Can You Believe I Am Working On The Kestrel Again?

| Comments

I don’t want to call my Kestrel a failure. I didn’t think it was a good idea to attempt to commercialize it, and I’m quite confident that improvements in hardware and software have minimized or completely eliminated the need for the Kestrel’s vibration-damping suspension. You’re going to get clean HD video recordings on any modern 3” quad. You don’t need fancy magic!

That said, though, I am confident that the suspension helps. I’ve managed to cinewhoop with my 4” Kestrel using a GoPro Session 5. You normally need to soft mount those to get good results, but I just strapped it right to the top and it worked quite well!

I have a local test pilot!

A few months ago, my neighbor stopped by the house. He had a bunch of components, but he didn’t have a usable frame. The motors Richard had were a bit under-powered for something as heavy as the Kestrel, but we figured it would work well enough, so I went through my box of prototype parts to see if we could assemble an entire Kestrel out of spare parts.

A few weeks later, my 3” Kestrel’s Caddx Turtle died. I wound up sending Richard home with that entire quad too. He had a spare camera and receiver, so he was able to get it back in the air, albeit with no HD camera.

He’s been crashing Kestrels ever since.

Richard's Kestrel

Richard has only been flying FPV for a few months. Maybe more than a few by now, but he’s definitely still at the point where he’s more likely to end a flight with a crash than a landing. Maybe. Much more often than I ever managed to crash either of my Kestrel builds, so he’s finding all the weak spots!

If the Kestrel will never be a product, why are you working on it?!

It is still an open-source project. Maybe someone else will want to cut it. Maybe I’ll build another for myself one day. Fixing the weak spots in the model is quite easy and rewarding, so why not do it?

Beefing up the arms

Richard managed to break several arms in the last spot I ever expected a Kestrel arm to break! He has been snapping them between the mounting screw and the dog bone.

This is the spot where I broke my first arm, but that weak spot is nearly three times wider now. I’m impressed that it still breaks right there!

The Kestrel arm-mounting points are spaced 30.5 mm apart. This means that if you really wanted to, you could squeeze a full-size 4-in-1 ESC in there. At the time, this seemed like it might come in handy if I ever scaled this up to 5” arms. Today, this seems a bit silly, but it is still a reasonable setup.

Kestrel Arm Upgrade

I have to cut a chunk out of each arm to make room for the center stack’s screw heads. This is somehow the weakest part of the arm!

To remedy the situation, I made the dog bone 2 mm longer, and I made the ends of the bone 4 mm narrower. This combination puts more meat in the weak spot. I added a little more meat to the arms to make up for the dog bone getting longer.

I think we’re in pretty good shape now. We’ll be cutting these arms this week to get one of Richard’s Kestrel builds back in the air. I bet he breaks these somewhere else!

He has managed to break one of the arms near the base too.

Why not just make the arms a lot wider or thicker so they never break?

Everything here is a balancing act. When you make one part stronger, another part will wind up taking more of the force in an impact.

If you hit something really hard, would you rather snap a $4 arm or bend a $14 motor? Forget about the money. You only need a hex driver to replace the arm. You need a soldering iron to replace the motor!

Richard flies heavier batteries than I ever intended!

I had a particular weight range in mind while designing the Kestrel. My goals from the beginning were to build just about the lightest 3” freestyle drone that I could manage with individual arms, vibration damping for the split-style HD camera, and have no props in view of the HD footage. I hoped these goals would push things heavy enough to have a huckable freestyle quad without being ridiculously heavy.

My 3” build with 1306 motors came in at 225 grams, and my 4” build with 1606 motors is 278 grams. Both builds are measured with the same 650 mAh 4S battery, and that’s the battery I use most of the time. I expect the 4” to be less durable, but I did compensate somewhat by giving it wider arms.

My Four Inch Kestrel

Every time I decided how thick part of the frame needs to be, I would ponder whether it would likely survive a 60-mph crash if it weighed about 230 grams? I didn’t do any science here. I was just using my gut.

Richard has been flying some older 850 mAh 4S packs. I’m pretty sure he’s been cruising with some even heavier 3S packs as well.

Richard is putting more batteries through the Kestrel frame than I ever did. He’s flying heavier batteries. He’s crashing more than I ever did. He’s doing a good job finding all the weak spots!

