Servers by jDrones

This should be interesting

(Bill Geyer) #841

Yeah, now that I have the X3 flying again, I am going to tune using that technique and then look at the frequency response. I also want to look at the frequency response with using mostly FF and I gain. Mainly in stabilize mode. Just curious if it provides the same frequency response as one with more P and D gain. Keep me posted on your progress and hopefully I can get a version of software out that has automated sweeps (chirp) in it.

(Chris Olson) #842

I’m another one that used Leonard’s tuning method on my GX9. It worked perfect and flew at 65 mph out of the box in Auto with no issues at all.

However, some differences from what Tim notes; I couldn’t get the Accel and Gyro filters to make one squeak of a bit of difference. So I left them at default. I ended up with WAY more VFF than what I have traditionally used on FBL heli. More in line with the amount of VFF I use with flybars to get rate roll and pitch to match.

On the stabilization adjustments for the rate PID loop, I could set pitch D to some ridiculous value and never did get any shakes in the pitch axis. So I used roll to tune and applied the results to the pitch axis, except making them a notch or two more aggressive on pitch than roll “just because”. No real scientific reasoning to what I used for the pitch axis, other than it is typically a slower-responding axis than roll, so I made it a little sharper than roll.

I didn’t bother with the overshoot thing. It was fine without messing with that.

I know from experience that the P gains I ended up with, and setting I gain equal to that, does not work in high speed flight with a helicopter. I set I gains equal to VFF instead, and it worked “out of the box”.

The rate loop filters I set to what I’ve always used, and they “just worked”. So didn’t change them.

The results I ended up with on the 900 just so folks can see some real-world results using this tuning method on a big heli.



ATC_RAT_PIT_D , 0.003
ATC_RAT_PIT_P , 0.09
ATC_RAT_RLL_D , 0.0025
ATC_RAT_RLL_I , 0.18
ATC_RAT_RLL_P , 0.08

(Timothy Bacon) #843

It definitely is intereating to see what you used on the 900. I really need to take a moment and reduce the mechanical rates on that Trex. :confused: If I had a value of .2 on VFF and P near .1 it would be snappier than my 3D helis with a pile of overshoot to boot. I think all along I have been working with super quick rates, and the cyclic servos that can double for tail servos if need be likely make matters worse?
I really just did this quickly as I noted above and just wanted to see what results shake out.

(Chris Olson) #844

Oh yes, no doubt, and the 900 was the same. That’s what you use these for:

I set the VFF to get the DesP and DesR to match actual and didn’t pay any attention to “feel”. Once I got those to match I slowed it down with the above params because the default of 110000 was way too fast. Those params turned it from a snap demon into UAV-type handling in Stabilize.

When I did the D-gain procedure (with P and I set to zero) I tried pitch first. I think was at like .02 and it never did do anything different. Instead of wasting time adjusting these values one notch at a time, I went to high extreme on roll D. I think I started at .007. Lifted off and the heli was totally uncontrollable. it started to shake and settled right back to the ground - didn’t even respond to the collective. I set to .005, lifted off, and it was fine. Flew it sideways and scraped a skid on the ground and it immediately went into a gentle shake. I figured that was “close enough”, cut it in two and set it to .0025. Then set pitch one step more aggressive at .003.

Did the same thing with P gain. I think I started at .15 on roll and got a gentle, but quite pronounced shake right away on liftoff. From experience, I know that’s the “edge”, cut it in two, and instead of making it .075 I just set it to .08, with pitch one step sharper at .09.

Since I gain don’t make one hoot of difference in hover anyway with a heli, nor does it make any difference in “feel” I hovered it in Loiter with I gain set to zero and the above settings. Pulled the logs and saw perfect tracking of desired vs actual attitude in Loiter.

So I set the I gain equal to VFF on a “hunch”.

Set up an auto mission with 7 laps in a 1 mile long figure 8, with a Do Jump command to repeat the figure 8, and increase speed 5 m/s on each lap. Turned it loose and just stood there and watched it fly, ready to take over if it did anything bad. it didn’t. Pulled the logs at the completion of the auto flight and very pleased to see less than .5 second response time on desired vs actual pitch and roll at all speeds.

