Archive for January, 2010


Cold Arbor Update 4: This, that, and the other thing

Jan 13, 2010 in Bots, Cold Arbor, Project Build Reports

Unlike most of my other robots, Cold Arbor doesn’t really have a cohesive manufacturing plan, mostly because I’m waiting on too many shipments. So I’m hitting the machines as components tickle my fancy – not necessarily working on the same assembly until completion.

This has led to the robot table becoming a hopelessly cluttered mess, but at least things are getting done. One month and a week to Motorama – it’s time for SRS BIDNESS.

I managed to mechanically complete Deathrunner as well as start on the drive motors and the frame.

I put in the radially symmetric features of Deathrunner using my haute usinage utility, a rotary indexing head. It’s like my very own 4th axis, if I had a CNC mill where the “C” didn’t stand for “Charles”. Here’s the set screw holes on the motor hub being power-threaded.

The Petri Dish of Motor Mounting (+6) has a slot cut into it in order to pass wires. How did I drill those cool elliptical holes in the center?

Unfortunately not with a rotary broach. I drilled once and discovered that I had transposed two numbers in my dimensions – 0.695 instead of 0.659.

Oops. So I went back over them with a carbide endmill. The extra-large clearance holes won’t affect stator mounting.

All the parts of Deathrunner are prepared. The endcap is my favorite piece of machine work so far.

Look! An assembled Deathrunner!

No windings yet, nor any case screws, just a proof of concept. The case runs true and everything is smooth.

Let’s move on to the drivetrain. Remember when I said I couldn’t find any more 24:1 drills?

Reading back through old Überclocker build reports of 2008, I remembered that I had purchased two additional 24:1 geared drills for spare parts, but they were never used.

They had to be around somewhere. A bit of excavation in Überclocker’s parts box revealed the gears of that drill pretty evenly distributed across the bottom.

So here’s the spare 24:1 first stage that I’ll be incorporating into Arbor. That means the fancy  2 speed drill hack has been ditched, and I’ll just be making two more Clockerboxes.

Every build report of mine seems to have at least one picture of a pile of drill parts. Here’s this one.

I began parting out the Handiworks gearboxes to use the ring gears and spindles. While I have a reserve of spare drill parts, I decided to use Handis anyway because their parts matched. There’s no guarantee that any two drills spindles I take out of the pile are actually the same size.

In a moment of inspiration, I used the chain tool as an improvised gear puller. This worked great for the 700 size motors that came with the Amazon drills.

Protoforms for the gearboxen emerge. I popped these off using the Bitch Chuck and some 2″ square aluminum barstock.

The Handiworks ring gears turned out to be different sizes by a matter of thousandths – one was 1.493″ diameter, the other 1.497″. That much difference is enough to throw off a press fit, however, so I actually had to make two different parts. They are paired with their respective ring gears.

Both bearing and ring gear will be secured using the freeze-and-burn method.

… and here we have it, Round 1 of waterjetting for the frame!

Arbor’s frame is all 1/8″ aluminum plate with selective 1/4″ reinforcements. This time, the tab and slot puzzle doesn’t have T-nuts.

I’m dead set on either zinc-aluminum brazing or real, legit TIG welding the frame together.

How am I going to TIG weld the frame? With the NEW MITERS TIG WELDER!!!!!

Okay, so it needs some love. As in, it needs a correctly fitting torch, a gas supply, electrodes, rods, and a 230 volt plug of a type that exists in the build space. As usual with Harbor Freight, the fittings and connections seem to be all proprietary, or extremely rare. The cheap torch we purchased turns out to use a different style of weird proprietary fitting.

Since the unit IS cheap and Chinese, I have no qualms about hard-wiring things or chopping something togethe, and that might just happen with our cheap replacement torch.

