APM:Please 2.75 has now been released.

Have a look at the release notes in the forum for lots of details on the changes in this release.

I'd also ask that you not reply to this posting, but instead use the new APM forums for discussion of the release. We'd like to use the 2.75 release as a trial of the new forums for release discussions, in the hope that it will provide a better way to host the discussions.

Cheers, Tridge

Views: 2735

Reply to This

Replies to This Discussion

Hi Mack,

In both your 2.74 log and your 2.75 log your servos are moving much too rapidly, with considerable oscillation, especially in the elevator. I think both your P gains on roll and pitch are too high, and the D gain is too high as well. That oscillation makes the control algorithms unstable, and I also would be concerned about overheating your servos.

I think you could also raise TECS_TIME_CONST to 10 to smooth out the altitude changes, but first I'd concentrate on the roll and pitch controllers.

I'd suggest the following changes:

RLL2SRV_P from 3 to 2

RLL2SRV_I from 0 to 0.1

RLL2SRV_IMAX from 500 to 2500

RLL2SRV_D from 0.2 to 0.1

PTCH2SRV_P from 3.3 to 2

PTCH2SRV_I from 0 to 0.1

PTCH2SRV_IMAX from 500 to 2500

PTCH2SRV_D from 0.3 to 0.1

TECS_TIME_CONST from 5 to 10

WP_LOITER_RAD from 9 to 50

These changes are rough guesses based on your logs. You will probably need some fine tuning.

If you can try it again and post a followup with how it goes (using 2.76) that would be great.

Cheers, Tridge

thank you for the help.. I will follow your suggestions and let you know how it goes

I had a very similar, almost catastrophic, experience yesterday.  I was running 2.73, and ungraded to 2.76, and confirmed the same 2.73 settings were set.  Went out for a test flight...  hand launched in FBWA, flew for a bit testing the response, seemed fine, maybe a little more 'loose' than previously.  Switched to Manual and flew for a bit, and then executed an RTL.  The plane sped up and down as you described, started heading towards the home location, started banking hard overhead, losing altitude, and continued going down, down ,down, until I was able to switch to manual and recover....

Going to go out today and try to fine tune it a bit, using Andrew's suggestions below.

Warren

Why do you guys post these alarming experiences without any logs. You are forcing Tridge to beg for the logs so he can do you the favor of solving your problem.
I'm very interested to see of there is a trend here, I just loaded 2.76 but haven't flown it yet. I have an important mapping mission coming up and might not have time to test fly.

Hi Iskess,

yes, logs are always appreciated, although if logs are not available then I still like being told of issues, as it sometimes lets me know what to look out for.

I should also mention that the dataflash logging in APM:Plane has had a substantial upgrade for the 2.76 release, and I have developed new tools that make it easier to analyse dataflash logs. I'll be continuing to update the dataflash log code for the next release.

To get useful dataflash logs for the DCM problem you would need to enable the IMU logging which uses dataflash very fast. With IMU enabled (along with the other default messages) the APM2 dataflash will fill in about 20 minutes, so make sure you switch off the plane after getting interesting data (the dataflash logging wraps).

On PX4 using a microSD card this isn't a problem, as even a tiny 2GByte card would last 100 hours.

Cheers, Tridge

Maybe all of us guys aren't as experienced as you and didn't know Tridge would want the files.. I did not ask him to "solve my problem", but he offered to help and I surely accepted...

Hi Warren,

Going to go out today and try to fine tune it a bit, using Andrew's suggestions below.

Note that when I suggest a tuning change based on someones logs those suggestions are specific to that airframe. Don't apply tuning changes I suggest for someone else to your plane or they may be way off.

Also, if you think you are seeing the DCM attitude failing then please enable full dataflash logging (set LOG_BITMASK to 32767) as raw high rate dataflash logs are the best tool for analysing that particular problem. The APM2 dataflash will only hold 20 minutes of logs with that enabled, but it will give the detail needed to really look into the issue.

Cheers, Tridge

@iskiss
My apologies, newbie first post mistake I guess, but... as Mack stated, I also didn't want to be presumptuous and post the logs 'expecting' someone to help... I would rather wait until it might be requested by whoever might be willing to help. That said, I am quite new to this, and from the little I've poked around the forums it seems Andrew is extremely helpful to all, so in future I will be sure to post the logs with any questions/concerns.

@Andrew,
I should have stated in my original reply that I am also flying a Penguin airframe, as was Mack, so I figured weight and power aside, it was probably close as a start...

I just got back from 3 flights of tuning where we followed the fine tuning guide from here: http://plane.ardupilot.com/wiki/roll-pitch-controller-tuning. It was nice to be able to be in the field with someone to adjust the parameters 'live', and see and feel the changes right then. We just went through Roll, Pitch, and Yaw tuning, as there is no airspeed sensor (yet) on this plane. After the tuning, it successfully executed the RTL and some simple WP missions. It seemed to respond well and maintained altitude fairly well during turns. I think we have it close, but it would be appreciated if someone with more experience might have a quick look at the logs.

Attachments:
Sorry for my last post, I've had a rough day and I guess my frustrations bled into my post.
As a general rule, if you have a question about your plane and you possess some logs, just post them.
Tridge is wonderfully generous with his time, and obviously more patient than myself.
Tridge, what does DCM stand for?

Hi Warren,

Your roll response looks ok, but your pitch response is a bit too wild for my taste. I think you have too much P gain and not enough I. This is what your pitch response looks like after you tuned, with the elevator channel on the right axes:

I think you should make the following changes:

drop PTCH2SRV_P from 1.7 down to 1.3

raise PTCH2SRV_I from 0.005 to 0.1

raise PTCH2SRV_IMAX to 2500

then see if your elevator moves a bit more smoothly and the pitch doesn't oscillate so much.

I also notice you are not using the defaults for AHRS_RP_P and AHRS_YAW_P, possibly you loaded parameters from an old release? Change those back to 0.2.

You might also like to adjust PTCH2SRV_RLL so that the plane holds altitude in FBWA when turning with no elevator input (it is a bit hard for me to work out the right value from the log).

If you continue to find that the plane oscillates in altitude in AUTO and LOITER, then raise TECS_TIME_CONST to 10.

A square or rectangular circuit AUTO mission would be a good next test for getting the tuning right.

Cheers, Tridge

Hey all,

I am going to use the latest firmware (2.76) to fly my RC C17 (http://www.hobbyking.com/hobbyking/store/__17784__usaf_c17_globemas...)

This airframe is less responsive compared to many others. So, first, I wanted to tune this in HIL  before i go out there and crash it!

Actually, I flied this using very old firmware (2.4 I guess), and had couple of crashes til I got it right.

But now with the all new release 2.76, I wanted first to test this in X-Plane. So I built a fairly accurate 3D model for this.

Now I am testing this in X-Plane, and I know even if I got it flying nicely in x-plane, it's just 50% certainty. But still much better than 0%!

So my question is:

1-  During the takeoff stage, I'd like to use the frontwheel to keep the plane on the center line (not to go off road), so I decided to put it in the rudder channel and use the ground steering new feature to solve this problem. Is this approach correct?

Thanks in advance.

Tridge, what does DCM stand for?

DCM == Direction Cosine Matrix. It is a 3x3 matrix representing the orientation of the aircraft. The "DCM algorithm" is a method of updating that matrix using gyroscopes, accelerometers, magnetometer and GPS.

Cheers, Tridge

Reply to Discussion

RSS

© 2017   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service