Params check please

Hi
I 'm trying to fine tune my Trex 550E helicopter. I did a autoflight yesterday, which went quite well, But would like to know if I can improve on the setup? Any help would be welcomed

Attached bin file

I looked at your log. If you look at the desired attitudes compared to your actual attitude in the ATT message, you will see that the roll axis tracks pretty good. However looking at the pitch axis, the trend looks similar between the two but there is quite a bias. IMO, This is due to the I gain not being able to close the gap. I noticed that your ILMI is 0.02 for both pitch and roll. If you look at the PIDP message, you will see the I value goes up to about 0.02 to 0.03 and levels off. IMO, this is caused by the leaky integrator leaking off the I term before it can build to help hold attitude. Now I thought the leaky integrator would have been turned off because the aircraft is moving at 5m/s however I looked into the code and it seems that the leaky integrator is not turned off until the aircraft speed is above 5m/s for 2 seconds. So I don’t think leaky I was turned off.
So initially I would recommend increasing ILMI in both pitch and roll to at least 0.05 if not higher to maybe 0.08. Your takeoff behavior may seem a little different where you will have to make some attitude corrections as you are lifting off but it will be manageable. Ultimately the changes will help the flight control system hold attitude much more tightly.
Make those changes and do another flight and post the log. I would like to see how the pitch axis is doing. Thanks!

Thanks Bill, I will try adjusting the ILMI. The weather is not looking good, but will do test flight as soon as I can, and let you see the log.

hi Bill
to-day, was perfect for flying, no wind and warm.
The two bin files use ILMI of 0.060 and 0.080. I also increase the speed to 550 between weighpoints. It looks a lot better on the pitch. As this copter is going into a Bell 222 Shell, I like to fly at a scale speed. Is it possible to reduce below 500 and still get the I param to work?
Thanks again for your help.

Yes things are definitely looking better with ILMI of 0.08 on pitch. You can probably leave roll ILMI at 0.08 or slightly less if you so desire. Pitch is not bad at all but if you want it looks like you could use just a little bit more ILMI on pitch. I think you could set it to 0.10 and that would be best. Let me know what you think and post another log with pitch ILMI of 0.10 if you like.
Thanks,
Bill

I would like to point out that when you have IMU aliasing like this, and vibes on the y-axis reaching up into the 20’s it is going to cause deviation from desired attitude that can’t be fixed with tuning params. How the vibration affects the aliasing of the IMU’s will depend on the frequency. The PH2.1 is quite good at filtering high-frequency vibration, low frequency vibration affects it more.

Hi Bill & Chris
Sorry for the delay. To cut down on the vibrations, I fitted another new main shaft, both supposed to new. As you can see this one is better than the first. will they be OK now Chris?
I altered the two ILMI as you suggested Bill, putting pitch up to 0.10, and roll down to 0.06. I did a test flight in the garden hovering about 4ft. also in POS HOLD. I think they make DesPitch and Pitch line up a lot better.
Again thank you both for the help you have given me.

Geoff, your IMU aliasing now looks good with whatever you did. The Y and Z vibes seem a bit high but it’s not affecting the IMU’s that much.

Just getting the IMU’s so the signals aren’t aliased helps attitude tracking accuracy, and how nicely the heli handles.

For the most part it’s really hard to compared desired to actual attitude in hover. It’s kind of like, if the heli handles good in Stabilize and Acro, and holds position in the wind in Loiter or Pos Hold, who cares if it tracks perfect? It’s just not that important. Where it becomes important is in dynamic flight where the heli starts to fly.

