Archive for the 'Twelve O’Clocker' Category


The Dragon*Con 2013 Complete Roundup, Part I: Operation GIVE ME A BRAKE and A New Surprise Antweight!

Sep 08, 2013 in Bots, colsonbot, Events, mikuvan, Pad Thai Doodle Ninja, Twelve O'Clocker, Überclocker ADVANCE

I’m back.

Somehow, and not broken down in western Maryland or something. The past week has been so chock full of adventures that I didn’t even have time to post it day by day like I originally wanted to. The Dragon*Con party got back into town at 1:30AM Tuesday, and now that I’m done unpacking everything and catching up to the last week of shop shenanigans, it’s time to spew it all out before I forget. This post is going to be the length of a small novel and will have 4 official subdivisons with this being the first half. If I start dividing something up at the start, then you know it’s gonna be bad. High energy food supplies and plenty of water are recommended.

A flurry of things happened in the week surrounding 12 O’Clocker construction. Besides working on the bot, I was also racing to make sure Space Battleship Mikuvan could make it 2500+ miles without breaking down or being patently unsafe outside of reason (with me, just the qualifier “unsafe” is insufficient). And on top of all that, I was designing on-and-off an entire new bot.

Here are the four parts. The first two are in this post, the second will be going live later and the two bottom links will be updated accordingly.

  1. Operation GIVE ME A BRAKE: Brake system and inspection all-around on Mikuvan!
  2. Pad Thai Doodle Ninja, an Antweight 4-bar pushybot I designed and built in like 72 hours!
  3. The trip down, the con, and how the bots did at the event!
  4. The links and documents associated with my two panels at  Dragon*Con.

 Operation: GIVE ME A BRAKE

In continuing the tradition of naming major van work after very bad puns, the brake system inspection has been designated GIVE ME A BRAKE. I’ve known for a while that the brakes on this thing were “functionally obsolete” – meaning, nothing bad was happening, and it could definitely stop every time, but it took more effort than any other brake-booster equipped vehicle that I’ve driven and the pedal was on the soft side. For bumming at rather low speeds around the city collecting its own parts, I had no reservations. But before a 2500 mile road trip where the option of breaking down is not available, I decided to at least give the system a visual once-over, and replace some of the major components. At the very least, even if it cannot go I should still be able to stop.

It helps that months prior I had picked up the majority of a new brake system on Rock Auto on some serious discount. New rotors and drums were had for basically $10 apiece, and I also bought new shoes, pads, shims, springs and hardware, and other goodies all on clearance. I’m hoping this doesn’t mean I’ll never be able to get parts again, but for the next few myriad miles it should be all set.

Because I’ve already been surprised multiple times by the severity of mechanical degradation, I also bought a bleeder vacuum pump kit and like a gallon of brake fluid. So this was going to happen eventually anyway, and I took the impending Dragon*Con trip as an excuse to use some of these parts and tools for which I was beginning to feel a bit of buyer’s remorse.

The plan was to work from the rear and move forwards. I’d already gotten visuals on the front disk system in Operation: LOST BEARINGS, and they were serviceable, albeit heavily scored. The rear drum? Never looked at them. All I know about drum brakes are that they are this carefully balanced arrangement of springs and punched metal levers and this weird ratcheting thing that will explode if you touch them, or so everyone warns me.

I spent a while on the Internets watching videos of drum brake repair, and I keep wondering to myself who ever thought this was a good idea. Like, I’d have figured cable-and-cam actuated disk brakes (like almost all scooter and bike brakes) would have been way easier a solution at the beginning of it all.

Anyways, let’s begin. One night I decided to just dive right into it and started by removing the rear wheels.

With my trusty Harbor Freight impact driver (this whole thing is basically a Harbor Freight ad, by the way), I removed the lugs which have clearly been impact-gunned on like you’re totally not supposed to but everyone does anyway. Mikuvan is RWD, so when the wheel comes off the drums are kind of loose on the wheel studs already.

Or they’re supposed to be. I guess years of cyclic fretting causes these things to become stuck together. Someone’s helpfully smeared a layer of antiseize grease onto the wheel contact surface already.

The drum has a M8 tapped hole in it specifically for you to insert a bolt and use it to jack the drum away from the hub.

So here it is. This is the thing. Now what??

When I tapped the drum off, a small mountain of brake dust fell out (the piles on the ground to the right). There were more cakes of it in the crevices by the dust shield, and way more behind the axle hub. After an extensive cleaning and soaking with brake cleaner, the above pictured setup emerges. Before, it was all sort of this even black color. I’m sorry, Earth.

As dirty as it might have been, everything was remarkably new and in good condition. This suggests to me that the drums were serviced (relatively) recently, and rear brakes tend to wear far less than front ones. The lining thickness was almost original – maybe less than half a millimeter thinner than the brand new brake shoe linings.

I played around with this mechanism for a while and got to see finally how the parking brake links up to the shoes, and most importantly how the damned self-adjuster barrel works. Self adjusting brakes are one of those automotive things that I sort of hand-wave and accept that they work and exist, and to not try and figure it out. The other items on that list include manual transmission synchromeshes (“some kind of coney thing bashing into another coney thing and it all works”) and all automatic transmissions (“insert analog hydraulic computer, get different speeds”)

I determined at this point that the rears most likely do not need any parts replaced, if the work was done symmetrically.

Well, was it? I ran around to the other side to see:

This drum took quite a bit more effort. I did eventually get it unstuck with a large gear puller, but not before I thought that maybe some pressure was still remaining in the lines, so why not try and bleed the system to relieve it and see if that would get the drum off?

(Spoiler: The rear right shoes seemed to be adjusted out more than the left, so it was grabbing onto the small wear lip inside the brake drum. The puller just sort of munged everything over that lip.)

Harbor Freight, I’m counting on you to save the day. More scarier words have never been said.

This thing attaches to the bleeder valve and allows you to pull a vacuum before opening the valve, so nobody has to be at the brake pedal to pump it in time with your opening and closing. Create a vacuum in the canister, open the valve, a small amount of fluid (or air bubbles) is extracted, and close the valve before the pressure approaches ambient again.

I’ve noticed that this van is great at 3 things:  raining bearings at me, dropping little flakes of rust everywhere, and emitting brown and black mucus when I least expect it. I knew that brake fluid degrades after a while, but eww. Armed with a jug of new brake fluid, I decided to perform a full rear system flush (the fronts would wait until I have them apart). Out come the Gatorade bottles…

The bottle on the left doesn’t really capture the blackness of what came out for the first few minutes, since it’s diluted out with some newer stuff. I used the rear right wheel’s bleeder valve, which is the furthest point in the circuit, so both rears were cycles. Check out those deposits in the right bottle…

Anyways, here’s the right side assembly after some cleaning. Looks identical to the left one, so I decided to put everything back together. Since I messed with the adjuster on the left side, I decided to rough-adjust both sides using the brake drum as a guide (“Just a little drag”) and let the self adjusters handle it in the parking lot later.

The next day was dedicated to the fronts. I’d already removed the front hubs and calipers before to replace the front axle bearings, but had not tried removing the caliper slide pin or dismantled the caliper in any other way.

I spent the better part of half an hour trying to get the slide pin loose to swing the caliper to the shown position. Why? Because some fucker who serviced this before definitely impact-gunned it on. With a MUCH bigger impact gun. It took me 10 seconds of straight impact wrench bashing to get the damn thing off.

Blame it on weaksauce Harbor Freight wrench or whatever, but stop impact gunning my shit.

After removing the caliper body, the rest of the steps were fairly intuitive.

And back on. The C-clamp shown was to reset the piston to clear the thicker pads.

At this point, I could remove the caliper as a whole in order to take the front hub and disk off.

Here’s the left front hub removed, showing the nice and scored rotor with a giant ugly wear lip on it.

The disks are bolted onto the hubs, and I removed them by clamping the disks in a vise and impact gunning the bolts out. These used discrete nuts – the hub wasn’t threaded or something, so it was an adventure trying to apply back-torque with a breaker bar to some very corroded nut threads. Was it too hard to thread one of these things, guys?

All new disk mounted and torqued not with an impact gun. I cleaned out the grease cavity and bearing races completely because cleaning the hub caused a ton of grime to fall into the bearings, so they had to be cleaned out and repacked.

Front left wheel buttoned up. Now that I have a vague idea of what I was doing, the right side went much more smoothly.

This time I was a little smarter and made myself a shop rag seal for both sides.

This is the scene at the height of entropy, when I had all the doors open and all my tools out. I was convinced someone was just going to come by and steal everything while I was working inside.

But they’d be stealing Harbor Freight tools – so am I really worse off, or them better off?

Time to complete the system flush. Hey, did you know I had front air brakes? I didn’t know either!  The first thing that happened when I opened the valve was a small riot of air bubbles. That would explain the soft pedal for sure.

(I guess it’s more “air over hydraulic”, eh?)