Total procedure took maybe two hours, most of that time spent pulling the SD card and looking at logs during the VFF part of the tuning to see if the Des vs actual rates were matching. If I recall, I started at .1 VFF and turned it up .02 at a time until they matched. I must admit, I was a little surprised to end up with what I fly with flybars for the VFF.

As an added note, this tuning method could probably go into the Trad Heli wiki for tuning. It works. And it is far less dramatic than the current wiki because tuning the VFF first with everything else set to zero insures that you always have control of the heli during the rest of the tuning, short of getting a severe shake like I got tuning roll D. But if you’re only a few inches off the ground, even that’s not dramatic or dangerous.

(Timothy Bacon) #845

Ill have to re-read your post when I have more time, but its note-worthy that what i described above was with ATC_ACCEL_R_MAX 20,000 and ATC_ACCEL_P_MAX 20,000

(Chris Olson) #846

That would be strange. That should limit your angular acceleration rate to 200 degrees/sec squared. I think I tried it as low at 36000 and it was a little too slow. I never did adjust the RC feel. I left that at 50.

At any rate, I was happy with Leonard’s tuning method. It was quick, simple, never had an instance where the heli was being flown with no, or inadequate control, and it worked. It kind of gave me a better understanding of how the rate controller works. The feedforward is used to match the actual rate request. The PID loop is the electronic flybar and dampens it. That was kind of the "light bulb over my head’ realization when it was explained to me. It is exactly how I have tuned my flybars - set the VFF so it responds properly, then adjust the paddle weight and pitch for the damping. The only difference is, with flybar the damping is mechanical, with FBL it’s electronic.

(Timothy Bacon) #847

I have RC feel way down there as well. I think it likely is a combination of my chosen servos as well as mechanical rates? The servos I have on cyclic are .04 at 8V which is what I run them at. Even though they are fast servos they should only move as fast as commanded so I am at a loss as to why its so darn snappy?
Only other thought i have is blades. They are very wide chord blades which can make it very poppy and floaty so that could be contributing?

(Chris Olson) #848

I’m sure it has to do with disc loading, and personally I would not run the blades you have for a UAV helicopter. I have used asymmetric blades on my 700 since day one. I had Sungwoo’s on that. I just got a set of A-VII 620’s for the 600 about a week and a half ago from Empire RC. They increase your flight time over symmetrical airfoils. With the Spinblades and another 5A battery on it, it increased the flight time to 34 minutes @ 30 mph cruise @ 1,550 rpm headpseed, making that helicoper a serious player. And it weighs 11.9 lbs takeoff weight with 13A 6S on it.

(Timothy Bacon) #849

Hmm, dunno Chris. It very well could be, im kind of leaning towards blades myself.
Heres the thing, all of this has been using those big blades while also running it very light which is different than I would normally fly it. I have to mount the extra battey trays on the frame sides so i can load it up a bit more, plus another 6000mah has its bonuses for auto missions I guess.
I know from testing long ago the blades I have on this frame are good for carrying weight. I had tried quite a few other blades, including semi symmetrical and settled on these wide chord 800’s for using the gimbal and getting the best flight time and smooth movements when carrying it. They especially shine when running lower headspeeds.

As i noted awhile ago, this helicopter had been purpose built for low headspeed work really. Super soft dampeners, wide chord 800mm blades, tail gearing is sped up to compensate for lower main rotor speed and longer and wider 116mm tails vs stock Align 105’s.

I have latched the gimbal on now and tested with that a few times and it flew well, mostly hovering with some mild FF. Looking back it was much more sluggish with the added weight, but not bad really. I had a couple slow oscillations if i let it sit there and dig in hovering, but thats not really usefull as i did not have the camera on there so CG was off by quite a bit, I have to muster up the courage to throw a DSLR and SSD deck on there and get a better test under my belt.
I kind of pushed all of that to the side though as I have been doing testing and data collection mostly with the frame to date. The filter work with Bill was a boon for using the controller to eliminate instabilities vs making mechanical adjustments. Basically, every flight and every tweak represents more knowledge and more data and through the use of the tuning knob I have a much better idea of how P, I and D terms affect flight behavior as well as a better understanding of the implications of adjusting the Gyro and Accel filters.
I have other frames so its not like im down for the count or anything.

