Fast Forward: Evolution
The Fast Forward is an A Tail racing quadcopter with radically tilted and rotated rear motors for maximum forward thrust.
I’ve been obsessed with V Tails and A Tails from the moment I saw the Hunter VTail, and the first quad I tried to design was an A Tail. I had it laser cut by Ponoko out of delrin because I didn’t know about CNC and G10 yet, and it sucked ass but looked kinda cool. Someone said it looked like a person so I named it Eddy Propellerhands.
At the time I could barely fly, and had no clue on tuning. And while it had moments of graceful swooping, it mostly threw itself roughly to the ground until the delrin split. I made a couple more versions of this until I gave up on the A Tail. But there was something about this design that stuck with me. With the angle of those rear motors, some thrust was being pushed back, so that it naturally wants to fly forward when level. At the time I considered it a flaw, and tried to reduce it by making the rear arms more parallel.
My obsession with the A Tail was so great I even made one to carry a GoPro gimbal, but it was a miserable failure.
Then everyone started flying mini’s and racing faster and faster, and I began to think that A Tail thrust was a feature and not a bug. After all, pointing two motors at each other is just wasting thrust, right? Why not turn them back and get some speed? Let's call it a 'Thrust A Tail'.
I drew up a prototype with 30 degree rear arms and motors tilted at 35 degrees. I was worried about ground clearance for the tilted motors so I made it two 3mm carbon plates, with the lower one incorporating the front arms and the upper one the rear arms. The rear motor mounts were interlocking 1.6mm G10. The pdb and fpv cam plate was scavenged off my QAV250.
I flew it at first with the peg landing gear from my QAV 250 (yes it shoots forward on takeoff), but I knew I was going to put wheels on it. Once I had it tuned and flying well I cut landing gear out of sheet metal and took to the field. As you can see the rear wheel pivots, but without a bearing it didn’t spin smoothly enough to work properly. Not that it mattered – the tail is so light while taxiing you can steer even with a locked rear wheel.
It was everything I hoped for: sweet sweet dumbassery. I loved rolling takeoffs, touch and go’s and soft landings with a multirotor. Here I am in a moment many of you must be familiar with, trying to fly while a lookey loo peppered me with questions.
With all the clearance from the landing gear, I moved the rear arms down to the lower plate. The upper plate was now 1.6mm G10 to save weight. I drew up landing gear as well, with M2 axles. Most importantly, with the concept proven, I rotated the rear arms another 30 degrees for even more thrust. I’m sure this layout would fly with the rear motors facing straight ahead, but that would result in lesser yaw response.
With all this extra thrust Eddy was now quite a handful. It was constantly shooting forward, and flying it mostly consisted of pulling back on the elevator and hanging on for dear life. Hovering landings were difficult with its nose up attitude, and rolling landings were high speed adventures that usually ended in a rollover and broken props/antennas. But it was a hoot to fly, turning more like a plane with its nose level, almost every stick input equating to more speed. As you can see I had some trouble handling its extra speed (the good stuff is at the very end).
Then I put Kiss ESC’s and Oneshot on it, and something very strange happened. With the same PID’s it was now capable of being as docile as a normal quad, while still able to tear it up when I wanted. I could land on a dime or a slow roll. This was my first experience with Kiss and Oneshot and boy what an eye opener it was. We flew at Ferry Point Park in the snow, setting up in a 10' by 15' patch of clear concrete. Some spectators showed up and left us even less room to land, but I was still able to slowly circle and drop it on a dime. Hard to believe this was the same craft I flew into a lightpost at speed.
After breaking a few unibody frames in the name of testing I decided that I had to redesign the frame with separate arms. I turned the rear motors forward another 10 degrees for yet more thrust, made the overall shape swoopier, and made 50 degree motor mounts for, you guessed it, MORE THRUST.
I built up this latest version the night before the Motherboard shoot for the Blast. I'd been told we'd be flying in an enormous construction space, so I brought it along to maiden. I was so confident flying it with 35 degree mounts I didn't think it an indoor maiden would be a problem.
Our space turned out to be a lot smaller than expected, but I wanted to try the steeper motor mounts so badly I gave it a go anyway. As soon as it lifted off it shook so violently I cut power and let it drop. What I didn't realize was that yaw authority increases with motor angle (more horizontal thrust, more yaw force), and now it was now so responsive it had violent yaw oscillations. It was eventually tamed with some outdoor tuning.
As I flew it more I found a strange and interesting flight characteristic: it accelerates forward with just throttle, no need to drop the nose, almost like a plane. Most of all it was just freakin' FAST, always taking off with a mind of its own. At times it felt like it needed more pitch I, as it had a tendency to drop its nose a few degrees and settle there. It wasn't an aggressive correction like auto leveling, more of a feeling that the craft had an angle it wanted to fly at. After some experimenting I figured out what was going on.
