Servers by jDrones

This should be interesting

(Bill Geyer) #781

Yes I know this can be difficult. The non Traditional Helicopter Parameters (ones with out H_) are likely the multicopter defaults which in many cases don’t work well with the helicopters. So in looking at the param file you posted on the other discussion thread, I see that your setting for PIDs in Pitch and Roll appear to be in line with Chris has shown to work. The IMAX parameter is not the one you want to change for what you are seeing. It is already at its maximum so it will allow the integrator to provide as much input as it needs to hold trim attitude. But it doesn’t sound like holding trim attitude is your issue but that you are oscillations due to disturbances. I’m not sure if this is a feedback thing or just the aircraft not able to resist disturbances. Since your P values are very low, I suspect it is not caused by them. It could be your D value in pitch. You might try setting that to zero and see if this issue is still occurring. If it goes away then it could have been a feedback instability caused by the D gain which only became an issue in FFF. You could also try raising your I gain a little more but I’m not sure that this will help.

So another option might be raising your P gains to determine if and at what value you see oscillations start in a hover. I think if you could raise these a little more, it would go a long way to helping reject wind disturbances. But once you see the oscillations, you will want to back off by at least 0.01 and make sure you can’t excite them by making sharp inputs. Then you can build up in forward flight speed and verify that you aren’t seeing any oscillations. If you raise the P gain, then you may want to lower the VFF by the same amount as you might feel the aircraft is too responsive to your inputs.

Hopefully this helps.

(OzzyBob) #782

Hi Bill, I initially found the opposite to what your saying worked for me!?

As I started going faster and faster in FF, I had to lower my Pitch P, and raise my Pitch VFF… my understanding was that in FFF the VFF was to compensate for the “pitchyness” of the P term, so I thought I should raise it, not lower it???

I’d love to strap the Pixhawk to my heli that has the Futaba CGY750 FBL gyro on it, record a flight, and log what comes out of the futaba Vs what the Pixhawk would’ve done, the Futaba is absolutely sorted at 150km/hr FFF… might strap the GoPro to the heli and record what the swash is doing during high speed FFF.



(Timothy Bacon) #783

Well… I will finally get some time to myself tomorrow, so I have the gimbal waiting to go along with the helicopter for a little testing. I am saving the current known good parameters then as Bill suggested dropping back the gains to lower levels and nixing the notch filter and raising the LPF to 20hz again. Ill go back through channel 6 tuning P gain although I will have a bit of VFF and P initially for control. I have my notes from previously as to which frequencies the rapid strong ossicilations cropped up, I am very interested to see what happens with all the extra mass and how and if it changes the frequency.
I have some weight to simulate the camera on the gimbal as well as other hardware like the SSD recorder. Only worry I have is that I approximated the cameras weight but its just a block of lead and as such the weight is distributed on the gimbal camera platform quite centralized and usually alot of weight in the form of the lenses hangs further out and dampens the movement of the gimbal. I am intending on tuning with the gimbal live as before with testing I saw a fair bit of coupling between camera movements and the helicopters movements and as that will always be the case it does not make sence to do otherwise. I have PID’s saved for a few different cameras I fly for the gimbal controller, but nothing for using a lead weight as I am going to now. The gimbal was a bi^$h to tune the PIDS on so I am praying this works with the lead…
So here goes nothing . guess… weather permitting.

(Bill Geyer) #784

please let me know if I’m restating this correctly. You found that as you increased the P gain, the aircraft became too responsive even if you reduced the VFF a fair bit. Did the aircraft want to oscillate when you made inputs with high P gains? or did you have a tough time targeting a pitch attitude to maintain your speed? Please explain what you mean by “pitchyness”.

So I like to think that VFF and P Gain do the same thing in the absence of distrubances. However when there are disturbances the P gain works to reject the disturbance where VFF will just apply the same input regardless of the disturbance. This is why I find it difficult to believe that there is a control problem. Is this a controllability issue or you observe the aircraft being twitchy. I fly airplanes with the pixhawk and when I’m flying in windy conditions you can see the aircraft is a little twitchy as it responds to combat the disturbances especially in the roll axis.

I would love to see logs that show a higher P gain and this “pitchyness” versus a log of low P gain with High VFF.

