I have an experimental aircraft that I have flown for many years. I know that the autopilots discussed here are for model aircraft but.... The idea is to have the auto pilot drive small trim tabs on alerons to gently level the wings in flight. It would be a system that woud be easy to overide even if the servos failed, due to the small size of the trim tabs. This is a project that anyone could comment on and all (good and bad) I'm not taking this as joke! My butt will be in the seat.
Lets discuss this
Mike Adams

Views: 261

Reply to This

Replies to This Discussion

I moved this to the hardware section
This is certainly within the purview of an Amateur-built Experimental aircraft. You would need to do the engineering to size the servos and tabs, and ensure that there isn't any mechanical failure mode (like say, a broken servo gear) that could permit a tab to flutter, driving the surface and wing into oscillation. Same for the software; you would need to work out a slimmed-down version for which you could confidently verify the functions.

A lot of time would go into failure scenarios -- you know what to expect when it works, but what happens when it quits? How about hard-over in multiple axes? What if a board shorts and catches fire? You may never get them all, but forethought in your design will minimize the chance of a nasty surprise in the air.

Once you understand the changes you want to make, you would then need to put together a test series, something like the original test series the aircraft went through when it was built, but focusing on how the new system behaves and interacts with the rest of the airplane. And of course, don't be doing test flights with any more than required crew in the aircraft.

If you enjoy the engineering process, this would be a fun project! Done with the same care as the design and construction of the original airplane, there's no reason it couldn't be done safely.
Hello Micheal, I think this could be done. I would use the best(reliable) servos I could get and they would have to be big enough, how big? I would use a full 6dof imu and gps. The gps just keeps the heading correct. If you have an aircraft gps or vor you could use the +/- 50 mV output into the Ardupilot and now you could navigate as well. It could be done, but your on your own with the FAA. Any help would be "for educational use only" .
Hi Michael,
If your aircraft does not yet have trim tabs, an easier method may be to just have two small actuators on the stick (ELEV & AIL), automatically trying to level the aircraft. This would greatly reduce your pilot workload. Also, if something screws up with the servos, you can just rip them right off as they're right in front of you.

SLOWLY turn the gain up ;)
Hi Guys
The trim I'm looking for is a wings leveling trim system. I have a manually controlled pitch, elevator/rudder trim on the stick to reduce stick pressure.
The wing leveler would be a plus for longer flights. If I have my pitch/yaw trimmed leveling the wings would be even less work, and give me more time to scan for traffic and monitor systems.All of the thoughts will be addressed. The fludder issue is very important. My existing trim for pitch and yaw are controled by servos that are used by a large experimentialuser base.
They are lenear servos that lock in position if a failure occurs pressure on the stick but, not a in flight failure emergency due to the size of the trim tab size Fire is also addressed by fused link like all my electrical equipment.
Test period will also be required when any change to the flight control systems that were signed off by my FAA inspector.
My hope is someone can give a lesson on the auto pilot capabilities related to a fullsized aircraft (or giant scale) Also the avability of code for this purpose.
Thanks for your expertice
Thanks for the input

You might want to start out with a passive system that simply collects data, estimates your roll and pitch angles (maybe yaw angle too?) and then displays that in real time as you fly (or maybe draws the result over the top of a video image for later evaluation.) You will probably want to spend some significant time watching and evaluating your attitude estimate so you can build confidence in it before handing control off to an autopilot basing it's actuator outputs from this attitude estimate.

There are a myriad of subtle things that will (or could) make this whole experiment less satisfying and less perfect than you might originally hope.

Your gyro and accelerometer sensors are going to be less accurate and more noisy than you like. If you buy a cheap IMU, these sensors might be connected to a 10 bit ADC ... the long and the short of that is you might only be able to resolve your sensor readings down to around 3 deg/sec step size with a 10 bit ADC. You'll be reading the sensor at a fast rate (like 50 hz) so this 3 deg/sec step size isn't quite as bad as it first sounds, but at the same time, it's not great.

You probably have to move up into the $500-1000 range (like a vectornav or isensor cube) to get the kind of IMU sensor quality that you really would want.

But these are just raw sensors. You need to feed this raw sensor data though a complex filter to fuse all the data elements together and estimate your actual roll, pitch, and yaw angles. Here again the result is not going to be as good as you hoped for. There are simpler filters that honestly have a lot of error in their estimates. There are more sophisticated filters that do better, but you are limited by many things such as sensor calibration errors, sensor resolution, update rates, etc. A good filter needs some external source of data to correct errors introduced when the apparent gravity vector changes due to motion.

