Archive for February, 2010


Cold Arbor Update 10: The Week of Terror, Part 2

Feb 17, 2010 in Bots, Cold Arbor, Project Build Reports

With Deathrunner cooking on the radiator and the robot mostly assembled, I started attending to the small but significant mechanical details. Making parts line up has been a challenge in this build because of my inability to fuse metal accurately. What made things even worse was my phenomenal ability to design fastener holes in places that they cannot physically be installed.


Well, this was certainly bad, because several of these locations were critical to the structural integrity of the robot, such as…

This long cross bolt holds together the base of the saw arm and attaches it to the front half of the robot frame. It also bolts through the gap in the middle. Without this, the saw has very little rigidity, and probably will just bend and break at the gap.

To install it, I had to thread it through the base and into the set of mounting ears next to the gap, then bend the aluminum flanges, install a nut, then bend everything back and lock the base into its T-nutted slots. Tightening the bolt straightened the ears out again.

The braze joints did not break when I did this, so it at least holds some promise.

The saw and its hub are fastened to the worm gear by a Giant-Ass Bolt, namely a 5/8″-11 grade 8 bolt. There is no mechanical coupling, only the bolt pressure, so it should provide the first line of clutching action.

This bolt weighs ALOT. I might consider hollowing it out some.

I dropped the rear actuator motor into place, routed the chain, and backed the linkage nut on.

The leadscrew in the clamp actuator is a live rotating shaft, and the sprocket is retained by set screw and flat.

The clamp links themselves are sections of 1/4-28 threaded rod with ball links on the end.

I made some fat Delrin washers to keep the claws in alignment.

Here’s a view of the ball link attached to the claws. I ended up having to space the ball links away from their respective mounting surfaces a bit. As it turns out, the ball links did not have that much angular flexibility. The threaded rods also ran into the frame.

And both links are installed.

Now, there is technically nothing preventing the nut itself from rolling side to side because of the torque exerted by the actuator. I’m sort of depending on the clamp arms hitting something evenly. When only one arm is loaded, or if they are run with no load or into the ends of their travels, the nut does tilt far to one side. I have yet to get the system to bind, but if it becomes a problem, the nut will get a “slide plate” or some other planar joint to keep it in line.

With most of the mechanical subsystems in place, I dropped the EBays into the robot to test fit and placement.

Result: Spot on, but the T-nuts sticking out of the side of the polycarb caused some alignment issues. No problem – I just drilled out the mounting holes in the baseplate slightly larger.

The last random part that I needed to make was the shaft coupler to link Deathrunner and the saw blade gearbox. It’s a relatively simple round aluminum part with two holes.

The two 5/16″ set screws seat in flats on both shafts, and are liberally Loctite’d. I think they will be fine – but I need to test the saw to check the integrity of the gearbox.

Deathrunner has been (hopefully) permanently attached to the gearbox. The saw now turns whenever I move the motor!

I decided to try actually mounting the top plates to see if those holes will line up. In another example of my unbeatable fastener placing skills, I shoved a screw right next to the saw, so close that a drill with a screwdriver bit can’t reach.

I couldn’t find the the long reach modular screwdriver either. Or a ratchet extension bar. So in a time of desperation…

Hey, it worked. The top plate holes are mostly aligned without incident or interdiciton.

With the robot superficially complete, it’s time for…


At this point, the robot is only missing electronics and the four custom sprockets, which I need to either waterjet or hack around Real Soon™

I’m still waiting on Dealextreme to deliver my ghettodeans connectors so I can fill out the electrical bus.

Otherwise, the robot isn’t that unfinished, but it’s certainly further behind than I had planned, and very much untested.

Here’s a size comparison with Überclocker. Overall, Arbor is smaller in footprint and a bit taller.


Cold Arbor Update 9: The Week of Terror, part I

Feb 16, 2010 in Bots, Cold Arbor, Project Build Reports

I realized I’ve accumulated almost 40 pictures over the past week or so. That’s quite a bit too long for just one build report, so I’m splitting the more recent work into two halves. The bottom line is that the robot is done.

… no, not really, but it’s close! You know, just like the last 4 or so build reports where it was “close”.

Let’s start with a picture of machine abuse.

Actually, this is just a picture of possible machine abuse. The big 10″ milling saw I bought can in fact fit on two arbors: one Cold Arbor, the other an R8 arbor for the mill.

…so, if I ever wanted to slit something serious, or just blow up the Bridgeport spindle, then I have the tool to do it with!

