That’s it! The event is over once again. If by my lack-of-enthusiasm-induced time delay is any indication, I didn’t win anything. The robots performed about as much as I would have expected given the effort that I put into them this time around (read: like none). In the antweights, Unmodeled Dynamics…no, that’s not the name of a robot, though it could be… ended up causing Pop Quiz an early loss after winning one match. And in the 30lbers, Überclocker proved to be a little behind the times now in terms of speed, ground clearance, and maneuverability, winning only one match.

I didn’t even get to twirl anyone this time around. That is how bad it was. No con pictures this time either, since I really only stopped in for some of the Robotics and Science talks and otherwise didn’t have a good enough camera to feel like taking pictures of people. Also, without a consistent videographer this time, I don’t have complete video of both robots’ matches. So, no highlights video just yet.

The events, though, were phenomenal. The insect-class event ended up attracting 36 entries – 19 beetleweights and 17 antweights. This makes it the largest insect-class event on the east coast so far, and many of the robots were from new builders. Because of the number of entries, though, the tournament had to be converted to single-elimination (historically has been double-elim) to fit in the given convention time slot.

Pop Quiz

As the testing session video reveals, Pop Quiz has a slight “sudden unintended takeoff” problem stemming from the frame’s flexibility and the low blade height. I’m interested in seeing this under some serious high speed video equipment because the development takes milliseconds. I’m guessing the combination of robot movement, slight nonplanar mounting of the blade axis (due to floor bumps or varying wheel instantaneous heights) causes gyroscopic forces to flex the motor mounting area, which because of the very low blade clearance causes it to strike the robot corners. The blade then quickly destabilizes and hits the ground due to the bot’s very low height.

Either way, sort of annoying. There’s nothing wrong with Pop Quiz right now. It in fact works just fine, so I’ll probably retain it as a demobot. However, I’m definitely out to build a better antweight – of what style I am unsure of at the moment (but I really, really do want to name it Unmodeled Dynamics)

I have one complete video of Pop Quiz’s second match, which ended in a PopQuizCopter-induced loss.

Überclocker Remix

Oh dear. Where do I even start with this one? Clocker’s now 3-year and 4-tournament record is starting to show – the design flaws of late 2008 are slowly getting worse with time, compounded by my apathy. This year, the stage has pretty much become so damaged (purposefully, since it was historically a good wedge inhibitor) that Clocker was being hung up everywhere. The 2.5″ colsons have worn down alot, so much that I had lots of trouble maneuvering. Additionally, the lack of (functional) chain tensioners resulted in the chain being caught and damaged by the stage surface for the first time. To mend this, I made some quick zip tie tensioners (continuing the tradition of fixing Clocker at an event with zip ties):

The zip tie wrapped around both sides of the chain and prevented it from sagging and touching the ground. This worked – right up until it caught on the master link retaining clips and ripped them right off. The result was a broken chain in the bot’s third match. By that point, I really had stopped caring and just wanted to continue the show, so I elected to run one-sided (and without the infamous Zip Tie Ratchet even!)

Past that, the robot’s lack of actual forward-pointing sharp things (The ends of the fork are actually fairly blunt) meant that I was easy to get under from the front. Finally, opponents were finally able to take advantage of the bot’s 23 inch long flat broadside.

At least the motors worked this time!

Clocker version 2 (“Remix”) is now retired. I’m going to be rebuilding the bot from the ground up to emphasize manueverability and terrain handling – terrain referring to bumps and inconsistencies in the arena surface of up to 3/4 inch. Yes, this thing is going to have a boatload of ground clearance, will be much faster, and also with more “grabber clearance” so I can latch onto the more irregularly shaped robots. In order to fix the huge broadside problem, the sides of the bot will be rounded like Cold Arbor (it has prove hard to attack from the sides due to its tendency to “roll” into or away from the pushing force).

I haven’t designed out everything yet, but have been making a few “concept sketches”. One of which is below (though it’s unlikely to be used)

After obtaining 4 set screws, I put PQ2r2 together for the final shakedown test.

Okay, so it turned out to be more of a shakedown than I thought. The blade is just off balance and possibly just tilted in-robot enough to cause the sudden instability and “popping” in the video. To mitigate this, I’d have to spin up just before approaching an opponent.

Because of the missing battery half, the robot only weighs 13 ounces. It’s fairly maneuverable with the four Sanyo style gearmotors and soft wheels, so I hope if it’s too light to be an effective spinner it can still get under people and shove them around.

Microbattles event in about 10 hours!

The past week or so has seen me throw 500 miles on a rental car as I dash back and forth between home base, the Invention Studio, and Freeside Atlanta (and a hardware store or two, among other places) working mostly on Pop Quiz.

Thank Robot Jesus for free mileage.

Technically, Dragon*Con has already started (today being the first day of the con), but the robot events aren’t until Sunday and Monday. That’s fine, because Pop Quiz is done.

Well, it’s missing four set screws. But that’s another trip to the hardware store.

Step one: Clean up all of those 3d printed parts from last time with a bit of crafty knife work.  The small surface imperfections in the plastic print mean the motor mounts are extra-cozy – a bit of pressure is needed to seat the motors into them. This is good, since I’d rather have that than wobbly loose motors.

Holy panel gaps. After laying out the frame pieces on the baseplate, it’s clear there’s a small mismatch in boundary sizes. This is probably due to shrinkage of the plastic from the hot build environment to room temperature.Either way, about a millimeter on all sides will need bridging with epoxy, thick CA glue, Goop, or some other substance.

