Servers by jDrones

ArduPilot New Sensor Test

in-tree-sensors
beta-drivers
(cala2) #1

First test: a Robocat 270 with 2350 kv motors, 30A esc with oneshot disable 2200 3s batt and in process to tune; I upload latest vers and try as it.
Yaw I know that it was untunned, pitch and roll looks little worse that previous 3.4, but flyable, then I try alt hold and looks worse, than previous, move up and down and clip0 looks bad. If It’s usefull for you I can do newer tests with your suggestions.
(at begining I didn’t modificate compass, It’s looks good as it, I hope. check filters if they are adecuate for this vers, I modificate in the past )

Calling all testers! New ArduPilot sensor drivers
(Thorsten) #2

Hi Cala,
yes, the virbrations on this one are (too) high. Do you have a log of this copter with AC 3.4 to compare with?
I am wondering how much the higher sampling rates will help with such vibrations. It seems the code to filter vibrations has not yet been updated in this respect - I might be wrong but I havn’t found anything related in the recend bunch of code changes.
I’ll try to test it asap as well.
@tridge Should RAW IMU logging be enabled for these tests?
Best regards,
Thorsten

(cala2) #3

I was surprise about vibrations, the only change I did was to adjust motor bell that had up/down movement, but with no experience perhaps I ruin an axe? :flushed: , I’m going to return to 3.4 release and test again to be shure, GLUP!!

(cala2) #4

2nd test: now I return to stable release and no other changes, more windy, things return to normal FIUUU!!!.
something change with latest I hope, but it’s no possible a firm issue I can try again if something loose in my frame? (I don’t remove the batt, only open and close the FC cover to plug usb) Let me know if I can help with more tests, thank’s

(tridge) #5

@cala2, thanks for the test data! I’ll re-check the filtering on the icm-20608. That seems to be getting a lot more clipping than in the 2nd log.
I’ll also re-fly my pixracer based 250 and see if I can reproduce it.

1 Like
(tridge) #6

@cala2 I’ve pushed a new version of the code to master. It should be on the build servers in a few hours.
The new version uses a new filtering strategy which should help a lot, and also uses DMA for SPI transfers.
Can you possibly try the new version and post a log?
Please set INS_FAST_SAMPLE=3

1 Like
(cala2) #7

Downloaded, ready to test, at night I upload results :slight_smile: .thank’s Tridge

(cala2) #8

@Tridge: today a windy day too, I test before and after change firm because I autotune yaw and pitch again, roll is still manual tune yet (not finish), Ins_ fast was =3 in the two tests. Copter looks better but still little nervous, more flyable now, perhaps needs different tune?.
First one with standard Fw


The latest