The total amount of brakerade generated. It’s interesting to see the different shades between rear and front. The next day, I took this to the local auto recyclers for disposal, where they presumably lit it on fire in the back or something. By this hour, all the traffic in the area had totally cleared out, so I took “wearing in the pads” as an excuse to take the longest, most convoluted possible way back to home, starting with gentle low speed stopping and progressing into trying to see how fast I could stop before a red light while not locking up or doing a stoppie. Brake responsiveness and pedal stiffness were greatly improved by the work, which I suppose was the goal.

Continuing on the trend of extracting brown mucus from various places, I decided to change the differential oil since it’s probably another one of those things which was last serviced 153,000 miles ago. This process was relatively painless – untighten the drain plug, unscrew with your hand, then feel the viscous brown goo envelope your hand as you wondered when you went wrong in life and became a van mechanic.

The smell was horrid. Old gear oil additives seem to decompose into various phosphate and sulfide components over time and it was actually like 20,000 eggy burrito farts at the same time. I refilled the diff with some Mobil synthetic 75 weight gear oil. I’m actually not sure if this entire rear solid axle is oil-flooded or not, but it takes like 2 liters of the stuff and the bulb volume under the fill hole is not that large.

While I had my waste oil bucket out, I also changed the engine oil completely and installed a new filter.

Look closely at the picture of utter chaos a few lines back and you’ll notice I have little devil horns up front. They’re a set of these things that I turned into an adjustable roof rack using some spare 80/20. There was a point a month ago or so when I was extremely concerned about cargo space – when we possibly had like 5 robots and up to 3 large props travelling down, so I took some recommendations for roof racks. These little things seem to be convenient if you don’t want to drill and rivet into bodywork, and so long as I have a 10 foot long rain gutter on the sides, it can be slid anywhere.  I can bolt entire Chibikarts to the roof now. This might get exciting.

So, that’s the state of the van on last Monday night before our scheduled Tuesday night departure. It ended up that said large props and numerous large robots weren’t happening, so this is decor for the trip, but will surely come in handy some day.

Working roughly in parallel with this was the design and (mostly) fabrication of an entirely new bot.

 Pad Thai Doodle Ninja

I some times take interest in how people name their projects and builds. For myself, I began it all by building Test Bot which literally was a test bot to see if I could put together parts in a meaningful fashion, and the name just stuck. I tend to be very direct with names – for vehicle type projects at least, it’s usually [noun][thing] or [adjective, usually a size or qualifier][thing]. Melonscooter, Kitmotter, Johnscooter, Tinycopter, Chibikart… even Mikuvan.  it’s a naming method which I see as sort of idiosyncratic of my stuff, and which also spread to some of my former students or MITERS peers.

It’s harder to call for other things. It’s easy to see where LOLrioKart came from (if you’ve been living under a rock since 2009, it’s like Mariokart), but not so much Überclocker. I myself have even forgotten where I got the idea to take overclocker and turn it Über, and 12 O’Clocker was a jocular offshoot of that since it was a 12 pound bot. So I guess I name things by “least resistance” – I’ve never spent hours or days thinking of a name for a project. Nor do I do that for products: RageBridge was originally “Ragetroller” because I was enraged by the lack of good motor controllers in the robot universe, and DeWut!? was only a short step from DeWalt, whose drill motors I unashamedly press into duty doing things their engineers would have never suspected.

So of course what I’m saying is, I have no clue how the hell the name for this bot came along except for this image:

Look at the very bottom left.

This modern art example came about because somebody brought a bag of Internet-themed word magnets into the shop, and shenanigans ensued on the local Rancid Dragon (a greasy spoon Asian takeout place) restaurant menu. Pad Thai Doodle Ninja just had a good floooooooow to it. This bot was named before I ever started the CAD, which is rare.

So what is Pad Thai Doodle Ninja? I started itching for a new antweight right after finishing 12 O’Clocker the week prior. I could have re-entered Pop Quiz  from 2011 with a new, one-piece 3D printed frame, but that thing had a tendency to take off without warning (protip: long blades on horizontal bar bots are awesome but impractical). At the same time, in conjunction with my sentiments expressed in the original 12 O’Clocker intro post, I did want the return of Test Bot in some way. I miss driving a bot that’s 100% drivetrain, or mostly drivetrain with a single degree of freedom weapon. Not since I built Überclocker in 2008 has this been the case with one of my entries.

So why not make a tiny Test Bot?

It would come together quickly, once again being a 3D printed frame, and would only use parts on-hand and from McMaster (which is basically next day turnaround). I sort of rushed into designing this, so there are no early CAD pictures. Here were the goals:

  • Four wheel drive using two motors, some 20:1 Fingertech Sparks I had on hand, rear motor in a fashion similar to Test Bot 4.5.
  • Servo actuated 4-bar lifter using unmodified servos so the stick position is arm position (using some HK939MG mini servos I had already from the thrust-vectoring deathcopter project)
  • Sloped front with embedded lifter, possibly a short hinged wedge. Armor to be made with 0.015″ spring steel shim stock overlaid on the 3D printed frame
  • Able to self-right.

This last one is kind of tricky with 4-bar lifters. You really have to take into account the center of gravity of the bot, and the length and extension of the arm, in order to facilitate this. Generally, 4-bar lifter bots flop onto their backs and come to rest on the arm whenever it is then deployed, as the CG is too far forward, and no self-righting is possible. Check out this classic video of former Battlebots heavyweight Biohazard to see how a 4-bar could self right.

Notice how its center of gravity is far enough back that the bot hinges on its rear edge and does not come to rest on the arm. The arm’s retraction then keeps the CG within the line drawn between the arm’s contact point and the bot’s rear edge, and it gathers enough momentum to push back over. Making the bot able to do this meant making the arm extend all the way back across the bot. Notice also how Biohazard had a ‘tang’ at the very back of the arm, a part that sticks up – this aids in the maneuver by making the contact point with the ground further forward, so the ‘line’ is longer.

This goal meant that I was continually watching the bot’s center of gravity in autodesk Inventor, and also continually modifying the linkage to suit. The arm had to have a certain amount of extension to make sure the CG was in the right place, and that extension had to jive with everything else’s placement. Here’s an example of a 2D sketch linkage I used (many times, with different lengths) to check the arm geometry:

Notice the nonplanar attachment points for the arm – meaning, the pivots aren’t all on flat lines with each other. So the virtual arm (the top link) actually doesn’t sit flat whereas the real arm takes the mounting point shift into account and does.

Making little sketch linkages in CAD programs is one of those things which distinguishes a geometric modeler from a parametric modeler. The former just treats your lines as a drawing, and if you move an endpoint or something the line length and orientation changes, with no effect on other neighboring elements. In a parametric modeler, you can add things such as dimensions (exact lengths, regardless of orientation), and geometric constraints (this line must always be perpendicular to that one, or this point must lie on that line, etc.) and these constraints are dynamically solved as you force the elements to move.

This is the frame of the bot about 1/3rd through design. I modeled the basic proportions after Test Bot, but shifted the rear motors out such that the wheels could touch the ground if the bot were tilted up. This necessitated mounting the motor much differently than in Pop Quiz (2 piece top-down clamp mount) or in most of my other bots (face mount) – the motor mounts are actually C shaped and slide in from the back.

Also modeled in this early picture are the two metal gear miniservos and the battery, a 3S 460mAh lithium polymer pack left over from one of the copters. The choice of wheels was going to be my insectweight default: O-rings stretched around a custom 3D printed rim. The outer set of rings will double as power transmission to the front wheels. O-ring drives are pretty popular in these smaller weight classes, but as I learned early on, there’s a catch – O-rings have to be stretched over their wheels, or else they’ll just roll sideways right off! Typically the stretch is 25% or more. The same is true for O-rings used as drive belts.

About 50% done, and a few hours in. I’ve kept the center of gravity marker turned on (the yellow ball) to check that at all points in the arm retraction, it lies between the arm’s contact point (just barely behind it) and the bot’s upper rear edge. I’ve also now put in the mounts for the servos – a top down clamp.

A drastic change from the previous snapshot to now is the addition of solid wedges. I’ve historically not been a fan of solid wedges, but I think hinged wedges would have been too fragile in an antweight when faced with modern weaponry. It would also let me use a very thick section of 3D printed ABS, which would increase the strength of the frame. At this point, I was also extremely underweight, so the thicker the better, right?

The 0.015″ spring steel shim will be inset into the side wedges and front, and be retained by infinite #4 self-tapping screws. Attachment of armor to the substrate is just as critical to its effectiveness as what material you use. If an extra hard steel with good backing is used, weaponry will tend to glance off and not catch and rip the material.

Spring steel bits added. This arrangement of top armor leaves the servo and drive motors serviceable without removal. The front armor slopes down further than the bottom of the frame to complete the front wedge.

In retrospect, it would have been better to leave the front armor also stopping at the bottom of the frame, so there’s only one point of contact with a potential opponent – the lifter. During the event, any bending of the front armor caused the bot traction problems.