(Chris Olson) #850

While I do not know the methodolgy used for your testing, and measuring power consumption vs lift, symmetical airfoils have the worst lift/drag ratio of all types, other than swinging a flat plank. The blades you have are designed for 3D. You are wasting ~30% of your power in drag just to have the luxury of high manuverability in inverted flight. In the real world, virtually all heavy lift helicopter rotor blades are narrow chord, high aspect ratio cambered airfoils because they maintain laminar flow over the blade, creating lift. Symmetricals do not. This is the type of airfoils used on helicopters outside of the RC 3D world - and even the scale RC guys use these types of cambered airfoils made by SpinBlades, Aerodyne, Sungwoo, etc…

Putting on real blades is an instant 25-30% increase in flight time and available power. Properly designed ones, like the SpinBlade A-VII’s, typical German engineering, have a few degrees of washout from root to tip and have been tested and retested to make sure they have proper flapback characteristics. And another advantage is that the asymmetric blades are relatively quiet-running and their CG is out further on the blade station, which gives them inherent stability and less vibration problems.

(Timothy Bacon) #851

Unfortunately the truth is that there was no perceptible difference in flight time running asyms vs syms both being 800mm in length. I totally get the aerodynamic principles involved that “should” net a gain in efficiency but alas, it just wasent so.
The test was simple, take the heli with gimbal and support gear, bring it into a hover out of ground effect and fly until pack cells reach 3.78v under load. I also compared mah consumed vs time and in the end there was no difference.
What I did see was a large improvement in handling qualities when carrying weight using the 800mm wide chord blades vs everything else.
If there was going to be some big gain in flight time by running them I could have over looked the stability issues but it just was not there. There is another guy in the southeastern US running a very similar frame to mine, although using a front mounted camera, who landed on similar components and setup.
There could be a multitude of reasons why things went as they did for me, headspeed, AUW vs rotor efficiency and AOA, not adding a significant amount of effeciency from forward flight because its mostly hovering work and slow passes, i dunno?
I still have the asym blades and might dig them out again when it finally comes time to do some faster mock mapping runs as I totally agree they “should” be more efficient, but maybe i just did not see it becasue of the reasons cited above?

(Timothy Bacon) #852

Chris, while I am thinking of it, the blades are Rotortechs. I had also thought about trying Spin blades and the SAB asyms.
I also had problems with ballooning in a hover in any sort of wind with the asyms and it was not nearly as bad as the blades I am running. I might try them again once I get into mapping style missions, as I said, i understand the reasons why they should provide more efficiency, but the way I was using the helicopter the bad outweighed the good running the asyms.

(Chris Olson) #853

I have no clue. Even the special Halo blades on my GX9 are narrow chord, tapered and twisted cambered airfoils. Not impressed with the rest of the helicopter, but it is stable and very efficient, considering the 32.8lb takeoff weight full of gas.

This is the first time I’ve tried SpinBlades on a helicopter. I had Sungwoo’s on the 700. It took 3 weeks to get the SpinBlades. I was getting right on 20 minutes at 30mph ground speed with stock blades. Adding a 5A battery only increased it to 25 minutes. Which is not too impressive considering a single 5A will fly it for 12 minutes @ 1,700 rpm. Putting the SpinBlades on and reducing headspeed to 1,550 got me to 34 minutes @ 30mph. I got more flight time from the blades than I got hanging more battery on it. It’s Vne speed is 55 mph airspeed due to blade stall @ 1,550 rpm. But I never fly it that fast anyway. At 30mph groundspeed it will handle 20mph wind, flying upwind, without issues. And that’s all I needed from it to make it a serious player for flying surveys with an electric platform.

(Chris Olson) #854

I have never had that problem with either the Sungwoo’s or the SpinBlades. But my heli’s are also considerably higher disc loading than yours. I’ll bet the takeoff weight of my 600 is higher than your 800 - 13.3 lbs. I flew the SpinBlades ~ 6 hours this past week. One difference is that all my testing is done in flight. I have zero use for hovering a helicopter other than to take off and land with it. I typically take off, switch into Loiter for maybe 10-15 seconds to check everything over, switch to Auto and she’s gone. And doesn’t come back until it’s done flying its mission.

