tested with copetr and plane and adc test:
someway z seems coupled with y in a strange manner.
when a negative y value is present z acts in a stepped way, at certain values it drops by about 500 than it continues as expected.
HOW TO REPRODUCE THE PROBLEM:
slowly roll the apm to the rigtht from 0 to 90° (y: 0 to -1000 , z: -1000 to 0 )
EXPECTED
z axis going slowly from -1000 to 0
PROBLEM
as soon as y start to be under zero z goes immediately to -600
than slowly to -1200 where it sudden goes to -500 (at this point y is -500)
than slowly from .500 to -1000 (y to -1000 too)-
when y is positive all works ok, x axis is not playing into this problem.
I cleaned the board with no results. Checked for electric coupling with no results.
When tiltin g the board several time it freezes, this can explain my crashj on first flight, I didnt checked the sensor before flight because the dcm was working, now noticed that dcm works ok even with acc broken but if I give it the time to correct the drifts with acc I can see the wrone attitude returned by acc that correct in a wrong way. I will never know if it was faulty from start, only thing that I can say is that it was not a so bad crash and that I'm sure it was an apm freeze problem.
wandering..
Ideas?
Replies
after all trials I excluded everything else and found the issue is accelerometer z axis.
I took some log and found the acc problem was there since the first log. I talked to another guy here in my country and he told me that 3 oilpan over 5 he bought recently had and acc or gyro problem, in the past there where no problems.
Is it a bad batch problem? how should I proceed?
the problem is not the acclelerometer, all axis checked with a voltmeter shows correct values.
adc works for everything except this y z acc blend, I never saw a problem like this with adc.
considered the jump are about 512 my tought is about some software error, it happened a similar thing to me and is was an error in recontructing a 16 bit value so a bit that should be on 2nd place was on the 9th giving the jumps in values.
with adc test I found this:
when y pass at 2048 I have a 256 jump in z axis, this repeats at y= 2304 (that is 2048+256).
it looks to me as a soft bug in dealing with the 9th bit but really I can't imagine that I'm the only one who noticed that.
can someone confirm that mine is not a common behaviour? just to be sure. still wandering in the dark.