We’ll see where this goes…

Cold Arbor Update 3: Deathrunner

Jan 09, 2010 in Bots, Cold Arbor, Project Build Reports

Certainly the centerpiece of Cold Arbor is the 10″ milling saw, but circular saws are useless without motors. So I have commissioned the construction of a custom-wound brushless motor (hereafter termed Deathrunner) to drive the saw. While I certainly could use a commercial outrunner, high performance and solid motors cost some serious money. Factor in IAP’s excessive amounts of usable build time (for those of us who are enthralled with anything but academic work), and I’m better off pumping out a motor of my own design, while maybe learning a thing or two about said motor design.

The reference motor is the HobbyKing 70/55 outrunner. Deathrunner’s physical layout is practically identical, with key areas made more supple to take the impacts of battle. In this latest generation of  Inexpensive Chinese Brushless Motors (ICBMs), the distal end of the can is supported by its own bearing. This ensures that the can won’t begin resonanting at high speeds, or just… fall off.

I never trusted motors which hung a huge bell-can off a single set screw connection to a shaft. The original Snuffles motor suffered from can-wiggle to the point that the magnets started falling apart.

We begin with a Giant-Ass Steel Pipe (GASP – is someone keeping track of these random acronyms that I make up while writing 7AM build reports about the night before?)

Hey, speaking of the original Snuffles motor… here it is! I intend to harvest the stator, more or less the only functional part remaining in the motor, for Deathrunner. This motor has been completely trashed. The can is physically out of round, the shaft connection has been stripped and Loctited countless times, the bearings are crunchy, and the chrome plating is scratched, tarnished, or completely eroded away in places.

The steel pipe is 3.5″ OD and 2.875″ ID, 5/16″ wall thickness. I picked this pipe as a compromise between several motor designs that I want to build, all requiring slightly differing can diameters.

Unfortunately, this means that I had to turn down the pipe to make the 3.25″ OD can for Deathrunner. We’ve seen what happens before when I try to make a big steel pipe into a little steel pipe.

You know, the exact same thing.

The difference is that this time I had a real (sort of) boring tool. Optimally, a piece this large is finished off on a chunkier machine, but I was already at MITERS, where we’re limited on tool sizes, and didn’t really care to seek out another shop.

The consequence of hanging a tiny bit way the hell out beyond the holder is CHATTER. You can observe my amazing finish on the inside here.

I decided to just put up with the chatter because nobody should ever be seeing the inside of this motor anyway, and if nobody’s looking at it…

Plus, all those ridges ought to make the magnet epoxy stick better, right?

I cracked open the old Snuffles motor. Check out the scrape marks on those magnets – and the missing chunk.

Beginning the uninstallation of the windings. The motor is wound “DLRK” style, a symmetrical winding that results in three easy to terminate connections.

Easy to wind also means easy to unwind. Here’s the successfully recovered stator. It is integrated (epoxied, pressed, you name it) onto an aluminum mounting post.

The second custom part for Deathrunner is this… can end? Bearing holder?

It’s BOTH!

The Integrated Aluminum Post (IAP) is conveniently 40mm in diameter at its widest point, the bottom of the motor. A 6808 metric miniature deep groove bearing is 40mm on the bore and 52mm diameter.

This just screamed can bearing!!!! to me, made all the better when I discovered an Ebay bearing supplier that was even sketchier than VXB, who supplied me with…

Google Earth bearings? Kiss my leg? Kill my lizard? Lower Tanudan Kalinga?

Whatever. They’re steel, round, and slightly greasy. That counts as “bearing”.

Making the REAL can endcap. Why am I using a Bitch Chuck instead of the more convenient 3 jaw? Because the 3 jaw couldn’t hold the pressure of me cramming a boring bar full width into the face of the aluminum round. The piece jammed and rotated in the chuck, throwing off the established axis of rotation forever.

Well shit, there goes about 5 minutes of work, but it’s a big chunk of aluminum to just toss and start over. Using a 4-jaw chuck, I dial-indicated the piece back to 0.0005 runout in a moment of totally uncontrolled OCD. It’s a number that I have a feeling the Old Mercedes can’t actually hold any more, but hey.