(Bill Geyer) #785

I understand what you are saying regarding the simulated weight and the actual camera where the CG of the camera is lower on the gimbal. Certainly this is a difference from a flying qualities perspective but I think the gains you come up with here will be a good starting point if not the final answer for when you mount the camera. Now the difference in cg of the simulated vs actual camera and its affect on the gimbal is a good question. So I may have missed this but what is the weight of the heli without the gimbal/camera set up and then with the gimbal/camera setup?

Were you going to try some auto missions in slow forward flight before mounting the camera? Also please post your last log with the settings you landed on that removed the pitch bobble.

(OzzyBob) #786

Hi Bill, No you’re statement is not what happened, VFF had no impact on my responsiveness,

I tuned it by raising P till I got the oscillations, and backed it off until it was stable, then set the D to 0.001 and it shook really bad, so I took it out and raised the I to about 0.35 it was then that I reintroduced the D, and found I could leave “I” 0.001 on pitch but not roll,

All my FFF tests were done in Auto, So I couldn’t say I had trouble “targeting” a Pitch attitude, I just found that as I flew my missions, if it got disturbed, my hover P was too high, I would reduce P and increase VFF simultaneously.

By “Pitchy” I mean Pitch Oscillations escalating to where I had to take over. I would let it build and build to see if it would recover, but you could see the amplitude increasing,

Maybe I start again, and instead of simultaneously reducing P and increasing VFF I just leave VFF alone?



(Timothy Bacon) #787

Firstly, yeah I am going to plot out a slow shorter auto mission before I start with the gimbal. If it looks remotely good I will save the settings as good/workable.
As to the weights, the heli is 13lbs with 2 packs, 16.5lbs with 4 packs.
The gimbal varies depending on camera choice. Smaller mirrorless camera like the GH4 or a Nikon ASPC DSLR are roughly 1.5lbs, the Canon 5D more like 3lbs. Another 1lb for the SSD recorder and 2lbs for gimbal itself, mount and other support gear. So AUW with gimbal and camera plus extra packs is rougly 21-22lbs with most cameras, closer to 24lbs with a 5D or something like a Blackmagic Production camera.
And yeah, I will get you that log file, dident forget just haven’t had a chance to get to my desk…

(Bill Geyer) #788

So the gains I looked at were your starting point or something you had landed on after doing some tuning in a hover? I would say for now, just zero the D term in pitch and see if that gets rid of the pitch instability in FFF. What was the P gain for the pitch axis that caused the oscillations in a hover? I’m just curious. I wanted to see the delta from your setting of 0.02 in the param file and what caused oscillations.

So zeroing the D gain is my going in position. See what that does and go from there.

(Bill Geyer) #789

Ok. great. Thanks for the weights. look forward to hearing your results

(OzzyBob) #790

OK, Thanks Bill…

I had Pitch P on CH6 Knob, I think I got up around 0.04 (didn’t bother to save a param where it was shaking) it sometimes hovered OK for a while and then you could see an oscillation start, I think I found 0.03 was safely below oscillations, but then I progressively reduced it, as my missions got faster… to 0.02

I’ll remove the D from Pitch and take it from there… cheers

Is there any way we can get the D term to another decimal place? or adjust the scaling so we can actually get some D term in? if 0.001 is the lowest you can get, and thats too much, the scaling is obviously out?


(Bill Geyer) #791

So you weren’t able to get much P gain at all So I wouldn’t worry about raising the P gain. As for the D gain, I have entered values to the 4th decimal place but you will have to do it from the Full Parameter List screen. Write the params and then read them back to be sure it took the value.

(Sunit Pal) #792

Finally I was able to make a maiden flight using my small heli after facing a bunch of issues.

  1. PH2 would not load 3.5.rc7. Looks like the firmware is bad as it is able to load other firmware but not this one. Even if I download the firmware and use custom firmware option to load it, still has the same issue. I tried it on a different PC but same issue. Even though I have flybar I decided to use Bills custom firmware because I was able to upload that. It will be of use later when I swap the FBL head.

  2. I calibrated the compass but still I keep getting the Prearm message “Compass not calibrated”. I can see the compass param values have changed. I calibrated it 3 times but same message. So in the end I set ARMING check param to 250. This ignores the compass check.

