Pixhawk Feedback?

[dazzab]

Now, DJI may effectively mask this issue from you, by artificially making the copter feel like it's not underpowered. They do this by using auto-throttle for most modes. And also limiting the pitch/roll angles and rates. Our system is capable of flying much more aggressively, and so it reserves more throttle for stabilization. This isn't a bad thing, it's just a choice.

EXACTLY the point I was attempting to contribute to the project when I was shouted down. Choices have been made that may make the Pixhawk less suitable for this specific use (heavy high throttle aerial video rigs) than the choices that apparently DJI has made. As you point out, there is nothing wrong with that. I think DJI may know 'something' that the Ardcopter developers either haven't figured out or don't care to figure out. We will never know given that DJI is closed source. Let's not forget that the Wookong M is specific to multirotors and the Pixhawk is a general purpose flight controller. A general purpose FC that in my experience is very heavily weighted towards fixed wing. But anyone who flies a heavy copter with a Pixhawk and then a DJI will probably notice that the DJI sounds and feels smoother. Why is this? Why do DJI arrange all the top props to go in one direction and all the bottom props to go in one direction but Arducopter alternates them? This is the info I was trying to contribute. But God help the person that brings up DJI in an Arducopter forum. You will be invited to go fly one and have fun. Which is _exactly_ what I did. Turned out to be good advice.

 

[dazzab]

I must admit the Naza is probably the best FC DJI has produced. It's relatively simple in user functionality. That may mean it fails less because there's fewer functions to fail.

Isn't the Naza the FC associated with all the fly aways?

 

[Old Man]

Dazzab,

I realize I'm the FNG in this forum and you don't know me from Adam but I feel a need to say something here. You brought up some pretty good complaints in your last but you may not realize you hammered down some strong positives as well. First, I work and have worked for a long time for the UAV arm of a major aerospace outfit. We have more logged mission flight hours than any of the competition. let's hope that suggests I at least somewhat understand autonomous flight systems.

Without going into specifics I'll venture our lowest end autopilot runs close to $70,000.00 and once in awhile it still has a hiccup. The closest aerospace grade autopilot I know of that would work for a multirotor runs upwards of $15,000.00 - $20,000.00 and troubleshooting with the maker involved will cost you in the area of $250.00/hour. So price wise the Pixhawk comes in pretty good and topping that it has more safety features than the ones I work with every day. Based upon actual flight experience with the Pixhawk and our AP I believe the Pixhawk is spatially more accurate in all but one flight mode than ours. Where software is concerned the Pixhawk is far less complex.

You mentioned you made calls to 3dR and was mistreated. I know how you feel because I had some issues along similar lines with them not that long ago. But they made some personnel and evidently some policy changes at the same time. Now they respond with alacrity and courtesy and that's the best I can hope for with anyone. Call the "industry leader" making "professional grade" FC's (DJI) sometime and see what kind of a response you get. Also take the time to read their warranty policy all the way through.

Next you mentioned you discovered an issue that got their attention quickly, which 3dR near immediately passed on to their other users in order to prevent crashes. Try that with the industry leader, maker of professional grade FC's DJI and see how fast they act on the information. You say Pixhawk is not ready for prime time but my experience with the most expensive "professional grade" multirotor FC made by a hobby vendor indicated their $1,000.00 lump of electronics is a P.O.S. and seriously ripping off consumers

The point to the above being that 3d Robotics did something after you had issues while the other maker(s) typically and habitually avoids any involvement or responsibility when their stuff takes a dump. Their policy is to never, ever admit to knowledge of any defect issue with their products and to let the users sort out final development via consumer beta testing. That's a standard part of the Chinese hobby industry business model btw, and something I've repeatedly experienced with numerous vendors from that area for the past 20 years. Deny, deny, deny. At least 3dR lets people know a software revision might be beta and permits them the choice to upload or not. DJI just dumps upgrades out there hoping for the best, and recently issued another upgrade after one released less than a week previous to cover up some mistakes. That was for the exalted A2 btw, the so called creme de' la creme...