But the other advantage of the Bitch Chuck is the ability to get the exact same centering after you take a part out and flip it around. Ordinary self centering chucks only self-center on one specific orientation of the part. If it moves, as events transpired, or you have to take things out… all bets are off.

The  Bitch Chuck requires the use of an indicator and is thus excessively bitchy and inconvenient, but I’ve become pretty adept at wiggling the jaws to the right position.

Above is a picture of the other side of the can endcap being finished.

Test fit. Because all the mechanical connections from can to hub will be by cross-drilled screws, nothing needs to be a press fit. This is great, because it means I can disassemble my test fits. It was tantamount to me discovering that I left the internal shoulder 2mm too long. A return trip to the machines took care of that.

Most of the protoforms of the motor are assembled here. Technically, this is the entire motor, but…

… it needs a mounting flange to adapt the 6 hole bolt circle of the IAP to the rest of the robot.

I’ve been anticipating Deathrunner for a while now – it was originally going to be an EV motor for the next wacky scooter. So I invested a chunk of cash in custom full-circle magnets from Super Magnet George. The thing I hate worst about motor construction besides winding 9000 turn stators is installing the magnets, because inevitably they want to snap together onto either eachother or your fingers. A full magnet circuit comes with a few electromagnetic penalties, but installs so much cleaner.

These custom magnets are N42 composition and I ordered a full circle of 14 plus two spares. They are perfect – as in, I’m sure they were made right to dimension, but I bored the can a hair too large, resulting in the ~2mm wide gap in the circle.

I suspect just the thickness of the epoxy alone will close the cap. Else, that’s what they invented index cards for.

What’s left to do on Deathrunner? All secondary features – holes need to be drilled, slots need to be cut, and threads need to be tapped. Those are all mill operations. I wanted to quickly pop off the protoforms of the parts before continuing.

Oh, and the thing needs winding. I’m settling for running the entire robot on 23 volts (7S A123s), so the motor will be wound rather hot. The original motor boasted a 176 Kv, which actually puts it in the right speed range to drive the saw to design speed. I’ll probably seek double that.

Cold Arbor Update 2: Drivetrain Shenanigans

Jan 08, 2010 in Bots, Cold Arbor, Project Build Reports

In the last Cold Arbor episode, I figured out that I really like giant sawblades and designed a robot around one. Since all the mechanical aspects of the design have now been hashed out, I set upon gathering parts and fabrication.

Lately I’ve been running into a bit of a bind with respect to robot drive motors. By that, I don’t mean I cut the gearbox cases too short, but I simply cannot locate appropriate components. Essentially all of my robots larger than 3 pounds have used cordless drill gearboxes in one form or another for propulsion. They’re (very) cheap, easy to work with, and relatively standardized.

The vast majority of “cheap” drills under 18 volts use 36:1 planetary gearboxes – 2 stages of 6:1. Problem is, there’s only so much power you can put through a cheap drill motor, or any motor of similar frame size, before it just explodes. For robots larger than 12 pounds, more power is needed for maneuverability than the average 550 size drill motor can safelyprovide. I ran into this problem in Überclocker and had to trade up a motor size to compensate.

The issue is that the solution I used in ‘clocker is extremely difficult to find now. Typical 36:1 drill gearboxes use a 9 tooth motor pinion, which at most can take a 3mm motor shaft – the size of all 500 frame motors. But again, you can only put so much power through that 3mm press fit before the pinion vaporizes.

There exist single speed drills with 24:1 gearboxes. These use a 15 tooth motor pinion in the first stage, which can be bored out to 5mm, the shaft size of more substantial 700 frame motors and the venerable DeWalt drill motors. That’s what I did for Überclocker.

Problem? I can’t find another 24:1 drill. All of the ones I have unearthed through exploring the expanses of the Internets are now horribly expensive – $40 or more. Harbor Freight’s #66965, an alleged “900 RPM” drill, is in fact incorrectly labeled. They probably don’t know the difference anyway.