Oh, OK. One thing that bothers me with RC helicopters is that little engineering work is ever done. I always look at the full-sized helicopters to see what works because actual engineering and testing is done. All the engineering is done in RC heli’s by the marketing dept. Maybe we can change the size, shape, chord, logo, whatever, and since it’s different it must be better and everybody will have to buy it.

I do not know why you did not get the results that most scale people get with decent blades. Maybe either the blades you got are marketing junk, or assuming they are actually good, maybe you didn’t use the proper setup for them. They work on full-size helicopters, and they work on RC too. In RC I tend to spend more time picking the brains of the scale guys than 3D guys. The scale guys know how to actually engineer and build helicopters that fly decent.

(Timothy Bacon) #855

Chris, with 2 packs it is close to the AUW of the frame I have been tinkering with. Its about 14.5lbs with 2 packs and 18.4lbs with 4 packs. Been adding sensors to the helicopter for my FrSky telemetry link. I have a telemetry link for mission planner/tower of course but on this heli I deceided to outfit it with RPM, dual temp sensors, RX pack voltage, main pack voltage, a gps sensor for lost frame retrieval as well as individual cell voltage across all 12 cells for determining when to land while doing video work. I looked into the Mah consumed sensor for the Pixhawk, the Mauch 200amp sensor, and once i have a definitive mapping rig I may go that route, but when I am flying with the gimbal and its support equipment I feel much safer knowing the health of the entire pack. There is always one cell that dicharges quicker than the others and if you just use mah consumed there is risk in over discharging that cell. Just way too much money in the air to be risking everything on one sensor.
If I get into longer range mapping missions where loss of the radio control link is a possibility, then RFD900+ radios will likely be in order, but even then I will give FrSky props in that their radio link is solid at some pretty extreme ranges, and if I put a 5db on the radio it gets pushed even further.
But I digress, back to blades. The RotorTechs are manufactured by FunKey I believe, they hsd a presence in the industrial UAV world and I talked to a few people using them hence my decision to purchase them. Maybe they will be the star of the show once I set up for mapping? I dont know?
As to my results in testing, they mirror what others had had under similar conditions. I think it likely came down to the simple fact I was hovering a heavy bird vs really flying it? At the time I had done a bunch of research and had indeed talked to some of the scale guys and was thinking there must be something the other guys were doing wrong to not get more effeciency out of them? So I pulled the trigger on the over $200 set of blades and in the end put the 800x72mm symmetrical blades back on and went to work.
The conditions in which those blades work well for me are quite specific, heavy AUW, lower headspeed and not much forward flight.
Thinking about it a bit deeper, I changed one of the metrics that was working for me, high AUW, so I should have expected it to be touchy like it is flying it 8lbs or more less than I have been. :confused: Light bulb face palm moment.

(Chris Olson) #856

Well, in hover there is little you can do. A helicopter is least efficient in hover. A multi-rotor is actually better in hover than helicopters are. It’s the dynamics of forward flight where helicopters shine. In hover you have a specific amount of weight, and it takes a specific amount of horsepower to move and accelerate air equal in weight to the aircraft to make it stay airborne. The blade planform used on most multi-rotor blades is far more efficient for hover than typical helicopter blades. So even though you have tip losses and more losses in motors and ESC’s with a multi, they will usually still either match or exceed a helicopter in hover at the same weight and power.

You have to test a helicopter in its most efficient flight envelope - forward flight.

Watching this whole thing it is evident you cannot run normal gains for a UAV because your machine gets too wild and unruly. My opinion is that you would be better served putting 700 blades back on it, set up your gearing so you’re running at least 75% throttle where your motor and ESC are most efficient, and get that disc loading up to where most of use run our UAV heli’s. Even my little 600 weighs 11.9 empty, 13.3 with typical payload, on just 15 ft^2 of rotor disc area. And it flies superbly stable in the wind.

I dunno, man. But I think what you got is the highly popular 700 -> 800 3D stretch mod, and not a UAV setup.

