# FBW

Discussion in 'Instruments / Avionics / Electrical System' started by Jay Kempf, Feb 5, 2013.

1. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
I was just reading Karlolina's excellent blog: http://designaplane.blogspot.com

And this question of large leaps of innovation in aviation starting with available fly by wire is a great thought. So I thought I would like to have a discussion based on what ifs.

What I know is that the rest of the aviation world from Drones to RC to Airbus and all the others are working on FBW solutions for everything. What I also know is that things like Arduino are making high speed progress along the lines of progression in the personal computing hardware and software sort of speeds. So all the building blocks are there.

It seems that digital proportional computer driven controls are not just possible right now but actually everything needed to do it is off the shelf and cheap right now. So instead of talking about why it can't be done why not talk about how it can be done. What would someone have to do to design and test and open source type of system in order for it to become commonplace?

Some examples in the automotive world that are already intervening between operator and vehicle are traction control and throttle and brakes by wire. Steering by wire is coming very soon to the US. So it is happening. Again, Airbus especially and many others are already there but those systems and the development are astronomically priced.

Take a look at a project like the Megasquirt fuel injection system. It is not harder technically to take that same development environment and scale it to a control system with redundancy.

So what would it take?

2. Feb 5, 2013

### cluttonfred

#### Well-Known Member

Joined:
Feb 13, 2010
Messages:
6,366
2,231
Location:
World traveler
I can see a definite need for full authority digital engine control (FADEC) for lightplane powerplants by why would you want fly-by-wire flight controls in a light aircraft? I think it's safe to say that cables and pushrods are generally reliable and not terribly heavy or expensive. I can see the need for FBW in intentionally unstable-military aircraft or to save weight and mainentance cost in airliners, but I don't really see the utility in a light aircraft.

3. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
It is not the case that mechanical controls are less complex or reliable than digital electronics. For the same reason that you cited FADEC, FBW gives all kinds of advantages. You put one servo per moving surface and you have one central processing center with redundancy. You can also have redundant servos. All that is significantly easier to build and install than a standard control system. One need only to look at modern RC digital transmitters, their ability to mix functions, the ability to store and modify control algorithms, etc... All real time, dirt cheap, available. So for instance, right now you could bolt an RC transmitter to the side of the cockpit and use it to control larger servos and fly and airplane right now. There are RC airplanes that are big enough to sit in so all that exists. An RC transmitter is $35. Servos like$10. I am not suggesting using the cheapest stuff to base a full scale system on but I am saying that is the progression of development. There is no need for a massive control system like in an Airbus. Just a simple input output system to start. Merging an autopilot or turbulence and/or flutter damping is free once you have the basic control system in place. That is just more sensors and software. Batteries to operate a system like this for hours are small. Redundancy cheap.

My particular interest in this is to get rid of automatically hooking up controls on removable wings (complexity) and mixing of controls like V tails and camber changing flaps with multiple moving surfaces. There are a ton of advantages there for digital mixing based on single surface servos. I like electrical retracts for small planes too. The integration of the glass panel stuff is also a goal like autopilot, GPS way points, yadda... Already exists on the drone side with Arduino level computer hardware software.

One interesting mixing mode is ailerons up and flaps down for landing drag control. This combined with camber changing flaps and with a mixed V tail is a mechanical control nightmare. All modern fighters have this sort of mixing of functions like the tail and ailerons mixed during different flight modes/speeds. Changing control volumes for high/low speed. Mixing mostly coordinated controls with neutral rudder for most of the flight can be done too. Anything can be mixed. And it all exists, and has been tested on models, drones, modern fighters, Airbus sorts of implementations. Just general aviation is missing.

4. Feb 5, 2013

### Hot Wings

#### Well-Known MemberHBA Supporter

Joined:
Nov 14, 2009
Messages:
6,385
2,308
Location:
Rocky Mountains

Someone that really thinks it can be done at the home-builders level, and the time to do it. I only fit in one category right now.
I'm one of the few people I know that has been left stranded by the side of the road in the middle of nowhere due to a faulty FBW throttle. The ONLY reason I was left stranded was because the manufacturer in question didn't have robust software in place.