So what to do now? I continued hunting for these things by looking for clues in the description found on shopping sites. Meanwhile, I started exploring other options. I recently hacked apart a DeWalt 3 speed drill gearbox, the type used in their newer lines of power tools. DeWalt being DeWalt, it was  built rock solid. All the gearshift mating surfaces are steel! Sintered steel, but steel nontheless.

Unfortunately, Cold Arbor’s design didn’t allow the use of these gearboxes because they’re too long. It would have caused the robot to be excessively wide, which adds weight. Additionally, I would have had to design a very creatively shaped housing to keep all the gears inline – the stock plastic gearcases have no easy mounting solutions, being specially designed to go in a drill body.

But examining the DeWalts got me thinking – cheap two speed cordless drill also abound on the market. These generally advertise 350RPM in low speed and 1300 to 1400 RPM in high speed, and can be had for $20-30. I decided to investigate using their gearsets, possibly locked permanently in one speed using my own housing.

My search hit paydirt (in the sense that I had to pay to get parts) when I found these widgets on Amazon for only $14.50.

It looked like they were cleaning out inventory, since the “normal price” was advertised at $40. I immediately snagged two samples for investigation. These drills joined my table of parts, along with its other cordless brethren.

Hey, are those Handiworks? Indeed they are. I bought them way back in 2004 when WalMart had them on sale regularly for $8-10. They began disappearing around 2006, though. Handiworks were the staple drive motor of many 12lbers and the occasional 3lb bot. I excavated them from my garage back in Atlanta and brought them with me back to Boston. They won’t see use on Arbor, but who knows when I need a screwdriver motor attached to a drill gearbox?

Let’s have a look inside the “Denali” drills. This thing is pretty average construction for a cheap drill, except it comes with a beefier 700 size motor. I found the crude slip rings on the battery pack receptable interesting. The 2 speed gearbox is filled to the brim with blue colored grease.

Next step is a full teardown of the gearbox.

Well this doesn’t look very promising. Check out the size of those 2nd stage gears. They are 9 tooth, and wouldn’t surprise me if they are just repurposed pinions from other lines of drills. In “low” speed, all 3 sets of gears are engaged. For the record, the gear ratio for the first stage is 3.33:1, the second is 2.81:1, and the output stage is 4.5:1.

In the high speed range, the nanogears are bypassed by sliding the toothed carrier  on the left into its ring gear. This means the gearbox probably won’t last too long in low gear if used in robot duty. Small planet gears are substantially weaker. Worse, there were no parts immediately swappable with my existing stock of drill gears. In these drills, the motor pinion is 18 teeth and the ring gears either 40 or 42 teeth. So it looks like I was left with either two duds or two half-assed solutions to my mobility problem.

Naturally, I take the half-assed way out.

I started rearranging gears in the hopes that I can come up with a meaningful ratio using only the parts in the stock gearbox. Conveniently enough, I found that the first stage planets and ring gear could be paired with the second stage carrier. This allowed me to cut out the nanogears of the second stage completely.

The result is pictured above – a hardware-level bypass of “low” gear. The resulting ratio from this arrangement is a cool 15:1, which, while a little fast, is totally adequate for the job at hand. Using the stock drill motors, I calculated Arbor’s theoretical top speed to be 14 miles per hour. That’s fast for modern 12 to 24 foot arenas.

I am still concerned about durability of the gearbox – the pins on the carriers are only 2mm diameter, which is smaller than the usual 3mm pins found in other drills. This could ruin my day if they shear off, but for now, I’m willing to put up with the risk simply because I already dropped the money on these suckers.

Later investigations I might conduct include seeing what additional gear mates are possible with these carriers. They have a set of 3mm pin holes (but no pins) on a slightly smaller circle, which indicates that they might be able to cross-pollinate with other drills.

First up, however, is to design a casing for the “hackbox”.

Hey, that looks sort of like the gearboxes I came up with for Überclocker. That’s because they essentially are, just slightly modified to accept the new internal arrangement.

