Replies to This Discussion

Permalink Reply by Limebear (Rob) on June 8, 2011 at 5:04am

HI guys, here's a general question related to drift:

 

How critical is it that the quad is perfectly balanced in both axes? If you balance it on the centre of its arms, should you add some small counterweights as necessary, (once you've done the best you can by fixing the battery position.) How much error is acceptable (in terms if grams counterweight near the motors)?

 

Secondly, how critical is it that the motors are exactly vertical. After a minor crash i noticed that one motor was a bit off, so I corrected it by twisting the frame manually. It's quite easy to sight a vertical with the quad on a table and using a try-square to look at the motor-shaft.

 

Thanks

Rob

• ▶ Reply

Permalink Reply by Kristian on June 8, 2011 at 5:14am

I don't think it's that critical. I sometimes fly with a camera(300grams+) hanging from the front arms. Barely notices it.

 

I also have a quad with another kind of controller(multiwii), that have had it's share of crashes. I tried to fly this once with one arm bent 10cm upwards and one arm twisted, no problems, but it needed a litt stick trim. :-)

• ▶ Reply

Permalink Reply by Kristian on June 1, 2011 at 10:34am

I enabled RAW logging, flew around a bit until I almost could not control the quad anymore. The logfiles are available here:  http://dl.dropbox.com/u/19126845/ArduCopter%202.0.20-logs1.zip

• ▶ Reply

Permalink Reply by Vorgin Smorgin on June 3, 2011 at 3:01am

Had Another go today.

Disabled compass,reloaded the code and still the same.

Starts off hovering in one spot,than after a bit of maneuvering ,

hands off the controls and the quad still moves around.

here are the accell and gyro offsets,How do they look?

 

Gyro offsets  -35.00, -45.01, -68.91

Accel Offset  14.37, -19.07, 18.61

 

Tried changing the imax from 50 to 100 on both pitch and roll ,which seemed

to help.

Will try more testing tomorrow???

• ▶ Reply

DeveloperPermalink Reply by jasonshort on June 4, 2011 at 11:22am

Some people don't seem to have this problem, but if you land and wait 30+ seconds and have full control again it's the DCM's drift compensation. Doug is working on a new concept so that we can compensate drift dynamically. We'll be able to slow the adjustments during flight maneuvers and speed it up when the sticks are in neutral. Hopefully this will solve the issue!

Jason

• ▶ Reply

ModeratorPermalink Reply by John Church on June 4, 2011 at 12:26pm

That's good to hear- thank you, guys.  :)

• ▶ Reply

Permalink Reply by Greg Fletcher on June 4, 2011 at 1:12pm

This drift problem seems to happen during aggressive maneuvering. Could it be that some airframes/PID settings can cause the rates to exceed the gyros limits. Gyro saturation is data lost and the soloution gets tilted.The DCM then re-levels its self when it is sitting still on the ground. There's not much you can do about that except to control the maximum rates somehow.

• ▶ Reply

DeveloperPermalink Reply by jasonshort on June 4, 2011 at 1:24pm

Yeah, two things can happen, you exceed your gyro rates, or you accelerate over a long period. Anything that will make the accelerometers not point to the ground on average will cause an error. The more you do that the worse the error. When the accels are truly stabilized, you can correct that error. Hard to tell in many situations, but quads I believe are easier than Airplanes which can fly in perfect circles. 

 

• ▶ Reply

DeveloperPermalink Reply by jasonshort on June 4, 2011 at 2:50pm

Oh, just thought of something that may account for some folks not seeing the issue. Noise and vibration. Be sure to dampen the APM with some foam.

• ▶ Reply

Permalink Reply by Limebear (Rob) on June 4, 2011 at 2:51pm

That sounds like the right solution. Just like with an echo canceller, you don't adapt the filter coefficients when both ends are talking, here we shouldn't adapt the drift offsets rapidly when more than a little user input is being applied or the vehicle is accelerating or rotating significantly. There will be plenty of time to adapt when sticks are near neutral AND current calcuations indicate low accelerometer and gyro outputs in all axes. I agree that slow/fast is better than on/off so that you can't get stuck in a situation of persistently calcuating an erroneous high offset.

Rob

• ▶ Reply

Permalink Reply by Steve Smetters on June 6, 2011 at 7:20am

Up until now I had not noticed the re-centering issue but I was concentrating on loiter and simple flying. Yesterday I put the quad through much more aggressive maneuvers with forward (and backwards) flight, steep turns with dramatic yaw rotation and the re-centering issue was very apparent. Landing and waiting a bit did seem to help the issue.

• ▶ Reply

• ‹ Previous
• 1
• …
• 3
• 4
• 5
• Next ›
• Page