(Timothy Bacon) #857

The one thing to take away from all of this is that the frame has already been proven and used for 2 years now with a simple hobby grade FBL. All of the issues I have been experiencing are unique to the Pixhawk 2.1 only as it applies to this frame. I had no tail bob, or crazy fast rate issues etc etc when using a simple AR7210BX FBL. It just plain flew and i used it for many videography projects and site inspections. The goal of a Pixhawk was to replace the only other option out there, the DJI.
In already went through testing blades, motors, gearing combos cell count changes etc etc.
I already know the frame will hover or do cable shots or what have you just fine with little to no balooning in the wind and when flying with my gimbal i already know what flight time to expect based on its prior use.
The only issue I have had all along is getting the Pixhawk to replicate what I already have done before.

My entire goal of this project is to be able to produce a repeatable frame with good parts support that uses Pixhawk and is auto mission capable.
Some parts of it are proving challenging, but really employing filtering has immensely helped with things like the pitch bob and it appears the attitude controller handles things a bit differently than other FBL controllers so its just a case of me adapting how I do things to come out the other side.
3D FBLs are likely designed with fast mechnical rates in mind, furthermore it has been determined that Rob was right and they are indeed pre-scaling the PID’s through a step in the setup process so they are able to handle a wide range of mechanical rates while also keeping the gains similar across many difderent frames. This week i am going to lower the mechanical rates and re-setup the Pixhawk 2.1. My guess is it behaves alot more UAV like and if indeed that is the case it wont be hard to repeat the process on my other 800’s.
I just really want to clarify things here in that the helicopter was already flying and working 100% fine. I got 15-18min of flight time carrying a canon 5d mkiii and an atomos ninja 2 ssd recorder and support gear on the 3 axis gimbal. Flew up to 50ish mph in follow shots or hovered great with awesome wind disturbance rejection, much much better than the tarot t1000 octacopter i use occasionally.
Any and all issues are because of integrating a Pixhawk into the mix, that being said, its proving that through some work it can all be tuned out. Its just different, and now i have auto waypoint capability and gps position hold. Those features alone are worth the work in my opinion.

(Chris Olson) #858

Oh, OK. You probably fixed all that since the last param file and video I had seen. The last param file I had downloaded, which was some time ago now, you had a really low throttle setting and I noticed considerable “bogging” in your headspeed when you made the videos we watched, when you fed it collective. And that was with no payload. Which is not normal at min takeoff weight with a governor. When I saw that I went, “hmmmm…WAAAAY too much rotor for that drivetrain…” :wink:

You must’ve fixed the gearing in the mean time, as what I saw there would cook a 800MX motor (and probably the ESC) in short order carrying any amount of load with the heli - unless you really cranked up the throttle to get that motor up to speed.

Curious what gears you finally went with.

I got a problem with the 600 since I put 17.09 gears in it and reduced the headspeed to 1,550. My tail is a little weak. I’m maching a mount on the mill to put a direct drive tail on it to fix that.

(Timothy Bacon) #859

Headspeed is around 1500 currently, although I am shooting for 1400. As to the gov holding, havent spent much time tuning it. Likely not going to as I have the Kontronik 160 I have to put on it and it has a much better gov as well as active freewheeling.
Gears, off the top of my head I cant tell you. Head gearing is around 9.xx. The auto and tail gear are the fastest helical gears available for Align 700/800’s. Ill look tomorrow and verify the ratio. The tail is rock solid all the way down to 1100, below that, like when I threw 6s at it running 900-1000rpm hard collective inputs resulted in a small tail kick, but that is getting down there headspeed wise.
The tail gears and going to 116mm rail blades made the tail solid. Thats one part of this that has been great from the get-go. Still running almost the exact same parameters you gave me some time ago to start with.

(Chris Olson) #860

There’s a param you have to adjust to counter that - H_COLYAW. It’s a feedforward for the rudder to compensate for collective yaw torque and the range is -10 to +10 (depending on direction of rotor rotation) in increments of 0.1. Clockwise turning rotor should use positive value.

If you are using the 9.33:1 gears you are geared for 2,100 rpm with a 800MX motor on 12S :wink:

Edit for addition:
A fellow in our club has a beautiful Bell 222 on a Trex 700 frame with Aerodyne’s on it. His heli weighs about 15-16 lbs. He uses two 10A 4s in series and flies it for 15 minutes no problem at about 1,100-1,200 rpm. On 8S you’d be geared for ~1,500-1,600 rpm at 90% throttle with the motor and gears you got. And you wouldn’t have to buy new batteries for it. Just take them apart and turn two 6S packs into three 4S packs.