Overhaul vs. beta: The Post-Match


A little early, no? I’m going to be at the Detroit Maker Faire this weekend with all of Friday the 29th being a travel day so it’s either now or next week. If you’re in the area, stop on by the Power Racing Series track where I’ll be acting as technical inspector, Grim (penalty disher-outer), flag bro, or really whatever needs doing. Chibi-Mikuvan isn’t in a battle-ready state, so I’m foregoing bringing it.

For this match, there’s not an extensive backstory and tale of preparing, so I’m going to just jump right into the analysis! So be warned, don’t hit the “Read More” line unless you actually do want to Read More!

Read more “Overhaul vs. beta: The Post-Match”

Overhaul 2 vs. Cobalt: The Post-Match; Plus the Lead-up to the BattleBox


Okay, you get the idea. That’s what I’m going to say about every episode of BattleBots. I was there, man.

This post will dive into the details of the Overhaul vs. Cobalt match that was aired on 6/30, so what I’ll do as a favor for those who haven’t seen it – as people who haven’t gotten to see it live will probably see it with a DVR or via streaming over the next few days – is begin by talking about the event itself. Basically, there was a few days before the first matches even began where a lot of preparation was being done, and robots being, umm, finished. That way, we have a sizeable spoiler gap before getting to the match itself, and lastly what I’ll do is discuss the lead-up to the Cobalt fight, what happened during, and the results & work after. So don’t scroll too far down if you don’t want spoilers!

Here we go! After landing in Los Angeles around 2PM on the 14th of April, it took over an hour to get a rental car at LAX… because apparently everyone was also in town for Coachella. I heard from builders who arrived later that the situation only got more desperate, with some waiting over 3 hours for their allegedly guaranteed reservations. No free upgrades for you.

Now, given that it was getting close to 4PM by the time we finally high-tailed it from the airport, it was only natural that we…

…sit in traffic. The distance from the airport to the venue was about 45 minutes during a good time, and now was not a good time. Let me make it very clear that I’ve basically forgotten how to drive a normal, modern car. If you saw me make my way over from airport, you would have sworn I got my license yesterday. Mikuvan: Elevated view, infinite visibility, super short turn radius, and middle wheels only, bitch. That often means an extra 2 to 3 feet of maneuvering space since I can overhang the curb and eat a sidewalk momentarily. Modern car: I can’t see shit, I can’t turn worth shit, I have no clue where the rest of the car is but boy can I go fast. A I R B A G S.

Having to remind myself that I have an extra 5 feet ahead of me that can’t touch anything is a very demanding task, and it keeps me paranoid. Hey, that’s defensive driving right!? Plus, the first time I missed a turn going to dinner and had to swing a U-turn, I didn’t even think twice before almost plowing the fuck perpendicularly into the opposite sidewalk….. because Mikuvan would have made that turn fine, but apparently that’s too much to ask of a 2015 Nissan Maxima. Kids these days.

We actually didn’t end up going to the venue at all on Thursday, as they were still setting up everything, and apparently word on the street was that only international teams were permitted on site during this time. So we hit up the event hotel, another 15 minutes east, and decided to settle in for now and formulate a, umm, battle plan for the weekend. There was a trip to Home Depot to pick up spraypaint so we could finish off the bot, which was unpainted when shipped.

Fresh and early Friday morning!

The venue was a series of rounded warehouses in an industrial corner of east L.A., which calls itself the cutesy name of L.A. Hangar. No, none of those buildings are hangars, they’re round warehouses. You’re no where near the airport. I promise, because I just came from the goddamned airport.

The neighborhood was “interesting” to say the least – primarily Hispanic and touted as “that part of town” to me, but hanging around made me wonder if this was actually the case, or *side-eyes* watch out for those Mexicans, man. Okay, fine, there was that whole manhunt which had the whole venue on lockdown and a police perimeter around the neighborhood and all…  but the weekend corner taco stand beat out any of the food trucks there by kilómetros.
This isn’t a photo of the venue, by the way, but of the artists nest in the building right next to it. This building will come into play later.

Hello neighbor.

Our bot crates were lined up neatly in front of our pit tables already, so the only thing to do now is unpack everything!

Ah, we meet again. That’s right, I did build one of these… I thought the last 3 months might have been a hallucination, that or someone drugged me and stole $17,000 while telling me I built a robot and that everything is okay and I was still a grown-up.

That number is not even close to being a lie, by the way, and I was by far not the most machining-heavy bot there. Oh shit, this is real.

Step 1: Pile everything on the pit table first, think about your life decisions later. After we were done unpacking, two dudes with pallet jacks double-teamed my giant 8 x 4 foot crate into the storage area. Everyone else took just one! I took two guys! AT THE SAME TIME! Wait, what were we talking about again?

Step 2: Take the whole bot apart, and have Paige and Cynthia do the very important task of painting it while I set up the battlestation.

No, this is not “have the girls do the art” – after Overhaul 1 showed up naked to BattleBots last year, and with us having built the robot brand already around the Miku-blue, there was no way they would let me fuck it up. Because don’t you all know that adding paint also adds about 27 to 33 points to your Tail of the Tapeworm score? Overhaul 2’s getting AT LEAST 65 in the DASHING GOOD LOOKS category here, guys!

After putting together the toolbox and lining up my EE tools, I went outside to check out how they were doing.

There was a van!

Outside the Artsy Building was a blue 1985 Toyota 4×4 Van. I didn’t know who it belonged to, but it was someone who worked in that building. Van bros!

I was totally scouting Craigslist the whole week trying to see if there was a cheap van I could buy on-the-spot in SoCal and then drive back across country. Based on the Mikuvan Extraction Mission, there was ABSOLUTELY NOTHING that could have gone wrong with that plan. Nothing.

Late in the day on Friday, and the primer & paint’s dried enough for us to bring everything back in. We would let everything fully dry overnight before putting the bot back together. Not shown here are the pontoons, which were still drying at the time – they’re painted flat black.

SATURDAY!!!!! It’s the day after Friday.


A view of the pits from our station facing the entrance of the venue (beyond the wall is the Artsy Building). Mutant Robots and Don Hutson to the left, and in center stage is Blacksmith. The New and Improved bunnyforce Bite Force is in the distant center.

180 degrees from that shot, and Hypershock is under a blanket, lying in wait. Behind that row is a contingent of British bots, with Cobalt on the left and the Crunchy Candies Creepie Crawlies across the aisle.. There were more pits to the left and right of this, obviously, but I’m gonna skip that for now because Hey, a robot! Look, another robot! This is a robot too! We’re all robots! YAAAAAY!


Saturday is “pretty pictures” day. So we assembled Overhaul in full, with the painted parts.

Obligatory sponsor shot!

Hobbyking, our future overlords, provided the majority of the important parts for OH2 that were not machining. Like we’re talking motors, controllers, my radio, batteries, chargers, and support equipment for all of that. They legitimately handed me around $7K of parts. Do you know how much $7,000 of Hobbyking is? For those of you who might not have purchased things from Hobbyking, that is a lot of Hobbyking.

MarkForged dropped me off a Mark Two printer specifically for this event, which I put to work immediately – you’ll see what it’s printing soon. Beyond that, if you ever don’t feel like machining something, we’ve got your back. You’ve seen MarkForged parts in action here on Chibi-Mikuvan primarily in the steering assembly. You’ll notice that many parts were printed in a black material – this was Onyx, a carbon fiber filled nylon, which was still secretly in beta testing at the time of the tournament. Check it out! This stuff is actually ridiculous.

PRETTY PICTURE TIME! We begin with the 360 degree bot shot, which is where the hovering robot icons in the background of the hosts come from. This is shot in front of a greenscreen that is actually gray. We’ve somehow gotten to the point where computers can automagically tell between a gray piece of cloth and a gray robot. May Robot Jesus save us all.

These are edited pretty much on the spot and the result piped right to the production floor. They need to assemble all of this by the start of matches, after all.

A visit to the arena, still having final light, camera, and action rigging done.  This is in a separate roundbuilding which you actually have to move your bots a fair distance – probably around 500 to 1000 feet – beyond the Artsy Building to get to.

