Dangerous divergence: FIXED and understood

The last Bastion to fine tune your FC matching your airframe & setup

Re: Dangerous divergence issue: NOT YET SOLVED

Postby afernan » Thu Feb 14, 2013 8:48 am

I see now that many folks have experienced this problem, and I agree the "key" looks around "Tilt_rate_D" but probably not only. Anyway needs to be verify by test (I´ll try as soon the weather let me).

"Tilt_rate_D" is in fact the second derivative of the error (or derivative of the rate). It "slow down" the first derivative or error speed (or rate). If too high, It will push in the same direction than the error (easy to see if applied on a curve like error=sine(angle)). But the truth is outside, in the flight field...

@Max: yes, this´s the same quad than have yaw problems.

Angel
afernan
 
Posts: 1214
Joined: Sat Jul 07, 2012 5:24 pm
Location: Madrid, Spain

Re: Dangerous divergence issue: NOT YET SOLVED

Postby LPR » Thu Feb 14, 2013 3:02 pm

Angel

The only proof I have for using all default values is the videos of the octo and quad flying. No one is more surprised then me by the fact that the default values work for both a 9 lbs octo and a 3 lbs quad.

http://www.youtube.com/watch?v=_7im1ExXEug

http://www.youtube.com/watch?v=mMRxmmTCNRo

The string you see connecting the motors helps keep vibration under control. The aluminum arm causes vibration but their cheap and the string stop the problem. When I'm done with beta testing I'll use the carbon fiber arms I have.

I do have some problem with yaw while taking off with the quad.

Larry
LPR
 
Posts: 1323
Joined: Tue Jun 26, 2012 1:29 pm
Location: MN, USA

More flight test on divergence

Postby afernan » Fri Feb 15, 2013 1:32 pm

I´ve done two more flight test forcing this divergence issue.

First I did with r76. I had a lot of problems with this release. I can´t say now if it´s the release or my copter (I really thought my FC board was broken). Ended with a crash (you know, "no a day without a crash...).So, for the moment r76 is under susspect for me.

I came back to r65. And tested again today.
Started using full defaults. all basic things were OK.

Reduced the CTRL_TLT_RTE_D to 5000, then to 4500. Things improved a lot. Looks the right direction. Forced more and more the divergence but did´t appaired yet, at least laterally (ROLL).
Testing then in PITCH it came back and ended (again) with a crash (Only one arm + 2 props).
(Main difference between Pitch-Roll is that in Roll it has mor inercy due to the battery is set left-right direction.)

See the kml file attached and play in GE. Zoom on it to see how the copter behaves. It´s very teaching. Crash is at the end of "Pitch-Crash" track

Conclussion of the day. The problem is still there. Apparently TLT_RTE_D is playing an important roll, but not yet enough.
Still to test:
- use the 100/100 mix table
- reduce CTRL_FACT_P/ROLL to .4
- reduce CTRL_TLT_ANG_D to 900

Angel
Attachments
DIVERGENCE -Speed TESTS.zip
(41.18 KiB) Downloaded 825 times
afernan
 
Posts: 1214
Joined: Sat Jul 07, 2012 5:24 pm
Location: Madrid, Spain

Re: Dangerous divergence issue: NOT YET SOLVED

Postby JussiH » Fri Feb 15, 2013 1:58 pm

Angel, you can always switch back in manual mode when you see it begin to happen...then it should be able to level the copter again.

And, I am still convinced that this is tied to uneven distribution of power (eg unbalanced ESC´s or motor mismatch)

Pls post a screenshot of your motor log in a hover!
JussiH
 
Posts: 2103
Joined: Thu Jun 21, 2012 9:25 pm

Re: Dangerous divergence issue: NOT YET SOLVED

Postby LPR » Fri Feb 15, 2013 2:23 pm

Angel

One test you could try to see if your motors are balanced is to, in manual mode, try to very rapidly increase altitude and see if the quad tilts one direction consistently.

It sounds like your getting closer to finding a solution.

Are you using a 50, 50 pitch and roll motor mix.?

Larry
LPR
 
Posts: 1323
Joined: Tue Jun 26, 2012 1:29 pm
Location: MN, USA

Re: Dangerous divergence issue: NOT YET SOLVED

Postby afernan » Fri Feb 15, 2013 4:05 pm

@Jussi:
I get the divergence flying in manual mode. Obviously the effect increase in PH or RTL. When I get the divergence in manual some times I can recover by counteracting with the stick the rotation, but it´s very difficult.

