LSM303AH self-test fails

[Eetu Nevalainen]

Hi,

We're not able pass LSM303AH accelerometer self-test on our products. Please see one of the results below. The datasheet doesn't describe what may be wrong if the self-test fails so could you guide us with this issue.

As far I understand the allowed difference on LSM30AH is: 70 <= |st avg - nost avg| <= 1500.

nost x: {56, 76, 68, 84, -40}, avg: 48.8

nost y: {-180, -172, -220, -148, -204}, avg: -184.8

nost z: {-8284, -8284, -8296, -8284, -8284}, avg: -8286.4

st x: {3116, 3128, 3240, 3228, 3164}, avg: 3175.2

st y: {2460, 2460, 2504, 2468, 2488}, avg: 2476

st z: {-5588, -5608, -5596, -5588, -5588}, avg: -5593.6

x diff: 3126.4

y diff: 2660.8

z diff: 2692.8

// Eetu

[Eleon BORLINI]

hi @Eetu Nevalainen​ , which FS did you set? Please note that these values are for 2g. And did you write 01h to ST [2:1] of CTRL3_A to enable positive self test? Regards

[Eetu Nevalainen]

FS is set to 4g and positive self-test is enabled. Here's I2C traces from the self-test procedure. I didn't notice anything out of ordinary but maybe you can spot something.

[Eleon BORLINI]

​if FS is 4g you have to double the self test limits (the ds ones refers to FS 2g), changing from min 70mg, max 1.5g to min 140mg, max 3g. In this way, y and z values falls inside the limits, not the x value, which is too high... are you moving the LSM303AH during the self test? Regards

[Eetu Nevalainen]

Thanks for the clarification. So are the limits triple if 8g is used and so on? The board is stationary on my table so the sensor should be not moving.

[Eleon BORLINI]

>> So are the limits triple if 8g is used and so on? Yes, you are right

But is the x value repeatable? Are you facing issue in reading DATAOU_X value? Eegards​

[Eetu Nevalainen]

Nothing indicates that there's an issue in reading the values. I changed the board to another one and this one has a problem with y-axis even though the orientation is the same as before; x=2992, y=3748 and z=2974. I repeated the test over 10 times and the results are pretty much the same.

[Eetu Nevalainen]

Okay so maybe the values themselves are incorrect. I'm using 14bit resolution so should I right-shift values by 2?

[Eleon BORLINI]

​Hi @Eetu Nevalainen, yes, consider that in normal mode case OUT_X/Y/Z_L/H_A data have to be computed on 16 bits and in two's complement (although the two lsb are 0). Regards