zmr250 & EKF strange behaviour

Log Analysis -- deprecated Copter Logs -- deprecated
1

Hi all,

I’m tuning a small zmr250 build controlled by a Pixhawk Lite.

[youtube]https://www.youtube.com/watch?v=tVc4ktyg9N0[/youtube]

As you can see on the attached picture, I see divergence between AHR2 and ATT (EKF) after a fast forward move. After I release the stick, the quad is still leaning forward as shown by AHR2. However, EKF thinks the quad is flat and therefore no correction is made to flatten the quad. What can be the cause of this behaviour? Notice that AHR2 and ATT converge again after a while.

[attachment=0]ahr2_vs_att.png[/attachment]

I have high interference on the internal compass and can’t use compassmot since my current sensing is not working. I think this could be related to the compass and therefore increased EKF_MAG_NOISE but this does not help.

I have no GPS on this build, only the Pixhawk Lite. Notice also that my CoG is a bit at the back and I initially thought this could be the problem and I tried to increase IMAX on pitch but after investigation, I’m rather convinced that my PID are not too bad and EKF gets somehow confused :confused:

https://drive.google.com/file/d/0Bw0ExtElIgOXOUp0R0tsd3Jja1E/view?usp=sharing

Thank you.
Cyril

2

Hi lostcontrol,

We are having a look at this. If you are up to it we may get you to test the new ekf that should handle this better.

3

Hi mate,

Can you get us some logs with your bitmask set to:
485373

Chat soon,
Leonard

4

Hi Leonard,

Thank you :smiley: Sure, I can give EKF2 a try and give you as many logs as you want. Unfortunately, it’s getting dark early in the evening these days so you may have to wait a few days before I can flight again.

I will increase logging with 3.3 and do some fast forward moves and then try 3.4-dev. Can I take one of the daily builds?

Don’t you see this on your small quads? Do you have an idea what causes this behaviour?

Maybe some more details on my setup: ZMR250 (from ebay), Pixhawk Lite (from GLB), SunnySky x2204s, AfroESC 12A, Gemfan 5040, 3S 2200mAh.

[attachment=0]IMG_20151022_230008.jpg[/attachment]
Thank you and Regards,
Cyril

5

I will directly switch to 3.4-dev and enable both EKF (but flight with EKF1 first) as described by Paul here: groups.google.com/d/msg/drones- … R8InOJDwAJ

6

I did a (very) quick flight with 3.4 just before it was completely dark to compare EKF1 and EKF2.

[attachment=0]ekf1_vs_ekf2.png[/attachment]
In this case, EKF2 seems to better follow the real attitude of the quad. Here, despite the offset, I guess EKF2 would have done a better job. I did it in a rush so maybe EKF2 would benefit from a longer “warm-up” phase!?

I will actually try to flight EKF2 tomorrow if I find some spare time.

Full log here: https://drive.google.com/file/d/0Bw0ExtElIgOXOW5YQ29mendIbm8/view?usp=sharing

Thank you
Cyril

7

So I flight with EKF2 on Saturday and had a very bad crash :frowning: I unfortunately hit hard ground and many things got damaged included the microSD which is not usable anymore. Therefore, no logs… My impression was that EKF2 was not much better than EKF1 and even had more drift in situations where EKF1 was performing not too bad before.

I shouldn’t have tried to do flips on the first flight :cry: Actually, I’m to blame at 99% here. Switching to stabilize didn’t flatten out the quad quickly enough to recover. The roll/pitch were maybe not accurate after the flip!? Hard to tell without logs though.

I now ordered a Naze32 for this small quad since I will likely crash a lot and only need basic features :wink: I’ll keep using APM:Copter for my larger builds.

Cyril