IMHO software is where the real challenge is going to be. The hardware is already very good and the weight and cost of hardware redundancy is at, or close to, the point of making it feasible. Developing the software to a point where we don't get the aviation equivalent "blue screen of death" is not trivial. I've got a couple of notebooks full of scribbled flow charts for "what if" situations. Writing code to make a plane fly where you want it to is actually quite simple. Getting it to do the same when there is a glitch, especially if the pilot is the source of the glitch, looks to me to be another couple orders of magnitude more complex.

snip>>by why would you want fly-by-wire flight controls in a light aircraft? <<

Because there are options made available through software that would be too heavy or too complex to implement in hardware. A variable ratio mixer for a V-tail, ailerons, or elevons are very obvious ones. Also software gives you the ability to move the pilot out of the control loop (by whatever amount you desire) to keep them from doing stupid things. Software can also do things far quicker and more precisely than any human can. The fact that it often doesn't is the fault of either the engineer or the bean counter, not the technological state of the art.

Thanks Jay for starting a Fly By Wire only thread! It's been needed for a while. The roadies have theirs. We need one too :ban:

Now, do we have to decide if there should be a distinction between a system that only moves things due to pilot input via an electronic link, and one that inserts software between the pilot and mechanical bits as well?

5. Feb 5, 2013

### Dana

#### Super ModeratorStaff Member

Joined:
Apr 4, 2007
Messages:
8,625
3,011
Location:
CT, USA
It would be quite possible and even easy to build a simple FBW system today with currently available components... look at the autopilots they have right now for R/C models, or the controls for R/C quadcopters. The issue now is reliability. Sure, you can have redundancy, but a mechanical system is so utterly reliable that you don't need redundancy. In large aircraft there are significant weight savings, and for unstable military aircraft it's the only way to go, but is the performance gain from some form of esoteric control mixing really going to be significant in the real world? Perhaps... or perhaps not.

The biggest problem, though, may be adequate debug. My day job is automation, so I know how long it takes to debug computer controls even with simple two position actuators and on/off sensors, and we can baby step it through one motion at a time, expecting crashes... proportional controls with infinitely variable positions on an aircraft in flight (where you can't just hit the emergency stop and reboot) are a whole 'nother affair.

This is also a subject near and dear to me because my college roommate was killed test flying an aircraft when engineers made what were supposed to be minor adjustments to the FBW force feedback system software (the aircraft oscillated in roll several times before rolling inverted immediately after takeoff, crashed and burned). And that's with the engineering resources of a major aerospace company (Bombardier), not an amateur homebuilder.

-Dana

No trees were harmed in the transmission of this message. However, a rather large number of electrons were temporarily inconvenienced.

6. Feb 5, 2013

#### Moderator

Joined:
Jul 8, 2009
Messages:
10,732
2,542
Location:
Rotterdam, Netherlands
The very fact that it is done mechanically and the designers are willing to take the significant complexity and weight are maybe the most obvious advocate for FBW.

Flaps that vary their deflection are a massive tool with great value. Simply deflect full-span flaps with 20 degrees over the whole span during TO. Do the same during landing, then start adding drag by deflecting the inboard flaps more and more, while deflecting the ailerons upwards.
Gust allevation. Dump that Vra where it belongs, in the trash bin.
Spin prevention (if the tip stalls, automatically put ailerons full negative, in sailplanes for example this is far more effective as lowering the nose etc).
Perfect landings. Couple a sensor to the LG and raise flaps upon touch-down. Works like a charm in sailplanes nowadays.

I'm not so sure the software is the real problem. Essentially it's a fairly simple system, taking as input:
*Desired yaw, pitch, roll, drag (flaps)/current aoa, slip and yaw angle, IAS)

Seems relatively simple? Keeping the software relatively straightforward first seems like it can cut down complexity a lot...

As for redundancy, does having 3 or 4 separate systems that each take an alternate control surface (System one takes elevator control 1,5,9,13 etc) remove the complexities of redundancy? If one displays major deviations, simply shut that one down?