A quarter section view. I spent an afternoon taking calipers to all the relevant internal gear components and modeled the whole deal in Inventor.

The next step is manufacturing. Stay tuned for more!

The January Bot: Project Cold Arbor

Jan 05, 2010 in Bots, Cold Arbor, Project Build Reports

Since the beginning of televised robotic sports history, combat robot builders have been repeatedly asked by fans, enthusiastic onlookers, and bright-eyed newbies, a variant of the following question:


Good question. Why haven’t we? Sawbots are conspicuously missing from the combat robot hall of fame. Everybody wants to see them, but there have been comparatively few saw weapons built in all of the history of robot combat, and even fewer ones which actually achieved its end goals. Minion, the most famous saw-wielding Battlebot, replaced that weapon with a kinetic energy disk starting in Season 3.

So why is this the case? Saws need a delicate balance of torque and speed actually cut something. Too much of one or the other, and you either blunt off the teeth or jam it in the opponent. Essentially all saw weapons to date have been directly driven or geared very fast from a (comparatively) low-torque electric motor. That means they can scrape and spark nicely, but actually laying into an opponent will stall the blade.

The other element missing is controllable pressure. It’s difficult to cut consistently in the same place on something that’s flailing around or moving. A saw that sticks out the front of a robot but is otherwise fixed has no control over how quickly it can advance teeth through opposition armor. If it’s simply jammed into the opponent, then the torque required to actually crank the blade teeth through the material becomes phenomenal. Most saw weapons extant today rely on the stored kinetic energy in the saw to deal out damage.

This particular aspect was what inspired me to think about the Sawbot Conundrum when I was watching old episodes of Robot Wars featuring the house robot Dead Metal. Observe how effective the saw is on Dead Metal from the 2 minute mark onwards here. Dead Metal works because it is able to trap the smaller robot in its pincer jaws so they can’t move. The saw also advances on a linkage at a modest pace and has plenty of horsepower to back it up. It can actually cut things.

I immediately wondered if I could replicate the function in a smaller class, say, 30 pounds. Control of the opponent is the dominant trait of Überclocker, but it doesn’t really have a means of doing… well, anything else. I wanted a complementary Überclocker which trades off some of the control for more ownage.

In an alignment of mechanical planets, I was swiping some more slitting saws for MITERS off Ebay when this mutant blade showed up in the listings. I didn’t even know they made “slitting saws” this large. It’s a 10 inch blade, 3/16″ thick, and weighs a solid 4 pounds. The bore is 1.25″ and keyed.

Now I was practically obligated to build a robot around it.

To drive a saw this large would require immense torque. I investigated planetary and spur gear reductions before deciding that neither could satisfy the torque transmission requirement, be minimal in weight (spur gears were totally outclassed here) and also be easy to build or cheap to buy (there go all planetary gearboxes in existence).

Luckily, I remembered that I had…

…this worm gearbox that I scrounged out of a pile of discarded lab cleanup materials. It came of the base of a nicely waterjetted and assembled robot arm that looks like something heavy ran into it – or, more likely, it ran into something heavy. Either way – free 30:1, 12 (ish?) pitch single-enveloping worm gear.

While worm gears are notoriously inefficient, it’s much easier (in today’s world of brushless one-upsmanship) to dramatically over-motor a weapon to compensate for that. Worm gearboxes are probably the most common things in industry, also, because of their simplicity, compactness, and durability. If you don’t mind the less-than-peak efficiency – which \m/assive \m/etal generally doesn’t – then why bother with something else?

There do exist machine tools which use worm gear driven circular saw blades to cut metal, generally competing with bandsaws in situations where edge finish and accuracy are important.

They are known as cold saws, named for the fact that they use a toothed steel blade to generate chips of material instead of a disc of abrasive rotating at high speeds. I like them better than bandsaws, because I can hang off the handle on the largest cold saw in the shops and take down a 4″ solid round of steel in under 30 seconds, and that’s just too awesome.