All kidding aside, let’s drill the wheels out. These have a bolt circle so they can be fastened to the hub.

I was concerned that fixturing to half of a gumball-hardness rubber wheel was going to result in it flying across the room after ricocheting off my face. But, since polypropylene machines like air, the drill did not seize or twist the wheel out of the fixture. All four wheels were done In A Jiffy.

I turned my attention to assembling some of the clamp actuator parts. The actuator is quasi-integral to the frame, so I had to perform some interesting sanding and grinding maneuvers to get the assembled structure to slip into the back frame rail.

A little bit of frame assembly, now with clamp actuator bracket and all four wheels.

…and now with both actuators.

I machined some little round Delrin spacers to take up the axial slack on the swinging saw arm assembly. The hinge point itself is just aluminum on steel, but honestly – nothing here is moving fast enough to warrant a real set of bushings or bearings.

I made an end block for the actuator leadscrew and went ahead with a test assembly of the saw arm. Due to using a longer bushing than I intended to originally, the saw doesn’t swing back into the robot quite as far as I anticipated. This can be corrected, but wasn’t critical enough to warrant taking everything apart.

Here’s the actuator test mounted in the robot.

I guess this is still a Pretend-o-Pile, since there’s not enough there to be a robot just yet.

Alright, time for the last round of waterjetting! With the frame itself complete, I needed to start filling out the interior. Here are the electronics mounting facilities (the Ebay, so to speak) cut out of 1/8″ and 1/4″ polycarbonate, with plenty of T-nuts to put it all together.

The single 1/8″ aluminum plate will mount the two actuator ESCs, some Dimension Syren 25s and provide an additional heat sinking surface for them. They mount over the Victor 883s which I have on tap for drive controllers. The rounded cutouts in those plates allow me to access the wire terminals of the Victors without taking apart everything.

Finally, the top and bottom plates!

These are made of 1/16″ FR4, which I seem to favor across all my bots. I like it – it’s relatively cheap and stiff, and also nonmetallic. The one downside is that it does not waterjet very well. The layers tend to delaminate on a pierce due to the jet pressure forcing abrasive between them. To mitigate (but not completely avoid) this, I usually cut sheets of laminates over a waterjet brick or some kind of scrap backing material.

Delamination was especially bad on Überclocker’s plates because I wasn’t able to find backing material in time to catch an open spot on the machine. However, for Arbor, they came out much better. I clamped the 2 foot square FR4 sheet to a corresponding 2 foot square piece of 1/2″ plywood first. There is only minimal blowout around holes and locations where the material width is small.

That also yielded some conveniently robot-shaped 1/2″ pieces of plywood. Hello, supplementary anti-hammer armor.

Some preliminary assembly of the Ebays. This is the right side, which houses all the important plushy electronic bits. The left side houses only the battery and relevant electricals. If you’re wondering what the big gaping hole in the top plate is, that’s a battery access panel which will have its own 1/4″ polycarbonate screen over it. This is so I don’t have to take out 9000 screws just to replace the battery.

Back to random mechanical bits. For the longest time now, my 28mm Banebots gearmotor from Überclocker’s actuator has been sitting unused due to thrown motor windings. Kind of a shady failure mode, but it was convenient to have on hand for Arbor’s clamp actuator. To use it, I had to replace the motor.

I found a 380-size motor that was wound for 18 or 24 volts – an unusually high voltage for such a small motor. Generally, motors like that are wound for 3 to 12 volts.

Running a 18 volt motor would bring the clamp speed and power down to reasonable levels.

I pulled the pinion off the stock motor and smashed it onto the new motor. Since I took the brute force method of disassembling the gearbox to begin with (due to the low quality Philips head screws, which just sort of stripped the instant you tried to untighten them), I had to find replacement screws.

Fortunately, I was able to locate the #2 screws in a stockroom, so I was in business again.

Alright, time to actually build Deathrunner. I’ve just been twirling an empty shell up until this point. I broke out the epoxy and custom segmented magnets and cleared the work table of all ferrous things.

This was the quickest magnet installation ever, in part because the magnets are already shaped in a circle. So it was just a matter of dropping them in while covered in epoxy.

I put the magnet can on top of a radiator to cure.

Instead of watching glue dry, I made the motor shaft. While a 10mm shaft has been pictured in just about all Deathrunner photos so far, it’s the stock shaft from the original outrunner motor, and is not only too short, but very dinged and scratched from countless slipped set screws.

