Frenzy - Embracing the Chaos

[Revsmin (Ari Callister)]

Part 1 - Taking out the trash

I’m not a superstitious person, but I truly believe that Ultra-Violence was a cursed robot. It never completed a single event without some sort of random failure, usually a new one that had never happened before. I spent 2 whole years trying to persevere with the design but early last year I finally had enough and decided to dispose of the thing finally. To be sure the curse was dealt with I burned it:

Ultra-Violence burns to death for being awful

…and set about designing something new to replace it. In the meantime, Matt Smith convinced me to get back into antweights, and to be honest this probably single handedly saved me from taking an extended break from robots entirely. I had so much fun running an ant again it rekindled the love I had for this sport that Ultra-Violence had been slowly killing, but also helped me find the inspiration I was missing.

Antweights physics are weird; they’re tiny and have so little mass behind them that things like centre of mass and momentum etc don’t matter and they drive very precisely as a result, with no drifting or acceleration time. I wanted to see if I could translate this sort of feel up to the beetle class, using very powerful motors and a heap of magnets. Now I’d avoided brushless drive in all my bots for a very long time, I always had concerns about reliability and fine control, but with the introduction of AM32, and with how anemic the brushed drive on UV had been, it was finally time to make the jump.

I looked at all the options and decided on the Repeat Max, mainly because I wanted to run this bot on 6S for the extra power and the Max are rated to run that high, while the other options like MegaSparks only state 4S as their maximum voltage. Originally I planned to run a pair of Repeat AM32 ESCs, but while looking through the Palm Beach store for a group buy of ant batteries, I spotted the Repeat Dual ESC. The integrated BEC and onboard current limiting sold me, and I got one imported along with the batteries.

I liked Ultra-Violence’s weapon, when it worked, which wasn’t often, but was the part of that thing that had the least issues, so I kept the setup largely the same, although I changed the motor mounting since the aluminium angle bracket I had been using had issues with bending, now using aluminium standoffs into the side of the frame which was much more solid.

The weapon motor is a Propdrive 2836, which is only rated up to 4S, but I know has been run at 6S in some US beetles. I still wanted to be sure the motor didn’t die from too much current so I put out some feelers for an AM32 ESC with enough guts for a motor that size and current limiting. Ellis was testing out some ESCs that fit the bill, and I agreed to test a couple out.

Originally the design was heavily inspired by Manta from the US, with a much wider more drum like weapon, but it ended up so heavy I ended up redesigning it into more of an eggbeater. At 80mm across it’s still pretty wide for a beater, but the proportions of the bot ended up much more square, although a lot of the style of the Manta inspired design stayed the course:

The entire robot is 3D printed, much like Ultra-Violence before it, but rather than using TPU for everything I was looking at different materials for the chassis since the TPU was far too flexible and caused a ton of issues. I’d tried nylon on UV at its last couple of events, printing the chassis from a reel of Taulman 910, but even in the handful of fights it had the nylon chassis had already cracked. So I needed something more rigid than TPU, but more flexible than nylon. Gus turned me onto OBC, which has a shore hardness of 55D, still somewhat flexible but a lot more rigid than TPU, as well as having the added bonus of being a solid 35% lighter.

The catch here is that OBC is a polypropylene variant, which is infamous for being a total pain to print well due to its tendency to warp. Supposedly OBC is less prone to warping as it prints at a lower nozzle temp, and that is true, but it’s still a lot more warpy than even nylon and printing a part as large as Frenzy’s chassis still leads to warp around the corners:

I tried a ton of different things to help alleviate this issue but no dice, whatever, it’s not bad enough to render the chassis totally unusable so I just ended up accepting it was part of using this material. The chassis was printed with 6 walls and 15% infill and came out super light, about 90g all in.