Once again, this is real. It didn’t feel real last time, at least not RIGHT UP UNTIL the worbly light and sound mindfuck they do right at the beginning of the match when the countdown starts. That, last season, for me, was when it finally hit me that I was actually at BattleBots and holy crap Donald Hutson is coming at me. I had a miniature internal panic at that point.

For those of you who might be coming to #season3: Prepare for the Worbly Light & Sound Mindfuck. Fear it.

After this arena visit, I ran back to the pits and….

There are two vans.

Hold… Hold on just a second here… earlier, there was just one van. A second identical 4×4 Toyota Van showed up, and along side it a lowered classic pickup truck.

I actually ran inside and hunted down the owners this time, like some van-themed serial killer. It was a married couple, and these were fucking HIS AND HERS MATCHING 4X4 TOYOTA VANS. Okay, some times people call Cynthia and I nerd-cute for having matching hair and the like, but I’ll never convince her to have matching vans.  This is… I dunno, grease-cute. Gear-cute. Faruq, King of Dubious Robot Puns, I call upon thee to help me the fuck out.

Alright, universe, you win. I’m going back to work on robots. Like a nerd.

Recall that before we shipped, I had prepared all the remaining DLUX 250A controllers I had to be installed when we got on site. This is now the time.

The plan was to run the 250s on the drivetrain only, since they were known to be reliable, and leave the DLUX 160 controllers on the lifting forks. That way, I maximize reliability for the highly pivotal first match – either you win, or you have to vie for the wildcards, which could mean it’s your one and only match.

In the background behind the extracted ESC box is one of the parts I set printing “live” on the Mark Two – it’s a “four 250A and two 160A” mount. Another one is attached to the lid already in the foreground.

An ESC swap sounds simple, but in reality, it was a lot of time-consuming rewiring. All of the 250A output wires had to be remade, as did the extensions to exit the box, since the spacings were different. This took the better part of Saturday evening and into Sunday afternoon…


Work continued into Sunday on the wiring job, which is shown here basically finished but not yet well packed. Up until this point, we haven’t even gotten a chance to drive the thing in the test arena yet, and we were all getting a bit nervous.

I moved onto closing up some unfinished business, which was the P90X. I designed and had printed earlier a servo mount that you see attached to the right side of the actuator. This holds a standard-size helicopter servo from Hobbyking.

Unfortunately, this was as far as the P90X implementation got. After some rumination, I decided not to chance this part in battle if I haven’t already tested it and found that it works reliably. The P90X work fell by the wayside as the robot build moved on, and I never really came back to it.

This will be a science experiment for another day. For now, I locked the ring gear into high (even though it spring-defaults to high, having something hold it there is better) so I prioritize grabbing speed over holding force.

Here is the aftermath of our first test drive in the box. This was when I realized that the front wheels were likely going to cause a lot of trouble, and that I have no clue how to glue things together. The “urethane adhesive” from McMaster did not adhere all that well. Lacking bail options on this part, I just swapped a front wheel (6 were made, so there were plenty of spares) for now and decided to see what happens.

I also discovered at this point that Overhaul took a bit more effort to self-right than I anticipated. It took me a few tries, first of all, to actually gather enough bravery to run the forks hard backwards such that the clamp hit the top plate. It clutched itself as designed, of course, but there’s that nervousness that comes with an untrusted and untested power system. Once I just beasted it into itself, Overhaul could self-right fine, but as known from OH1 and the new CAD model, could only do this if the clamp arm were somewhat open – above about halfway open. I had to practice going “up and to the left” on the radio stick, which from OH1 as well as Uberclocker was “Lift up forks and open clamp”. This was going to be an artform if I were in a self-righting critical match, such as Bronco or a vertical spinning weapon that caught a good bite on the frame’s underside.

With the wheel replaced and self-righting mostly conclusive, it was time to call it a night.


This was it.

We get to find out who we fight first! Monday was “practice day” by the tournament schedule – the “Alternate” entries get to run around in the box with full introductions and ceremonies, so everyone has a chance to run through their lines and get in position, and the camera crew learns how to not zoom in too hard on your robot’s acne.

Producer-bro gives everyone the “don’t impale yourself” speech before everyone moves over to the bleachers, where Aaron, one of the executive producers, primes new builders on how things went last season and how they were expecting to do things this time.

I didn’t get any pictures of this process, but as they dramatically announced each robot, the team would stand up…..then they announced the opponent, and the other team stands up. There was then several seconds of awkward searching for each other before locking eyes and usually doing something intimidating as best as robot builders could possibly, which is to say….. not very. I really, really hope a “BattleBots Outtakes” package becomes available somewhere on the black market, because this was quite priceless. Hell, I even forgot what we did once Cobalt was announced, but it was probably something cringeworthy as par for the course.

Alright. Strategy time! First, we went back to the pits and

There are three vans.

THREE! THREE VANS! IT WAS THREE!! AND COCO WAS ITS NAME! I ain’t goin to heaven no more… I already sold my soul to Trey Roski.

I quit life.

note: I did not know if the 3rd van was actually named Coco, but if you get this reference, congratulations.


We switched Overhaul to its short arms. Wait, what?!

These didn’t make an appearance during the design and build, because they were a bit of an afterthought. They are essentially the arms but cut off about 5″ shorter and at a 45 degree slope, leaving the bot a little stubby looking. They are specifically designed to rest behind the armored pontoons. This is my “spinner configuration”, especially for horizontal weapon.

My basic plan? Use the pontoons exactly as designed – fend off the blade, use Cobalt’s energy against itself, and go in for the grab once it’s disoriented or neutralized. The specific goal was to get the blade to ride up on the pontoons and then hit the second angle level, which will ideally destabilize an already rotating bot. Keep the front pointed the right way at all times, and let the rubberwubbies do their job.

Did I succeed? Spoilers lie ahead.

Overhaul vs. Cobalt



This is fun!

Everything worked quite according to plan. If you tuned in and watched the (sadly, just in highlights – damn you, ABC!) match, you saw Cobalt sail end over end repeatedly, bouncing off the pontoons. Geometry! I didn’t approach as squarely as I wanted to on a few occasions, which results in damage that will be analyzed shortly, but for the most part, this match was highly entertaining.

After 3 or 4 rocket-jumps, Cobalt began smoking significantly and losing power. The impact forces caused the weapon motor, an original ETek* motor, to begin self-destructing internally. There were several more smaller-magnitude impacts after this, and we were only 30 seconds into the match. I kept the forks down and clamp all the way open as long as it kept spinning.

*Robot nerds will point out that Cobalt used a Manta motor, which is a reconstituted freeze-dried ETek. This company bought Briggs and Stratton out of remaining ETek parts after production ended and assembled a short run of their own, advertised as wind generators, but they work okay as motors too.

Unfortunately, you probably also saw the end result: Overhaul loses a drive side, then kind of sits there and flails while Cobalt contemplates existence in a more mobile, hence less county-outy fashion. Overhaul was declared knocked out around 1:30, but really I had mobility problems starting around 2:30 remaining (Note: These times are rough mental estimates, and I will not remember more until the full match gets uploaded).

EDIT: BattleBots has uploaded the full match video! My memory wasn’t that horrible after all. Around 1:31 in the video (30 seconds into the match) was the last time the left side drive was working.

Well, that part didn’t quite go according to plan… Let’s see what went down!

First of all, damn. DAYUM. Cobalt packs quite a punch – enough to straight up pierce the AR400 5mm plate weldments, in several spots.

One of the first connects actually machined off a sliver of the left-side pontoon. The big smear to the right of that is “everything working exactly as intended” – the blade hits the pontoon, glides up, and exits stage right. That hit I’m pretty sure actually sent Cobalt somersaulting over ….haul. I will need to slo-mo the fight to be sure.

Overhaul didn’t look too bad exiting the arena. Nothing was missing, or cleverly collapsed into itself, Ghost Raptor style, but the extent of the damage was discovered as soon as the pontoons came off.


Oh boy. Those are two 1.5″ section frame rails that now look like a real-life FEA simulation result. That’s a whole lot of energy, people.

