Pitt,
I will say this, I know a stable helicopter should be possible with the Pixhawk and the high tune-ability it seems to offer. I believe more filtering would definitely help with instabilities, and getting the right mix of gains etc for your specific helicopter I think plays a huge roll.
In my experience tuning FBL units, its usually quite easy to get a flyable helicopter3, but to get to the point where it will just sit in a hover like its resting on the ground and the tail stays rock solid with not even a hint of movement regardless of the wind or the speed and direction of the helicopter requires a lot of work and tuning through many flights. I have a 3D heli I just finished tuning with a Spirit Pro on it and it took about 30 flights to get to the point where iām at now. But in a hover it has no drift, wind doesnāt move it and every movement is smooth and fluid like a simulator. I was using a regular old Beast X FBL before on this particular 800 and was getting very nice video from the platform and the BeastX has very little if any advanced tuning available. Thatās system was designed to be user friendly and easy vs tune-ability.
So that tells me the helicopter is capable of flying great with Pixhawk, I just have to find that āperfect mixā of settings and maybe some advanced filtering is in order? I donāt know. It doesnāt fly terrible now and watching it sit there in loiter in 20mph winds like a rock in the sky was very encouraging. Just need to keep putting the hours in an hopefully get it the last 10%.
It would be nice to see the gap closed between Flybarred heads and FBL though when it comes to the speed at which it corrects unwanted movement. Every year processing power grows exponentially and with todays CPUās having their power requirements lowered while processing power goes up itās a matter of time before flight controllers have access to supercomputer like processing at 5V.
And iām sure the firmware will keep advancing as well, just need to keep interest in the helicopter platform and itās advantages so it doesnāt get trampled anymore from the multirotor mania. 
Tim
Bill, I will do that. Iām also quite interested to see if thereās a combo of settings to get the autopilot to work well at low rate loop speed. When I first started messing with FBL I set flybar mode to 1 and worked up a rate and expo to fly it with no gyros with the radio.
What I eventually found that filter does is like two methods of flying no-gyro with the FBL. You can make slow deliberate stick movements and the heli wanders around like its drunk. Or you can learn to do the stick dance and fly it quite accurate with no gyros. The filter set at 4, I found out, is like the slow deliberate type flying. Setting it higher is like doing the stick dance.
What I have tuned for recently is only to get ultimate stability in forward flight for auto mode, trying to match the flybar. I havenāt really looked at any of my logs. Itās all been done by making over 120 flights and tweaking and adjusting until it gets closer to what I like. Iāll set up an auto flight and fly it repeatedly 3, 4, 5 times making a tweak on every run and observing the heli and looking at the video from the run to see how stable I got it.
So there is no āscientificā method that can be applied to what Iāve been tuning as of late.
Tim,
I worked on incorporating the notch in 3.5.0-rc6. I have a compiled version ready but I realized that it might be nice for you to be able to set the leak rate on the Leaking I feature. This would help the integrator retain error in a hover. Once you get into forward flight the leaky integrator turns off and the integrator is allowed to build up to IMAX. I suspect Iāll have something for you tomorrow. Iāve got a 4 day weekend so I hope to get some other hobby stuff done like getting my X3 closer to flying again. The honey do list is wreaking havoc on my hobby time 
Tim, thereās a lot of garbage marketing that goes on in RC helicopters to get people to buy the latest and greatest. Itās like belt drive vs torque tube tails. The torque drive is touted as being more efficient and yadda yadda, none of which is true. Same with the FBL rage. It was all brought on by what sells with the 3D crowd.
I do not crash my helicopters on regular basis. In fact, I have never crashed one on a commerical flight. I had a mishap with the 600 testing firmware. But using the fact that FBL heads are ācheaperā to crash is bogus. There is none of the maintenance parts that I canāt get for an Align helicopter flybar head, and thereās nothing in them really that wears out. The bearings in the mixing/washout arms and flybar cage are all commonly available bearings used in many appplications like brushless motors. The link ball ends are identical to the ones used for FBL. Feathering shaft thrust bearings, blade grips - all the same as FBL. Even the swashplate is the same as a washout driven FBL head.