View from the front. One thing that is missing from this image, but made it into the final “production” arm, is a little tang in the back of the main arm link similar to Biohazard’s. The “ears” are both for adorabu and as a front stop to prevent bots from just driving right over the top, since this bot is so short (about 0.9″).

Sunday before the departure, construction began on Pad Thai Doodle Ninja by waterjet cutting the steel armor and aluminum arm parts. I also started the build of the one-piece frame on a Dimension 3D printer. Pop Quiz was originally slated for such a one-shot print, too, but I elected to use Make-a-Bot (when it was still a thing) to keep the resources ‘local’ so to speak.

Tossed in with the build were the auxiliary components including servo and motor mounts, and the little o-ring wheels.

I thought I had a set of 10:1 Silver Spark motors, but it turns out I either gave them to someone without thinking (This happens more often than it should…) or never had them in the first place. Instead, these 20:1 Gold Spark motors will have to do. It means my top speed is only going to be about 3 feet a second, which is quite slow for my tastes.

The o-ring wheels have the D profile already in their bores, but also have a cross “drilled” hole that I’ll tap for a 4-40 set screw regardless. In Colsonbot, I had trouble with the D bores stripping in the soft plastic.

The waterjet-cut pieces were out of 1/8″ aluminum for the arms, and my 0.015″ spring-temper steel shim stock for the armor.

I heated up the spring steel shim with a torch while it was in a vise in order to make these bends. The area of bend will be weaker than the rest of the steel, but I tried to keep the heat local as much as possible.

The holes are sized such that they’re just about .01″ too small for a #4 countersunk screw to pass through. This ensures that I have a reasonably flat surface up front, but is much stronger than if I had actually countersunk the screws fully. As will be seen, the screws stick up just a little bit.

One thing I forgot to do was mirror the last set of outside holes to the right side. Whoops…

There will be 3 standoffs between the inner and outer frame in those hole positions so I can mount the rubber O-ring drive without having to cut it every time. To make these new holes, I had to turn a 0.2″ peg that stuffed into my 0.2″ counterbored hole in one of the positions, use that to establish a coordinate system, then countersink the rest (though with 0.25″ cutters). The servo mount backs up the plastic material from sinking down due to cutting pressure, and the elaborate clamping prevents the plastic from fluttering.

This was the status of the bot before we left on Tuesday night. I was going to wait until we got to the Invention Studio and set up a forward operations base of some sort.

Bright and early on Thursday at the Studio. I packed Colsonbot and the semi-retired Pop Quiz; Colsonbot was actually going to be entered, but Pop Quiz was only along as spare parts if needed. On deck were machining some arm standoffs, modifying the lift servos, and then wiring the whole thing up.

Normally, I’d use some custom-machined spacers in these kind of applications, but the GT machine wasn’t very well suited to producing small stuff. It’s large in swing, gearheaded (and noisy), and the tooling was not in the best condition. So, to speed-finish the bot, it’s time to resort to plastic washers! This wasn’t as bad as I make it out to be, mostly because plastic does have some ‘give’ so I could tune the friction and slop of the joint using a threadlock-glued pivot screw.

The front link attaches directly to the servo output arm. I was preparing to run 2 servo lift on this bot in order to get more force – with 2 servos, the calculated max lift force when the arm is fully retracted (therefore in the worst mechanical advantage position) was 1 pound. So in other words, it can dead-lift an entire 1lber from the lowest position. Now, typically, when an opponent is lifted an edge, you’re lifting somewhere around 50% of the weight.

As I found out, these servos aren’t very well matched in how they handle the same range of PWM pulses. In fact, one servo traveled about 10% more than the other, while Y-connected to the same radio channel. This meant that the servos fought each other when the arm was at either extreme of extension. Digital servos would be far better matched.

In making the 2-servo version, I also had to “mechanically reverse” one of the servos since they faced each other across a mirror plane. Normally, Y’ing each servo to the same radio channel meant they traveled in the same direction while looking at their own outputs. But I needed them to travel in the same direction in a global reference frame, so one servo had both its 3-lead potentiometer feedback reverse, and the motor wires reversed.

Doing only one of the above would make the servo run straight into one end stop and smoke itself.

At this point, the bot was about 0.9 pounds, so I could as be as liberal with giant wires and solder blobs as I wanted.

Still with two servos, and getting through the wiring now. The black amorphous blob at the top is a small 3A switching regulator that gives 5V straight to the servos. I wasn’t about to try and hitch the servos directly onto 11.1v volts, because they would just grenade almost instantly.

The bot is mechanically together at this point. Notice the standoffs in the center between the frame rails that attach the outer wedge ‘flaps’ to the main body. If this thing were actually one piece, I’d have no way to actually mount and dismount the O-ring belt besides cutting it each time.

Completed bot on the googly-eye scale at 0.88 pounds. The extra amount down from 0.9 is presumably made up of wiring that I trimmed short or something, because I definitely added more screws…

Drive testing of this thing caused it to burn up and strip one servo, mostly due to them fighting themselves with the arm fully down. Going to one servo would have meant losing the ‘dead lift’ margin, but getting into a situation where the bot had to dead-lift an opponent seemed far less likely than a normal edge lift.

The left side servo was gutted, leaving only the output gear to act as a bearing.

The bot was a full 0.12 pound (or about 2 ounces) short at this point, and it was failing to self-right because the CG wasn’t far back enough. It would some times get in the right position with a forceful actuation of the arm, but with one servo a forceful thrust was out of the question. So I bought some fishing weights and melted them down, an ounce apiece, to append to the rear of the bot on top of the motor mounts.

Here’s the “before” shot, the pretty clean bot (no weights have been added yet).

And with the 2 extra ounces in the rear, the bot could self right every single time!

I handed PTDN off to Cynthia to drive for this Microbattles tournament. The event report for both big and little bots, and match videos, will happen in the next half of the post.

12 O’Clocker: The Completion, and Überclocker Upgrades for This Year; Dragon*Con panel information

Aug 26, 2013 in Bots, Twelve O'Clocker, Überclocker ADVANCE

What a week! 12 o’clocker was completed early on, then I began to focus more intensely on upgrading and repairing Überclocker for this year’s competition. 12 o’clocker is currently working and undergoing test driving. It handles enough like Überclocker, but I hope to be able to second-nature the peculiarities of its operation. Plus, it’s fast. Just how fast? Find out now.

Continuing the manufacturing of the “solid state reactive outriggers”, I needed to drill some 1/4-20 clearance holes through the Rockwell C44 hardness spring steel. The week before, I purchased two solid carbide drills off eBay for a cool $7 each – if you have the patience, eBay is a wonderful industrial and machinery supply house – to do this one task. These are straight-flute bits designed specifically for drilling shallow holes into very hard metal. The straight flute maximizes stiffness, but of course does not allow for very deep holes.

Carbide wins over steel any day, and this process was totally straightfoward. The curls that came off this thing could cut you open in an instant – to prevent this from happening in MITERS, I hammered them all apart and trashed them before anyone could discover them on the mill.

Here is an assembled leg. The aluminum end blocks are also drilled clearance, and a bolt holds each end block together. I elected to take this route over threading the aluminum because the bolt could impart much higher fastening forces without stripping the thread.

The legs mounted. I unfortunately didn’t have any lower profile nuts, so the big locknuts on the end kind of function as 2nd-order fish-hooks. Strange how a normal 1/4-20 nut – just about the most common possible nut – couldn’t be located this time of day. I bought an entire bag of 1/4-20 Keps and grade 8 plain nuts from McMaster in retaliation.

With the bot now mechanically done, I turned immediately to electrical work. Remember the original CAD picture and how it had the cells in the middle and the RageBridges above the motors? I found that arrangement wasn’t optimal. Instead, some arranging of parts in real life found a better arrangement which kept the battery mass in the rear. This arrangement did however force me to mount the Ragebridges vertically. That, and the batteries shouldn’t be lying directly on top of the drive motors because that would be unfavorable to the motors in high impact applications… like combat. I needed to come up with a vertical rack for the controllers, and some kind of shelf for the batteries.

I went back to the 3D model to generate these required components. This is the result, after a few minutes of staring. The black vertical guides hold a hardboard (HDF, pegboard) plate upon which the RageBridges are bolted. Two auxiliary motor supports keep the battery from jouncing on the motors directly; these were originally slated to be made from Delrin, but I only had enough of the material left to do this right exactly once….and didn’t. So they ended up being made of HDF anyway.

The little squiggle at the bottom is a flexure joint to keep these plates tightly stuffed onto the motor. Unfortunately, they were made pretty much useless in HDF.

I decided to make the battery first, since it could take a first balance-charge overnight while the Ragebridge mounting end guides were 3D printed on my Up? machine. The first step is to select a group of “relatively” matching cells out of the cell medley. I generally reserve the highest voltage, most closely matched ones for EV use. This is because robot packs may be frequently swapped out and charged externally (with the implication that you’re usually using an R/C style balance charger), but an EV pack needs to be buried inside the vehicle and out-of-balance or weak cells can degenerate from lack of charge monitoring very quickly. It actually took a little while to gather enough middle-of-the-road cells. The majority of the A123 Collection hovers at 3.29 volts these days, and I was looking for cells in the 3.20 to 3.25 range. The camp is pretty well split 90%-5%-5% between good ones, shady ones, and ones that are totally dead (defined as an OCV of under 2 volts).