Surprisingly, this side never bound up, and was working up until the end of the match. Hurray chain slop!

I surmise that the hard side-blows that Cobalt dealt pushed the pontoons sideways until the shock mounts bottomed out, at which point steel met aluminum. The linear gouge to the left is the side of the pontoon hitting the frame rail. However, the actual cause of the bending is probably the two hits (which line up with the two holes in the right pontoon) which seems to have connected AR400 straight to aluminum, with the energy of the blade behind it. I will need to review the match to see exactly which hit this was.

Cobatl definitely hit me going across the front at least once, with this nice dig out of the metal here.

So. We lost our first elimination match, but there is still a shot at the Wildcard picks. Wildcards are usually picked by a combination of factors, one of which is how ready you are when they come around to stare at your robots. Therefore, this thing is getting fixed up, and now is where my investment in two full frames’ worth of spare parts pays off.

We begin ripping the sides of the bot off – some screws had to be coerced back out, because duh.  The right side was the most damaged, but the front left tire fell off during the match, so it has to be removed. That means the left side had to come off also, and that’s when we found it.


Can you see the FUUUUUUUUUUUUUUUUUUUUUUUUUCK in this picture?

When Overhaul’s left side stopped moving, it wasn’t totally dead, but was capable of moving back and forth maybe 2 inches before stopping again. I was pretty sure one of the P80 drive gearboxes was stripped as a result. But something didn’t quite make sense when I was moving the drivetrain back and forth by hand in the pits. Those motors could spin more than a full revolution. Those gearboxes are fine…

Can you see it yet?


Oh dear. There’s a set screw caught in my drivetrain.

Yes, ladies and gentlemen. A single 1/4″-20, cup-point set screw fell into the intermediate chain and locked both drive motors up. But where did it come from? There was only 1 place nearby that I use 1/4″-20 set screws – the motor’s double sprockets. But I used custom flat-point set screws, and all of mine were still there.

That means Overhaul picked up and ate a set screw, which fell off someone else’s robot in a previous match, which landed in exactly the correct location to cause total drivetrain failure. My guess is the front wheel kicked it up and over the middle wheel, it landed in the intermediate chain and sprocket area, and a chance impact jostled it into the motors. This screw was forged into the chain and needed prying to remove. It also took out the sprocket teeth that made contact with it, as seen in the example above.

So in the pits, they have mini production teams running around filming interesting things. You could also flag one of the production staff down if you were doing something interesting, and they made an effort to divert camera and sound crew over to what you were doing. They really did go out of their way to try and capture the frantic robot work in the pits, make no mistake. I exercised this privilege and actually got a team over to film a very quick but very angry rant about how you can be utterly … screwed… by something totally out of your control and up to chance, no matter how much you prepared for it. I was pissed. In an alternate universe where BattleBots is filmed and produced my way, my angry rant against the robot gods would have taken up the entire minute long highlight.

By the way, we think we know which robot it was – and its builders are also fairly sure of it. But we’re not going to spoil it quiet yet!


Overhaul here seen at its Point of Maximum Entropy on the repair path. I eventually put back together the damaged side on its own, and if you visited any of the places I brought the bot, or the Artisan’s Asylum, this is the Science’d frame rail I displayed. Now you know what happened.

I posted a Facebook photo not long after this match with the caption “The front fell off”. It attracted a lot of speculation and attention, but was completely innocent. We in fact just slid the whole fork and clamp assembly off to open up more space to work on the frame rails.

Within about 2 hours, the frame rails had been replaced with spares. In total, it took about 4 hours of work to get everything back together, and we finished Wednesday evening. The only thing to do was to wait for the Wildcard Ceremony the next day (Thursday).

Thursday was designated as a “Repair day”, which means McMaster-Carr made bank off all of us. There seemed to be a continuous ant trail of builders going to the L.A. McMaster warehouse, and… oh, I’M SO HAPPY THERE’S A L.A. McMASTER WAREHOUSE.  Now, the plot twist here is that we had nothing left to repair after the frame pull, so we spent much of Thursday hanging out with the other builders and relaxing.

If you watched BattleBots S2E3 until the very end, you’d have seen the (much more hurried) Wildcard selection, which concluded with……………………….us. This happened in real life, too, and I think it alone contributed 50 years of aging to my cardiac health state.

Now we’re up against BETA, which is…. another British bot with a friggin ETek driving its weapon. I can NOT catch a break this event, eh? What’s going to happen? Well, I don’t know which episode it will be on yet, but for sure there will be another post-fight analysis! So stay tuned to BattleBots on Thursday nights at 8/7c, because we need those sweet, sweet ratings.

The Overhaul 2 Design & Build Series, Part 12: The Final Beast-Smash

And finally, we are here. The last build report for Overhaul 2! Just in time for Episode 3 of the season, where you’ll hopefully get to see it take on Cobalt, billed as the Tombstone of the U.K.! I guess that makes me the Complete Control of New England, or perhaps, at least the Stewbot of the Atlantic Seaboard.

So now we’re in the last week of march, and the April 5th shipment deadline is looming. Everything is in process, but nothing is particularly done. Most of the work remaining at this point was final mechanical integration, and then electrical work can begin. Electronics is usually the make-or-break of a large electromechanical project, but I think this time there was enough pre-experimentation such that I knew what had to be done. Not to say I didn’t have trust issues with the DLUX 160A controllers, though.

It’s an unfortunate effect of having a long to-do list, but the photo documentation does start getting a bit more sparse here. I usually try to stop and take photos as much as possible – usually way more than people I know – but even I occasionally get pulled into something attention-demanding and forget to get the camera until it’s done. So I’ll try to chronicle the remainder of the build as completely as I can, with the explicit caveat that there might be some things magically appearing.

As of the last episode, the rear most #35H drive chains were being hooked up. It was time to also connect the front chains.

Often times in my robot designing, I model a crude chain path that doesn’t take into account the actual number of links needed to make the chain loop. I figure that half-links do exist, even if they’re evil, and otherwise a tensioner will take up the difference. The tensioner here was “designed” after the frame was sent for machining, but it was just a matter of drilling two holes.  It’s a MarkForged nylon print, but with the wall thickness turned up to “all of it” so there was a thick layer of nylon to support the chain. The mounting holes were slotted so it was adjustable over time. Not really a traditional sprocket-on-a-bearing like you might think of chain tensioner, but it’s more of a “chain guide” similar to those found in modern car engines which don’t fail to scare the crap out of me are just blocks of tribological plastic that the chain glides over. They do work, mind you, but that doesn’t stop them from scaring me. I tend to prefer rolling tensioners whenever possible, but it was far easier to just hit “print” here, and it gets the job done.

This tensioner serves two purposes – besides making sure the chain isn’t floppy, it makes sure the bottom-most strand is not hanging out under “robot frame level”. It would suck, for instance, to have a solid aluminum and titanium frame bot slide over the arena saws and the only thing that gets taken out is the two exposed chains! These things therefore also keep the chains above frame level for a good portion of their travel.

Here is what the finished chain path of one side looks like! After this, it was repeat for the other side, and then close everything up with the outer rails.

Also visible here is a test fit of the upper clamp arm.


One of the other “machining kibbles” was finishing out the P90X gearboxes. Here is what I was talking about when I said “chamfering” the shifter ring and the new planetary carrier. This was a lathe operation, and I actually took a spare P80 shaft, bored a 1/4″-20 threaded hole into it, and used it as a mandrel to hold the planetary carrier. The chamfer (angled edge) on the carrier mashes into the mating feature on the ring gear when the gearbox shifts -acting like a crude synchromesh of sorts.

With the bot ostensibly ready to drive, it means I could no longer delay the inevitable: Electronics. Oh boy, here we go. Can’t I just hand this one off to someone else?! Why do I have to develop my own custom motor controllers again???

A few weeks before this, Paige and Cynthia had already cut out two sets of electronics box plates from 1/4″ polycarb. All that was left to do now was to dress it out – the last update had a little bit of the preparation, but now is when we start filling it up for real.