And if they try to wipe it completely from the market eventually, Iāll build my own heads and parts. Iām probably the only one that flies flybar helicopters with ArduPilot. But that doesnāt bother me either because I like how they work, and theyāre way easier on servos.
Awesome! Thanks Bill!
Chris,
I agree about the marketing hype, thats what drives most industries and is unfortunatly not unique to model helicopters. I understand some of the advantages to both systems, hype aside there are plusās and minusās to both designs.
Ive been playing around with speed helicopters lately and have been more than considering a speed specific model, namely the Henselit TDRII. I dont know if youāve ever seen one, but I assure you they are a model of German engineering and are filled with many speed specific features and little hype. Incredibly tight tolerances requiring the ability to use heat and cold to assemble, not your average Trex 700 or Goblin. Not that those arent great models, i own both, it this is a heavily engineered speed specific model.
As I said, if youve never seen one its worth a look at their site as youll notice some things they do very differently than the rest of the heli manufacturers. Its funny youve mentioned torque tubes and how they claim they are effecient, well with the TDRII they could have used anything they wanted but went with a ābelt driveā. The drive train is multi staged as are most german designs and built to take power systems that would strip the gears on any other model before it even spooled all the way up.
Servos are linear, no servo arms and varying geometry based on travel position. Most FBL systems cannot be flown with it unless the built in algorithms to compensate for non linear swash movement can be turned off, it results in severe over gain symptoms. (This also had me wondering about things with FBL units, servos and the like).
. The head is FBL, and the helicopter is purpose built for one thing, to be incredibly fast and incredibly stable at high speed. Its worth a look if you havent seen one, I think you would appreciate some of the departures from the norm they have employed.
Anyways, Flybarred helicopters, as you have pointed out, are inheriently stable by themselves dependent on flybar weight/paddle size at the sacrifice of increased mechanical complexity and more stringent setup skills. Truth be told these days, if you can somewhat read and have a ruler good to at least 1mm and can look for a few flashing lights on you shiny new BeastX, youll end up with a helicopter that will fly. It may not fly great, but whatever setup errors youve made will mostly be cancled out by the fbl.
I totally agree that flybarās will more smoothly dynamically balance the head provided they are setup properly compared to an FBL head, but I am also starting to see that when it comes to flybarless there are many, many variables that will either result in a stable well flying platform in the end, or a jittery mess.
Variables like setup skill, servo choice, blade choice, FBL choice, ability tuning the PIDās and other parameters and tolerances of the helicopter head in general.
For the time being, im stuck with my choice of helicopter to employ the Pixhawk 2.1 on, good or bad. I have multiple of the same airframe and a large cache of spare parts to support them so its logical i make this work. I am 100% interested in my next Pixhawk 2.1 build being a flybar helicopter and am exicted to see the differences in setup and flying style. I appreciate the information you have given in defense of them along with clear cut reasons why you choose to use them vs FBL and that is largely why I will build one next. I honeslty would have considered it novelty until only recently as the market is full of FBL and flybarred helis are like hens teeth now. Combine that with my heli experince only going back some few years in todays market, i just dident get the exposure. Yeah my Kyosho Concept wasent FBL, but that was a long time ago and things were very different back then.
Tim
Well, lets put it this way - my flybar head for the 500 is just sitting on the shelf right now, and itās going back on the helicopter as soon as I get done experimenting with the FBL head 
https://goo.gl/photos/472BcBFkk9iuwn8j9
The FBL, for me, is just an experiment. There is no way I will use one on a commerically flown RC helicopter. When I started the experiment I thought I might. But what Iāve learned along the way is that itās just not as reliable under all situations and condtions. So Iām going to stick with what I know works, and doesnāt have all the quirks.
I certainly am not familiar with that speed heli you mention.