The end-to-end architecture of the pack meant I had to fold over the interconnecting battery braid once I was done with a joint, and I had to do these one “string” of 3 cells at a time. Usually, in a professional setting, packs like these are made by end-to-end soldering using a “hammer head” iron tip, or with spot welded tabs which are then similarly folded over.

A side effect of folded tabs is that they are well exposed on one side to attach the cell taps for balancing to.

Many of my robot and rideable things have featured custom A123 packs. If you’re interested the basics of assembling a custom pack, here’s a short writeup getting the basics of the process by a former duckling student of mine. With some… uhh, more detail and rigor, this is essentially what I train people regularly to do these days – after all, if we’re going to have a pile of donated batteries, it is better to train people to safely and correctly use them than to try and hide everything. In my opinion. If you’re a safety bureaucrat, you may disagree.

Balance taps added. These days, I usually don’t bugger with soldering and crimping my own JST terminals, but buy balance harnesses off eBay en-masse and cut them up. These are professionally crimped and molded, and usually only cost a few dollars.

While the battery pack balanced, I uploaded the print job to the Up printer and by the next morning it had produced the RageBridge rack in question. The HDF parts were laser cut.

RageBridges pulled “from stock” and mounted to the rack. I’ll totally have some of these at Robot Battles.

In the process of wiring it all up. Notice that I’ve bussed the wiring together on the RageBridges – this is one shortcoming of the current style design which I openly acknowledge. Recall that these things were designed originally to replace two Victor 883s in Überclocker. It had 2 sets of power wires already, so it stayed like that in the “production” versions.

The battery is just coated with 2 layers of heavy polyolefin heat shrink, and then unceremoniously stuffed in place. There’s just not enough room to pack all the wires in – a bit of stuffing has to be done. Unfortunately at this point I decided to scrap the glowing 12:00 idea because I didn’t have any space left in this thing to add an EL inverter.

The last few bits of distribution wiring added. Notice the small yellow and brown wire on the right side. This is a direct battery tap for whenever I decide that I have the capability of adding gaudy lighting.  The master power switch for this bot is just “disconnect the battery” – as such, the battery wires are just run outside the bot. When on, the wires are stuffed into the protective confines of the front right frame rails. The RageBridges are set to 25 amps (drive) and 20 amps (weapon motors), which is a pretty egregious waste of their capacity. I’m strongly contemplating a RageBridge “mini” version that retains the Hysterical Current Limiting scheme but can be more finely tuned for amperage at the low end – say 25 amps continuous and 50 for short terms.

With the wires complete, it was time to function-test and pack everything up! I had to play “change the motor wires” a few times to get 12 o’clocker’s control directions to match up with Überclocker

Here’s the ‘press shot’, so to speak, of the completed bot. The weight is 12.05 pounds.

Yes, it’s over, but I’ll remove one of the standoffs from the clamp arm if it’s really that bad. Or just wear the wheels down a little. Worst cast, I’ll replace the steel drive sprockets with aluminum ones.

Remember that part where I said it’s fast? It’s like, unnecessarily fast. Here’s how fast it is:

The calculated ideal top speed with these new motors is somewhere north of 20mph (30 ft/s). In the narrow hallway I wasn’t confident enough of its straightlining ability to push the throttle to full. In the narrow arena confines, it will never be able to hit this top speed, but the acceleration (also shown in the video) is blazingly fast. I think 12 o’clocker is going to warrant a slightly different strategy against opponents than Überclocker just due to it speed. The extra speed overhead will give me some room to maneuver around and behind people instead of strategically avoiding head-on engagement like with Überclocker.

I found out in sparring matches with Turboencabulator that I needed some more current on the lifter and drive, so the adjust-o-knobs were turned up to 30 amps (drive) and 25 amps (lift). Those extra few amps made for substantial improvements in the lift.

Here’s the D’aaaaaaaaaaawwwwwwwwwwwww size comparison shot with 30lber Überclocker. 12 O’clocker is currently ready for action. Speaking of which…


The big change I wanted to make to Überclocker this year is making a more robust clamp actuator. As recapped before, the issue with the Cold Arbor harvested actuator is that it was slowly destroying itself from being made of untensionable chain drives, and the lead screw was damaged pretty badly at Motorama. The shallow pitch leadscrew also made it prone to “bolting” itself, tightening to the point where it couldn’t loosen again. The reason I purchased the IGxx gearmotors was to see if they can act as a commercial replacement for my custom hacked drill gearboxes. That post concluded that yes, they could, but not in a way I would like, so I made the hybrid all-metal gearbox from two of the ones I purchased.

The gearbox was later wrapped into an actuator design:

The architecture of this is very similar to my other custom actuators. The gearset is from Vex (I’m obliged to say every time I mention Vex: When the hell did they get so legit?). In the center of the top gear is shoved a machined down McMaster fast-travel ACME leadscrew nut. The two bronze bushings supporting it are flanged to also handle the thrust loading, and not shown in the model is the gigantic loogie of grease that this whole thing will be basked in.

The lifter model refined a little and inserted into the robot model. This actuator orientation better uses the leadscrew length (instead of having like half of it as filler length just to reach the clamp arm. One thing that I don’t like about it now is that it does not actually shield the motor that much better. With a standard-can speed-400 (RS-385) type motor, it’s okay. But, I only had “long can” motors (RS-395 type) that won’t grenade at the 26v electrical system, and those stick out a little further. The standoff is there as some degree of protection and as a lower travel limit for the leadscrew.

It’s totally fine against blunt-sided bots, but a good high speed wedge drive can probably reach the motor and bend it. I almost want to re-engineer this whole system from scratch instead of just patching parts onto existing structures.

This is what the reach looks like. If I *am* facing a blunt sided opponent, then this actuator is also a nice “hard stop” for the back of the fork. It doesn’t really stick out that far.

It was time to start cutting metal. Here’s the ACME nut about to lose about 50% of its volume! The threads had to be machined off and the part turned to press-fit diameter, then cut off at 1/2″ long. I was wasting 2/3rds of it by mass, but until McMaster makes raw Acme nut stock…

The nut turned down and the gear bored out. They are to be combined using copious amounts of green Loctite and an arbor press.

Pictured behind it is one of my “troll drawings” – in which I put about 10 parts on one page with just enough dimensions to remind myself of what I’m doing. Some times it even fails at that.

The lower gear is attached to the motor by… well, not much. I machined a hex adapter from some steel hexagonal stock. It’s set screwed to the motor, and the gear simply sits on the hex. That’s it – that’s FIRST Robotics engineering for you. It just has to last the season.

What everything looks like when put together. On the bottom (motor side), the hex adapter is supported by another one of those R1212 bearings.

The new leadscrew cut to length and with attachment flat machined.

And finally, everything mounted in the bot. See how far that motor sticks out?

When the clamp is down (in a position ready to close on something), it isn’t a problem since the motor is well shielded by the aluminum body of the actuator. I guess one thing that I have to watch out for now is someone taking a high speed run at me when the clamp is all the way up.

I extended the umbilical cord from the bot to account for the new actuator placement, backing up the wire the entire way with heat shrink.

Fully installed, wired up, and with a healthy dose of lithium grease. I ran some tests where I practice “clampbot yoga” using Null Hypothesis as a chew toy:

This new actuator is great so far. With the actuator motor’s Ragebridge half set to 15 amps, I can literally run it from endstop to endstop, full speed, and still release in either direction. The RB’s fast current limit algorithm helps here, as does the high lead angle of the screw so it can never “bolt” itself together. The clamp speed has also increased to about twice what it was before. Clocker should miss less grabs now because of this upgrade.

I sparred Clocker briefly with Null Hypothesis (feat. Jamison as the pilot of Null) just to shake everything up from Motorama and to try and catch problems early:

Guess this wasn’t really that representative of a match – the floor is super slick polished concrete, first of all, instead of hard outdoor carpet. However, I did confirm my vulnerability to someone getting stuck between the springy legs and the fork. Careful maneuvering will be needed to make sure this does not happen.

I put the bot away for a few days to work on 12 o’clocker, but with that affair now complete, it was time for some pre-competition preventative maintenance.

I tore down both drive halves of the bot completely to check for issues. Are the DeWalts still in one piece? Is anything really, really worn down?

I mean, besides those wheels. They’re pretty destroyed, and only got more destroyed after Motorama from demos and sparring. I did make spare wheel stock before Motorama, so I decided to give Clocker 4 new wheels and keep the half-worn front tires as spares for now. (Due to the way the fork can load against the ground, Clocker goes through back wheels much faster than fronts.)

All I needed to do to swap wheels was to use one of the spacer rings holding the sprocket away from the wheel as a template to drill clearance holes into the new ones.

