So the idea behind a slower leak rate would be to allow more integrator error to build. I set ILMI to zero and put the leak rate to 0.001. This should get rid of the aircraft wanting to lean prior and during take off and also will help with maintaining attitude in a hover
From this log, I found the frequency to be between 5.5 and 6 Hz in pitch axis
So I set PIT_NTCH to 5.8
Result:
2017-05-28 12-57-41.bin (4.2 MB)
The heli oscillate in a different maner. Slower, softer oscillation, got worse with higher P gain. At stable P gain without notch filter (0.130), the oscillation is only bounce when giving pitch jab.
In short: Notch filter seems to make things worse in my case.
Just curios, what happens when lowpass filter (FILT) cutoff freq is higher than notch filter (NTCH) freq?
About I_LEAK, its seems like I didnāt include it in my code.
Is it in 3.5? I will be moving to 3.5 once It supports this frame.
Pitt,
Iāve found that the notch will work in some instances and not in others. Since your aircraft has intermeshing rotors, it may be more difficult. We may have to look at just notching the gyro signal. That way we remove it before the mixing of the controls and command signals. I was working on an implementation of a notch that only filters the gyro signal. Lucky for you I incorporated it into 3.4.6. But I have not flown it in my heli. It seems to work in the SITL and I even injected a oscillatory signal on the gyro to check to see if it was working properly. You are welcome to incorporate it into your code. The commit removes my previous notch and adds the new one. Here is the link
https://github.com/bnsgeyer/Copter3_4/commit/009ab6f3a38df254254346777be246f207626fb2
As for the I_LEAK parameter, it will not be in any version of released code as I have not done a pull request on that change. Iāve given it to Rob but he is still testing his latest changes and Iām not sure what heāll push to master.
Finally figured out how to use git cherry-pick
Which commit will I need to pick?
- Leak rate of I term (Jan 22th)
2)Notch filter on gyro signal (8 days ago)
Do I need to apply all other previous changes?
Todayās tuning, with high head speed, it holds attitude much better
P gains are 0.15 for Roll and 0.110 for Pitch 0.220 for Yaw
PIT/RLL_FILT = 4, *_NTCH = 0, YAW _FILT = 22
Good enough?
No, you donāt have to apply any other changes to implement them.
As far as you settings, I donāt know. How well is it holding attitude? To within 1 to 2 deg? 0.22 for yaw is good. Is the yaw response better?
It is less likely to wander around and not wobble in wind.
I will post a log and video if there is no rain tomorrow.
For yaw, I can keep adding more P gain. If it starts to show sign of bounce/oscillation, adding D gain will stop it.
But with too high D gain, yaw response becomes too soft (when centering stick, it would slowly slow down before it stop yawing)
Iām staying around 0.008~0.010 D gain. A little soft but since I couldnāt make it rock solid, this is the sweet spot.
However, this high gain may be the result of low H_YAW_SCALER (0.4 means 40% servo travel of differential cyclic compared to pitch scaler)
Yes, I was thinking the same thing. You may be able to reach these high gains because the differential long cyclic does not have as much authority.
Anything missing?
Log: https://www.dropbox.com/s/saeomfijvrcby6z/2017-05-29%2014-12-37.bin?dl=0
Yaw scaler changed to 0.5 from 0.4 (20% increase) no sign of oscillation
Edit: video:
No. It looks like everything transferred in fine.
Havenāt had a chance to look at the log but the video looks great!
Pitt,
I looked at your log and you still have some oscillatory stuff going on in the background but I think the higher I gains with the non-zero ILMI have gone a long way to helping keep the attitude tracking the target. So if you want you could back off a little on the P gains to remove the oscillations and if you feel it isnāt snappy enough then add some VFF instead. These settings are more along the lines of what I learned from @ChrisOlsonās tuning. With this setup I would not recommend using acro or sport mode or any mode that is rate command in nature. If you want to explore the use of the notch to increase the P gain, you can. It is up to you.
Havenāt flashed the new firmware yet.
I might move to 3.5 altogether.
I donāt feel like using those anyway but would like to know the reason.
How about loiter or auto?
This tuning should work fine with loiter and auto since they are all attitude command based. Another reason for backing off on P gain a little is that the loiter and auto modes may be more likely to excite the instability frequencies. So that is why I recommended bringing up VFF if you arenāt getting the response you need after lowering the P gain.
Before I flip the switch to loiā¦alt-hold first?
What other tuning/parameter check will I need?
In my opinion, the low P gain values would not provide sufficient damping and unless you are an experienced pilot that is good at flying lightly damped heli. The other thing is that I think the high I gain would make the aircraft lag quite a bit behind your inputs. If you have ILMI = 0 it might be a little better. I played a little with these settings when I started with arducopter last year. I did not like acro with my settings which was P gains around 0.06 and I gain around 0.1. I had not D gain and VFF was around 0.6 if I remember correctly.
I might have a chance to go to an open field this week. Can you help me set up for Alt-hold and loiter?
Iām still using CH8 passthrough for throttle which shouldnāt work but I canāt find a proper guide for that.
I might not be the right man for that since I havenāt flown heliās autonomously. Iāve flown multiās but not heliās. I think the big thing that you have to look at is the ACCEL_Z_P value to be around 0.3 or less. I think that is the biggest concern.
Forgot to update 
Had a successful Loiter. 12mins = 3.76V/cell
Log: https://www.dropbox.com/s/b7akxqsbquuhhoa/2017-06-29%2014-58-44.bin?dl=0
Now trying to mount a camera 
Also found this 420mm blades interesting http://www.mscomposit.net/CFC-main-blades-42-cm-6-3-d66.htm with 6mm root, It fits my 6.5mm grip. However, Iām not confident my rotor can spin it safely, the grips and spindles should be fine but 6mm mainshafts and M2.5 spindle screws worry me.
Tapping M3 maybe possible but a step up to 8mm mainshafts will need a lot of work.
Is it safe to stay with 6mm mainshafts?
Thatās really neat. Love that machine.
Pitt,
What size rotor system was your design based on? Iām assuming that the rotor hub, shaft and blades were all based off of the same rotor design. It sounds like you are going to longer blades? I would base the decision of your shaft diameter more on the weight of the helicopter than the size of the blades. I say that because the shaft is primarily transferring the rotor moments into the aircraft. I think for most 3D helis that are turned into UAVs, the shafts are way overdesigned because the sizes are more based off of loads in 3D aerobatic flight. The angular rates/accelerations are much higher in 3D flight and put more strain on the shafts for the same weight helicopter.
Take @Chrisolson or @timbaconheli, they are running 700/800 size helis at nearly double the design weight with the extra fuel/batteries and camera gear and I believe that is acceptable because they arenāt doing 3D aerobatics. Even my X3 is nearly double the weight of its original design.
The other thing to remember is that you have two rotors, so each rotor is only imparting half the required moment to accelerate the aircraft.
I might be more concerned with the blade grip bolts. I would think they are design more for a centrifugal force based off of blade mass and rotor speed. so I wouldnāt exceed the maximum mass* rotor speed^2 value for the original design of rotor system with your new setup.
Just my thoughts.