The reason it’s so hard to say it’s great in hover is because heli’s can correct with only a few degrees of cyclic. They are way more inherently stable than multicopters. A human pilot can fly a FBL heli with no gyros or flight controller at all, and one that’s set up good mechanically will basically hover by itself in no wind with no gyro corrections needed. Full size helicopters are flown that way all the time - just direct stick to swash. There is no gyros making corrections unless it’s a FBW helicopter. The reason people think they are hard to hover is because novice pilots tend to over-correct and end up “chasing it”. So then the reaction is, “whoa - this thing is like trying to balance a beachball on top of a basketball.” When the real problem is the loose nut on the stick. This is a thing we teach new helicopter pilots - demonstrate how stable the machine is by letting go of the cyclic in a stable hover and watch the look of disbelief on the student’s face when he/she reaches the realization that the pilot is the problem.

That is impossible to do with a multicopter - at least to my knowledge there is none that can be flown without a flight controller.

The autopilot has some of the same problems as a student pilot in a full-size in a hover. It tends to “chase it” a bit. But once the heli goes into translational lift and starts to fly, now if the autopilot calls for 15 degrees of bank the heli better respond. Now we’re calling for real control inputs where over-controlling is a not as big of a problem as under-controlling. So seeing a log where the heli is flying at 7-10 m/s tells me a lot more than hover.

@ChrisOlson and I have a different view on attitude tracking.

I believe desired and actual attitude should track within a deg or two regardless of hover or forward flight. For the flight control logic to work well, it expects that these are tracking. Granted, satisfactory handling qualities may be achieved with these loosely tracking but this is not desirable from a flight control design point of view.

This is only due to the flight control system not tuned properly. A properly designed and tuned flight control system should fly just as well in hover as it does in forward flight. Also this statement implies that we can’t tune a heli as well as a multicopter in hover. This doesn’t make sense if multicopters are inherently less stable than heli’s.

Many people use helis for their great performance in forward flight and maybe are willing to sacrifice less than optimum handling in a hover. However I’m sure there are users out their that may use them for hover and I think it would be unacceptable for the aircraft to be chasing the attitude.

Multicopters just aren’t as quick to respond as helicopters are. I’ve based my observations on watching what the flight control system does in Stabilize vs turning on flybar mode and hovering the same FBL heli manually in Acro.

If you remember quite some time ago I was doing some flight testing and tried out Acro with (I think my 626) and was almost overjoyed because I thought they had “fixed” Acro flight mode because it worked so nice. My joy turned back into sorrow when I discovered I had accidentally left flybar mode turned on.

The autopilot is not perfect. It does exceptionally well in forward flight. But in hover, my opinion is that it “chases it”. It tends to overcorrect and always seems to want multicopter desired attitudes for the correction, when heli’s only take a tiny portion of what it seems to think is “desired”. And actually, from what I can tell, this doesn’t affect the hover performance or wind fighting capability at all. It’s just not quite as smooth as a skilled human pilot. It can be seen in the logs. But it’s most evident if you mount a fixed FPV or video camera to the heli and play the video back. On autopilot the heli is constantly making corrections that a good human pilot won’t do, which appear as oscillations in the log. But what it is is over-correcting by the autopilot - it’s “chasing it”. And there is no combination of tuning I’ve found that will stop it. In a stable hover, the desired roll attitude no wind, for instance, should be a straight flat line. But it’s not. The heli makes a little tweak to port, the autopilot makes a bigger tweak to starboard. Repeat, and it’s “chasing it”.

If your heli is set up good mechancially, turn on flybar mode and hover it in Acro and try it. You’ll see the difference immediately. It is suddenly very smooth and graceful and loses all the robotic type “feel”.

And I’m not saying this is a “flaw” in the autopilot. It is a limitation of any autopilot. It’s a robot. It doesn’t have the skill or computing power of a human pilot.

Yes I would agree that the autopilot hasn’t been perfected yet for the heli or we haven’t figured out how best to tune it. Certainly it does well enough to be able to hover and hold position but it doesn’t have the smoothness that I see in multicopters. We need to fix that as well as smooth air/ground transitions. I hope to learn a lot from @Leonardthall regarding how he has done the I term management with multicopters to help improve this and potentially get rid of the ILMI and leaky I implementation.