All parts printed, wiring done (thanks to Dave) and all fixed together, Frenzy was ready for her first event:

[Revsmin (Ari Callister)]

Part 2: Trial and error

Not a bad result for a first event eh? A really solid performance and some big hits, but all I could think about was how it just wasn’t quite how I wanted it to be. The weapon had inconsistent engagement, it gyro’d like crazy, it stripped the teeth off the drive pulleys:

…and oh yeah, the weapon fell out:

What I think happened was the big hits against Propane had cracked the bulkhead around the weapon shaft, the fight with Chucky had worsened that crack and then when Gizmo flipped me over in the semi final and the weapon hit the floor, the bulkhead finally split. So obviously, some work to be done on the chassis strength. For the following event the bulkheads were redesigned to be a bit bulkier around the weapon shaft, and I added some stabilisers to the side armour to try and reduce the gyro:

While I can’t talk much about the event, the stabilisers worked, kinda, and helped a lot with the gyro. Unfortunately the bulkheads did not survive once again, and I had to rethink again. I’d also adjusted the pulley centre distances for the drive to try and tension the belts more, hoping it would stop them from skipping the pulley teeth and stripping them away under load. This also didn’t work, and I lost yet another couple of perfectly good tyres to stripped pulleys. The cost of 6S I guess!

BBB Summer Showdown now, and the only real change was to try and beef up the bulkheads yet again. This time, the frame bulks up to 15mm thick around the shaft, and more importantly I sliced the mesh at the front to bulk up the print around that area. While changing the wall count between sections when you slice a mesh in Bambu slicer causes the walls to no longer be continuous, they wrap around the seam and the parts just snap apart with no effort, what I did find you can do is set the infill to 100% with the concentric pattern, which essentially turns the infill into more walls while keeping the continuous walls between the parts:

Where the material was solid the bulkheads completely survived, but where they weren’t, they were crumpled by Sniper:

The TPU wedge didn’t help, he just cut clean through it. Back to the drawing board on that, I was really hoping to not have to buy titanium, get it fabricated etc.

Anyway so it turns out the Repeat Peter Bar Kit wedge fits the dimensions of Frenzy almost perfectly so I don’t have to do any of those things:

…and I now have a proper, proven in combat wedge setup! For the next event, Minibeasts in the East, I swapped the steel woodscrews I was using for the lid and base for stainless plastite screws, which are also holding the wedge to the printed TPU mounts. They have a much stronger hold than the wood screws and are more durable to boot, can’t complain (though there was an unseen negative effect to the switch that will come up later).

For this event I made more of the chassis solid infill, from where the back of the weapon reaches into the frame forward is now completely solid. That’s a fair bit of extra weight in the frame, and it had to come from somewhere, so here is where things get expensive! I swapped the drive shafts from steel shoulder bolts to grade 5 titanium shoulder bolts, and lowered the side armour infill from 15% to just 5%. This saved enough for the extra frame bulk, and everything was ready for MITE.

…oh no

In the first fight with Flick!, the chassis cracked along where the infill transitions from solid to 15%, which seemed to be due to leverage since that point was right along where the frame splits off adjacent to the bulkhead. That leverage would’ve made it easy to bend along that point and crack the two sections apart, annoyingly right along where the mount for the front left wheel is as well, leaving the shaft loose and dropping the frame down on that side.

Frenzy spent the whole event grinding along the arena floor, deafening the judges and destroying what was left of an already badly worn tooth:

Starting to think I maybe should’ve sprung for the heat treat… Nah that would be too expensive right?

So at this point I’m getting sick of going through a whole frame per event, so instead of printing another in OBC, a material you can’t actually get anymore, I switched to a regular polypropylene instead to try and perfect the design before using any more before BBB Champs. This time I went wild with the chassis, and moved the point where the frame transitions from solid to low infill further back until it was a solid 2 inches behind the point where it cracked before, and I sent it. I had some weight left over from the saving I’d done for MITE and burned through that here, but fingers crossed this frame doesn’t break

The frame did not break

In fact, it was flawless. It was so strong in fact that instead of the bulkheads breaking on big hits, the 8mm “grade 5” titanium weapon shaft was bending pretty aggressively:

Dave was able to get it bent back at the event thankfully, but it did happen again, to the spare and then again in the final.