The box now has ESC mounts fitted. The mounts are attached to the box using #4-40 hexagonal NFPC “Pemserts”, which add strength to the joint by gripping more plastic than what threading the plastic can manage.

Paige also finish-machined the bus bars, which are cut from alloy 6101 aluminum. This aluminum alloy has almost the conductivity of pure aluminum, while maintaining something resembling rigidity and machinability. Here they are shown attached with some flathead screws. There is a small split in the upper one on the right side – that’s because the bar to the left is for the 4 drive controllers, and the one on the right is for the 2 lift controllers. After last year, BattleBots mandated separately switched circuits for weapon and drive.

I decided during the design phase to avoid copper here for weight savings, even though that would have made the bars smaller  – they still had to accommodate the ESC’s own wire size and screws to lock the wire in place anyway.

One of the gifts we received from HobbyKing is a small artisanal IED box o’ lithium batteries. This took the longest to get here, for obvious reasons.

Overhaul will use 4 of these 6.2Ah, 6 cell batteries. They’ll be run in 12S2P configuration for a total of 44V and 12.4Ah.

While arraging the packs for their intial charge, I noticed something funny. They felt different. Like one pack just didn’t quite feel the same size as the other. Was I going insane or something? Is it too early in the morning for robot building?

no. they actually were diffrent sizes. In fact, 3 differet sizes! What thell!? Here they are lined up across the top of the Mark Two for your amusement.

I, uhh, just picked the four shortest ones to integrate them into the first battery being made. To be clear, all of them took in the same amount of Watt-hours +/- 3%, if I recall, being first charged. I don’t claim to get what is happening.

Here is a test population of both battery box and electronics deck before I start commiting expensive wires.

I realized I’d run out of 8 gauge flexible wire at this point, so I found a local car audio place. Besides hobby shops, audio places are one of the only other suppliers of silicone-insulated high strand cound flexible wire (Welding shops would offer heavy gauges, like 4 or 2 gauge and up). This led to one of the best conversations I’ve ever had. The lead-in was me asking about “power wire” which is NOT QUITE THE SAME as “speaker wire”, which I didn’t realize – the sales guy was telling me about their selection of large gauges like 2 gauge and 1/0 and up, and I was looking for more tame 8 and 10 gauge. It turns out the smaller gauges are generally used on the speaker side of things – go figure. It didn’t help that I also showed up near the end of the workday, so this guy was probably done with people not being sure with what they were looking for. It went something like this:


“What are you using this for anyway?”

“Have you ever heard of BattleBots?”

“No, I don’t watch cartoons.”

Damn. What a burn. To his credit, he brightened up after I showed the BattleBots website and the big clamp & fork hub, with 16 3/8″ studs sticking out of it, that I happened to have in the back of the van. A small discount was obtained for 20 feet of 8 gauge red and black wire, plus a baggie of ring terminals.


To join the battery packs, I made a “octopus cable” which ended in a larger 75A Anderson Powerpole connector, and otherwise had four XT-90 connectors for the batter packs. After one was verified, Cynthia cloned it for battery #2.

Here’s one battery, after stuffing everything in!

Next came primary wiring of the bot. This is the octopus cable for the robot side. The cables divide back up into two 8 gauge wires, one circuit for the lift and clamp, and the other circuit for the drive.

Getting my other ducks in a row, the 160A HV controllers, trimming the leads and inserting them into the busbars. The holes for wires were purposefully made oversize, since I knew some of these were getting more than one wire stuffed into them. After all, the power input has to feed from somewhere.

In the interest of preserving these more delicate DLUX 160A controllers, I decided to make a soft-start circuit. Also called a precharge circuit, it is a parallel switched circuit to the main switch, which has substantial resistance in it to “slowly” fill up the motor controller inptu capacitors. Without this, I was 1. making a spot welder, and 2. potentially causing catastrophic ringing surges every time I turned the bot on, which I have personally detonated controllers with before.  When testing Sadbot, I discovered that the 13,000uF of capacitance I put on the 3 DLUX 250A controllers was actually welding shut a Hella switch when turning the bot on!

Without much time to make this circuit more integrated, I accepted that Overhaul  might just have a small bundle of wires sticking out in an awkward spot next to the master switches. In the near future, I want to revise the 3D printed switches to incorporate a precharge feature.

The order of ops for turning on the bot was therefore: Connect the two precharge pigtails, then crank the two master switches shut. An extra layer, but worthwhile for not causing premature stress to controllers of unknown reliability.

Here’s the electronics put in as a test fit. Everything goes together great so far!

I decided to perform an intial power-on and test of the system to determine motor directions. This was done with some skinny little alligator clips between the battery and ESCs, just in case one of the DLUX 160A controllers decided to go haywire. With this test, we were able to label all of the motor leads with where they should go to.

This test was conducted using one my cheap Hobbyking 6-channel radios, because at the time, I could not make heads or tails of the 9XR Pro. Full disclosure: Up until this point, I had never owned a “nice” radio – by which I mean, computerized and programmable. I scrapped by on a 2nd-hand Futaba T4VF from 2001 until 2006, then I upgraded to a Spektrum DX6-not-i, the original one. After that, I basically went straight to the cheap HK 6-channel T6A as my staple radio. (We ordered dozens of these for the 2.007 class sessions from about 2010 onwards, so I got reeeeal comfortable with them).

The idea that EVERY CHANNEL IS CONFIGURABLE TO DO ANYTHING I WANT was therefore super fresh to me. Adding to that was the 9XR Pro’s nonconventional channel order as it comes. Note the highlighted “AETR” – this means Channels 1 through 4 on the receiver are Aileron, Elevator, Throttle, and Rudder respectively.
I was super confused initially when playing with the receiver because I think the stock order is “RETA” or something similar – Channel 1 is Rudder, and so on. What? Anyways, luckily, this can be changed to any of the permutations of the 4 channels, so I restored it to what I knew the best.

I set up a bunch of servos to make sure what I did was sane.

After setting up the stick modes and receiver channel order, I started messing with mixing and travel limits. This is a typical “Elevon” mix to allow single-stick driving on the right side stick. Channel 5 is assigned to the 3-position switch to talk to the shifter servo for the P90X, with the center and travel limits being 1st gear, neutral, and 2nd gear.

One thing that happened after this but I didn’t take photos of was swapping a stick from the spare 9XR Pro transmitter I requested and planting it in the left side of the working radio. Operation of OH2 like my previous clampy-grappy weapons requires two fully spring sticks, whereas you buy a normal hobby radio with one stick that has a “friction” channel which stays in place – this is the throttle channel for your plane or helicopter.

The 9XR Pro did not seem to use identical sticks between spring- and non-spring-return sides. They were mirrored parts! Often, manufacters just spin the stick 180 degrees and call it the other side, which seems to hold true for the stick body, but not the spring mounting pieces.

Lacking the time to figure out a non-hackish solution, I decided to simply swap a right stick to the position of the left stick, which is a 180 degree rotation, but all the wires came out the wrong side. That was fine – I just extended one potentiometer cable to seat in the conenctor properly.

After this radio was fully set up and readied, I went on a drive base test, which is recorded in this video.

And suddenly, ROBOT!!!

Is it done yet?! No, this was largely a test-fit of everything in place – pontoons, arms bolted in, and top clamp in place. Boy, it sure LOOKS DONE.

There was one element missing, and that was probably the most important thing on Overhaul if you asked members of the public – the ears. They aid in self-righting by forcing the bot onto a stable position on its back, from which I can power back over.

They were cut from the same 4mm AR400 plate, but as a flat part which needed bending. To do this bending, I used an oxy-acetylene torch and a giant box-and-pan brake. Heat intensely along seam until its dull red, then smash it in the brake and go for it. The faster the heat is achieved, the less the metal is tempered and loses strength. Obviously the bend was going to be weaker, but the rest of the geometry should save it.

The ears were bent separately, and I attended to other matters while Skunk welded them fully in place on the clamp arm.

I began assembling the linear actuator that pushes the clamp up and down. The assembly was very straightfoward – press in the angular-contact bearings on either half, mount the motor, slide the sprockets on to the keyed shafts (the end of the ball screw shaft being keyed according to my drawings by the seller), and there we go.