I'm truly saddened to learn you had issues and those issues cost you time and money but from where I'm sitting it looks like 3d Robotics is about the only maker of FC's out there that responds in any way to reported customer issues. At least you can call them on the phone and obtain responses to e-mails. IMO, not one of the FC's that are currently on the market meets the level of redundancy and reliability that is necessary for commercial ops in civil airspace, but from what I've seen so far Pixhawk is the closet to that level that is also affordable. Could it be better and easier to use? You bet but until something that is actually better comes along I have to go with it.

Thanks for listening.

 

[Old Man]

Isn't the Naza the FC associated with all the fly aways?

No, Wookong and A2 are also guilty. i believe most all of them have occurred while in a flight mode slaved to GPS.

 

[Old Man]

Regarding auto throttle compensation, I was trying out the "Drift" mode in Pixhawk this weekend and I discovered there is some level of throttle compensation present when using the flight mode. The discovery was made when changing forward thrust angles without having to adjust throttle position. In other modes I've used when you change the thrust coefficient from hover to forward or other lateral motion there is some loss of lift apparent that require a minor throttle adjustment. No so when increasing forward AoA in Drift mode. The altitude remained constant until I increased or decreased throttle, at which time an appropriate altitude response became evident.

 

[dazzab]

No, Wookong and A2 are also guilty. i believe most all of them have occurred while in a flight mode slaved to GPS.

Egads! What's a boy to do? Now I am very scared.

 

[dazzab]

Dazzab,
I'm truly saddened to learn you had issues and those issues cost you time and money but from where I'm sitting it looks like 3d Robotics is about the only maker of FC's out there that responds in any way to reported customer issues. At least you can call them on the phone and obtain responses to e-mails. IMO, not one of the FC's that are currently on the market meets the level of redundancy and reliability that is necessary for commercial ops in civil airspace, but from what I've seen so far Pixhawk is the closet to that level that is also affordable. Could it be better and easier to use? You bet but until something that is actually better comes along I have to go with it.
Thanks for listening.

I hear you and agree with you. But you have confused one important point. It's not 3DR that I had issues with. They only make the hardware and their only involvement with the software is to support the Ardupilot open source project by paying a few developers, some who are on their staff, some who are on retainer. All others are volunteers. I have no issues whatsoever with 3DR or the hardware which actually was developed elsewhere and just manufactured by 3DR. 3DR never treated me poorly at all. I love their products and I'm a huge fan boy of the company.

Don't confuse the open source hardware, the open source software and the commercial company that is gluing it all together in a very unique and exciting manner.

 

[jfro]

Regarding auto throttle compensation, I was trying out the "Drift" mode in Pixhawk this weekend and I discovered there is some level of throttle compensation present when using the flight mode. The discovery was made when changing forward thrust angles without having to adjust throttle position. In other modes I've used when you change the thrust coefficient from hover to forward or other lateral motion there is some loss of lift apparent that require a minor throttle adjustment. No so when increasing forward AoA in Drift mode. The altitude remained constant until I increased or decreased throttle, at which time an appropriate altitude response became evident.

Curious if you are doing your testing with 3.15 or the 3.2 beta. I put the 3.2 beta rc4 on my 550 quad and it seems to be performing pretty good.

Somewhere I think you referred to testing some Pixhwawks on larger MR's. Have you any findings you can share?

If I'm halfway succesfull with my pixhawk testing and mapping tests, I"m going to be looking to upgrade to a longer flying quad an/or larger MR. If I can get real confident, I may even try it on my 18-20 lb x8 for video work, but that's a lot of money for me, so that will take awhile. I'm a semi retired hobbyist with experience in the video and computer field who is spending too much time and money on this addictive hobby.

Sounds like you are going down the same road for both flight time and maybe some heavier units, so I look forward to more of your reports. In my forum readings, I haven't found much in the way of experiences on the larger MR's & Pixhawk. Not sure if it means it doesn't work or just hasn't been tired that much.

 

[Old Man]

I don't think has been tried that much yet. It sounds like Dazzab did and had a bad day with one but the software is progressing in development and capability. yes, I'm on a similar path and looking to use the Pixhawk on heavy lift and large sized M/R's. I've given up any hope of using stuff from the industry leader, it's just too unpredictable. At the moment I'm following a systematic path to establish what can be done with the Pixhawk and explore the limitations. I'm hoping to develop some form of relationship with those that have been must influential in the software development. There are more than a few large sized configurations i have to work up for specific payloads so there's some work ahead yet. I know the Pixhawk will easily handle a 700mm quad or octo but the next larger version is still on the bench in development.