In sailplanes the size and weight of control systems (not to mention flutter) are limiting today. Much higher performance would be possible if it weren't for the controls. (think 80 ft span, 100 sqft of wing area)

Is there any hardware available off the shelve nowadays that is reliable, low-power and light enough? I've done a student project with piezo-electric strips a few years ago for a trim tab, what was then available was good enough to fly.

7. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
As far as force feedback and all that, no to start. And as far as development of flight software and redundancy I am literally talking about taking the software out of an RC transmitter and duplicating it. These things already have significant development. Have been hacked so are programmable and have been in development in the millions for decades. They are already single board USB access programming computers with all the mixing and combining functions figured out.

My thoughts were to put large servo tabs on a mechanical system and have the RC transmitter input next to the side stick. That way you could just have an autopilot drive through the trim tabs and leave the flight controls alone. You could auto trim the main controls that way too. with that I would have a choice of inputs. One would work well for high speed low control volume cross country and the other would work well for everything else. I thought I would have an override for using the small joystick for trim vs. flight mode. Simple. But I see no real reason not to make the leap of faith towards just building a highly redundant system I can buy really nice RC transmitters for like \$100 and steal all the parts and populate them onto a flight control stick. Or I could get a gaming input high end device and interface it with a current flight sim software kernel and output to actual servos. Either way there is a lot of work done already. In the end Arduino is the way to get it done for a full product. There are also linux single board computers that are dirt cheap now and all of the daughter board stuff is fully mature for moderate autopilot stuff. It is not ready for inertial navigation but for a relatively simple flight control system it is all there.

8. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
Yes, there are large scale digital servos that could drive the controls of a small aircraft right now. And they are inexpensive enough to put more than one per moving panel. I am talking about a small test bed vehicle. One of the advantages of building in this direction is that you could actually test a new design as a drone before subjecting a human to it. Yes, you would have some beaurocratic nightmares to hurdle but they are possible. The military has test ranges for this. There are deserts to test in like in the salt flats...

9. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
So another reason I have always designed around full flying tails is triming the tail to the amount of flaps in various modes. FBW allows those combos too as well as differential steering between highly separated dual rudders, ailerons. Flaps and ailerons can be multiple small panels and work as a team. Deflect the inners more or less than the outers etc... so that you get a roll profile not a widely spaced roll input at the tip. The combinations are only limited by the imagination.

10. Feb 5, 2013

### Hot Wings

#### Well-Known MemberHBA Supporter

Joined:
Nov 14, 2009
Messages:
6,385
2,308
Location:
Rocky Mountains
Software can be simple. But is that the best way for a man carrying aircraft, especially if the pilot is in the loop?

It's what happens when things start to fail that are going to make the software more complex. Pull the stick back and the plane doesn't nose up. Very simple software is just going to add in more up elevator command and not know why the plane isn't doing as requested. This is fine for RC planes and RPV's over open areas but I don't think I'd like to fly in one.

Add in a few subroutines and it's possible for the control software to figure out if the plane is not responding as desired because a sensor failed, a servo failed or maybe the plane is just in a strong down draft or even inverted. A couple more subroutines and the control software might be able to predict that the particular sensor involved was/is about to fail based on data scatter, too much delta since the last read, or an increasing rate of offset during cross checks.

This does add complexity to the software but it also adds the robustness we are going to need. *edit

Example:
I've had more than a few throttle cables break on me over the years driving VW Bugs but none of them left me stranded the way the FBW throttle failure did. I could always implement some kind of a manual "limp mode" for the Bug's broken cable, even if it was as simple as blocking the throttle open and using the ignition key to control my cruise speed. There was no such limp mode in the software that left me with an engine that would not start or run. A limp mode for a broken throttle wire should have been included in the software. Being stranded by the FBW throttle wasn't due to any inherent problem with the FBW concept, just a poorly executed implementation. Another few milliseconds of processing time and maybe a hundred lines of code doesn't seem like too much complexity in this case.

