Zero G Offset

[beakie]

Been doing some tests with my Chronos EZ430 watch's accelerometer and I am getting some unexpected results.

When I work out the value for 1g and the zero G offset, I get different values for each axis?

(Rotating the phone as slowly as possible) I recorded the maximum and minimum values returned by each axis. See below.

Min Max
X -52 52
Y -44 60
Z -7 99

I looked in the datasheet for the accelerometer and it says there could be an offset shift on Z due to the soldering process. What about X and Y? Shouldn't they be the same then?

Thanks

 

[Sceadwian]

What is your test setup? Even minor out of balance spinning or tilt will skew your results.

By the way I love you forever, I want one! If only I didn't have a hundred other things that need to be done.

 

[beakie]

What is your test setup? Even minor out of balance spinning or tilt will skew your results.

By the way I love you forever, I want one! If only I didn't have a hundred other things that need to be done.

Haha. Couldn't resist when I saw the price

I rotated the watch whilst capturing the max and min value of each axis. I then confirmed my results by getting the same results whilst the watch was lying flat.

I am sure my results are right... but is this normal?

 

[Sceadwian]

That's not what I was asking beakie, I was asking HOW you rotated it, did you just spin it in your hand or were you using a properly fully level surface and a mechanical turntable that was RPM controlled and slides? You may be slightly off axis from the true axis of the accelerometer.

 

[beakie]

That's not what I was asking beakie, I was asking HOW you rotated it, did you just spin it in your hand or were you using a properly fully level surface and a mechanical turntable that was RPM controlled and slides? You may be slightly off axis from the true axis of the accelerometer.

I simply rotated it in my hand. BUT after doing this I confirmed the results by getting the same results while the watch was lying on the floor untouched. Surely the maximum and minimum values returned are those aligned perfectly with each axis?

 

[Sceadwian]

Considering the X and Y axis are separate sensors inside the chip it's probably just normal chip variance. These types of sensors can't be used reliable for dead reckoning, drift dominates after any period of time. You're also assuming the axis you're moving the watch on is it's true axis, the chip may not be mounted perfectly straight.

 

[beakie]

You're also assuming the axis you're moving the watch on is it's true axis, the chip may not be mounted perfectly straight.

I didn't move it along an axis.

Firstly, I turned it very slowly in random directions for a long period of time. I appreciate there is no accuracy in this method but this was only a means of finding rough values to help me with the next step.

I have the watch output showing on some software I wrote on the pc which allows me to see what I am getting realtime and I lined the watch up so that I got the same min/max values I had before. I then balanced the watch in this direction with some books and took my hands away. This proved the values were not affected by slight hand movements.

Surely the maximum value I would get on any axis is going to be +1g and the minimum would be -1g... as long as the watch wasn't moving?

Cheers for ya responses btw

 

[beakie]

I tested it and when I now display it as a graph I am seeing what I hoped.

When I rotate the watch 45 degrees, the 3D object isn't turning correctly but that is a separate issue which I will look at now.

Cheers for ya responses

 

[Sceadwian]

Do NOT try to process the data on the watch, offload the raw data as fast as you can to the PC, and use PC software to process it.

 

[beakie]

I am not doing any watch-side processing.