Originally, I was planning on using an offset link in the chain here, and designed the sprocket spacing to accommodate this. However, offset links are evil, so I cut the chain at the next whole-link up. This necessitated another chain guide part, which I modeled and made from plain nylon.

Assembled actuator, also showing the assembled P90X. This thing worked quite well, but the P90X needed a few minutes of gentle running to wear everything in. I’ll likely do a more detailed build on this thing, since I like it. Up until this point, however, I had not yet designed in where the shifter servo would go. It was therefore left in high gear, which its “failsafe location” if the linkage fails, as biased by the small springs visible to the upper left of the photo.

Alright, so we’re staring April 5th in the face now, and it’s time for a weigh-in. We piled everything that could possibly still be mounted onto the bot, with some extra frame bolts that had not yet been installed, and so on. Final weight here? 247.50 pounds.

This is as close as I wanted to cut it. It was a little under my final expectation of 248 pounds, in large part because of Skunk’s excellent welding that did not in fact require depositing a full 25lb spool of MIG wire into every welded assembly, like I had anticipated in my weight calculations!

After a few more hours of assembly, I’m proud to present….


The somewhat bare looking but complete Overhaul 2.0!

We decided to hold off on painting anything, because if you painted it now, I wasn’t going to touch it it again until the event, and sure as hell wasn’t going to run it into anything to save my paint job! Instead, a couple of minutes of (unfilmed) test driving was had.

My first thoughts upon the test drive:

  • oh shit this thing is fast. Exactly as I intended. It’s fast, yet stable. Still a little front heavy, but having 95 pounds of steel right up front will do that, and the 4 wheels up front (middle + fronts) made the handling much more stable than OH1 was, at least on concrete.
  • It otherwise drove like I intended, and I thought I had it down fairly well after a few sessions. To avoid premature shitting of the unknown-reliability controllers, I decided to save the rest of the driving for the test box at the event.

Speaking of the ESCs, I made sure to update all 5 of the remaining DLUX 250A controllers I had from Sadbot to the latest settings, and packed them. My electronics contingency plan was to swap four of them in as soon as I got to the event, leaving Overhaul with 4x 250A and 2x 160A – driving the lifter only. If any of the 250s let go, I would back-fill them with 160A controllers. This maximizes my reliabilty for the first matches.


The last machining operation to do was to shave down the tooth. Waterjet-cut from a slab of 4142 prehard steel, it was still reasonably soft enough to be machined with carbide cutters fairly easily. I wasn’t out to actually pierce other bots anyway, so I decided to save time and money and not have a “legit” tooth made from heat treated S7 tool steel.

Pack it up! Pack it all up!

It was now the 4th, and the whole day was basically devoted entirely to packing. One of the initial conditions of a good robot event is decided when you re packing parts. We made an inventory – spare mechanical parts, spare electrical parts, all hardware that the robot used were packed in these sorty-bins* with priority. Other spare hardware that could conceivably be used in an emergency repair was brought along, too – like half of my hardware bank came along for the trip. These were packed in different bins.

I even relabeled my toolbox with REAL, HUMAN-READABLE LABELS instead of my bullshit. That way, any of Paige, Cynthia, or Matt could immediate grab a tool without me defining what “Hoodrat Shit” contains.

Furthermore, and most importantly, all of the service tools used on the robot were put in its own sorting system in the top drawer, which was labeled. One mini-bin had the impact driver bits needed to dismantle the frame for two people working in parallel – all of the hardware was all 3/8-16 cap screws unified for a reason. Another one had the tools needed to pull the front armored pontoons, and so on.

That way, even if we were systematically fucked, the robot can come apart in an orderly fashion.

Don’t let this make you think I am some kind of experienced pit crew chief – this is just the culmination of years of mistakes made in robot competitions, and the same years of learning, finally written down at the behest of the others and implemented.

Hell, if you left me at this myself, I would have just brought my entire damn red toolbox.

*several people will hurt me if I keep calling these “SORTY BINS”, so I’ll point out that they are called ALCs – Attached-Lid Containers.

Know what else is getting sorted? Parts and hardware!

We made separate bins for primary drive & frame mechanical hardware, auxiliary hardware (electronics, actuators, etc.), and P80 parts.

There was another one for electrical system parts, such as more wire and connectors. Each of these things had a label saying exactly what they were and what they did.

When you are guaranteed only 1.0 hours to recover from a Tombstoning, every second literally counts, so digging for hardware is out of the question.


Actually not so magic – realizing that we were going to be working up until zero hour, I hired out construction of a shipping crate to a few Asylum wood nerds. This was a beautiful job – they measured all the “big things” going into the crate, and made internal shelving to accommodate!  All of the spare metal parts, for instance, will be living on that little shelf, and the robot with containerized spare parts occupy the lower floor. It had a door that doubled as a loading ramp, hinged at the bottom, and was accessibly by forklift from all 4 sides.

Wood, man.

Oh, did I say “I would have just brought my entire damn red toolbox”?

Well, I guess I did… it was part of the plan…

Oh, also Herpybike I guess. It was right there, man. I figure at the least, if my robot career doesn’t take off, I’ll go on a high speed chase in LA with it.

The robot crate was picked up on April 5th around 1PM. We breathed a momentary sigh of relief, and went home to sleep for a bit. But the story doesn’t end there.

The intervening week between shipment and us physically leaving was filled with spare parts work. This is a spare clamp arm and spare fork parts being cut out. I asked one of the MIT FSAE students to tack all of these parts together for me, such that if we HAD TO at the event, they could be fully welded up and put into service. I also converted four more DLUX 160A controllers – ones HobbyKing sent me as spares – and these were brought along with us when we departed.


So that’s it! Wow, what a journey. I honestly still am not sure how it was all pulled off – any one of several things could have gone catastrophically wrong, and I would not have had a functional robot in time. Truth be told, I didn’t get to test OH2 nearly as much as I wanted. This left me very uneasy going into the tournament, and I made one or two “bailout” solutions for problems I anticipated might occur.  What they were, and what went down at the tournament, can only be revealed after the episodes air.

Overhaul 2 was by far the most intensive and involved engineering project of any type I’d ever been involved in, for work or otherwise. You could say it’s a “magnum opus” of sorts. I ultimately take whole responsibility for how the tournament goes: I was the chief designer, engineer, purchasing office, project manager, fabricator, QA (#ZEROSIGMAS), AS WELL AS operator/driver and field ops chief*.  Basically my decisions alone will potentially dictate the direction of the next few months, if not years, of my life. This is an enourmous weight to think about for those who might be less experienced or just starting out, and is very difficult to explain or describe. Honestly, you just kind of fall into it when the time is ready and your experiences means that a lot of things are second-nature. I’m simultaneously stoked about doing well at the tournament as well as accepting of failure as a chance to do better next time. I know I made mistakes during the design, build, and testing process.

So what’s going to happen? Tune in Thursday, 6/30, 8pm/7pm Central, on your local ABC station to find out! My goal is to have the Cobalt post-fight analysis posted a day or two after the match, so people have a chance to watch on streaming services also.

*This is honestly what kicks ass about the sport. You experience a microcosm of every step of the product engineering process. Many, many young builders have gone straight into successful engineering careers with the insights and experience they gained putting their product in the field and, uhh, “using it”. Robots is cool, kids.

The Overhaul 2 Design & Build Series, Part 11: The Race is On


If you don’t know what’s going on, obviously I’m talking about the BattleBots Season 2 “premiere”, which I put in large airquotes because I don’t believe in “teaser episodes“. It’s episode two, dammit!  Help ensure there’s a #season3 by finding your legitimate source of TV shows!