I ordered a short section of 10mm precision ground shafting from McMaster for the new motor shaft. All I really had to do was put 3 flats on it, which was a short operation.

Look, it’s an assembled, fully magnet’d Deathrunner!

Installing the stator was originally going to be controlled and gentle, but the magnetic attraction between magnets and stator was so strong that it sucked the stator out of my Controlled Stator Installer (i.e. a set of channel-lock pliers) and just sort of crammed itself together in what must have been less than 100 milliseconds.

The impact moved the ring bearing 2mm into the motor, but this isn’t enough to disturb motor operation.

So, I have Deathrunner fixed to a quasi-static surface. What’s next?


This totally legit testing rig involved two battery packs and all of Überclocker. I didn’t have a spare receiver, so needed one to test, and the large controllers I have require 5 volt power to run. The only system that satisfied all these criteria was Clocker.

Conclusion: I approve.

The estimated motor speed is around 3,300 RPM. This puts the saw at a no load speed of about 110 RPM, which, while fast for a cold saw, is good enough for a robot weapon.

Alright, enough random fun. Time to get back to assembly. Here’s the clamp actuator motor situated in its mounting points. The clamp actuator also uses sprockets in lieu of gears.

I made an aluminum block to embed the leadscrew nut in. Eventually, the clamp linkages themselves will attach to this block, so the nut may actuate them both at once.

The pin was put in after I realized I failed at making a good press fit, and the nut spun in the block.

With most frame-dependent components in place, I started mounting the bottom cover plates.

This was actually a bit nontrivial. Variances in fixturing while I brazed the frame pieces together coupled with inconsistent sanding and possibly even heat warping meant that NOTHING LINED UP THE WAY IT SHOULD HAVE.

Alright, so at least half the holes could be used. The other half? I had to clean out with a clearance drill for #6 and #10 screws first, then…

…bend some of the frame pieces into shape with a spreader clamp. Apparently, the front and rear rails are very slightly curved inwards, which supports the thermal warping hypothesis.

That’s fine, I bent them back out a bit, and all the holes lined up the way they should.

Both baseplates attached, now right side up. It’s starting to look a bit more cohesive now.

I popped off the saw blade hub using a 2 inch aluminum round. The offset round hole is for a dowel pin, which will keypeg the saw blade.

The orange appearance of the aluminum is due to my choice of layout fluid, namely Sharpie markers.

A wild Pretend-O-Bot appears.

I threw all the frame pieces together and mounted the saw. Very few things are actually fastened down right now, so this will all be knocked down again for more piecemeal work.

The Pretend-o-Bot from the other side. Deathrunner was still cooking at this point, so it’s not mounted.

Stay tuned for the next episode, where I recap all the events of…

… yesterday.

Cold Arbor Update 8: It’s Getting Close

Feb 06, 2010 in Bots, Cold Arbor, Project Build Reports

The event, that is. Not the robot.

With term now well under way, the time I have to actually work on the robot has become more limited. Fortunately, there’s not that much grunge fabrication left to do on Arbor. It is, however, nowhere close to actually running. Major assemblies have been completed and most of the remaining tasks are just connecting the dots. Pictures of the dot-laying itself are below.

While waterjets are good, they’re not perfect, especially for sub-thousandths sensitive applications like ball bearing mounting. Ball bearings don’t like being compressed too much by interference fits – they get tight and a bit crunchy. Both bearing holes on the saw gearbox side plates were about 4 thousandths undersize, which is far too excessive for a press fit.

Normally this would be a simple boring operation to resize the hole, but it was greatly complicated by the fact that my brilliant design skills left no two continuous parallel edges to fixture to in a vise.

So the 85 pound milling vise had to come off. I elevated the piece on a scrap piece of 1/4″ aluminum and used that as a landing spot for the boring head. Otherwise, it was a smooth operation.

After cleaning up the bearing bore, I hopped on the 30 ton hydraulic press and slammed the bearings in. They are R14 double shielded ball bearings. Here’s the gearbox assembly just slipped together for visual effect.

I threaded the perimeter holes for a #8-32 screw. This was actually a hackaround for my neglect in laying out the 2D file for the threaded vs. nonthreaded side.

The threaded side, obviously, was designed from the outset with smaller pilot holes. Unfortunately, they all ended up the same size, which was a #6-32 clearance.

To remedy this without using nuts, I had to slightly enlarge the holes in the other gearcase parts and then use the existing #6-32 clearance hole to tap a #8-32 thread.

I fell into the trap again.