I was able to fly fine. In fact I am very happy with the flight characteristic. It is just great. I am powering the servo with 6 volt. I plan to switch it to 5V and see how the params change. Experiment 2 is to attach the FBL head and see how the flight characteristic and prams change. This thread already has lot of great info, hope it will add something to it. I am sure I can tune it a lot more as there seems to some oscillations in the log. I will continue to improve upon this.

2017-06-06 12-23-47.bin (2.8 MB)

(Timothy Bacon) #793

2017-06-01 17-05-24.bin (3.0 MB)

(Timothy Bacon) #794

That parameter file was one of the last flights I did where the pitch bob had subsided substantially. It was not the tightest feeling flight but the goal at the time was to get rid of the pitch bob through the parameters if it was possible. The headspeed was still low where I have had it set, as I stated earlier the heli flew with an FBL with even lower headspeed I am running now and it had no pitch bob. I honestly am not running that low of a headspeed, 1350-1450 is the range I am shooting for possibly 1500 with the goimbal, but once I am trying to achieve endurance 1350rpm is what I am hoping to run as it is a sweet spot for stability vs flight time at least historically using a standard FBL.
In any case, it is far from the best looking flight comparing desired vs actual but that was not my goal, getting rid of the one-two bob after any aggressive pitch input was. I will upload the parameters used for that flight as well.

(Timothy Bacon) #795

reduced pitch bob parameters timbaconheli trex 800 pixhawk 2.1.param (13.8 KB)

(Bill Geyer) #796

So looking at your log and parameters, it seems like you landed on a little higher D term of 0.003 and you lowered your P gain values slightly, now down around 0.08. It appears that you have some small oscillations going on when I look at the RATE message. I wonder if it is due to the D gain?
It still looks like you had the I_LEAK_RATE at 0.02. Hopefully you were able to change that to 0.001 to close the error a little better in steady conditions. Any luck flying today.

(Timothy Bacon) #797

Yeah, I had increased the D term all the way to .007 at one point. I also saw the small ossicilation in the log, although it was not evident in flight to me. Im sure it likely is the D term causing it as it resembles a couple logs back the larger ossicilation I encountered with D term at 007 only smaller. I am going to change the I leak to .001, lower D term to .001 and try a quick auto flight hopefully tomorrow morning. Sadly it rained yet again today. :confused: This weather pattern im telling ya has been something else. I feel like im in the Pacific Northwest. Its June and its been so cold and rainy some of the leaves have turned red/yellow.
In any case, I will try the quick auto flight and hopefully I am good to save the settings, clear the parameters and start with the gimbal. I did manage to get the gimbal ready for the test flight though so it wasen’t a total loss.

So I have been thinking and reading and looking at logs… I just can’t help but feel like I am not running enough P and D term. The D term does in fact ossicilate quite strongly if I increase it much further, although when flying the heli it does feel a bit like its missing the shock absorber(D term). Im assuming the instabilities are still preventing me from getting meaningful values there. As I have said before, I have always used D term to eliminate bounce back and just plain make it feel like its flying better. Just not with a Pixhawk running trad heli, its with multirotors, whether a large octa or s small overpowered race quad. I should say that especially the little 250mm race quads with the ridiculous Scorpion motors and 20amp escs running 4s need it. Seems the faster response, the more D I run. Just not here. :confused:
It feels better with higher P and lower I though, much more balanced. Just wish I could get it up to .15ish on P without any ossicilations at all. It felt the best, reminded me of my other helis when I had P at .15+. Having higher P and lower VFF just plain flew better sans the here and there ossicilations. Nothing builds and gets continuously worse with P under .17, over that it gets real bad real quick.
I tried to get the LPF over 4hz, no luck there although I hadent played with the notch when I tried 5hz, 8hz and 10hz so there is likely some improvement.
I wonder, how much of a chore is it to adjust the depth/width of the notch filter?