As with most things finding the "best" compromise is going to be the real challenge. Of course most of the above is moot if all you want to do is put an electric servo/sensor between the stick and the control surfaces that simply mimics what a pure mechanical system would do. Sensor and servo data logging to track condition and predict failure would still be a good idea.

Is there any hardware available off the shelve nowadays that is reliable, low-power and light enough? I've done a student project with piezo-electric strips a few years ago for a trim tab, what was then available was good enough to fly.

I remember reading a paper about using piezo to move very small Gurney flaps on trim tabs for control. Was this related to your project? It looked like a good idea but the lack of off the shelf hardware that could be adapted seemed to be a problem.

*edit
Military FBW may not need this kind of robustness but we don't have ejection seats!

11. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
The throttle by wire in my audi still has a throttle cable. All fuel injection systems have some sort of input. It ain't connected to anything though other than the throttle body butterfly and has a position sensor that does all the real work.

I am not talking about a flight system that intervenes to start. So no level of the system deciding what you mean. Just raw input and output with some mixing functions and add the complexity later. All the reprogrammable fuel injection software that I have seen has stock map vs. override with a stalk combination. You could do that as well. So lets say you have a selectable mode screen or switch where you just get simple input/output from the system for failure mode or to recover a system that wasn't responding. That is possible. So what I see is a development environment where you build the system to be electric with redundancy and it lets you fly the airplane with a joystick input. You can have varying levels of added complexity up to and including servo loop pilot overrides and inertial navigation. It seems there should be a let go of the stick mode for every airplane regardless of mechanical or electrical base system. In mechanical systems they call that a trim system and built in neutral stability. To start I wouldn't build an airplane without neutral stability. To start you could build the entire aircraft with regular mechanical controls and redundant electrical controls to test.

12. Feb 5, 2013

#### Moderator

Joined:
Jul 8, 2009
Messages:
10,732
2,542
Location:
Rotterdam, Netherlands
Just a quick note, more to say later.
I'd cut up the system in many actuators and sensors. 1 ft span flaperon sections (with a single actuator each) and both your elevator and rudder also split in at least 3 surfaces or so. That way your reduncancy doesn't involve any mechanical stuff (one flaperon section that goes Berserk is easily compensated with the rest).

A simple test-bed is replacing the flaps of a conventional design. As long as you make sure you have enough roll authority when a panel goes hard over, you can safely test-fly it without extreme precautions.

13. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
Yup, flaps and elevator trim mixing with flap deflection would be a good start. We already have electric flaps with a dumb up and down button on the panel. Servos and a digital controller isn't much of a leap beyond that.

14. Feb 5, 2013

### Hot Wings

#### Well-Known MemberHBA Supporter

Joined:
Nov 14, 2009
Messages:
6,385
2,308
Location:
Rocky Mountains

So we are thinking 2 completely different levels of development? Even if the system is just pilot move stick, control move in response, you are still probably going to want a bit of build in self diagnostics.

If you start adding mixing functions I think developing some feed back routines so the software can do at least a bit of self correcting will be easier than developing a bunch of maps. Kind of like letting the software run in closed loop with the O2 sensors means that we don't have to have maps for every possible combination that the engine sees. All we need are a few basic ones and then let the system tune itself.

The FBW throttle that died on me has the pot on the pedal, wires to the ECU, wires to the throttle servo. (I'm pretty sure the software drives the servo using pedal position only as one parameter in the final throttle setting). Wire broke on the pedal, engine quit. Inexcusable IMHO.

Yup, flaps and elevator trim mixing with flap deflection would be a good start. We already have electric flaps with a dumb up and down button on the panel. Servos and a digital controller isn't much of a leap beyond that.

More like a little stumble than a leap. :gig:

15. Feb 5, 2013

### Jay Kempf

Joined:
Apr 13, 2009
Messages:
3,673
937
Location:
Warren, VT USA
Who made your throttle by wire setup so I can avoid buying that brand... SHEESH! My Audi has the pot on the side of the throttlebody. Never had one fail, ever.