Thus the inspiration for Cold Arbor was complete. The name is both a play on cold saw and the Battle of Cold Harbor, one of the “bloodiest, most lopsided battles” of the American Civil War.

Anyways, I love my found object gearset, but that janky plastic case has to go. Why put such nice gears in a plastic case?

Oh, right – nobody is expecting me to hang a 10 inch sawblade off one. Using the existing mounting dimensions, I designed this preliminary gearbox case, to be made of aluminum. Some 7/8″ bore ball bearings will take the place of the bronze bushings. Since the shaft is hollow, a large bolt running through the center makes for easy blade mounting.

So here’s the first try at the actual blade assembly. This is more inspired by a sliding compound miter saw. The premise here is to reach the blade as far into an opponent as possible. Using a plain swinging saw, the maximum “cut depth”, so to speak, is limited to the saw radius minus the radius of the gearbox. This design was an attempt to remedy that by allowing the saw to “reach” further by sliding.

The prelim design more progressed, showing the prospective linear bearing rails. This was about as far as the design ever got before I realized it was going to weigh far too much, and would suffer from stiffness issues. So that was scrapped.

The truncated circle floorplan begins to make a return here.

Crazy design number 2 uses a six bar(!) linkage that allows the saw to retract fully into the footprint of the robot, but extend about 6 inches forward. The swinging motion bring the saw over, then down upon the opponent. The whole thing would be moved by 1 actuator, mounted at the bronze bushing and causing that part of the linkage to move about 70 degrees.  The motor would remain parallel to the ground at all time.

A much more practical idea than the sliding design, but ultimately, it also faced some pretty hefty weight issues, especially after I designed in the actuator.

Alright, big leap of faith here. What the crunk is going on?

I decided to give up on a cool swinging linkage and just made  a conventional linear actuated pivoting assembly. This linkage swings through just over 75 degrees of motion to bring the saw all the way to the ground in front of the robot.

The motor mount is integrated into the body of the gearbox to make it easy.

One additional component visible here is the actuator for the clamp. Both clamps are connected to the single linear actuator at the rear, so they should move (sort of) together. They are linked to the actuator nut with a ball-jointed tie rod.

A clearer shot of the clamp actuation. Notice the top claws have been replaced with a fancy to-be-welded structure.

While the claws actuate together, both clamps (top and bottom of each side) are capable of independent swinging motion. This should allow me a bit more flexibility in approaching opponents.

A bit more progressed now. Top and bottom cover plates will be the usual order of 1/16″ Garolite reinforced in select locations.

The saw in full deployed position. The bot can actually reach far enough to cut itself. To prevent this from happening, the leadscrew length in the actuator will be tuned appropriately.

The frame of the bot will be sort of a unique first try for me. It will either be zinc-aluminum brazed or, with luck, be legitimately TIG welded. Either way, my first foray into “permanent joinery”, so to speak. We’ll see how it turns out. The robot just didn’t have the space to put T-nuts int the frame.

The frame is four pieces – front, back, and the two sides.

Filling out the internals a bit using AUTODESK INVENTOR 2010!!!! on my NEW LAPTOP!!!!. The fancy welded flaw is gone – I figured four fancy welded assemblies were enough for a first shot. They have been replaced by a single flat aluminum top claw. Less rigid, but also easier to make and assemble.

I plan to let this bot use the same batteries as Überclocker, since I already have packs made. 7S A123 cells yields about 23 volts nominal and 25 volts peak, which is roughly the tolerance level of the Victor 883 controllers anyway. I have leftover 883s and 884s from Clocker.

Electronics will be mounted on “cards” in the body. Not really for quick access and removal, but there wasn’t a way to mount everything horizontally.

A beauty shot of everything so far.

I discovered the animation module of Inventor and briefly played around with it. Here’s a cool CAD video of the robot mechanisms in action.

The schedule for completion is “ultra fast-track”. I managed to put together most of Clocker in two weeks, so this robot should be no different. Let’s hope I’m right.