Überclocker is now cleaned and buttoned back up.

At this point, both of the bots are totally ready to fight tomorrow if need be. However, the story doesn’t quite end there. In these past few days, I also went and upgraded certain parts of my 3200lb 2WD wedgebot to prepare it for the 1000-mile (one way…) trip to Atlanta and ideally back. That’s a story for later on.

Dragon Con 2013 Panel Info

I’m coming back as a speaker this year on two panels for the Makers & Robotics track at Dragon Con. See the full schedule here!

The first one, CAD and Maker Resources, is an extension of my talk last year on where to buy things for your mechanical whirlygigs and doobobs. Except this time, I’m focusing more on how to make nice things. By this, I mean exploring using free and free-ish CAD programs to better design out your projects before laying into a piece of aluminum with a plastic safety scissor, and taking advantage of modern rapid prototyping and digital fabrication resources. Basically, How to Build your Everything Really Really Fast for Kids who Can’t Build Good.

It’s been my observation that many folks, especially those already well-versed in the EE and software side of things, are dying to get into mechanical projects, so it will focus on such possible routes to start immediately instead of mocking things up with plywood and hot glue. Having been on all 3 sides of the proverbial coin (the knurled edge is a side, I swear),I’ll explain why hardware, and especially mechanical, projects are more involved in general and how they differ from the typical fancy PCB.

The second is in conjunction with my partner in hoodrat shit, Adam, and will concentrate on Electric Vehicles. The panel last year was a primarily discussion and Q&A driven session where we used peoples’ questions about EVs to dive off, and that will remain the same this year. Hopefully some hardware will be in attendance too.

I’ll at least try to have a set of slides made for both of these; no guarantees, of course, that they are actually representative of the panel. If the media resources are available, I will record the sessions and upload them after the fact.

12 O’Clocker Epic Post

Aug 19, 2013 in Bots, Twelve O'Clocker

Damn. This past week -and-a-little has seen 12 O’Clocker basically go from a plate of aluminum to an essentially complete bot minus wiring. I’m making good time for Dragon*Con 2013 after all, it seems. In a slight departure from my usual day-by-day- updates, I’m going to make one single epic post capturing the construction up until yesterday or so. I tended to stay later than usual in the shop in the past days, and writing up daily posts would have pushed my already marginally stable bedtime out to a level which would have made interacting with anyone else… difficult.

As I mentioned previously, by the time I made the introductory post, I had already frozen the design and put most of the parts down on aluminum. The first stages of assembly were doing finishing machining on these parts, as well as machining that which I could not figure out how to waterjet. These days, my ‘style’ is reducing everything to a waterjettable or McMasterable condition and trying to minimize manual machining operations. After mechanical assembly was complete, I spent time figuring out where exactly the electrical components – batteries and RageBridges – will be mounted internally, and this is the stage of the bot right now. Here’s forty pictures (from #24 to #64!) for your amusement. I recommend packing drinking water and high-carb food supplies.

This is the final series of parts made for this bot – previous parts, seen in the upper regions of the picture, were rolled in with a waterjet-cutting session for the SUTD summer go-kart class (which I still owe a full writeup on, before I completely forget it!). In most respects, 12 O’Clocker is a pretty simple bot, and there were (relatively) few parts to be cut.

One thing which mystifies most people who see me waterjet is that I manually path everything. I started this habit some time in 2010 when I was getting tired of losing entire plates to the shoddy Autorouter built into OMAX Layout. When I manual-path, I instruct the machine to make one part at a time (holes and all), then move onto the next whole part, and I try to route in a fashion which does not cross over anything it cut before. Because the path includes only whole parts, the plate can accidentally shift (such as in the all-too-common nozzle collision with a pre-cut piece) and I may only lose that one part. The autorouter likes to do all the holes and internal features first for some reason, so if anything messes up, you usually lose the entire sheet. It took only one $200 write-off to permanently scare me into manual pathing. Additionally, I use the Tab function like a maniac, on anything that looks like it can stick up and bump the machine, and on all plastic parts.

This way, I can route an two-hour-plus long job, then practically go to sleep and then come back to find everything all good. Okay, so I don’t really go to sleep, since the machine still needs abrasive feeding.

So here’s everything! The obligatory picture of the pile of parts includes all the aforementioned waterjet-cut puzzle pieces, a few different motors, and Delrin balls.

I machined the black Garolite, my choice of robot covering, on low-pressure to avoid substantial delamination. I further suspended the Garolite on a piece of waterjet brick, which offers a much higher support density. Without these two addenda to the process, I’ve always had massive delamination from Garolite. In this plate, the delam is limited to areas around the holes and initial pierces.

I design the holes in these parts to be already on-size for finish tapping without having to redrill (usually). The first step in assembling a puzzlebot is to tap with threads that which needs to be threaded. Because all of the holes are necessarily thru-holes, it’s easy to chuck up a gun tap (spiral point tap) into a drill and then just go at it. I mixed up a little cup of machining coolant (high-lubricity oil-water suspension – a.k.a the juice that flows in every CNC porno) that I dunked the tap in after every part.

In the background is one of my collections of random cordless tool motors. I think there’s like 50 drills (most of which are parted totally down) and like 10 cordless saws in there.

Next up was machining down the main lift shaft bushings. My approach to live weapon axles is to make them huge, and usually this implies buying metal tube stock to use as shafting. Now, the difference between metal you buy as shaft and metal you buy as tube is that the former is implied to be precision-finished to a size under the nominal (i.e. a 3/4″ shaft might be .748″) so it for sure will fit through your bearings and other driven parts. Tubing has no such guarantee, and in fact is usually oversize. The cheap 3/4″ 6061 surplus tubing I purchased was a cool 5 thousandths oversize.

To compensate, I bore out the bronze bushings on a lathe beforehand – the dimension needed, plus a thousandth or so for the bearing squish fit into the housing. These days, Tinylathe is my tool of choice if only because it’s like 5 feet away.

There was one problem I ran into right away. I didn’t have the right bearing to support the lift motor shaft. I ordered it, but it most definitely didn’t come on time, and I was sort of itching to put this together instead of waiting for the parts. That, and I couldn’t find the 13-tooth sprocket I originally specified. Since 12 O’Clocker isn’t going into space or (ideally) poking around inside an organ, I decided to make a design change on the fly to adapt to the parts I had on hand.

The bearing was going to be a 6803 type, 17mm bore bearing to basically envelope the hub of the sprocket. The 13 tooth sprocket’s hub was just the right diameter to shave down a little and throw into a 17mm bearing.

I had on hand 12 tooth sprockets left over from the EV class, and a 6802 type bearing (15mm bore). I decided to machine an adapter to turn the 26mm 6803 housing into a 24mm 6802 housing, and machine down the smaller sprocket hub to 15mm. This increases the reduction of the lift arm just a hair, which is  differentially speaking to my benefit, since it would both slow down the lift another few % as well as reduce the required current to lift an opponent.

….but in the end, I just shoved a machined Delrin donut in it and called it a day. Didn’t even machine the sprocket.

This is a very slow-moving part already, and drill gearboxen already have bearings! The whole purpose of this bearing is just to reduce the bending load (versus shear load) on the shaft. It doesn’t need to be sophisticated, and can just be Delrin, the venerable bearing-grade plastic I had a few rods of. After spending all of 3 minutes on this, I moved on with my life.

Assembling the puzzlebot now.  The part precedence I designed in means I fasten together the center towers, then the U-braces on the ends, then finally shove the sides on. If you’re thinking of designing a puzzlebot, and don’t read anything else in this whole post, read this: all of these parts were cut on an artificially close nozzle offset so all external dimensions are about .005″ smaller, and all internal dimensions .005″ larger. This is a cheat code to account for the waterjet’s taper between entry and exit sides, something inevitable unless you are really rich-ass and have a taper-free or 5-axis cutting head.

If you don’t have the ability to run a custom offset, then you must design all the slots larger and tabs smaller in CAD. Or, you’re going to be miserable filing and grinding everything.

These parts required very light belt sanding to take off the burrs and tabs, and only in a few cases a hit to the thickness of the tabs.

I’m sort of round-robin machining parts here – in one round, me versus everything. I put the puzzlebot aside and tested the fit of the Angerboxes. Beforehand, I threw these gearcase models on a Dimension 1200ES machine to be made from ABS plastic.

It turns out that the drill gearbox CAD model I had used a smaller diameter ring gear by about 0.5mm than the ones I picked out from the Tomb of Tool Parts. These drills are usually similar enough to swap parts into each other, but different enough that any design involving them is going to jump your precision-havin’ ass in an alleyway. I had to machine down the ring gears a little to fit them.

After a radial pass, I sliced off just enough to contain one stage. These ring gears are made via sintering, so it turns to powder when machined. This is actually the sign of a shitty sintering job, but hey, $20 cordless drills.

6801 type (12mm) bearings support the 12mm drill spindles.

No, I’m not using those plastic gears – they were just the first things I grabbed from the Tomb of Tool Parts to test thickness fit.

