.

Overall, for high precision applications calibration at least on room-temp is recommended after soldering the part. Remember that you can configure the offset trim registers for this purpose.

Some collateral in relation to this topic: 

AN-1057: Using an Accelerometer for Inclination Sensing | Analog Devices (calibration guidance at the end)

Choosing the Most Suitable MEMS Accelerometer for Your Application—Part 1 | Analog Devices

hope this helps, have a great weekend.

Mario SM

Hi MarioSM ,

thank you for your quick response.

Yes, we can observe the offset on different boards with different sensors from different lots. VSUPPLY is connected to a 3V3 source, so (if I understand the data sheet correctly) the internal LDOs are used.

Mechanical stress is for sure a possible causes, still I am not a 100% convinced as the top sensor is also affected. Also, I find a deviation of 10% or more/hundreds of mg) too high for just being caused by mechanical stress.

Hello LoremIpsum 

The drift is always positive and with the same magnitude between boards? (+0.1g)

Is there any major difference in layout, or placement of close by ICs as LDOs between both parts?

I also expect that both sensors are use the same script, are supplied by the same power source, and share the same GND.

is the drift constant in time ? Is there any major difference on the turn on output during power up for both sensors?

I will be OOO next week, expect some delay from my end, wish you a great weekend.

Mario SM

Hi MarioSM ,

thanks for your reply.

The drift is indeed always positive. The magnitude is not always exactly +0,1g, it can be slightly lower or higher. It was significantly lower (~ +0,05g) when the top sensor was not placed.

We do not observe a change of the offset over time, it remains the same over days, when the sensors are in steady state.

MarioSM said:

I also expect that both sensors are use the same script, are supplied by the same power source, and share the same GND.

Yes.

MarioSM said:

Is there any major difference in layout, or placement of close by ICs as LDOs between both parts?

This I need to check. I would assume so as they are on different sides if the PCB.

Temperature values are similar, qualitatively and quantitatively, this makes perfect sense to me.

Hi MarioSM ,

another thing we noticed, during our analysis, which might be completely unrelated, yet worth to add here:

In steady state, we observe noises peaks for all axes at around 100 Hz and some (sub)-harmonics, for sensor ODR of 500 Hz. 

This is perfectly repeatable, even for a battery powered device with no connection to the mains.

Attached is the plot for y axis, only the peak at ~75Hz is an actual disturbance we know of.

Also, we measured the VSUPPLY. It is way in between the limits, ~3.27 V. Only V1P8ANA/DIG (generated by internal LDO) are a bit of at ~around 1.72V. Still with remain valid with Vpp ~ 120 mV.

Kind regards

LoremIpsum