Remember a little while ago when I discovered that my 3/8″ (0.375″) aluminum was in fact 10mm (0.393″) aluminum?

I forgot that again, and designed everything around 3/8″ plate. Problem is, that just won’t do for the worm gearbox, which needs pretty precise spacing and centering of the worm gear to properly mesh. So the solution was to shave down the middle plates of the gearbox, which determined the bearings’ axial spacing.

A quick trip with a fly cutter reduced the thickness of the middle plates to 0.374″. I was concerned that the setup was going to be unsound, since I was effectly clamping a ring between two flat plates, but things turned out fine.

Uh oh.

What you see here is the center plates failing to fall in line properly. There are a total of four plates between the two large gearcase sides, two 3/8″ thick and two 1/8″ thick.

There are three reasons for this happening.

  1. The 10mm aluminum vs. 3/8″ aluminum quandary. While  I had fixed the middle two plates, by the time I discovered the discrepancy, I had already installed the bearings in the side plates, and didn’t feel like going back and uninstalling them. Additionally, trying to face down the side plates would have been an incredible pain.
  2. I reversed the bearing orientation relative to the worm gear. If you refer to the rendering of Arbor, notice that the ball bearings stick out of the gearcase side plates. This required substantial shimming of the worm gear between them, something that I did not account for when I placed my first McMaster order. To remedy this, I investigated reversing the side plates so the bearings pointed inwards. Each bearing then should have needed only 0.025″ of shimming.
  3. …if the aluminum plate was actually 3/8″ thick. Nope. The difference between 20mm of aluminum and 3/4″ of aluminum is about 0.04 inches. So, the interior of the reversed side plates were already 0.04″ too close. Now, while the two 10mm plates in the middle would have offset that and made things work again, I already cut them down to 0.375″. That means there was about 0.08″ less width than there should have been: for a grand total “leftover gap” of roughly 0.16 inches.

This meant that the two 1/8″ side plates couldn’t be installed.

I sat there scratching my head for a little while before deciding to just keep shaving the 3/8″ plates down until they accounted for the discrepancy. They had to be symmetrical in the middle to fit the worm shaft between them, so I couldn’t just use one or the other, etc.

Each side plate ended up at about 0.330″

All said, this took about 0.08″ of width off the whole assembly. Fortunately, the design is flexible enough to accommodate that change.

Here’s the first bolt-through of the entire worm gearbox assembly. I couldn’t locate 2 inch long #8-32 screws, so some dinky philips head deals had to take their place for now.

While everything was bolted together, I decided to bore the worm shaft hole in situ. I just stuffed the whole gearcase into the vise using a step block between the “ears” to hold the pressure, then drove a 0.5″ cutter straight through in the appropriate location.

Observe, the worm gearbox. The way the worm shaft is installed is slightly nonconventional. Each center thick plate has half of the 1/2″ hole in it. These envelope the 1/2″ shaft bushing that is stock to the worm gearbox. The rectangular extension of the gear casing is exactly the width between the thrust bearings of the worm shaft, so it fits in snugly and is captured both axially and rotationally.

With the worm gearbox essentially done, I went back to work on the other subassemblies of the robot.

Here’s the actuator of yesterupdate, but linked with chain and filled with gears. It worked like a dream after some selective shaving of material from the C-shaped output casing.

Essentially, using an even (14) and odd(11) tooth chain in the same powertrain will result in circumferential distances that are roughly one pitch too long or short. While they sell “half link” chain connectors, they don’t do half pitch ones, because those are physically impossible.

So, I was left with a chain that was just loose enough to occasionally grind and lock against the side of the casing because the chain links exhibited a limited amount of cam action.

Well, that isn’t any good, so I stripped off some of the wall thickness where the chain tried to bunch up. Problem solved.

I knocked off more protoforms of the wheel hubs using Delrin. The front wheels won’t be taking direct torque from the motor, so I elected to make them from plastic to save weight.

Using the leftover 3/8″ precision-ground 12L14 rod from Überclocker (the one that machined like aluminum), I popped off most of the small standoffs and dead axles in the design. Most, because I haven’t designed the rest yet.

I see where this is going.

Here it is. The 100th build picture of Cold Arbor and…

… what IS that? That’s not even robot-shaped. That’s a Pretend-O-Pile. It doesn’t even qualify for Pretend-o-bot.

I need one more trip to the hardware store and some more machine love before the robot can “stand” on its own, so to speak. Then, I need to cut the cover plates and electronics mounting facilities.

No, I need to design them first.

This will end well….