I honestly think the editing was way better this time than last, which is good to see that lessons have been learned. For one, the ratio of “builder hype” to fights was much better. I think they could have changed some of the matches being shown, but it’s likely that they need to conserve builder profiles for future episodes for the bots that did well. Another improvement is audio – last seasons sounded obviously “enhanced”, let’s put it that way, but whatever new sound recording setup used this year was able to capture the robot noises very authentically, perhaps best illustrated by the drum of Minotaur. Yes, it sounds like that in real life.  Furthermore, and perhaps most importantly, the “bot stats” – a.k.a the Tale of the Tape™, has been toned way down. Even during Season 1, I was a huge advocate of them being either removed or reduced in resolution to something like 1/5 or 1/10, because when you get too fine in resolution, people tend to ask why instead of going wow. This time, I only saw them bring brought out if there was need to drive some point home. Lastly, the commentary is way improved, with the interaction between the hosts seeming less contrived in my opinion, and more dynamic. I can only hope that this level of production is able to keep up for the entire season.

But enough about me! Let’s talk about me instead. Here we are in approximately the 3rd week of March. The machined frame was due to arrive any day now, so it was all guns blazing towards finishing the other bot elements!


The finishing step towards the intermediate shaft was to waterjet-cut some pressure plates. We simply threw these big keyed washers into the same AR400 4mm plate that the upper clamp arm was being made from. The plates are keyed in order to prevent torque transmission to the pressure-applying disc spring and adjustment nut – without these means, the nut is liable to be rotated by the torque transmission, modifying the clutch pressure.

In between the steel plates and gear are layers of clutch lining. This stuff, basically brake pads bought in Fruit Roll-Up format, is able to scrub and transmit force without tearing itself up, unlike running metal into metal. McMaster sells it with various coefficient of friction ratings – I picked a moderate one such that the clutch has substantial adjustability.

The last task was to broach the EPIC LIFT GEAR for a 1/4″ keyway. Now, with a standard 3 ton arbor press, I have to basically monkey bars off the end of the lever to be able to broach an inch or so of steel with a 1/8″ broach. This gear is 2″ in face width and the key is twice the size in every dimension. So this was done on a hydraulic press. Easy, just a lot of very rapid flicking of the pump lever was needed.

Around this time, some very heavy objects arrived.

Steel! So much steel! Look at my STEEL! This is my real steel! There is no toy steel! Shut up, Megabots. I know you’re judging me.

These plates are grade AR400, a popular and commonly used alloy in the mining, construction machinery, and bulletproof trucks industry. It’s been long favored by bot builders for its high resilience properties – not only does it have higher surface hardness in as-purchased form, but remains easily weldable. The alloy – named Abrasion Resistant for a reason – was developed specifically for tacking dumptruck bodies and armor plates together.

It only recently became more popular in robotland because of the rise of a supplier willing to do precision laser cutting for cheap – Discount Steel. All of a sudden, it seemed like everyone was using it, and Overhaul needed a stiff defense against kinetic weapons, so I jumped on the bandwagon.

Interesting to see that I have 3 different colors of plate from which my parts were cut. ARx is a specification, so many different steel houses make the material with different surface finishes and treatments. The green stuff appeared to be a primer or something, so you can paint your mining shovel right after blasting it together.

Enough cut plates were ordered to make four pontoons.

Oh, once the romance of the pieces wore off, it was time for the long drag of marriage. I was reminded of why I don’t build things from steel now. Steel needs welding, welding needs surface prep, and surface prep is what turns you into the fucking Dark Side. It took several hours of dedicated angle grinding to clean the edges of the ABRASIVE. RESISTANT. THEY NAMED THIS ABRASION RESISTANT FOR A REASON steel plates.

Out of a thinner gauge of AR400 that Discount Steel did not offer (Roughly 0.140″ to 0.150″, so I called it 4mm) comes the clamp arm parts! These were waterjet-cut in house, and once again, cleaned up using angle grinders and flappie wheels.

This is a cleaned up arm part. Now, it didn’t need to be actually this clean, but the job was passed onto the very excitable Matt. Biology bros are sure thorough!

I spent a while trimming the series of tubes from last episode into the proper lengths. In the case of the large 4″ arm base tubes, no one single saw in the place could cut either all the way through or cleanly, so they were finish machined. To prep them for welding, a quick pass across the belt sander to scrub the ends clean was all it took. Yeah, it sounds all simple, but this was 3 days or so of my life being filled with tubes.

Not just those tubes – the square skeleton tubes of the arm also had to be cut. They were first roughly sawn to shape with a…. Oh, by the way, here’s how you cut a 60 degree bevel on a saw only capable technically of doing 45 degrees. These arm tubes also needed bevels on both sides, so after the abrasive saw came the cold saw, which is more precisely adjustable by angle. Hot saw, then cold saw… Repeat about 12 times for two long arm pairs and one pair of short arms…


The first cohesive chunk of the bot emerges! This is a finished top clamp arm, minus ears.

No, those aren’t my welds.

All of the welding work on this bot was hired out to Skunk, a resident welding wizard and bike chef of the Artisan’s Asylum. I made the decision after the stark realization that it was going to be very painful for me to weld everything on this bot. What I discovered in my past experience with welding is that I don’t really have the patience to do it well. I can both TIG weld and MIG weld in steel (let’s not talk about aluminum), but the finesse needed to do it correctly escapes me. My preferred welding method is “giga-MIG”, as we tend to call it in my circle – crank the MIG welder up and insert approximately 1 spool of wire into your workpiece!

Hey, it worked for Sadbot, but for Overhaul’s more nuanced welds in thinner material…

You know, if you get Skunk to weld it, it might actually work!

– Rob Masek, when he was still Artisan’s Asylum Facilities Manager

Words of wisdom from Rob!


Skunk not only welded up the clamp arm, but all four pontoons, stitching the tab-and-slot edges shut. There wasn’t any finishing with a grinder needed at all. I was not looking forward to the hours of angle grinding that were going to be needed if I began farming ferrous banana slugs on this thing. I would have cornered the market in steel slugs, that’s for sure!


Then there were the arms. Skunk termed them “the most welded objects I’ve ever welded” and this was a completely accurate description – every possible seam was welded thoroughly. Once again, zero cleanup of any sort was needed.

Heaven forbid I ever get this good at something, eh? Honestly, seeing this work was one of my slap-in-the-head moments. I am definitely a Jack-of-Most-CAD-Programs, but I don’t think I have any outstanding skill that anyone else who builds stuff all the time wouldn’t have. You think I’m invulnerable? No, I’m probably one of the most insecure designer/engineers you’ll meet.

March 24th, 2016 was a joyful day.

It’s here. Oh my Coincident-mate, it’s here. The thing. The part that actually matters. It’s THE robot.

Without the frame, there would be no robot.

Full disclosure: At the 2nd week of March mark, my contract shop was having schedule-keeping questions and asking me if the delivery could be delayed. Again, this was a production CNC shop that I had make things like DeWuts and RageBridge heatsinks and the like, and here I was tying up their machinists with two dozen nearly one-off parts that required CAM, setup, and stock trimming for each.

There was some back and forth, and I ultimately accepted a delivery date slide of a few days (from March 21th to the 24th, basically). During this harrowing time, I blitzed out an emergency “bailout frame” to be done up by Discount Steel in AR400 plate. It had all the mounting points and features, but was just brutalist steel squares.

Luckily, this is as far as that design ever got before I received the frame parts. Remember, this is how shameful it could have all been.

Okay, that’s mean. I don’t want to knock on steel unibody bots; they work fine, but going from billet to gorillawelds to me was taking a massive step back, and addtionally, forcing the expenditure of another 4-figure sum to get the order rushed and air-freighted while also having to pay for the machining was an unpalatable scenario.

Rob being Rob immediately jumped into building my robot for me. Thanks Rob!  This was actually immensely helpful, because he discovered a few “Sigmas problems” with the frame, such as holes not being tapped quite deeply enough (likely my fault since I specified the thread lengths). I probably forgot to account for tap taper when doing so. The rest of the fitup was great.

Okay, I think I’m done. I’ll just stare at this for a while. Ignore me.

You can’t build a bot at the Artisan’s Asylum without everyone wanting to pitch in and help! Soon, I had a crew, and we put together a lot of test assemblies to make sure everything worked.