This could be a magnetometer, but those are very installation sensitive and noisy. It could be an airspeed sensor, but that could be hard to rig up accurately on a full scale planes without cutting holes where you probably don't want them. There is also gps ... that is probably the most general solution, but gps's have their own limitations. Some filters might tie yaw/heading angle estimate to the magnetic heading. Some systems might tie your heading estimate to the ground track of the gps. There are higher end filters that use a more generic math formulation to actually converge on the true heading of the aircraft. The downside to this is that you need a few minutes of flight (and manuevering) to converge on heading and when you sit still or fly or drive exactly straight, the heading will start to diverge. Even with a $1000 IMU, there is enough noise and inaccuracy that as you integrate rates to get position updates, you will diverge, and diverge a lot more quickly than you'd like.

There are also possible installation issues ... imagine a sensor that mounts on a board. That board mounts in a case. The case mounts somewhere in your airframe. How confident are you that the sensor itself is mounted exactly level with reference to your airframe? How do you know? How can you tell? If you are off by a degree or two you might not be able to visually see it, these things are hard to directly measure. But a degree or two error could lead to a constant turn when the system thinks it has the wings level. You could trim that out with rudder, but now you are flying slightly cross controlled.

So all that said, you can get an IMU + filter that does a pretty good job at converging on your true heading, true roll angle, and true pitch angle. There are more expensive commercial systems that do a pretty good job at this. I don't want to say anything negative about any DIY IMU system ... the people here are doing great work, especially factoring in the price, their systems are improving every day, and I haven't taken a close look (or any look in some cases) at all of them. Consider that a few years ago you'd have to spend thousands of $$$ to do any of this stuff, and now you can do things at least on the low end for a hundred (or couple hundred) $$$.

So I don't want to say anything negative here, my goal here is to educate on some of the finer points of IMU's and the issues you might face. These are the things you are likely to encounter if you push forward with a project, and it's the details and fine tuning that can take 95% of the effort (and cost.) Often it's the attention to these little details that can separate the systems ... you might be able to line up a matrix of features and even the cheap IMU's can check off all or most of the boxes these days. But there can still be a huge range of final results.

One more thing to consider is "truth". By that I mean that in most cases, most of us that are doing IMU development really don't know what the "true" roll, pitch, and yaw angles are for our crafts as we fly them. But we still build IMU's, develop software filters, tune and tweak, and in the end we come up with a result that we are happy with. We may talk about convergence, the quality of the solution, the quality of the final flight. But really, how close are we to the truth? Most of the evaluation is subjective or based on crude estimates. Does it look like it's flying level? But we don't have a real truth reference to get hard numbers from. Are we always within 2% of the truth? Within 5%? 10%? Are there times when we are way off for a few seconds until the filter pulls itself back together again? Are there types of maneuvers that kill our accuracy? One older filter I worked with could get 30-45 degree off in certain types of maneuvers, even though most of the time it was pretty ok and certainly it was good enough to safely get an airplane around the sky.

So anyway ... before you start connecting up actuators, you might want to spend some serious time evaluating IMU's and filter algorithms to make sure the end result is in the ball park of what you consider acceptable. And if you do start out at the very low end of the cost range, you'll want to evaluate the attitude estimate results very closely under a variety of conditions (accelerating down the runway, moderate or steep turns, continuous turns in the same direction, etc.)


Curt thanks for the reply..
Sounds like I would be working more on development than I could master. I would like to fly rather than tweaking a potentially distracting piece of hardware/software electronics. The information you supplied was very thoughtful and well described. Maybe the thing to do is listen to the members and watch from the air until
someone else can co-develop a viable option. I see all the fun that DIY Drone fans have and my goals are a little out there for a few more years.
I was at the Reno Air races this year and got to see up close a Predator what an aircraft! It uses the same Rotax engine as mine (914 turbo100hp). I didn’t know that!
Thank you Curt
If any one else has some practical experience with aircraft wing leveling applications yell at me…
Thanks all
What type of plane is it?
Its a S-19 using a "untested, as in not designed by the company" version of a 914,Rans uses a 912uls (80hp) like most light sport aircraft


© 2014   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service