A final I didn’t feel especially like I earned my place in, to be honest. Frenzy just wasn’t what I wanted it to be, still. But why? It won, shouldn’t I be ecstatic? But no, instead I’m wondering why it was struggling to engage properly, why the drive felt so much less than it had before and what I could do to really make it what it had been in my head before I built it

Oh that would do it

[Revsmin (Ari Callister)]

Part 3: Glorious evolution

BBB Champs was an alarm bell in my mind. Frenzy was just not the robot it was meant to be. I started to catalogue every issue and looked to find a solution them all, starting with the most pressing:

Poor engagement
There’s a number of factors contributing to this one, but the main one is the wear on the tooth edge. After one event using each side of the two sided tooth, both sides were so worn and rounded that there really was no more leading edge on the weapon, no huge surprise it couldn’t bite into anything properly. The other issue is how soft the weapon was. It was machined from 4140, but left unhardened, making it softer than any Hardox weapon blade and about the same as mild steel. At the time I thought a heat treat would be too expensive, but it turns out that Chinese machining companies will do a heat treat for $10 so really there was no reason not to! I also wanted to increase the weapons reach, so I started redesigning the weapon and settled on a more optimised US style S shaped beater, with a tooth on the counterweight, a more aggressive tooth and an even more single toothed design, with nearly 15mm of bite before the counterweight comes into play:

I’ve also dropped almost 50g from the weapon; 350g was excessive, and 300g is still on the heavy side for an eggbeater of this size.

The increased diameter means I can run the weapon at a lower rpm while still hitting a higher tip speed (increasing from 240 to the full 250mph), which should help increase bite as well. The new beater was machined from 4340 and heat treated to 48hrc, about the same as Hardox 450

Control and acceleration
When I first designed Frenzy, my thought process had been that speed was the key to bite. And it is, to an extent; a higher drive speed will mean more bite. But at 11.5mph, Frenzy was too fast to control well, and the gyro, despite being reduced by the stabilisers, was causing it to flip over when reversing and generally affecting stability quite badly. The first change to be made was reducing the drive speed, I changed the reduction on the drive belts to bring the drive speed down to a more manageable 8mph.

There was another issue with the drive being so fast: acceleration. Sure, at top speed the bite should be great, but Frenzy needed a fair run up to actually reach that top speed, and in the small arenas in frantic close quarters fights, it couldn’t reach that speed reliably enough to make the most of that speed. If Frenzy could reach it’s top speed instantly, it would always get maximum engagement, no matter how close quarters the fight became. The extra torque from the new reduction would help, but I wanted to go harder than that.

Frenzy always ran with magnets, about 1.5kg of downforce. Well, mostly. See when I made the switch from steel screws to the stainless plastite screws, the magnets became less powerful. As I understand it, having the ferrous steel screw through the countersunk hole was enhancing the magnetic force, and when switching it was lessened. So I wanted to run bigger magnets, especially since Frenzy had been running new Repeat Max Mk2s since MITE, with a larger 2006 brushless motor replacing the old 2004 motor, with more torque.

But was this enough? Frenzy had never truly lived up to my intention: to build a beetle that drove like an antweight. I needed to push the downforce more, and for that I needed even more magnetic force. I needed all the torque, and all the magnets. Instant acceleration, a robot absolutely glued to the floor. And honestly, to be rid of the tiny external gearbox screws of the Repeat Max that love to back out.