Chris,
I totally understand your position, and as I said, after hearing of your successās with a flybarred model, im keen on trying it out for myself. Task at hand though is getting this Trex 800 flying as good as I can with what ive got. Im hoping the notch filter bill has might help. Time and data collection will be the litmus test.
And you really should take a look at their āHenselitā site sometime. Quite a departure from the norm.
Tim
What are you going to try to do with that? Getting the rate P cranked up to some target value is not the answer. Multiās fly more on rate P and less on I. Helicopters fly less on P and more on I. And they need more I gain in FF flight than they do in hover. The software is already designed that way. And because helicopters canāt run as much P gain as a multi they have a setting that multiās donāt - the VFF. Used to be just Rate FF in 3.3.3.
So I dunno, Tim. If you ask me, you were pretty close the other day. Just turn that FILT param up so the rate PID loop can run a little faster and it would likely sharpen it right up. The default speed for that PID loop is 20Hz, and most people run 20. I wish you would try that. Set everything back to where it was when you made the Loiter vid the other day. Turn up those FILT params to at least 10 Hz, fly it and post a log. You posted those params as āthe best feel yetā or something like that. If I recall I thought your I gain was a bit soft yet, your P gains were not even close to being dangerous, and you were running somewhere around .12 VFF. That shouldāve been right in the ballpark except the desired vs actual was a little sloppy yet because you were running that rate PID too slow.
I know Bill disagrees with me on that, but thatās how I got my DFC helicopter to fly decent without having the ājittersā.
Just a correction for you Tim of the name.
The link: http://henseleit-helicopters.de/
I am to old for racers like those. Even I have no problems with the website - my native language.
There are YouTube videos regarding speed test with Henseleit Helicopters.
Chris,
Iām not convinced that your tuning strategy is the best tuning strategy. I donāt know if what I have in mind is whatās best. I will agree that you have had success in tuning for forward flight. And Iāve said that for attitude command this tuning may have acceptable handling qualities. But would most likely not be acceptable for a rate command control system like acro.
Iām interested in learning why you are seeing the unacceptable behavior in forward flight. I understand that you have a job to do and you tune for what works. In my opinion, the P gain should add more stability and help with the behavior you are seeing in forward flight. I want to understand why it isnāt. If my X-3 was up, I would check it out myself but I havenāt had as much free time lately to look at this. Do you recall at what speeds or above what speed you started seeing the unacceptable behavior?
By the way, I want to take a minute to plug my discussion on UAV handling qualities research that Iām involved in. There is so much experience on this discussion group and I hope you respond to my post below.
Fred,
Thanks for the link. I saw one at OHB and it was an incredible build. The goblin Comet also made an appearance and they did some speed runs with it Ive fallen for the TDRII though, itās something else to hear a 700- 800mm rotor turning close to 3000rpmā¦Sounds kind of like a wood power planerā¦
Building one of those is a substantial investment though, and your all but assured a crash running that close to the limits of everything including RBS. But its like all racing i guess, you think about winning, not what happens when things go wrongā¦
So your native German? Thats on the wife and iās bucket list once kids are grown, October Fest & German countryside etc.
Tim
Chris,
It was definatly flying good the other day, but there was definitely some work to do still.
Before i started changing parameters again, P gain was at .039 and .035, I gain was at .39 and id have to look up the VFF again. Ive saved the parameters from then just in case. My only reason for adjusting was trying to get desired vs actual closer together. It was still a ways off with those settings.
Since changing settings i will say it holds in wind better and I was able to get P gain over .05 and lower I gain closer to .3. What would be good from Billās filter would be allowing me to adjust the filter up towards 20hz and use a notch filter to target the instability while leaving the rest alone. As it sits now, I attempted to raise that to 10hz with less than stellar results. Had to lower P gain again from the low settings i am running now and every stick movement resulted in a small ossicilation.