After verifying that everything fit as designed, I put these gearboxes down and went to…

…the lifter gearbox. This is a totally stock 18v-native, 36:1 drill gearbox that I replaced the first stage gears with spare steel ones. The 12 tooth sprocket seen in previous pictures was bored out and threaded for its 3/8″-24 spindle. The long standoff piece at the end spaces out the reverse-thread M5 flat-head screw which typically locks a drill chuck in place.

Another blob of 3D printer poop with a drill gearbox shaped cutout will hold this motor in place. I basically made this mount to let me emulate the motor mount style used on Überclocker, which uses DeWut motors with side mounting holes. The generic drills don’t, so the sandwiching mount does.

Overly-long 1/4″-20 cap screws fixture the motor to its mounting cradle. In soft plastics (or at least, anything where $fastener_hardness >> $material_hardness), you generally want 3+ fastener diameters to take advantage of the material strength.

As a random aside, the recommendation for situations of comparable material strength (say aluminum threads with steel bolts) is one screw diameter, and materials of nearly equal strength just 4-5 threads. This is why steel nuts are usually made really skinny – count the threads on the inside of a typical machine nut some time.

The first pretend-o-bot! At this point, I was getting pretty damn excited to see this thing finished.

After the pretend-o-bot assembly, I moved on to the last item of the day, which was to put together the top clamp arm. This was a quick standoff assembly job.

I attached the end-effector, the rubber bumper, with a convenient chunk of 1/8″-wall aluminum square tube I snagged before getting to that part of the design.

The next window of progress was spent making some of the repetitive side parts. Seen here is my “sheet of every part drawing since they’re so damned simple anyway”. The majority of these are standoffs of various kinds.

The inter-fork spacers are designed to be made using some 3/8″ OD, .050 wall 6061 tubing. This was way better than the other option, which was to drill out all of those from a solid rod.

Pictured above are all the fork standoffs and two of the Angerbox output shafts in progress.

Axle pins attached. This is my first intentional foray into single-support wheels in a long time. Single-support wheels tend to last … not very long in combat, so I made everything here overkill. First, half-inch diameter axles on anything that weighs 12 pounds.

The 1/4″-20 bolts are unnecessarily long, to give the aluminum a stronger core, and the threads are actually embedded under a half inch of clearance hole as to preload most of the structure where the most stresses from impact will be seen.

It’s often these little unseen details which can make or break a design, and I’m really hoping for make this time.

Time to assemble the fork! Like Überclocker’s fork everything I build now? , this is just a pile of plates and standoffs…

…that looks like this. If there’s one lesson I have to summarize 2.00Gokart and the SUTD summer EV class with, it’s tighten your damn bolts. Untightened and unloaded structures are just flappy metal noodles.

Moving on now to the drive hubs, these are simple Delrin one-piece jobs. Since Tinylathe is too cute to contain a 1.25″ Delrin rod all the way through, I had to cut off chunks and sacrifice about 1/2″ of length off each to grab them.

The final machining step for these hubs is to purposefully chatter an endmill through them. Why? It’s once again the difference between buying shaft and buying metal rods. The aluminum rods I had standing by are just plain-finished loose-tolerance rods! They’re actually much larger than their nominal 0.5″ (more like 0.505-ish), because you’re supposed to machine them to something smaller with more precision.

Well, I didn’t have a .505 drill bit, nor an over-size reamer. Nor a boring bar long enough. The shortcut is to take a super-rigid cutter and purposefully guarantee it no center, so it just has to deal with it. I basically revved the spindle to full speed and manually shoved the thing through the center bore.

It made a terrifying screeching racket, but the end result was a bore crudely enlarged to about .510″ (varies greatly…)

Fits great though. Here’s 12 O’Clocker looking like someone stole its rims. In reality, I haven’t bored and machined the wheels yet.

The wheel assemblies were, again, designed to smash together quickly and template itself for finish machining. It has four parts – the wheel, a 1/8″ thick ring spacer, the sprocket, and the Delrin hub.

Here is one wheel as a test fit, and the clamp arm assembled on for looks. At this point, I called it a day for assembly because I discovered I ran out of 2″ long #4 screws. I needed long bolts to go through the wheel and spacer and into the threaded holes in the sprocket, acting as lug nuts.

I moved onto making the rest of the irritating small parts. Here are five blanks for the eggy-rolley-cam tensioners machined out.

Because they differed only in length, I lined them up one next to the other on the MITERS Bridgeport and drilled the off-center holes in one shot.

While downstairs, I revisited an old friend, the MITERS South Bend 10L, upon which many of my freshman and sophomore shenanigans were first performed. For how hard it gets wailed on by students, most of whom aren’t experience machinists, it’s never had any problems. 1950s style brute force tends to come out on top. I used the machine’s much larger swing to chamfer the tips of the main lift sprocket.

The MITERS lathe to me represents roughly the most useful machine size for the general hobbyist. It’s a nominal 10″ swing machine, with a 42″ bed, and a big through-bore. Tinylathe is too damned small, and the other machines in this building are enormous – 16″ and 19″ swing. I think the 10″ class has just enough feedback to enable really precise work on small parts, but can still contain sizeable workpieces like wheels and round plates. Plus, my favorite lathe (I have one, just like I have a favorite weird 1980s Japanese cargo van) is the Monarch 10EE, nominally a 10″ machine. Round dial, please.

Here’s a collection of little side parts. The shaft collars are for transmission of lift torque to the main lift fork, and you might notice on of them took me a few… tries. I tried taking eyeball and by-feel shortcuts on the Bridgeport right away again, but remembered about as fast why I only do that for machines I use every day and know exactly the recent alignment history of: because not everyone trams the head and squares the vise. Whoops.

Well, time to bust out the edge finders.

Fast forward to Day…. three? four? Whatever like 2 days ago was. I had dug out my box of 1.5″ long #4-40 screws, but they’re still too short!

Or are they? I decided to get this damned thing together, buy 2″ bolts later, and just counter-bore the shorter screws into the wheel for now, so they could still reach the other side.

So here we go. All 4 wheels counterbored up and installed!

Next comes the lift axle itself. The shaft collars are loosely scrwed into to the forks, the lift shaft slid through, then the screws tightened. The white residue on the shaft is a healthy dose of Teflon anti-seize paste. The reason is that this time, it’s aluminum shaft collars and an aluminum sprocket rubbing on an aluminum shaft. Without this intermediate layer, I would soon have a very curiously shaped solid aluminum modern art sculpture as all of the above galled into eachother and became one. The antiseize lets the clutching action of the clamp shaft collar still occur.

I maneuvered the lift chain tensioner into place and secured it for the time being. It will be adjusted out slowly as the bot wears in the lift chain.

Chains typically stretch a few tenths of a % shortly after installation as soon as they are loaded, since the burrs and high spots from manufacturing get worn down quickly. In my case, with so many custom sprockets, the net amount is a little more because I also have components besides the chains to wear in.

Here is a more detailed view of the eggy-rolley-cam tensioners. The ball bearings will directly push the chain (no intermediate sprocket teeth here).

The bearings themselves are type R1212 bearings, which is part of a series that is basically the 68xx of the English unit world. I’ve spent a long time searching for thin-section inch bearings that don’t cost $9000 each, and for the 1/2″ case, at least I have found the R1212. Apparently the other sizes (3/4″, etc…) are nonstandard and its code depends on the manufacturer. In contrast, searching ‘R1212 bearing’ gives me a world of generics.

With the tensioners ready for installation, I went hardcore after finishing the Angerboxes. The motors I ended up using are some unknown 12V (ish?) 550-class motors I found in a bin. The reason is because I’ve actually ran out of matching shitty 18v drill motors! I have a mishmash of gearboxes with motors, but I couldn’t find two matching 18v motors, especially not with the 36:1 standard 9 tooth pinions. Oddly enough, I have more 24:1 drills than 36:1, and this gearbox depended on the (far more common) latter pinion. I harvested some pinions from the Tomb of Tool Parts and pressed them on these motors.

Now that I’m running 12v motors, I’m actually going to drop back from the planned 8S (25.6v) system to a more sane 6S (19.2v) system. 30-lb Clocker does run a 8S-equivalent system, but with “real” motors which are 18v native anyway. A 12v to 24v overvolt in this case would have given me an unacceptably high top speed and would probably bake the motors in short order.

Three harvested, mismatching metal gears ride on the output pins, and the whole thing is closed from the back from harvested mismatching wear washers. I’m like the Iron Chef of making stuff from shitty drill parts. I did say that all of these drills tend to be the same enough…

Here’s one Angerbox closed up with another one awaiting. The shafts have milled flats (not quite visible) to aid in using set screws properly.

The first installation of all drive parts on one side is complete! This spent a few minutes just running from a power supply to seat all the bearings and run in the chain. A portion of the back wall is now covered in a very weird vertical black splatter mark, like a highly precise mechanical crime scene. The same process was repeated for the other side.

Lift chain in place! Now I could manually wang (that’s a strict technical term) the fork up and down and backdrive the motor.