The next version will be 1000mm to 1100mm, likely octo. I can tell you that Pixhawk defaults are pretty much set for a 500-530mm quad or X-8. As you start going larger the gains a too high. A 700mm quad with about an 8lb-9lb T/O weight is a bit twitchy in pitch and roll while yaw is good. t can be flown well enough to switch to auto tune and have no problems dialing itself in. As things go larger we'll start with the gains generated for the 700mm unit. Note that 3dR's quad and X-8 are only a few millimeters different in diameter, with the X-8 having a slightly different radial positioning. It makes sense the defaults are set to handle a unit that size best. Both fly extremely well "out of the box" so to speak. Note I didn't mention the Iris.

Lift capacity and duration ARE everything but stability and reliability have to be established before pushing too hard on duration. Duration is actually easy if stability and reliability are taken care of. Just throw a lot more money at it. There's a rabbit still hiding in the hat for after the point the primary requirements are met.

The largest negative to the Pixhawk I can see at the moment is the lower voltage capabilities compared to some of the other FC's. OTH, this may end up a positive if any of the issues other FC's have experienced have been due to excessive voltage or heat. The lower voltage requirements simply means some functions have to be handled differently from how other FC's do it. Perhaps if it was easy everyone would be doing it. Then again, those with the "easy" do have a lot of people doing it, and a lot of them are experiencing many unexplainable issues.

Oh, don't try what Foxtech is trying to push off on people right now with their 1300mm U8 unit. I can tell you from hands on with another custom unit that size with the same equipment their combination will never be stable when any wind is blowing.

 

[Old Man]

I hear you and agree with you. But you have confused one important point. It's not 3DR that I had issues with. They only make the hardware and their only involvement with the software is to support the Ardupilot open source project by paying a few developers, some who are on their staff, some who are on retainer. All others are volunteers. I have no issues whatsoever with 3DR or the hardware which actually was developed elsewhere and just manufactured by 3DR. 3DR never treated me poorly at all. I love their products and I'm a huge fan boy of the company.

Don't confuse the open source hardware, the open source software and the commercial company that is gluing it all together in a very unique and exciting manner.

Gotcha, and understand. I'm not a true fan boy yet, but what I've used has worked as advertised. There's still a ways to go before I'm fully convinced. But it has been looking good so far!

 

[jfro]

"The largest negative to the Pixhawk I can see at the moment is the lower voltage capabilities compared to some of the other FC's. OTH, this may end up a positive if any of the issues other FC's have experienced have been due to excessive voltage or heat. The lower voltage requirements simply means some functions have to be handled differently from how other FC's do it"

I shorted out their power module and while waiting for a replacement, I took a 5v signal and put it on the rails as they suggested for a backup power source for the Pixhawk. I did a 8-12 flights that way with no issues flying wise. I'd probably go that route for a while & if I need 6s batteries at some point. My next test's would most likely be with some 3520 or 3515 motors I already have. I"m afraid of all the new stuff with more poles as they seem to be dropping out of the sky way to frequently. I'm using simple old ESC's with the 3520 KDE's and 3515 Avroto's and they haven't failed me yet, although I'm mostly a slow flying for video.

 

[R_Lefebvre]

Yes, that's true. Enough to cause a lot of damage to my gimbal though. And why did it take three events before this was noted? I flew that copter with a dummy load several times without issues. We also determined that auto tuning it with the payload may have contributed to it as well. And let's not forget how poorly I was treated over the incident. No user should be treated like I was. So, it's not just one incident that I based my decisions on.

I'm not going to dismiss, nor apologize for how you were treated. I don't agree with it, but I didn't participate either. However, you do share some of the blame. As I recall, you refused to accept some of the issues being told to you, and compared it to DJI which you claimed flew it perfectly. Some people don't like that sort of thing, understandably. The fact that DJI flies your machine fine, is no indication that system is better. They just hide more problems from their users. Hover throttle over 80% is not good.