I think the autopilot does a pretty good job of hovering a heli, all things considered. The main difference between it and a human pilot is that the autopilot is a robot. It takes readings from various sensors and makes control inputs. But as a robot it is always one step behind what the heli just did. A skilled human pilot is always one step ahead. The human pilot is able to evaluate and execute with skill learned thru experience and training. The autopilot can only do what it is programmed to do. And this is why even in aircraft with the most advanced autopilots on earth, in mission critical situations where human lives are at stake, there is highly trained human pilots that ultimately have control of the aircraft. The autopilot is only a tool to reduce pilot workload.

As far as multicopters vs heli’s are concerned there is two issues.

  1. multicopters respond at a slower rate than helicopters. The heli can be tuned to respond at multicopter rates by adjusting the angular accel and rate (now in 3.6).

  2. helicopters are only able to hover because of their design. And this design has made them indispensable in modern aviation for everything they are used for. But the helicopter is more related to the airplane, as they are in their true element and become much more efficient in forward flight than they are in hover.

Ardupilot does a superb job of flying a helicopter on full autopilot in forward flight. It is the best there is for RC heli’s. It is where helicopters really shine for doing survey type work where you have to cover long distances with stability in the wind at steady ground speeds. They beat both fixed-wing and multicopters for being impervious to wind and/or turbulence. Even for point to point delivery work they are superior. Companies like Amazon that have this dream of using “drones” to deliver packages got the totally wrong idea. Use helicopters and then you got a concept that will actually work, that the FAA will take seriously.

So when it comes to flying a heli with ArduPilot, I can’t complain at all. It does a reasonable job of hovering one that may not be perfect but it’s good enough. I flight it’s so good I’d rather have the autopilot fly it than me.

Hi Bill & Chris
Well I managed to get to the flying field to-day. As you can see I’ve put the Trex 550E into a Bell222 body, and taken on it’s maiden flight. I seem to have got the shakes. The setup is as the previous flight. It. seem to have increased the ‘Y’ vibrations. I checked the main blades for balance with a new balancer, and one it slightly heavier which I have adjusted. Is there anything else I can check or alter?

Geoff, that constant bobbing in the pitch axis doesn’t look good. Yeah, the Y-axis vibes are pretty heavy, running between ~30-55. That could be part of it, but amazingly the IMU aliasing doesn’t look all that bad on the y-axis

If you want to try something, and I don’t know how well this work with a scale model, we now have some new defaults for 3.6 in the form a param file you can load with Mission Planner. You will find it in the dropdown selection for the different frame params you can install. It looks like most of your settings are what the param file has in it, but it will put a slightly different tune in your attitude controller.

With those winglets on the scale model it could be causing some interaction from downwash and airflow over them in forward flight.

If you would want to try these new defaults, save your current parameters first. Then load the new file called TradHeli-Copter36-Setup.param from the frame parameters dropdown in MP. This will only tune the position, loiter, nav and attitude controllers - it will not touch your other settings.

The tail settings are not too far from what you’re using, with the exception the new default params use more VFF. So after loading them test hover it first and see what the tail does. If it seems solid try your flight again and see what it does. If the tail is not quite right use your old params to re-tune the tail servo.

It would be interesting to find out if the new defaults fix that pitch oscillation. The new defaults use less P-gain and considerably more I-gain, and over-tuning P is one of the things I’ve found that causes pitch oscillation in auto flight mode. They also apply a tiny amount of D-gain, which may actually help with your scale fuselage.

Hi Chris.
This is interesting for me too.
After I changed my 700E and put it into a JetRanger scale body. I did had a terrible tuning session ( the Heli destroyed the tail and 3blades - to much D-gain at the start I believe) Now back to the 2 blade head.
The scale 700E Heli is now bobbing in pitch. Because I did not had the time ( tourist season) I left the 700E untouched since Nov. 2017. Winter is starting here soon and I will get the JetRanger ready for a new tuning.
I am also building at the moment a new TRex 600 to replace my RJX520 which is manufacture discontinued - no parts.
BTW, I have that tail destroying tuning and the before good flying with VFF only, PID-0, all on video.
I am watching what OAPpilot is coming up with and I will try the same.