16. Feb 5, 2013

### Hot Wings

#### Well-Known MemberHBA Supporter

Joined:
Nov 14, 2009
Messages:
6,385
2,308
Location:
Rocky Mountains
Let's just say it's a very common 1 ton diesel truck used by commercial fleets here in the US. It's the same truck I used an active noise canceling head set in when on the highway. Huge P.O.S. compared to my old and very beat Dodge that has mechanical injection.

One method of getting a bit of mechanical redundancy that I haven't yet seen mentioned here on HBA is to put 2 or more servos in series. This way if one fails the other can at least bring that portion of system back to neutral. Depending on the mechanism it might be lighter and cheaper than splitting up the load among several discrete subunits each with their own servo and position sensor.

17. Feb 5, 2013

### Dana

#### Super ModeratorStaff Member

Joined:
Apr 4, 2007
Messages:
8,625
3,011
Location:
CT, USA
For R&D, maybe, but an airplane without any control feel is not a good thing. There's a good reason why FAR 23 requires minimum stick force per g, etc. R/Cers get away with it because the planes are absurdly strong compared to a full size aircraft, so you can't overstress it... and the lack of feedback is one of the reasons that R/C is actually more difficult to learn to fly than the real thing.

Don't get me wrong, I can see that in something like a high end sailplane or something else where you're going for ultimate performance, FBW could be attractive, but for 95% of GA flying it's a solution looking for a problem. Still, in automation, 25 years ago we were still building mechanical cam driven machines and now nearly everything is computer controlled, so who knows?

-Dana

Back Up My Hard Drive? I Can't Find The Reverse Switch!

18. Feb 6, 2013

#### Moderator

Joined:
Jul 8, 2009
Messages:
10,732
2,542
Location:
Rotterdam, Netherlands
Would a strong spring suffice? Imagine a stick (and rudder) that are connected to strong springs, such that displacement and force are correlated and then take the displacement for the appropriate force you'd expect at a given q. You could simply correlate stick displacement to the appropriate amount of aero force feedback?

19. Feb 6, 2013

### Dan Thomas

#### Well-Known Member

Joined:
Sep 18, 2008
Messages:
4,858
2,070
Pretty much all the cars now will have the sensor at the throttle pedal. No cables or other mechanical linkage. My 2006 Hyundai has it, and Toyota is trying to settle a mess of lawsuits arising out of rogue throttles accelerating and causing accidents. The Hyundai uses a Hall-Effect sensor, which has many benefits over the potentiometer. Pots cause all kinds of trouble as they wear and carbon dust gets under the runner, and they're subject to contamination by dust and liquids. To experience this, you only have to fly an older flight simulator and watch the stupid thing make uncommanded maneuvers when those old pots start to introduce spectacular resistance changes. Cars that had the throttle position sensor at the throttle body used to have failures too, like my old Ford F-150. A pain, due to potentiometer wear.

Many RC servos use potentiometers for postional feedback. I wouldn't want those in any airplane I ride in. Pot failure means that the servo will continue to move while the computer looks in vain for indication of movement. A crash is pretty much certain.

To me (and maybe it's just because I'm getting older and have been around long enough to see lots of stuff fail) the idea of FBW in small airplanes is a solution looking for a problem. By the time one adds up the input machinery (using Hall-Effect stuff, I hope, or inductive sensors), the wiring, the force-feedback stuff, and servos big enough to cope with controls surface loads at speeds up to or 10% beyond Vne plus some safety margin, it will all weigh considerably more than a few feet of cable and a few pulleys. It would in my Jodel, for sure. The control surfaces just aren't very far away or too big to handle manually. In large airplanes that cruise near Mach 1 and have things like swept wings it makes more sense, as it can permit cruising closer to the coffin-corner on the performance graph and still prevent mach tuck and maintain stability as CG is very close to the center of lift in such machines, to reduce drag. And weight in a large airplane is much less critical than in a small homebuilt.

The throttle in my Hyundai has a tiny lag that is annoying. I sure wouldn't want any lag in an FBW system.

Dan

20. Feb 6, 2013

Joined:
Apr 4, 2007
Messages:
8,625