I maintain that based on my experiences, Pixhawk/Aruducopter is not at this point in time the best choice for commercial use with large copters.

That's your opinion. Very many others would disagree. It is not any kind of hard fact.

It's always pushing the envelope which is very cool, but not what you need when you have $10,000 of gear in the air.

What flight controllers aren't pushing the envelope? Do you know for a fact that DJI, or any others, conduct more testing that we do? The recent A2 debacle would seem to indicate they do very little.

 

[R_Lefebvre]

A general purpose FC that in my experience is very heavily weighted towards fixed wing.

That is completely untrue. What would make you even say that? If anything, it is the cross-pollination between the different branches which make the system so good. Some of our most important advancements came out of the plane branch. What other flight controller, can fly a copter at 140 km/h, ripping and banking, do a barrel roll, hammer head stall, and a couple back-flips, and them change seamlessly into a precise GPS position hold and, then autoland? You gotta wonder what we know that they don't.

But anyone who flies a heavy copter with a Pixhawk and then a DJI will probably notice that the DJI sounds and feels smoother. Why is this?

Because they dampen everything.

Why do DJI arrange all the top props to go in one direction and all the bottom props to go in one direction but Arducopter alternates them?

Simple preference. There's not much reason for one or the other. Could be just to make setup more simple. With Arducopter, you can setup a Y6 either way. And if you really wanted to do that on an X8, it could easily be done.

 

[R_Lefebvre]

Regarding auto throttle compensation, I was trying out the "Drift" mode in Pixhawk this weekend and I discovered there is some level of throttle compensation present when using the flight mode. The discovery was made when changing forward thrust angles without having to adjust throttle position. In other modes I've used when you change the thrust coefficient from hover to forward or other lateral motion there is some loss of lift apparent that require a minor throttle adjustment. No so when increasing forward AoA in Drift mode. The altitude remained constant until I increased or decreased throttle, at which time an appropriate altitude response became evident.

I don't fly drift much, so I'm not an expert. But I didn't think the throttle mode was different from Stabilize, which also has simple geometric angle/thrust compensation. I know Jason was talking about putting a basic, not-quite-alt-hold throttle compensation into drift mode, but I didn't think that happened yet. Could be wrong.

I don't use drift mode, because I tend to fly pretty fast with my helis, and there's an issue with the algorithm that starts to show up above 15 m/s. Things can get a bit jiggy.

 

[R_Lefebvre]

The largest negative to the Pixhawk I can see at the moment is the lower voltage capabilities compared to some of the other FC's. OTH, this may end up a positive if any of the issues other FC's have experienced have been due to excessive voltage or heat. The lower voltage requirements simply means some functions have to be handled differently from how other FC's do it. Perhaps if it was easy everyone would be doing it. Then again, those with the "easy" do have a lot of people doing it, and a lot of them are experiencing many unexplainable issues.

Do you mean what the maximum motor voltage is? There really isn't a limit. I'm using a Pixhawk with 8S on a helicopter. There's nothing really stopping me from using 12S if I wanted. Well, OK, that 8S heli uses a separate 2S for the avionics, but that in itself requires a special regulator because you can't use the PM on 2S. Not reliably anyway. I have another heli that is using 6S, and that is the supply for the Pixhawk.

It's just a matter of using the right voltage regulator to the Pixhawk. We should talk. I'll send you that contact PM that I've been neglecting.

Oh, don't try what Foxtech is trying to push off on people right now with their 1300mm U8 unit. I can tell you from hands on with another custom unit that size with the same equipment their combination will never be stable when any wind is blowing.

Do tell.

I'm very interested in this topic. We all know that theoretically, at some point stabilization via fixed pitch props won't work anymore as size grows. Meanwhile helicopters just get better as size grows. So I'm interested to know where things stand. I did notice last night, flying my mid-size quad with 15" props, it was less stable than others I've used with 10". Though I have yet to run autotune on it so we'll see.

 

[Old Man]

I don't use drift mode, because I tend to fly pretty fast with my helis, and there's an issue with the algorithm that starts to show up above 15 m/s. Things can get a bit jiggy.