One issue is that switching to a 12 tooth from a 13 tooth sprocket meant my original tensioner design doesn’t tension very much any more. I’m almost at the limit of travel and the chain will most likely slack beyond that limit. The narrow confines of these 3 rolling elements should not cause deraining problems, though. Worst case, I’ll put a spacing ring around the outside of the tensioner so it can push further.

Clamp actuator now installed, and run back and forth a few times with a dose of lithium grease. This motor will definitely be unhappy on even 18 volts, but I’m hoping the current mode of the RageBridge can delay the onset of magic smoke arbitrary far into the future.

After all this was done, I turned the bot over to install the bottom plate.

I decided to get a start on the very last item on the list of manufacture for the mechanical side: The springy legs. I segmented the 3/32″ thick spring steel bar into two 8″ lengths with an abrasive cutoff wheel and selective belt sanding. These will have two holes each put into them with a solid carbide drill bit, then have the accessories attached.

Here, have some robot googly-eyes.

These are actually the front rollers for the legs above. The axles these things ride on are 1/4″ diameter hardened steel shoulder screws, counterbored all the way into the Delrin ball, and extending halfway through the aluminum. This is to make sure that bending impacts are borne as much as possible by the aluminum and not the weaker threaded area of the screw.

This is the mostly mechanically done pretend-o-bot , with the only thing missing at this point being the legs. I cut out a “window” using 1/16″ LDPE sheet to cover the 12:00 and give it some contrast. I hope to have this whole thing back-lit by a big EL panel.

That’s it for now. I’m going to now compile the previous day-ish, and whatever happens today and tomorrow, hopefully into one more post to complete the construction of the bot. That, and reveal the upgrades to 30lb-Überclocker for this year!

It’s Robot Season! Time for a tiny Überclocker. The Announcement of 12 O’Clocker

Aug 10, 2013 in Project Build Reports, Twelve O'Clocker

Wow, how’s it almost the middle of August already? You know what this means?

That’s right. Basically, for the past few years, every late July and all of August has been spent rage-prepping robots for my annual robot party that I’ve gone to compete at since…


Holy shart, this one better be good then! And what better way to make it good (…besides having functional robots for once, I mean…) than returning to the 12lb class from whence I came?

It’s been a long, long time since I had a functional 12lb class bot. My parting shot at the class was in early 2008 with Test Bot 4.5 at Motorama 2008. At the time, the NERC Sportsman 30s were just barely becoming a thing, and after that event, I decided to participate in the class instead. The 12lb class at the time was just becoming infected with the “Brushless Penis virus”, in which builders rely increasingly on spinning larger and larger chunks of tool steel with larger and larger brushless motors, leaving the class designs polarized towards either those or heavily armored boxes without other redeeming features. Matches basically came down to who got in The Hit first and often ended with both bots being disabled or otherwise gimpy and hobbling.

It’s like being a dubstep groupie – always about The Drop, but with robots, so half the match you’re just waiting for The Drop and talking for days afterward about how brutal The Drop was while picking splinters of A2 tool steel from your forehead.

That game got boring for me fast. So, I spent my next 3 distracted years building dysfunctional versions of Überclocker, until last year when I finally seemed to manage something remotely interesting.

With Überclocker Advance seemingly performing up to snuff at Motorama 2013, and only needing very minor repairs and upgrades for this year (to be detailed soon), it was time to think of new robots. If I were actually good about sketching out my designs, what would follow is a small thesis worth of fantastic fun machines, none of which would actually qualify for

But what I really want is a tiny Überclocker. Ever since version 1 back in 2008, I’d wanted to scale it back for the 12lb class. This hasn’t ever happened yet since I’ve had enough fun getting the actual Überclocker to work properly. I’ve also thought of doing up a beetleweight or antweight version. Past that, the thing which has kept me from developing the idea further is that I was reluctant to build a bot for one event: Robot Battles is the only event I am aware of that has a “sportsmans 12lb” type competition – even though its own competition is much different and came way, way before the modern era of robot fights. Events like NERC Motorama would force me to run this elaborate contraption with the brushless pen0rs.

I’m not sure when I decided to actually pull the trigger on this design. Originally, I was hoping to make a grand comeback with Test Bot 5. I’ve completely designed Test Bot like 17 times over since 2008, but the basic flavor has been the same – a 4-bar lifter with a solid and fast drivetrain. Sadly, this is as far as I got on the last ‘redesign’:

Not even sure what I was getting at any more. Oh well. Those saw motors aren’t even available now.

I figured that all of my fresh-built bots will have some kind of weird teething issue if I start this close to the event, and that would be quite an embarassment to the Test Bot name. I’ll bring that back when it’s ready. Maybe having a 12lb Clocker would encourage the development of the 12 Sportsmans up here quicker – there’s been plenty of talk about it, of course, but we just haven’t quite hit critical mass yet.

Like Überclocker Advance, I jumped into Inventor and made up a quick sketch just to get the dimensions down and see what might fit where. This is where this entry will degrade entirely into CAD pictures!

This skeleton drawing was made about 2 weeks ago when go-kart season was in full swing. Part of what I wanted out of this bot was a bit of ridiculousness. After all, it’s a small Überclocker, and any good chibi version of an existing thing must emphasize its best features. This bot was going to be all fork – or at least the fork will prominently feature in the design. The body will also be shrunk into something with a higher aspect ratio for a bit of the goofy look.

I had in stock several 2.5″ “McMasterBots” wheels – the 40A type – for what I can only imagine was the next Test Bot, so I elected to start the design with this element, even though I think it would have also done great with 3″ and 4″ wheels. Some times, you just need to “ground” the design somewhere or you’ll never start. So this bot will be tall, with tiny wheels (but still plenty of ground clearance), and a giant fork.

One question I had which I tried to solve early on was what to use for drive motors. In the 12lb class, if you’re a pusher type, drivetrain-tensive bot, a set of 500-class motors is about the norm. When “hobby tool” sized cordless tools were easily available in the early and mid 2000s, four of them with their RS-380 motors made a typical first bot for many folks – and Test Bot’s third incarnation which last competed in 2005, and should still be around here somewhere, still uses them. I really wanted this 12lb Clocker to have a forceful drivetrain, since it would likely need to travel around carrying an opponent.

Speedwise, my choice of wheel size precluded using even the 24:1 single speed cordless drills – it would have been dirt slow. I was spoiled by Überclocker Advance’s near-20mph top speed and excellent handling with the 40A wheels, and I wanted something similar for this bot. I briefly considered a set of Vexboxen, but at my required ratios for direct drive (wheel-on-gearbox) it would have been too long with a motor, and would have made the bot like 16″ wide – almost as wide as 30lb Clocker.

Space-wise, an external chain reduction similar to the near 1:1 indirect drive of 30lb Clocker was the most feasible, since it meant I could keep to a single speed gearbox. To minimize size, I decided to make my own gearboxes:

The hell is that? It’s what I’m calling the Angerbox. Like Vexbox and Ragebridge, and other similarly pissed-off sounding robot parts. What it really is is a 3D printable case that holds half of a drill gearbox. Similar to what I did on Test Bot back in the day, cutting a drill gearbox in half and (preferably) using the all-metal output stage nets you a 6:1 reduction. What you see there on Test Bot’s motors is a 2:1 spur stage feeding into the 6:1 planetary output stage, creating a 12:1 gearbox that gave Test Bot a nice 15mph top speed.

I plan to implement that roughly 2:1 reduction using the chain drive this time. 6:1 direct on my 2.5″ wheels would give a top speed of like 25mph, which is a little over the top, but would burn out the motors quickly.

Here’s the Angerbox placed in the frame for first sizing passes. I also drew in various sprocket pitch circles to see what ratios made sense. In the end, I settled on an external 14:24 drive, yielding, with the 6:1 planetary and anticipating overvolting the 18v motors (since I have an asston of working motors with destroyed gearboxes from Null Hypothesis) to 8S (25v), to get me a top speed of 18mph.

Why a 14 tooth on the motor? It was the smallest sprocket that had a hub big enough to enlarge to 12mm bore (drill shaft size) and have enough meat left to hold a big set screw thread. Most times, “the math” doesn’t take these physical considerations into effect – this is one of the big lessons I try to get across in my Silly Go-Kart Camp, that you can math until the end of the world but your project still must exist physically, which implies other constraints.

A 12:24 would have been a clean 2:1, but the sprocket would have been so small that attaching it to the drill shaft would have been damn near impossible without machining the drill shaft down to size (weakening it greatly of course…)

I imported the old drill parts from Test Bot’s CAD files (after cross checking with my bag of drill parts to make sure I actually had the correct types), so here’s what the mostly finished angerbox CAD looks like. The stock drill shaft will be cut down to fit in the bot. The case will be a quick 3D printed jobbie (Hey, the drill cases are originally made of shitty plastic anyway, okay?)