Enter the Repeat Pro; stronger gearbox design and a huge 2207.5 brushless motor. Unnecessary? Maybe. Maybe if I wasn’t stacking four massive 20mm diameter, 4mm thick N52 magnets onto the base of the robot, giving an effective downforce of 6.5kg, 5kg more than the previous Frenzy at its most magnetic

Stripping pulleys
This was an issue I couldn’t solve. No matter how I changed the tension on the belts, it kept happening. So instead, I’ve changed the way the hubs work so the pulleys can be consumables without throwing out a perfectly good tyre:

The two part wheel weighs the same as the single piece wheel, but means I can remove a damaged pulley and swap it to a spare. To prevent the wheel from slipping off there’s a small lip on the top edge, which compresses into the bore then springs back out on the other side, locking the pulley in place:

It can be removed pretty easily with some force, which isn’t ideal for a fast switch but it’s better than the wheels sliding off the pulleys in a fight!

Ground game
Frenzy’s forks worked pretty good for the most part; the hardox forks were hard mounted into the TPU side armour, about a half mm below the wheels so that they flex up, levelling to the floor and being pressed down with some force. This works great, until the TPU deforms slightly and the forks start to get stuck up a few mm above the floor from leverage when the bot hits stuff.

To solve this, the new forks are loose mounted, and spring loaded more traditionally using rubber bands:

There’s also now TPU feet on under the bulkheads to stop the beater from hitting the floor when the bot tilts forward, something that kept happening when the forks came off.

Bent shafts
The titanium shafts kept bending, so for this new version the weapon shaft is now a steel linear shaft, induction hardened to 60hrc on the surface so it won’t bend, while still being soft on the inside so it won’t crack. It’s also bored out to 4mm all the way through, making it only 9g heavier than the old titanium shaft. Most of the strength from a shaft is on the outer edges, so this shouldn’t affect its strength much.

The new build is coming along nicely, and I’m going to be giving it a good test run at Chichester in a couple of weeks!

[James_Draper]

brilliantly laid out build-log!

defiantly borrowing a few ideas from this.

[McMullet (Sam Roberts)]

Great write up, thanks for sharing all this!

[Shooty (Eoin Ó Cionna)]

Excellent build log, and a lot of good ideas here! The drive power on display is ludicrous, and I really like the new hubs you’ve got, definitely need to do some similar experiments myself!

[IrregularJoe (Joe Earlam)]

I have similar drive goals for my bot (Fatal Deviation) - not exactly “drive like an antweight” because I’ve only ever built one and it wasn’t very good, but I having previous RC racing experience I knew I wanted a bot that was as fast as possible while still being predictable. After some early struggles it now drives great, and the biggest factor was implementing electronic stability control - it’s baked into my drive PCB, but essentially it’s the same thing as the gyro modules people install on RC drift cars. The difference between standard and stabilised drive is night and day, it’s a whole new bot. You can punch the throttle on any surface (even dusty polished concrete) and it’ll shoot off tracking dead straight, and it self corrects from all kinds of abuse; smashing the wedge into drain covers, bouncing off walls, getting kicked with steel toe boots…

I don’t yet have the arena time with the system to say how effective it is in combat, but I got to play a bit at champs and it felt fantastic. If you’re looking for a bot that feels super predictable and planted it’s worth experimenting with a similar setup. I might make a guide at some point, but I need to keep some of the sauce secret!

[Kraken4me (Alex Revell)]

Antweight adventure was a good idea, even if you dominating London for 2 months now is getting a bit boring. 2 Shrimp 2 Fresh could use a build log simply because that time it won a fight by holding onto the side of the arena instead of getting pitted.

[Revsmin (Ari Callister)]

Cheers all! I’ve been wanting to pull a build diary together for Frenzy for ages and thought that finishing up the new version was probably the best time to do it!

Bots basically done ready for Chichester, just waiting on the lasercut forks to arrive and some final bits to print. Threw it in the test box all assembled to see how the new drive works out all in and I’m pretty pleased with the results!

Frenzy new drive test

Feels amazing to drive, exactly what I wanted with tons of power and a much more reasonable speed. Even breaks one of the PLA fork standins at the end