Here, the drivetrain goes in for the first time.

guys it looks like the cad model omg

The bottom and top plates were cut from 4mm thick 6Al-4V (Grade 5, because fuck 6 and 4) titanium with a waterjet, and the bottom holes in particular were countersunk using a carbide countersink. The titanium curls that resulted were given out to Asylum members for artistic and FIRE! GLORIOUS WHITE TITANIUM FIRE! purposes.

This baseplate snapped into the fully assembled frame without any fuss. The magic of actually keeping track of your tolerances and anticipating tolerance stackup, ladies and gentlemen! I’ll be here all week month #season3.

Cynthia is seen installing the rubber shock mounts, or wubbies (that’s a technical term), for the final time. The AR400 plate is actually just a millimeter or so too thin to support a countersunk 3/8″ flathead cap screw. I knew the wubbies had chamfered edges on their holes, so I was hoping the two would line up and I wouldn’t have to countersink a big bag of rubber wubbies.

Well, turns out I did. That carbide countersink I bought for $60 to do all the AR400, titanium, and other difficult countersinking jobs turned out to be a great investment! If you look at the mounting holes in the wubbies, you can see they’re all countersunk deeper.

My pride and joy, the intermediate lift shaft, drops into place!

Adding lift motors and couplings now.

This is the “everything bracket” that was detailed in the Electronics buildout post. As it turned out, it was the longest piece ever done on a MarkForged machine! Because why else would you design something so ridiculously long!?  I printed one copy, then asked MarkForged to put a full set of spares on their print farm. The fiber backing here is kevlar.

Recall that I had to bail out to using the 160A DLUX controllers after finding out that the 250A model which I’d based my strategy around was no longer stocked. I redesigned the bracket in the electronics box to hold the 160A DLUX instead.

While other parts were coming together, the Mark Two was printing these parts out. Here, I’ve pulled the parts from the machine and am test fitting them to the waterjet-cut polycarbonate (Lexan) electronics casing. Paige and Cynthia took care of most of this waterjet-babysitting offsite.

The previous drivetrain test assembly was just wheels, nothing else. The final assembly meant hooking up chains! I bought a big box of #35H chain with the intent of making several swappable spare chains as we measured and cut each type. Four chains run the drivetrain per side, with three distinct lengths. There’s the center-to-front chain, the longest. Then the middle and rear chains were the same length, and finally the shortest one that bridges the two motors.

After this madness, it was time to start making the remaining details. Remember the nylon logs from the beginning? They were reduced to smaller nylon round things! I will point out that this material is quite possibly the most irritating substance I’ve ever had to machine. Short of hand-ground and sharpened high-speed steel tools, it would not really cut so much as smush annoyingly out of the way, making a large ring of deposited material in the way of your cut. However, I can’t fault it too much, as the parts were dimensionally accurate.

Other parts that were made in the same sitting included modifying the ball screws for the lift actuators, as well as the main lift shaft endcaps…

…which Shunk quickly attached to the ends of the lift shaft tube. This is 4140 steel endcaps on a 4130 steel tube with 1/8″ walls.

The big engine head studs were installed into the lift hub and the upper clamp hub checked for fit. It’s basically time for the arms to go on now…

In the next episode, the mad rush to the finish and the outlook at the event! I want to clear that early next week, such that once Overhaul’s match with Cobalt airs, I can post the fight analysis, ala Chaos Corps.

The Overhaul 2 Design & Build Series, Part 10: What Do You Mean It’s Out of Stock?

Previously, on How to Build an Overhaul…

dramatic over the shoulder camera angle here

charles stares intently at a lathe on autofeed


“this insert seems a little dull and i want some mountain dew”

Machining the parts for the lift clutch was well under way by early March, and parts were coming in almost daily. Our goal was still to have “the robot minus the frame” done by the 3rd week of March, when the CNC machined frame was to arrive. For instance, here are infinite wheels:

After confirming that one hub works, it was time to make all of them. Now, these Colson wheels are made of polypropylene with overmolded rubber. The plastic is so soft that it would rather mush around than be cut, unless it had no other choice. Broaching the wheels made all kinds of neat Colson art, like so…

The next step to assemble all the drive wheels was to chamfer the tips of the waterjet-cut sprockets. Chamfering the tips makes the sprocket more tolerant to side-to-side chain slop since the teeth become narrower, and so the chain has more wiggle room before it tries to catch between teeth and fall off. And I can think of no better application where your chain is likely to have a sudden amount of side to side slop!

There were 30-something sprockets to machine, and I needed a solution to make life simple.

Enter MarkForged

I designed a jig that the sprockets bolted into – they had the correct center bore and two keyways to drive the sprockets, and were made using several layers of Kevlar reinforcement for strength (This wasn’t a critical application that had to be Kevlar – it was honestly what I just had loaded in the machine at the time. Any of their fibers would have been as rigid, or more rigid). On the inside of this print was a hex socket that fit a 1/2″-20 nut. A 1/2″-20 bolt runs through the body, giving it even more rigidity, as well as retaining the sprocket. It was then easy to chuck the jig into a lathe, holding the sprocket away from the chuck, and quickly chamfer both sides with a cutter. Then simply swap the sprocket out and continue!

The other piece shown is a portion of the electrical deck. It holds the robot-side main battery connector onto the electrical box, and was included with this print because why not. I wanted to try out the dimensions in real life to see if any adjustments had to be made.

After broaching the 5″ and 3″ colsons, I began to become concerned about how easy the wheels were to broach.  While the 5″ Colsons had plenty of material at the hub, the smaller 3″ ones had a significant portion of the hub’s radial thickness removed when I broached them. This was concerning, because the reduced material ring beyond the keyway, coupled with the sharp keyway edges, and the fact that the front wheels would be under the most stress when lifting an opponent, meant that the wheel was likely to crack apart.

Not cool. So one of the first “dynamic redesigns” of this build is now needed – the front wheels. I decided to design a MarkForged printable hub, to which a polyurethane tire is bonded.

The 3″ Colson wheel also only covers about half the hub width, as smaller sizes don’t come in the 2″ wide size. The new front wheel would be equally wide as the rest.

The outer ridges are to increase the surface area available for the eventual glue bond.

Here are some “cheat code” features to optimize the part for a MarkForged machine. Right now (as of 6/2016), their slicing software doesn’t yet allow you to prioritize “internal” features, such as holes or bores. All fiber lay is from the outermost perimeter inwards. The reasons for this are many (including feature recognition challenges), but bottom line is, I can’t tell the slicer to fiber up the keyed bore itself.

So in a manner similar to the “hairline split holes” method of 3d printing parts with more perimeters, I added a few Slots of Persuasion to forcefully route the fiber layers into the places where I need it – the keyway area. The fibers will handle the majority of the power transmission stress, so I was less worried about the plastic necking down here.

I also could have made the wheel a “C” shape, with the ends of the C so close together as to be almost touching – and in real life, the plastic would bond together anyhow – but I wanted to avoid having a single linear weak spot in an otherwise complete structural loop. So, little guidance slits won out here.

I then added a few Holes of Persuasion to force some fiber layers towards the outside layer, and some towards the inside of the hub surface.

Here is a finished wheel! See the keyway being surrounded by fiber, as are the inner and outer surfaces for the most part. I could have put Kevlar into all the layers, but decided to be a bit more economical. Each wheel contains just barely under half a spool of Kevlar. The rest of the volume is simply densely filled.

I’ll be the first to admit that I am not a master of composites and adhesives. I used whatever glue McMaster called “urethane glue”, part number 7493A21. My criterion for selecting this adhesive was “Says Urethane on it, does not require an expensive-ass dispensing gun or carefully pushing out with a stick, and comes in the little cute bubble packet”. This is roughly the same process used to select a bearing grease for Mikuvan’s front wheel bearings (“Well, it has a picture of a car on it”) and those haven’t blown up on me…. yet.

Help me.

The tire material is a soft urethane tube, McMaster part number 87235K74. I cut rings off the long tube I purchased, 2″ wide to accommodate the hubs. It gets stretched over the hub after the hub is throughly coated in the urethane OH GOD IT DOESN’T COME OFF AND IT’S STUCK TO EVERYTHING adhesive. I wiggle the tire on and use the dribbles to fill gaps and create edge fillets. Then I leave the wheels alone for a day or two.