Tim
On another note, i like the idea of a notch filter from my pro audio experience. When i used to set up for bands, every room had different acoustics and was prone to feedback through the microphones at X frequency. Rather than use a graphic EQ or a lowpass/highpass filter, i would use a notch filter in the EAW UX8800 signal processor to target the specific problem frequency without coloring and affecting the rest of the audio spectrum. It was actively weighted so it adapted to internal gain on an as needed basis and prevented feedback loops in the system without a significant change in sound quality.
As ive stated earlier, i understand filtering a bit, but its all based on audio principles, and ive often wondered if any of it would apply to this application?
Tim
I thought your P gain was higher than that.
What worries me about your setup is tuning a low headspeed helicopter in hover, then having it break into a headshake in FF flight in Auto. It will result in total loss of control. I made this very mistake, man. I had the rate P up to .19 at one point. I was at about .11 or .12 when I just about lost it. Thatās when Robās warnings to me set in about running into instabilities in FF flight running lots of P gain.
You really need to take the helicopter out and fly at increasing speeds to know what you got. Tuning for little wiggles in hover just does not ultimately give you what you need to have a stable FBL helicopter. Youāre familiar with FBL units - you should know if you tune the rates for 3D and try to fly that helicopter at high speed it will wad up and crash almost before you leave the line. And what do they do? The heli first noses down, it rapidly noses back up, goes into a headshake and flares, rolls right inverted (CW rotor) and goes into a vertical dive. Pixhawk/ArduCopter is no different. And you pretty much totally lose control of the helicopter unless you realize whatās going on, get that rotor feathered right now, and get the nose pulled up before it proceeds without delay to the crash site. The autopilot is just bone stupid when it comes to something like that, and instead of flying the helicopter, it only tries to counter every move the helicopter makes from what it thinks is the target attitude, which makes the situation worse.
So I dunno. Mulitās were made to hover. Helicopters were made to fly. Donāt try to tune it like a multi.
Sounds like filtering might help here. How fast were you going when you hit this?
Iād be interested in seeing this log as well.
Chris,
Yeah, my P gain is still quite low. I had to back off the VFF a tad as well as it was still quite snappy. Once i added in the non zero ILMI it was a bit less snappy as well.
One thing the helicopter is not lacking in is rate authority, it isnt quite as snappy as my 3D helis, but to me it feels very quick for a UAV helicopter. It is limited in bank angle, which i guess is a good thing as it would cover some ground very quickly otherwise.
I have in fact flown it quite fast. The day i was beating the snot out of it in the wind abd had that ossicilation when flipping into loiter i easily had it up to 60mph, maybe more. It wasent a speed run by any stretch, but i was purposely flying it fast and erratically trying to tease out issues. Which i did when i was descending rapidly and flipped it into loiter to see how it handled it, which was not well.
Iām just trying to dig any issues out of it before i add expensive camera gear and gimbals etc.
I still have a ways to go i feel before im comfortable adding the gimbal into the mix.
I really would like to try Bills filter and see if it is possible to target the 5-6hz instability and get the low pass filter up closer to 20hz. I know if i just raise it now, even with the P gain at .04 I will have an ossicilation as i did befofe when i was bringing P gain up by itself with the filter at 20hz.
Tim
I think it was target speed of 15 m/s. That was probably 60-70 flights ago.
This is one of the things that turns me off on FBL helicopters. The constant coming up with patches, filters or fixes to get 'em to fly right. Thatās fun for my personal test heli. But for the commerical ones my insurance company will only allow one wad up. And theyāll pay the damages. But thatās the only one I get because I then become uninsurable. So thatās why I have to stick with what I know works.
Tim,
Had a little fight with git and compiling. I got it all sorted out and I have it compiled. I just havenāt had time to post it with the instructions. I should be able to post it tonight after everyone goes to bed. Iāve been fighting all day with installing interior doors. If I did this for a living I lose moneyš³.
Bill,
Thanks i really appreciate it. Totally understand about finding time, there is just never enough of it to go around.
Dont feel bad about the doors, my guys do it for a living and my jobs still lose moneyā¦ Hard to find good helpā¦
Thanks again,
Tim