Category: Überclocker Remix

Überclocker Remix: Round One

chorl

Oh SHIT

What do you mean it’s almost mid-August already? Why do I still have this big carboard box with McMaster baggies in it? Where the hell is the robot? Why haven’t I even finished the robot’s design? What the hell have I been doing all summer? I’m competing in a month.

This needs to be fixed.

The first round of parts! I’ve officially entered another game of “design the back half while I build the front”. Let’s hope this attempt actually produces something meaningful.

Here’s the main fr0k gear, along with a pinion.  The little sprue-y things hanging off the inside are tabs, or small discontinuities in the cut line to keep the material together so scrap or parts don’t float away.

Day 2, round 2. I only have so much access at a time, since otherwise I start getting in the way of legitimate research work. Most structural parts have been cut out here. Amazingly enough, I ran out of 1/4″ aluminum, after going through two whole 12″ x 24″ plates.

While I could have theoretically stuffed everything onto two plates, I got into the habit of cutting one part at a time (minimzing nesting) after losing half of a plate to “buoyant part syndrome”.

The automatically generated machine tool path tends to jump between closed loops that define parts, as well as cross over previously cut features. Great and all to shorten the path traveled, but on several occasions, a small piece of scrap (a hole core, the center of a truss, whatever) is pushed upwards by the nozzle’s powerful return current from the bottom of the machine. Of course the cutting head then moves near this already-cut feature, and runs into it.

One of three things generally happens at this point. If the stock is well-fixtured and the machine is performing a fast traverse, the nozzle explodes, you cry, and the shop managers get pissier than normal because nozzles are expensive. If the stock is fixtured and the machine is performing a slow cut, the axis servo becomes loaded down from pushing against the errant scrap and the controller indicates a fault. The machine stops operating, you become confused, and the shop managers become pissier than normal because resetting the controller is a PITA.

Or, if you’re me, and fortunately have only watched hapless peers experience the former two conditions, your stock gets bumped a subtle, unnoticeable amount. The waterjet then continues making the rest of your features using a completely cocked-up coordinate system, ruining everything.

I could have avoided alot of this trouble by manually routing, but that takes patience and effort. We can’t have that.

Parts of the puzzlebot come together. Tools needed for assembly: Allen wrench, belt sander, small needle file, soft-headed mallet.

Here’s a shot of the T-nut arrangement. Essentially, the orthogonal nut and slot imitate an end-tapped hole in the metal. The arrange is not as strong as a real tapped hole because of the thinning of material around the nuts and the sharp corners focusing stress, but man does it come together fast.

Slots and tabs take most of the physical loading, as well as positively locate parts. The nuts are just there to keep the thing together.

More detail of the (excessive, gratuitous, symbolic-of-laziness) t-nutting.

Waterjet weirdness. For one reason or another, it started losing cutting ability on progressive parts. I suspected an abrasive flow problem, but since it isn’t my machine to bum around with, could only continue on slower settings and call in the issue in the morning.

Even more weirdness. I could not explain some of the vertical roughness of the edges on the previous part until I saw exactly what was happening on the next one.

It turns out that my plate was bowed in the middle to begin with. Furthermore, I only clamped it down at the left and right edges. This meant the plate was a rigid body bridging two thin steel waterjet support slats, which is a form of spring-mass harmonic system.When the cut was parallel to a steel slat, the tiny sideways force of the nozzle quickly became amplified, and the whole thing started oscillating. I only noticed after little standing waves formed on the water surface.

The edge was fundamentally straight. A few seconds of sanding completely removed the undesired feature.

A bit more puzzlebot. Note the lettering on the side. I tested two “marking” functions of the waterjet – etch and scribe. Etch uses abrasive, scribe does not. While “etch” would have made the lettering more visible and made a deeper impression, the variation in nozzle speed when turning corners in the lines meant that there were little divots and deep spots everywhere.

Scribe did not have the problem, but instead, made everything too light.  But it emulated more what a legitimate scribe tool would do, or if I had blackened the area and put it on a laser cutter (light burning and evaporation of the metal surface), so I chose to use Scribe on the robot.

Day 3, round 3. Essentially all structural parts except for the back plate are done. The puzzlebot pile grows some more.

The same issue rears its head again. The first few parts go fine, then it went downhill from there. This is just a level of weirdness I have never seen before. This part is still good, however – just needs a pass with a 3/4″ reamer.