The evening I put these together shall forever be remembered as Urethanocalypse 2016.

Here’s a pile of infinite finished wheels! Well, semi-infinite. We decided that a 100% set of spares was sufficient to start. In the event of post-match damage, swap the wheel for a known good one right away, then deal with the repairs after

Next up, infinite spider couplings. Notice how there’s a lot of “infinite” on this bot? Again, a manifestation of the design goal that is making all of the spare parts in one shot, with the parts themselves designed to be easily replicable. Other builders prefer building 1.0 robots first, then working on spare parts.

All of the spider couplings are broached for 1/8″ keyways to slide directly on the lift motors.

And finally, we get to….

Infinite SK3 motors.

This is where I introduce Equals Zero Robotics’ 2nd sponsor for the 2016 BattleBots season…


Yes, that HobbyKing, a familiar sight on this website for many years!  I was in talks with them as early as Sadbot running on the dlux 250 hacks. Originally, it was just for more dlux 250 units, but I decided to go deeper. The robot community has a love-hate-love-again relationship with HobbyKing’s legendary Chinesium offerings, and many smaller bots have used HK parts to great effect. But can they play bigger?

Either way, some great exposure potential for HK, so they agreed to send goodies my way. One of these shipments was a nontrivial percentage of the world’s supply of SK3-6374-192 motors.

Another shipment was more running gear for the bot:


Critical subssystems and support equipment, all sponsored by HobbyKing.

Overhaul 2 will be running the 9XR Pro radio system with the long-range 433mhz R/C gear, to get me off the 2.4Ghz band. I wanted out of 2.4Ghz entirely – everything from other robots, to WiFi, to production running wireless lighting and microphones and the like, lives in 2.4Ghz. Last year, a few people had control difficulties (including yours truly) in the box, and I was not out to repeat that.

Next up are battery chargers and charger power supplies, as well as some BEC units to run the logic power (receivers, gaudy lighting, etc.) since the ESCs do not provide 5v power. Speaking of ESCs…

What’s missing is a boatload of DLUX 250 controllers.

Let’s rewind a few weeks here. What happened was that I compiled a bill of materials to send to Hobbyking, which included 12 dlux 250 controllers. Everything on the list was packed up and shipped to me, except the DLUX 250s, because…

I’ve gotten confirmation from the manager that this product has been completely discontinued and the factories are no longer making them.

Oh, snap.

After some highly intense back-and-forth over e-mail about possible replacements, I decided to spring for swapping the 250A model for the 160A model.

This was a very stressful decision. R/C amps are usually 0.2 to 0.5 of a real amp* and I was getting concerned about the real ampacity of these controllers. While I had familiarity with them from the BurnoutChibi project, the lack of large-package FETs (in favor of a sprinkling of small surface-mount FETs) and smaller thermal mass was the clincher. I also didn’t have a handle on their reliability – the only DLUX 250A units I’ve ever killed were due to my own stupidity, such as hard shorting outputs or leaving the bus capacitors off.

It was therefore imperative I get units in hand as soon as possible, as even the stock quantity of DLUX 160s was getting low (less than 15 remaining around the world in stock, as I recall). I really need to hand it to Hobbyking here for trying to round up stock in such short order.

*this statement is not backed by any form of science but is pretty accurate

The DLUX 160A controllers came a few days after the rest of the shipment. I immediate got cracking on tearing them open to see what I had to change for in the SimonK firmware. Luckily, the answer was not much. I had to redefine a few resistor values, and also set a correct deadtime (these switched slightly faster due to the smaller FET packages), but otherwise, they were the same pinout and layout. Go figure, but just to make sure.

I then uparmored them like I would the DLUX 250A, but these only had space for one big ol’ capacitor, which is better than nothing!

A set of three DLUX 160A controllers fully modded up. I was going to whip these things in Sadbot to ensure they don’t explode.

I removed the 250A controllers from Sadbot – now a valuable and rare commodity – and tied these guys in, then proceeded to drive around like a maniac, just like I did with the 250A, and lift & self right and do other things with the pokey stick.

Here’s a test video showing a little bit of it. The open floor tests were to try and stress the controller with many reversing and turning instances, and the sliding was induced by keeping the wheels spinning (drawing wheelspin current) as much as possible.

(RIP Frozen fans and normal people)

The controllers passed this test solidly, I’d say. I was beginning to smell SK3 windings before the controllers became too hot to touch, so that’s a good thing.

However, while powering up Sadbot for another test a few days later…

A minor smoke event occurred.

One issue with R/C controllers, especially large ones, is that they scale beyond their hardware designs that were originally put together for small ones. Without proper gate drive, and with the possibility of mixed parts in the power stage, if one FET blows up, it tends to start taking the rest with it. And some times, this happens almost spuriously. This is how a controller works for someone and not for the next person. Add to that the quality control generally associated with hobby parts, and….

Okay, let’s be clear: THERE IS NO SUCH THING AS A MINOR DLUX SMOKE EVENT. The whole thing is seemingly made of thermite, or the mix of materials acts as a rapid propellant, because it was over in 5 seconds, a foot of flame, and a huge plume of smoke. I was in fact fearful of a lithium fire with the batteries next to it, and was getting ready to shove the whole thing off the loading dock. The polycarbonate battery box from Overhaul 1 protected the batteries, luckily.

This was all that remained…. As you can see, the entire PCB is actually burnt in half.

By this time, I didn’t want to try to investigate yet another controller. And unless I were willing to pop out Brushless Rage™ in three weeks and get it right the first time, this is what I was going to have to deal with, and part of what I committed myself to doing – using big R/C parts in big robots – from the very start.

So it was time to put the controllers down and keep moving. I asked Hobbyking for a large stockpile of the 160A units to be laundered my way ASAP. I decided to formulate an electrical system continuity solution which used the DLUX 250As remaining (5) as much as possible, and substituting in the 160s if absolutely needed, but ensuring that I had a full bot set of DLUX 160A controllers right away.

While I was doing these tests, the others had finished machining the motor mounting blocks for the SK3s, and keying the shafts.

Ah, finally, all my ducks in a row!

The difference between a “left” motor and “right” motor is simply which direction the wire leads come out, to keep the wiring in the bot convenient.


The next step was to assemble all the gearboxes. Four left, four right, and four lift motors were to be assembled. Again, a 100% overhead (hehehe, overhead for Overhaul) such that we could pitch in a spare set right away and mull over the broken ones in the pits later.

Notice the delicious strawberry jelly on the left. Mmm, extreme-pressure additives.

Here’s a fully assembled drive motor, with retaining bracket!

Repeat like 20 billion times

Okay, so I lied about getting away from controllers, again!

Overhaul 2 still has that 1 brushed motor onboard, the A23-150 micro-Ampflow motor to run the clamping arm. To use this motor, I needed a high voltage (12S capable) DC motor controller.

Now, I had designed a 8S-capable motor controller of some type a little while ago, and in fact, one of the original RageBridge 2 plans was to have it be 12S-capable. But I decided to scale back for expense and usability reasons – most everyone runs these at 6S (around 24v) or in extreme, unsupported-by-manufacturer (AHEM, WILL BALES) cases, 10S (36-42v).

To change Rage over to the higher voltage specification was actually an intensive job, basically rebuilding the board. The main capacitors were changed to 50V types, the MOSFETs to 60v parts, and a lot of other little things moved along with them – such as changing the gate resistors to accommodate the new drive needs, and the logic regulator to a HV-rated part.

In the end, I whipped together two HVRage™ units for the clamp motor, and tested them (see the small blue wires used as scoping points) to ensure nothing was going to get TOO fiery.

And this is the conclusion of Overhaul 2: Actual Electric Boogaloo (sorry Orion). Next up? A Series of Tubes:

Side note: Overhaul 2 swag is now up on the BattleBots store, if you haven’t seen yet. All artwork by Cynthia! Proceeds of sales do go to the builders, so you’d be sponsoring supportng Overhaul for #season3!