In this last fly I´ve well re-calibrated the SimonK ESC (I´ve checked the PWM min/max of the TX to 975/1950 and then every ESC is calibrated with an external RX) . In fact now the yawing problem I had it´s solved.

I think this is a real problem, that´s more critical in ligth platforms. I´d like to see a video of others (or kml file) moving at high speed and suddenly release of stick. I´m sure any can get same problem.

@LPR. I´m ussing a 50/50 mixtable, which is the more standard. And I don´t feel we are close to solve this problem and the worst thing is that flight testing is very dangerous (I´ve got already docens of crashes testing this)

Angel
afernan
 
Posts: 1214
Joined: Sat Jul 07, 2012 5:24 pm
Location: Madrid, Spain

Re: Dangerous divergence issue: NOT YET SOLVED

Postby LPR » Fri Feb 15, 2013 6:26 pm

Angel

Here is a video of the my quad working in RTH. http://www.youtube.com/watch?v=mMRxmmTCNRo

Your quad is setup very much the same as mine is. I am using ESC32, 5010 motors and 15" x 5.5 props. That motor controllers are you using?

I changed to r76 today and the quad is flying great. I had a problem with r76 and RTH but when I used a different FDTI
and re-flashed r76 it is working fine now.

Larry
LPR
 
Posts: 1323
Joined: Tue Jun 26, 2012 1:29 pm
Location: MN, USA

Re: Dangerous divergence issue: NOT YET SOLVED

Postby afernan » Mon Feb 18, 2013 11:57 am

@Larry:
thanx for the video. Yes, I had a quad configuration pretty close the same you have: ECS32, RCT5010 motors with 15x5 carbon prop´s (although the weight was much lower: 800gr without batt). In this video you can see how it performs in RTH and PH against the instability : .

I´ve been testing a lot this weeked this problem in a new "heavy dutty" platform with thicker arms to protect it about crashes.

I can summary my experiences up to know as follow:

- this problems affects much more to lighter copters (< 1Kg w/o batt)
- this problem happens ONLY when flying agresivelly (much more than in lpr video)
- Divergence in PH or RTL are easyly controlled reducing "nav" parameters as described in my video
- Divergence in pure manual mode at "high horizontal speed >> stick release" I DON´T know how to eliminate today. It improves a bit by reducing:
- tilt_rate_D = 5000
- control_Pitch/Roll = 0.04 (this limit the max angle in flight, bud it´s superseed with the divergence, shoing angles up to 90º)
- (tilt_angle_P, I, D : don´t affect significantly)
But I can´t eliminate nor understand the behaviour of the code, really. There is a first overshooling that should not happen based in the limitations of the parameters.

For the moment I need to fly with a "LOU" (in aeronautical "Limitation Of Use"): "don´t release your stick suddenly if you fly at high speed"


Angel
Last edited by afernan on Mon Feb 18, 2013 3:04 pm, edited 1 time in total.
afernan
 
Posts: 1214
Joined: Sat Jul 07, 2012 5:24 pm
Location: Madrid, Spain

Re: Dangerous divergence issue: NOT YET SOLVED

Postby afernan » Mon Feb 18, 2013 12:25 pm

@Jussi:

attched is a plot of my motors in hover. I think they are pretty well balanced.

Angel
Attachments
PH motors output.jpg
afernan
 
Posts: 1214
Joined: Sat Jul 07, 2012 5:24 pm
Location: Madrid, Spain

Re: Dangerous divergence issue: NOT YET SOLVED

Postby afernan » Mon Feb 18, 2013 3:03 pm

In the plot below we see the pitch angle vs time (sec) during a real divergence it happened to me (see video attached). We see that we reach very high angles of pitch (+70º , -60º). At that moment the stick is centered.

It should never happen since CTRL_FACT_PITC = 0.04 that limits Pitch at about 30º. What´s failing in the AQ control? (at that moment motors are not saturated).

Angel
Attachments
divergence 004-AQL.zip
(661.47 KiB) Downloaded 806 times
Pitch vs time an a divergence.jpg
afernan
 
Posts: 1214
Joined: Sat Jul 07, 2012 5:24 pm
Location: Madrid, Spain

PreviousNext

Return to PID Tuning

Who is online

Users browsing this forum: No registered users and 15 guests

cron