The problem with heavier batteries

The Kestrel has a problem. It is still a problem with light batteries, but it is exacerbated when you fly with even more weight.

The pair of rails that ride along the top sides of the frame are long. These rails are rather stiff, but if you give the frame a good squeeze, you can definitely push the center of the rails down to your stack.

Cinewhoop footage from my 4” Kestrel

In flight, you aren’t going to generate enough forces to flex these rails. In a crash, you most definitely can. Richard tells me that if you’re moving at high speed and you connect the bottom of your quad to the ground, the momentum of the battery can push the side rails right down into your stack.

Bigger batteries exacerbate this problem.

It isn’t just the long rails along the top that flex

The side plates are cut from sheets of 3 mm carbon fiber. The long bottom plate is a wider piece of 2 mm carbon fiber. That bottom plate has flex too.

If you put your thumbs under the arms and your fingers on the battery pad, you would definitely be able to squeeze the rails down to touch the stack. You’d be pushing hard enough that you’d say, “Holy crap! I shouldn’t squeeze anything this hard!”

A little flex is fine. This isn’t a 700-gram 5” quad. It is only going to experience about one quarter as much energy in an impact. Any quad needs to be just stiff enough to absorb vibrations from the motors and withstand the force of the props while in flight. 5” props generate nearly 10 times as much thrust as 3” props, so I can get away with a more spindly frame.

I also did a test where I put my thumbs under the front and rear edges of the bottom plate while pressing down hard at the battery pad with my fingers. This is testing the rigidity of the canopy while not flexing the bottom plate. In this case, I have to apply significantly more force to get the top rails to touch the stack.

How much more? I don’t know. I’m not a scientist! It feels like a pretty scary amount to me.

A little flex is good!

My goal is to make certain that the Kestrel doesn’t deform in any meaningful way during flight. A 3” propeller is only going to generate 300 to 500 grams of thrust. If you could hang a one-pound weight off the end of an arm, and it doesn’t bend, you’d be in good shape for flight.

What about a crash? Everything that flexes is taking energy out of the impact. Imagine you’re flying at 60 mph and you manage to hit a concrete wall with the edge of your motor bell.

If everything on your quad is for all intents and purposes completely rigid, every bit of energy that your quad had will be absorbed by your motor. This is how you bend motor shafts and dent bells.

What if the arm flexes a bit? What if you’re using a TPU dog bone, and the arm is able to pivot a bit and stretch the dog bone? What if the rubber grommets holding the top of the quad on compress to absorb the inertia of the battery? What if the bottom and top plates also bend due to the weight of the battery?

Someone smarter than myself could do math to figure this out, but whatever energy goes into flexing the carbon, rubber, and TPU won’t go into bending your motor shaft.

It is possible to make arms that literally can’t break in a crash, but you probably don’t want that. They won’t flex enough in this sort of impact to save your motor. In fact, you probably want your arm to break before the motor bends. Breaking an arm absorbs significantly more energy than flexing it!

Too much flex is bad

My very first Kestrel prototype had the battery mounted quite a bit lower. All this flex I’ve been talking about managed to push the top rails into my flight controller’s USB connector, and it broke off.

I’ve since raised the battery by a few millimeters, and it hasn’t caused me a problem. Unfortunately for Richard, he’s managed to crash hard enough to break a USB port too!

The TPU dog bone isn’t great

Yesterday, I cut four arms for Richard and handed him a 3D-printed dog bone. The TPU I use is rather rigid. The dog bone feels sturdy in your hands.

On the quad, it is a different story! I’ve basically crafted a set of pliers with the TPU sitting in the jaws. If you grab on to the arms and put some leverage on it, you can squeeze the crap out of that poor dog bone.

Kestrel Arm Upgrade

He’s going to fly it and see what it does. The arms will only pivot in one direction, and even if you twist as hard as you can, you won’t be able to get the props to touch. Even in the worst crash, he will still be able to take back off.

It is a neat experiment, but I don’t think we’ll get much useful data out of it. I’m going to be cutting a new dog bone out of carbon fiber next time I’m in the garage!