…but this is just barely. Wow. This is what abrasive waterjets did in the 1980s. I actually had to hammer the part out of the plate – that explains the broken flashing around the edges.

However, a minute on the belt sander and a few more with a file and it was good again.

I’ve taken a liking to using the waterjet to create custom sprockets and gears. I’ve found that a standard .032 nozzle can, without taper compensation (i.e. an expensive but badass 5-axis head), cut down to roughly 12 pitch gears before the taper and surface irregularities get nontrivial. With a micro-nozzle, I bet finer gears can be made.

It is more a boon for sprocket-making, since sprockets are fundamentally plate-shaped anyway.  An added bonus is the ability to make custom bore features, such as the double-D bores here.

An early stage pretend-o-bot. The whole robot frame minus the fr0k assembly weighs around 5 pounds.

I finally got off my ass and started making parts. This is the basic configuration of the main fr0k. Instead of cool cutout plates spanning the tines, I’ve elected to just use large standoffs and high-strength threaded rod and nuts to bind the structure together. This should actually result in a stiffer structure as a whole because of the ability to put all the material in the middle in compression at once.

Top fr0k actuator leadscrew nut. This assembly will function like the last one, except everything is bigger. When I disassembled the actuator from Überclocker, I found out that the 3/8″ acme screw was actually bent slightly. This version upgrades the whole leadscrew assembly to 1/2″-10 Grade B7 acme screws.

I found some acme nuts for cheap on Ebay. Sure beat the hell of McMaster demanding 35 dollars for each bronze nut. As much as I love them, I prefer to be non-bankrupt.

The gear was de-hubbed and bored out, then crammed onto the nut with plenty of green Loctite in the middle. The leadscrew nut has a shoulder to seat the gear. Actually, it has two, because I messed up the first try.

Alright, that’s it for now.  Here’s a pile of robot-looking things.

Still left to do:

  • Make the drive gearboxen and hubs, including
    • Cut out the drive gears
  • Make the Integrated Dual Frankenb0xen
  • Finish the leadscrew actuator
  • Design the electronics enclosure
  • Make the electronics enclosure
  • Design the top and bottom plates
  • …make the top and bottom plates
  • Panic
  • Panic
  • Panic
  • Panic
  • Panic
  • Panic
  • Panic
  • Panic
  • Panic
  • Panic
  • Panic

Hey, what ever happened to Überclocker?

chorl

It’s late July. Why haven’t I seen a robot post yet, Charles!?

Well, too much LOLrioKart and lots of time taken up by my job building Terminators. But ‘clocker has been in the back of my mind for a while. With Dragon*Con and Robot Battles now a month and a half away, though, it’s starting to climb up the list really fast.

In the last Überclocker episode, we left off with the last known CAD drawing of the whole robot.

This was designed back in November-December of last year. Now that 2.007 has come and gone, and I’ve explored building puzzlebots more, I hate this design. There’s too much unsupported sheet metal. Too many things being bound together with a compression member that are otherwise supported, too many parts, and too many awkward assemblies. The overarm mechanism is also shady. But at least the clock logo on the top is cool.

Nevertheless, I needed to use material more effectively, so I opted to redesign the frame with large, T-nutted members alongside interlocking tabs.

So here’s a picture of the rework in process. The drive base part count was reduced 25%, and this goes together alot more like my other robots. Otherwise, everything remains the same – gear and chain-based 4 wheel drivetrain, and the Integrated Dual Frankenb0xen™.

I elected to switch from 4.2AH lithium polymer batteries to a small cluster of A123 26650s. A123 cells are renowned in the model and hobby robot industry for almost absurdly high performance (pulse discharges up to 60C and continuously to 30C or more) at the cost of a bit of energy density. The biggest upside is their comparative nonfinickyness.

The pack configuration I chose is 7S1P, or about 25.2 volts charged and 20 volts discharged. This is the closest match to my existing 22.2v electrical system.

A leap of faith later, and the bot is mostly done. The overarm is still from the last revision and is just there for visual effects.

The fork has been drastically simplified. Instead of cut plates, I decided to go with plain standoffs retained with allthread or long bolts. It turns out that using these standoffs with the binding pressure of good bolts is actually stiffer than the spans of aluminum I have been using in the real bot and the designs. It has the highest stiffness-to-weight ratio of the options.

But now it won’t look as badass.

The new overarm designed. Looks kind of like the old one, eh?