Using my 2D drawing as a visual reference, I started designing up the actual parts. Unlike all the Überclockers ever, this design will feature “overhung”, or single supported wheels. I tend not to favor them, but both weight and goofy looks might force me to use them this time. To make up for the single support, the axle pins are huge for a 12lber: 0.5″ aluminum to be made with my left over 7075 stock. A little bit of overkill, but the fatter the axle the less bending they experience for the same load.

At this point, I’m just trying various different layouts of parts to see what makes the most sense. The batteries, just a single string of A123 26650s, dominate the decision. I decided to make the frame tall enough such that they can stand fully upright, or else the square footage of the bot was going to increase drastically.

Hmm, I *could* make the bot longer and ditch the sloped rear to gain a bit more interior volume. This was, again, one of about 5 or 6 ways I arranged the parts before moving forward.

Another config again, still with a flat back. This was the frontrunner for a while…

Until I realized that with a flat back, 12lb Clocker was hopeless if flipped over, which it fuckin’ will be, most definitely.  That made me switch back to the sloped sides, and it also made the bot’s profile more symmetrical. This design would allow the rear wheels to touch the ground just a little to putter me around for a quick escape. I’ve moved on to the more important parts now – the fr0k. Other elements such as the drive chain tensioner (imported straight from 30lb Clocker, mostly for placement and looks) and the clamp actuator have also been rough-placed.

This clamp actuator has quite a history. I made it all the way back before Dragon*Con 2008 for the the original Überclocker when my previous design failed in testing right before departure. Between this and the imported 6:1 gearbox idea from Test Bot, this bot is bringing back some serious history! The original 20:1 (or something) small gearmotor on it was stripped, so I replaced it with a 9:1 HP Pololu motor. This thing was the first custom linear actuator I built, which made me go crazy afterwards with the Cold Arbor actuators.

I’m going to stick a stock 36:1 HF type gearbox on fork duty since I have many of them and they can be easily converted to all-metal gearing because of the identical 6:1 stages. That and because I need all the reduction I can get for this dumb fork. A 60 tooth sprocket is the output stage, driven by a 13-tooth sprocket on the drill shaft.

Why 13 teeth?! This size sprocket is small enough, and the spacing of the drill motor serendipitous enough, that I’m just tapping it for the 3/8-24 thread on the drill shaft and threading it the fuck on there. A 12 tooth’s hub is too small and a 14 tooth doesn’t get me as much reduction.

With the almost-5:1 reduction from the 36:1 box, I’m still looking at 150 rpm on the arm shaft. This is something like 10 ft/s of linear lift speed. Fast enough to be used as a really bad hammer!

The motor will be mounted by some 3DP’d clamps that conform to the shape of the gearbox. I designed this part to be serviced just like 30lb clocker – four screws and the whole motor can drop out the bottom. No more removing 75% of the bot to fix one thing like last year, ever.

Getting close to the final appearance here. I’ve added a first-pass geometric model of the clamp arm. I literally picked reasonably random numbers for the rise and slope of the portion where it attaches to the fork. Brought it in, see how it looks, and change if needed.

This is, of course, another lesson I try to teach in Silly Go-Kart Camp – at some point, you gotta start putting your design down and stop trying to solve for every dimension analytically before you open Solidworks. Once you have a starting base, then you can adjust and optimize dimensions and placements based on constraints and whatnot. That’s why I just started throwing rectangles at trapezoids at the beginning.

Too many times I’ve witnessed people sitting while staring blankly at sketches on a piece of paper and an empty CAD screen, not knowing what to start designing first. The answer: all of it. Right now. Or I will fail you.

I went through quite a few iterations of the clamp motor placement, too, before realizing the best position is basically where I had it first. In this position, the motor is well out of the way of the action and the leadscrew length can be kept short, which in this bot is critical for weight.

Notice the Clampy Shaft Collars of Unintended Power Transmission. I’d use the stock 3/4″ flange collars on McMaster, my favorite for go-kart steering linkages, but they somehow got EVEN MORE expensive. And they’re steel.

Can’t do that on this bot. I just bought some aluminum clamp collars and will just drill my own damned holes in them, because when you average it all out I make like $9 an hour anyway and I know drilling 6 holes is going to take less than 3 hours.

Getting a little more detailed up front now. I’ve added the leadscrew attachment joint and the lower fork standoffs. The leadscrew will be anchored directly to the fork this time – it will be a little harder to disassemble if I JUST need to get to the anchor, but the screw itself will be easily disengaged with one epic set screw. This is for simplicity and also weight savings.

Clocker’s “reactive outriggers”, or its spring loaded leggy things, are kind of its trademark. They’re also responsible for the various tricks it can do to opponents, like driving while clamping them or twirling them in circles.

On 12lb Clocker, a discrete ‘shock absorber on swingarm’ setup would be unnecessarily heavy for how rigid it needed to be. So I had a bright idea of making a flexure leg using a stick of spring steel as a leaf spring.  It only took like an hour, 2 FEA simulations, 3 Wikipedia pages, and a few MATWEB searches to do what amounted to a simple bending-beam problem, but using what’s available in the McMasterBots catalog, I found that either a 3/32″ or 1/8″ thick 1075 steel strip that’s 1/2″ wide will work. The 1/8″ was a ‘worst case’ estimate of a 12 lb opponent hung all the way out on the end of the fork and the legs deflecting no more than 1/2″ at the tip.

So what do I use? Well, I just bought 1 strip of both and will make both of them. I really want to see how this bot drives on one or the other before deciding.

Attachment-wise, aluminum machined anchors hold the spring to the body of the bot, and it passes under a standoff extension of the front axle pins, starting the bend at that point. Kinda-sorta – the “roller” constraint means it will actually start bending before then, but the distance from the standoff to the anchor is short enough that I don’t think it will contribute significantly.  At the tip, another machined anchor will be bolted to the steel and have a little ball roller riding on a shoulder screw coming off it.

To machine the Rockwell C44 spring steel, I snagged a cheap F size carbide stub drill off eBay since finding this around campus will be less than likely (nor do I want to ruin someone’s $50+ carbide drill by being too enthusiastic with it!)

Quick overview shot now that the major geometries are in place. Out of a practicality concern, the fork isn’t as stag-beetle-esque as I had originally hoped for.

At this point, I imported 30lb Clocker for a size comparison. Clocker is a huge 30lber – by footprint, it’s larger than some classic 60lb Lightweight Battlebots. This is due in part to the need to contain most of an opponent, and a long wheelbase/wide track make for greater stability once contained. So Clocker has never been very dense.

12lb Clocker is a little better in that department, but it’s still pretty big for a 12lber. It’s a full 14″ wide and 9″ long if you only consider wheels, and this isn’t too bad, but this increases to 17.5″ from back of wheels to front of fork.

With major parts complete, I checked the simulated weight: 11.4 pounds including the top and bottom armor, which are not visible in this picture.

Oh boy. This is without wiring, without modeled chains, and without most big hardware. It’s time to start “gothic cathedraling“, as I call it. Since these plates are going to be waterjet machined, I only have “cut through” or “not cut at all”, so things will look pretty spider-webby when done. First to go is the big sprocket.

Before starting on the “cathedraling” proper, I designed in all the corner gussets and final top/bottom armor attachment points. The torsion box up front and the gusseted C at the rear are very similar to how 30lb Clocker is set up.

The “cathedraling” this time is mostly X trusses, nothing too special. For kicks, I decided to add a simulated chain, and boy am I glad I did.

I totally forgot that, even though there is a pinch roller tensioner in the posterior run of the chain, the anterior (front) just goes straight down. And intersects my frame rails.

A bit of modification to that area later and I was all set again, only gaining back like 6 grams.

Cathedraling complete, and now with top and bottom covers mounted, including the huge obnoxious 12:00 cutout up top.

Why 12:00? Because it’s 12 o’clocker!

I really want to get a big EL panel kit and put it behind the cutout and have it blink 12:00 like an old VCR or broken alarm clock.

The weight as shown is 10.6 pounds. I’m now within the realm of possibility for making weight after factoring in the rest of the hardware.

The last detail I needed to address was the chain drive. The tensioner I imported straight from 30lb Clocker would work just fine, except the wheel placement would have caused the bottom run of the chain to hang like 1/4″ below the frame bottom! That’s just asking to get snagged on something and broken, leaving me stranded (again).

Not allowing that to happen. I rerouted the chain over two smaller roller tensioners such that for the most part the undercarriage is clear of chain runs. If weight allows, I’ll put a 1/8″ polycarbonate or other lightweight shield over this area to prevent accidental mishaps.

Here’s a better size comparison between 12 o’clocker, Überclocker Advance, and the 2008 version of Test Bot. So I guess 12 o’clocker isn’t really that big – TB was a full 12 x 12″, plus a few more inches of wedge and lifter arm.

Since the kart class caused my updates to run behind schedule a bit, I’ll put out a spoiler: The parts for 12 o’clocker have already been machined, and I can hopefully assemble most of the bot this weekend! Also on deck is a report on 30lb Clocker’s new clamp actuator, and more info on the D*C Robotics Track panels I intend to do this year.