(Bill Geyer) #798

I have a different view of what the D term is doing. From the aircraft standpoint, it is acceleration feedback. So it is anticipating the changes in rates. I don’t think of that as so much of a shock absorber (damper) from the aircraft dynamics point of view. The damping of that pitch bounce should be coming from the P term. If you remember from my instability post that it wasn’t until I increased the P gain substantially that I was able to damp out the low frequency lightly damped oscillation. So just like Chris said that he increased the rotor RPM for his flybarred heli’s when he sees this, effectively he is doing the same thing because increasing the rotor speed increases the power of the flybar to damp and the rotor also helps with that because it acts like a damper as well.
If you look at my set up, I had my I gain below 0.1 but I decreased the leak rate of the integrator as well which help a lot in the hover. My theory is that more I gain causes lag in the system. I don’t know if you tried this but my vote is to reduce or eliminate the D gain, reduce the I gain with the I_LEAK_RATE set to 0.001 and increase the P gain.
I think this might be better with a notch just on the gyro signal. I coded one up and have it ready to go but I have not had a chance to flight test it.
I would really like to tackle this analytically using frequency domain analysis techniques. However I don’t have everything in place to do the analysis. Ultimately this analysis could determine optimal PID settings but the Arducopter code is pretty complicated and I’m new to these techniques. On the flip side we could use flight data to determine your stability margins. This would give us an idea of how close you are to instabilities for a given set of gains and we could see how the stability margins change with change in speed. Its really frustrating having to hunt around for the right PID settings. My going in plan for my helicopter was raise P gains until you saw instabilities and back off enough just so you don’t excite them with sharp inputs. Then raise the I gain only enough that you see good attitude hold. My settings worked well on my heli up to 18m/s (36 kts) airspeed. Obviously that is well shy of the fast forward flight speeds that others are flying.
Well hopefully weather will cooperate tomorrow and you can get some flying in.

(Bill Geyer) #799

In re-reading your post, I guess I was saying something that you already knew. Sorry.:grimacing: How low did you get with the I gain when you had the P gain in the 0.13-0.15 range? I don’t know if it will help but getting it into the 0.1 range might help, especially with the lower leak rate.

(Timothy Bacon) #800

Firstly, I am going to lower the D term back to .001 as I noticed that ossicilation in the logs. As stated above, it does indeed make the heli feel a bit more normal, but without filtering to remove the feedback ossicilation I dont think it can be used effectively. It did seem to tighten things up with a value of .001, but then again it could also be placebo? :confused:
From what I understand through conversations with someone far more knowledgeable than myself is this. Tuning PID on a helicopter is exponentially more difficult than on a multirotor because pitch/roll/yaw changes happen far more rapidly and lead to feedback ossicilations much sooner than a multirotor. He said the speed at which the motors can increase or decrease rpm to acheive the desired pitch, roll or yaw by default acts as sort of a filter that allows a system to run much higher D term as well as higher P and I. He eluded the sucess of some FBL’s and their handling characteristics comes down to filtering out these feedback problems but as usual no one is willing to give up what exactly they are using for an algorithm/filter… Does that sound correct? Im someone who always looks at something objectively/skeptically as after many years on forums you run into alot of people who say things pretending to be someone they are not.
But I digress… To answer your question about I gain, I started at .13 with P being at .15 then worked the I up to. 2 seeing how it affected things. When I had done that I still had all that Expo in the radio from the initial flights so im sure it mucked things up. The helicopter has always been “very responsive” from the get go, so thats why I added in the Expo at first but I added in way too much I guess as I was in that territory where Expo starts doing some weird things to your controls. I realized it was messing things up when I started doing some nose down funnel like manuvers, when I hit the 4 corners of the radios gimbal it would have a weird sharp transition. Lowering the expo eliminated that issue and immediatly made it feel more natural. I am going to have to re-visit the scenario with the leak rate lowered, I term at .13 and P at .15 now that I lowered the expo. If I can rid myself of the feedback ossicilation with P at .15 in hovering and FFF I feel I would have a great flying helicopter. I intend on using the parameters Chris mentioned to reduce the “snappiness” I am getting vs trying to lower the P term and VFF to acheive the results I am looking for with UAV flight. It looks better in the logs and feels better all around with higher P so somehow I have to get there.
I have also been looking over the mechanics of the helicopter, I have some servo horns that I believe I can sucessfully adapt to use instead of the stock Align horns to get the ball studs closer to center and thereby reducing the mechanical gain inherent in the design. Although, if I can get where I need to be without doing that I would be happier.
I have the heli ready for a flight or two today and the weather seems to be cooporating so hopefully I can turn this speculation into data collection.