It’s slightly narrower and not as “huge”. The width is because it precisely fits the clamp actuator in the middle with little room to spare. Also, the overarm no longer hinges from the main fork axle, but rather is a fully independent assembly. This will save assembly time and simplify field repair, since I won’t have to rip the entire front half of the robot off to service one part.

And here’s the clamp actuator. Gee, it looks kind of like the last one.

That’s because it is. But it’s better. And won’t suck – I promise.

The difference lies in the fact that this one is substantially larger. It uses a Banebots 28mm planetary gearmotor driving a leadscrew nut, which rides up and down a stationary and larger leadscrew. While normally I shy away from Banebots equipment, their newer motors seem to be decent and the output shaft won’t be taking any direct shock loads, only gear torque. They’re also available in a reasonable voltage, unlike the next best option (screwdriver motors) and aren’t enormous (drill motors).

The leadscrew will be firmly (read: either threaded, pinned, or set-screwed-with-big-flats) embedded in its own hinge block, so we won’t have the problem of the clamp arm falling off and jamming the leadscrew under the robot.

I vouched to retain the leadscrew design because of the degree of isolation it gives between the arm being forced upwards and the force on the motor. The clamp is an enormous moment arm for anything on the end to push against, and if there were a direct 1:1 rotary coupling between the clamp arm hinge point and my actuator, I could very well blow something up.

Additionally, the fork is attached to the main shaft using Ginormous Death-screws. Like set screws, just more hardcore.

On the other hand, having the leadscrew means the only way the mechanism can be defeated is if the leadscrew bends, the threads in the nut strip, or the arm breaks in half. I’m ruling out the possibility of 5/16″ diameter hardened steel shoulder screws shearing. One of these shoulder screws might be replaced by a nylon or Delrin rod as a last-ditch overload lifesaver.

The frame, demonstrating gratuitous T-nutting. I purchased square nuts from McMaster expressly for fulfilling this role. It’s like an end-tapped hole, but there’s no drilling, tapping, or setting up the part which is inevitably impossible to hold properly as you drill it.

Another time-saver on this robot was my discovery of shaft collars with bolt circles. Seriously – shaft collars with threaded bolt holes in them, to attach something to a rotating shaft. That’s called a hub. If you only tighten the shaft collar a little bit, it’s called a clutch. If you combine the two, it becomes a torque-limiter.

So I have simplified the Great Cone Clutch Clusterfuck to use this one awesomecollar. Currently, the main fork shaft uses two shaft collars acting as clamps over a split tube to transmit torque to the fork, to great effect.. I expect a direct connection to be no different.

Covers. The clock logo is not yet laid in, but the main feature of the top plate is visible. It’s actually three plates. The center one is directly over the battery pack, so I can get to it and swap packs very quickly. Because the pack is only 2.3Ah, I’ll probably need to change batteries (as opposed to just charging in the bot) for quick turnaround times.

Either way, removing 9000 screws to get at one loose wire in Uberclocker was too much of a PITA to repeat.

And so it begins.

You’re looking at about $300 of McMaster hardware and $200 of other crap (wiring bits, controllers, motors). I don’t know how these things got so expensive (damn economy, etc.), or maybe now I’m just building real robots or something, because nothing I used to build cost this much.

Maybe this is just a symptom of my encroaching laziness. Hey, why build things when other, smarter and more productive people have figured out a way for me to buy stuff from them?

I got a nice enough deal on the 4 Victor 883s that I’ll probably end up using them in the bot, even though they are enormous and will take a pretty good amount of stuffing.

It’s time to excavate the robots and prep them for rework. For the past year or so, the bots have been on the bottom shelf of a multi-deck pushcart upon which I have heaved all my spare parts, random cruft, metal billets, and half-baked projects. They were really really dusty.

I’m actually considering keeping most of Überclocker 1 assembled, because there’s technically nothing wrong with the bottom half of the robot. The frame and running gear are still functional, and technically the lift gearbox still works – one motor just smells weird. Since I’m not really reusing any parts at all, it would make a good “audience bot”. Hell, if the upper clamp arm is removed, it works fine as a spatulabot.

Alternatively, anyone want a half-robot? I wouldn’t mind selling the current build to generate more parts money for the current one. Sans receiver, but including 2 Victor 883s, a Victor 36HV on the fork motors, and the LiPo batteries (with external balancer connection using a DB-9 connector).

The chassis metal is on the way, and with my waterjetly ways, the frame ought to be assembled by next week. Kind of good, considering next week is the last week of July.

Bot…on?