A Quick Diversion into FFhysics
Here's a simple schematic of an ideal Thrust A Tail configuration. Front and rear motors and props are identical, the rear motor is tilted at 60 degrees so that it has half the thrust of the front motor (cosine 60 = .5). CG is located 1/3 of the way back so that the rear motors have twice the leverage and half the lift. This configuration is very similar to a tricopter, with similar CG location, with the two rear motors taking the place of a tricopter's single rear motor.
Now let's tilt it forward 5 degrees. Now the rear motor has 42% lift (cosine 65), while the front has 99%. 42% x 2 = 84%, which means the rear has less lift and settles back down to level.
Tilt it back 5 degrees and you have 57% and 99%, 57% x 2 = 114%, so the rear lifts up to level again. In other words, when you have motors set at different angles, you can get them to self level much like dihedral on a plane's wings. (Thanks to Eric Murdoch of FliteTest for explaining this.)
Now, if I were to go into Cleanflight's CLI tab and customize my cmix, I could give the rear motors 100% throttle and the front 85%. Now when the craft tilts forward 5 degrees, the rear motors will still have 42% lift, but now the front will have 99% x 85% = 84%. Now the craft will self level at 5 degrees nose down (you can do the same by just sliding the battery forward – CG isn't set in stone with this design). As you can see, you can dial in a nose attitude by playing around with motors, props, cmix, and battery placement. You can make it fly neutral or you can make it go go go all the time.
This is what I lucked into. I had entered a cmix proportion that made the nose push down slightly with throttle, so that it shot forward without gaining altitude, almost like a plane.
Here it is Nascaring with the FPV Addiction boys at Coram. Keep in mind that the camera is mounted level to the frame, so it's zipping along with barely any tilt – the horizon's just above halfway in the frame most of the time. I'm accelerating out of the corners with just throttle input, no pitch adjustment at all. Left to my own devices I tend to fly easy and smooth, so it really takes a race situation to make me open things up.
The rear motor mounts were becoming a problem. With its slip-on design tolerances had to be perfect to avoid slop, and even then the mounts would eventually shake loose and start to vibrate. I CA'd them to continue flight testing but clearly I needed a different approach.
I'd just made a replacement gimbal arm with a t-nut configuration, and that worked so well I redesigned my motor mounts the same way. The stanchions were beefed up from 1.6mm to 3mm, and the rear arm was made to resemble the elevator of a jet. I try hard not to let form depart from function, but it just looks so much cooler.
My tester and I had both broken arms, and I broke the main body plates near the rear landing gear, where support from the top plate and standoffs ended. I found a good vendor via Alibaba and had one last prototype cut, with thicker 4mm arms (up from 3), no cutouts, and an extended shim between the lower plates to strengthen the rear half of the frame. I also made a few tiny tweaks so that the interlocking pieces in the motor mounts and landing gear would be held under tension to keep things from rattling.
Here's where I broke that arm. I don't intentionally crash test these things, but somehow I ended up breaking quite a few of them anyway.
Landings can be an adventure, especially with the 50 degree mounts, so I lengthened the front landing gear so that you can land slightly nose up and slower. I also think it'll be cool to have it sit nose up on the ground, like a Mustang or Spitfire.
Meanwhile, Ryan Gury of Dronekraft suggested that Eddy Propellerhands was too obscure a name. He recommended 'Fast Forward Flight' or FFF, and I shortened it to Fast Forward. The DVR triple triangle symbol made a perfect logo, so I used it for my top and bottom plate cutouts. My long suffering darling wife made it official with a real logo.
Like the Booboo, this frame has a lot of interlocking pieces. The Booboo was lasercut, which is much more precise – you can pretty much expect the cuts to be right on the nose, with .1mm of material removed. However, Booboo is made of bamboo, which varies in thickness more than carbon, so that production had its challenges. The FF is made of carbon of very exacting thicknesses, but it's CNC'd, and every CNC cutter does things slightly differently. In the past I've always had to do some filing to make things mesh just right. I was pleasantly surprised to find that the pieces from my Chinese vendor's first cuts fit together perfectly. The fact that the pieces are screwed together with some tension also helps relax the tolerances a bit – if there's a little slop it'll go away once bolted down.
Here's some video of the lastest version running 35 degree mounts. The first is me trying to chase down Ryan Gury, ending up with us colliding through a wicket like Archie Bunker and Meathead going through a door. The second is our club's top pilot Alex Walsh giving it a try. He taps it on the ground when his video is briefly jammed, then you can see him getting used to it lap by lap until he tears it up. I'm embarrassed to say his lap times were 4 seconds faster than ours.
Here's some shots of the final production version, which came in at 194 grams. I've spent nine months refining this baby and I'm very proud of how it turned out. There's lots of new innovations in racing mini's coming out but I think I've got them all beat. The Fast Forward has the snap of a V Tail without its inefficiency, it flies forward level without the weight and complexity of motor tilt mechanisms, and cameras mount nice and flat. Most of all, it's just plain fun to fly. I hope you'll try it out – the store listing is here, and setup instructions are here.