The new default parameter tuning file has flown a pretty wide variety of heli’s successfully. It is designed for kind of “middle of the road” settings for a 550-600. For bigger heli’s it has been necessary to turn up the VFF’s a bit. For smaller ones turn down the VFF’s a bit. And then adjust the ATC angular accel values for “pilot feel”. I have not found one yet where it needed any adjustment to the roll and pitch PID’s, although it has not been tried in a 450 class yet.

And then on the tail, the defaults have proved to fly every one. But every one has needed adjustment to the yaw rate P or D to suit the linkage and servo being used. It uses a little more VFF on the yaw than what most people are using, but I’ve found that gets rid of the “tail shaking” on all my heli’s running lower UAV/scale headspeeds. I found that the tail doesn’t have quite as much authority running lower headspeed, and the rate PID loop can’t correct it fast enough - hence the “tail shake”. Using the feedforward makes the servo respond faster and seems to smooth them out.

So we put that new param file in the Tools for 3.6, hoping it might make tuning easier for users, and only have to adjust a few things to fit the size of the heli and pilot preference.

Chris, I downloaded that param’s file, and did a test flight in the garden. I tried a PosHold at about 18" high. The garden is not very big, and the pilot not very good. It seemed a lot more stable. So when I can get upto the field where I can do the auto flight again, when I will sent another video and Bin file. I’ve attached a Bin file of to-days test… Thanks again, I would be lost without your help.

Hi Geoff, that sounds very good. That’s what the param file is supposed to do is provide a more UAV/scale type tune that fits a wide variety of helicopters.

So will be looking forward to your auto flight test. That will tell if it gets rid of that quite bad oscillation we’re seeing in the pitch axis with your scale model. I don’t know how your CG ended up with the scale build, but that “tune” in the param file should handle a pretty wide range of fore/aft CG quite well.

Just curious about your scale body - do those winglets on the tail boom pretty much mimic the real thing, with an upside down airfoil? On the real thing, the mast and main rotor is tilted slightly ahead in the frame. The purpose of those is to provide a nose-up assist in cruise flight to reduce cyclic loads for the pilot so the heli flies nose on the horizon. A Jet Ranger is the same. On a scale model where you likely have not done a mod to tilt the rotor ahead in the frame I don’t know what the effect of those airfoils would be in flight, but I suspect it might make it tail heavy and the autopilot would not like that.

The upside is that scale models are typically flown at low speeds for show and scale quality handling and characteristics, especially for scale competitions. So they might be small enough, and flown at low enough speeds, that it won’t affect it.

Yes I agree the horizontal tail might be affecting the tune in forward flight. In addition to aiding in the cyclic loads for trim attitude, it also serves to provide speed stability so if the aircraft was to be hit by a gust of wind the increased velocity on the horizontal tail causes more downward lift and raises the nose and slows the aircraft back down to its trim speed. The other purpose it serves is to provide pitch rate damping. When the aircraft has a nose up pitch rate, the horizontal tail sees upward flow, increasing its angle of attack, and causing a downward moment to oppose the pitch rate.
There are a lot of benefits to a horizontal tail in forqard flight. I feel that they are mostly for show on most acrobatic models but for the scale models, I think they provide the same benefits. However you probably will need to adjust the tune.

Geoff, the aileron and elevator look really good, actually, with the new params. But the rudder is a little shakey

It’s not using hardly any of the P or FF because the D gain is causing the shake. I’d back the ATC_RAT_YAW_D down to .002 or .003 so it will use a little of the FF and the P gain can damp it a bit better.