I did notice that a couple times after reviewing post flight aerial video. There were a few times the copter would enter a bank smoothly, then lay over a lot more to the point it exceeded the gimbals ability to self level. I was pushing some things pretty hard at those points though. I charged it off to not having exponential in the roll side of the control functions but your explanation also fits well.

There's a lot of flight modes available and the plan is exploring them to determine which would find usefulness in a given project environment. The auto take off, mission way point functions worked very well, as did the point flying. Both could be extremely useful for some projects. Did I mention the accuracy of the uBlox GPS? It's awesome. Drift mode not so much but it's fun to use in a chase type situation. Had the ability to manually input yaw been retained in Drift the functionality would be increased considerably but that would likely require a lot of additional software code to avoid confusing the FC.

 

[R_Lefebvre]

The issue with Drift is that, yaw is the master, and roll follows, to try and prevent "skidding". The problem is that the maximum yaw rate is such that above 18 m/s, 45° of roll is not enough to turn the corner. Basically, you need more than 1G of lateral acceleration to rail the turn, but you can't get more than 1G at 45° roll. So, it starts to skid sideways. If you aren't steady on the yaw rate, it ends up jigging the roll back and forth.

I had started work on a similar mode, but one where the roll is the master, and the yaw follows. I had called it "Yaw LookAhead". It just keeps the yaw always looking forward into the flight path. Pitch and Roll were fully manual as Stabilize mode, not GPS at all. I had considered using Alt-Hold auto-throttle. And you would be able to over-ride the yaw, force it to "skid" up to 45°. This would make it the ideal FPV flight mode, I think. I've just never had time to finish it.

If you think that uBlox is good, wait until you try the M8. Apparently it's like an order of magnitude better.

And then RTK GPS which will be coming... hopefully within the next year. Should be around $1000, and require a more complicated ground station, but offers centimeter level precision.

 

[dazzab]

However, you do share some of the blame. As I recall, you refused to accept some of the issues being told to you, and compared it to DJI which you claimed flew it perfectly. Some people don't like that sort of thing, understandably. The fact that DJI flies your machine fine, is no indication that system is better. They just hide more problems from their users. Hover throttle over 80% is not good.

What crap. There is no 'blame' to assign here from my point of view. I was just discussing the issue. There's no refusal about issues, just assumptions by cranky people that for some reason are sensitive about this, I just questioned a few things that didn't seem right. The whole thread is there for people to make up their own mind if interested. I'm not.

Believe me, I'm not happy that the Wookong works better for my use, but I have plenty of proof that it does. I'd do anything to get another FC to meet my needs and if it was open source so much the better. I think that I've made that point over and over. I'm not a user just bashing for bashing sake like some do. I was very methodical and put my money where my mouth was. I made mistakes, I learned and I contributed.

You know, the great thing about photography is that it doesn't matter what equipment you use or what claims you make. The proof is in the results. So, instead of claiming that the Pixhawk is suitable for the purpose of professional level commercial video how about showing me some videos that match the quality of the videos in the 'show me' threads? Give me the names of some professional photographers that use the Pixhawk and have results to show for it. That's the only thing that will ever convince me. Go ahead Robert, check out those videos done by the pros. They are stunning and a joy to watch. Then look at how many of them were done with copters using a Pixhawk FC.

Put up or shut up.

 

[dazzab]

That's your opinion. Very many others would disagree. It is not any kind of hard fact.

I never said it was. I've always been very careful to state that was my experience and others may have different experiences. So I'm not sure what your point is here. This forum is about sharing experiences and working together to get results. You are totally free as is anyone else to totally ignore my experiences and form your own opinion. I wouldn't have it any other way.

As for a lot of people disagreeing with me. Well, I said it before and I'll say it again. Put up or shut up. I've shown you evidence of my benchmark so now show me yours. Where are all these great commercial photographers using Pixhawk that disagree with me?

 

[R_Lefebvre]

When I talk about professional applications, I don't mean artistic aerial photography. Lots of people using the system for mapping, 3D imaging, SAR, etc.

I have seen some pretty good artistic videos, some of it even shot from an Iris. Surprisingly good actually. But the people doing it tend to be the "walk softly" type, so I don't have any samples on Youtube to show you.