Mitigating this risk

Richard already has a good fix. Instead of using tiny nylon nuts to hold his flight controller on, he is using a set of 3 mm or 4 mm standoffs. Nothing touches these standoffs while he’s flying, but if he crashes, there’s no way for the carbon below the battery to touch anything fragile.

After holding a Kestrel in my hands again and flexing it over and over again, I’ve decided to make a few more changes.

Kestrel Arm Upgrade

I raised the battery plate. The bulge below the plate used to make the side rails sit a little over a millimeter lower in the center than the rest of the rails. I also raised the top deck by another 1.5 mm. I believe this buys me nearly 3 mm of extra clearance between the top of the stack and the carbon lid.

I’m also not happy with the amount of material surrounding the standoff separating the side plates at the rear. I beefed that up. I probably beefed it up way more than necessary, but it won’t add much weight at all. I don’t want that to be a point of flex.

Almost every part of the side plates have been widened by around 15%. This should reduce its ability to flex by quite a lot, and it will make the nose sturdier in a crash!

Richard and I have both broken our noses

I put a lot of thought into the nose and camera mount while designing the Kestrel. How thick does the round part need to be? Will most of the force in a front-end collision travel through the round piece, or will it transfer through the camera?

I didn’t have real answers. I figured the supports holding the camera in place only needed to be strong enough to hold the camera, and that most of the force will travel through the round section up front.

I’ve only managed to break one Kestrel side plate. I was strapping a GoPro HERO6 Black to the top to get some cinewhoop-style footage. I squeezed down on the nose quite hard while trying to get the Velcro good and tight, and I snapped the tiny pieces of carbon.

After I did that, I figured I should beef up the front end. Sure, I couldn’t break them in a crash, but it is easy to manhandle these bits if you’re flying with a GoPro.

Then Richard broke those same bits in a crash. He’s been flying heavier batteries than I ever anticipated, but even so, that was another good reason to strengthen the front end.

What’s next?!

It is imperative that we cut Richard a fresh set of arms as soon as possible. I’m pretty sure none of his 3” builds are currently airworthy!

I’m hoping to do that this weekend.

The side plates will be a nice upgrade, and I’m excited about testing them. I don’t think we’ll be cutting any just yet. Richard has been talking about an upgrade to a Caddx Tarsier or Runcam Hybrid, and if he does go that route, I want to make sure his more delicate camera is protected.

If the suspension is no longer necessary, why are you still working on this?!

My 3” Kestrel shares quite a bit of code with my open-source 5” Falcon frame. I use the exact same functions to generate the arms and dog bone for each frame. The only differences are a single conditional statement and the measurements that are fed into those functions.

I expect to eventually design another frame with arms that are compatible with the Kestrel. Brian’s 3” Toothpick is running the BetaFPV toothpick board. This single board has both an F4 flight controller and a 4-in-1 ESC, and it doesn’t weigh much more than just the ESC board I’m using in my Kestrels.

Not only that, but these boards are capable of driving a light 5” build!

I’m not in a hurry to build another heavy 4” quad, but I can certainly see myself redesigning the Kestrel to make use of a board like this. It will probably look like a baby Falcon.


This has been a fun and unexpected detour! I expected to be working on OpenSCAD stuff for the CNC work for my network engineer toolkit. The next step on that project is going to be designing some sort of parametric box and lid combination. That should be fun, but I expect to have to iterate on it three or four times. Maybe more!

I’m hoping I’m still managing to keep the weight down. I remember the first prototype frames being roughly 12 grams lighter than the Ummgawd Acrobrat. I’m pretty sure I dropped a gram or two when I switched to the dog bone, but I wouldn’t be surprised if I added that weight back plus a little more while beefing things up.

This is kind of what I expected to happen, though. The plan was to make everything just a little weaker than I thought it needed to be. If it didn’t break, that’s fantastic. Anything that breaks needs to be strengthened, and that’s what I’m doing!

What do you think? Is there still a reason to fly a 3” micro with vibration isolation for the HD camera? Would the Kestrel be awesome carrying a naked GoPro way out in front? Should I keep iterating on this, or do you think I’m due for a redesign to fit upcoming components? Let me know in the comments, or stop by the Butter, What?! Discord server to chat with me about it!