By the way, this is the last good picture for a while. For reasons I can’t fathom, my usual camera has now decided to take pictures only in the ultraviolet spectrum (that is to say, mostly black with some odd color patches on normally brightly colored objects). So I currently am split between using my HD camcorder as a camera (something it is surprisingly terrible at) or doing it up Jed-style and using a cell phone camera. Most of the build from here on will use the latter option, since it’s portable and with me most of the time anyway.

Horrible picture #1: redoing the battery pack. I neglected to bring Pop Quiz’s anticipated battery, a 500mAh 2S from Hobbyking that is the battery used now in MIT’s 2.007 robot building course. It’s the only small lipo I’ve found with reasonable capacity that is 10mm thick, or as thick as PQ is tall.

I did bring the 3S lithium pack I originally bought for Ballcopter. Not sure what I was intending to do with it, but it was simple enough to knock a cell out of the pack. This 2S arrangement is now also 10mm thick.

Horrible picture #2, with my HD camcorder (the HD aspect ratio gives it away). This thing has even worse low-light performance but holds a sharper focus otherwise. I’ve decased the Spektrum BR6000 and test-fit it in its final location. That gap width was designed around the internal board, so it’s a snug fit.

The battery lead, also the main power switch, routes through a small cutout in the back. Hopefully it won’t fly upwards and get sniped by the overhead blade.

Back to the cell phone camera. That thing seems to have no white balance ability to speak of – the pictures only appear without a yellow tint as in the battery picture if I point it at a blue or green object, like this convenient receiver board.

I de-headered the BR6000 one pin at a time by heating up the solder joint while pulling upwards with pliers. I have never cleanly (reversibly) desoldered male header pins, or any high pin count through-hole component like them. This time, since I was going to hardwire the controllers anyway, I just removed the entire header block.

I took the VEX 29 motor controllers out of their cases. They are actually very compact – no more than 3.5mm thick, and approximately 14 x 25mm footprint.

They normally run directly off a servo cable using the 5 volt line. To get 7.4 volts to it, I had to break out the 5V and ground pins separately to the battery, and then solder a signal + ground wire on top of that. The white and black signal wires run off the bottom of the picture, while the appended battery inputs go to the nest in the back of the bot.

I gave the bare boards a wrapping of clear heat shrink afterwards.

A view  from above with all the wiring in place. The signal connections have been hardwired and everything turns on without exploding.

The Vextrollers do seem to work on 7.4 volts (actually closer to 8.4 volts since the lithium cells were fully charged), making me think that they are actually capable of higher voltage operation. There is nothing on the board which directly indicates they would survive voltages higher than 10 to 11 volts, but it’s worth a shot. Time to put one on a power supply.

The majority of small bots like Antweights and Beetleweights tend to run 2S lithium systems anyway – at only $10 each, the Vextrollers seem to be a great potential resource if that limit applies to your design. This makes them way better than the overpriced specialty robot controllers.

I put the top on for test fitting and to check that everything does pack down. With the top plate’s screws in place, the robot frame is forced to be totally flat again – otherwise, due to the very slight warpage in the frame pieces, the whole thing becomes concave.

Back at the Invention Studio, I popped the wheel hubs out of some 1/2″ aluminum round stock. This was a fast lathe and mill-facilitated drilling operation since the parts were all small and all the same.

The tires  are some very soft (40-50A) silicone flexible tubing I got from McMaster. The outside diameter is 0.5″, and the ID is about 0.25″ so I turned the hubs slightly larger than that. I’ll probably drop some CA glue onto the joint to really keep the tires on there.

After wrapping everything up, the final bot. The frame pieces have been CA glued to the base and the ‘panel gaps’ also filled with thick CA. The structure is overall fairly rigid, a fact definitely helped by the carbon fiber sandwich design. The final blade height is 0.65 inches.

The bot is definitely underweight – I have yet to weigh it, but since I designed for 4 lithium cells and ended up using only 2, I think its at least an ounce and a half underweight. This might be taken up later on by more steel bits added to the inside.

Otherwise, the only thing I couldn’t find to finish the whole thing and have it driving was tiny 4-40 set screws for the wheels. The blade spins up very quickly, however, and PQ has already successfully lifted off and batted itself once.

Onwards to Microbattles!

I’ve been slowly knocking down the amount of half-taken-apart projects that have been taking up table space at MITERS. The usual shipping delays and distractions means I’m a little behind where I want to be,  but it’s not yet concerning. In other words, nothing has yet gone horribly wrong.

As of today,

• Fix Überclocker!
• Repair RazEr battery!
• New motor controllers for LBS!

Pop Quiz 2

I received my shipment of carbon fiber from Dragonplate and proceeded to cut the top and bottom panels out on the same day. The CF this time is 0.023 instead of 1mm, so I lose some stiffness in the frame. I’m not quite sure why I chose the thinner panel actually – the scrapped PQ2 frame had 0.039″ CF top and bottom plates.

This was certainly one of the cleanest CF cuts I’ve made to date. I took several precautions this time to minimize delamination around holes and pierces, including fully double-sided-taping the carbon fiber to a solid wooden panel. Previously, I have either just clamped the CF to wood or taped the CF to a waterjet brick. The full support of the wood layer beneath the CF helped immensely – these cuts have almost no delam areas. The other methods would either allow the CF to flap up and down or still leave high percentages unsupported on the bottom.

The upside to this method is that I get a cute MDF billet Pop Quiz out of the process at the end. This was the backing for the carbon fiber after I stripped off the wet tape.

With Pop Quiz’s frame all printed and the CF panels cut, it’s mostly a matter of sitting down and assembling the thing. I got the VEX motor controllers in and also found my spare Spektrum BR6000 from who knows how many robots ago. Because PQ will be an experiment in using the VEX controllers, I’m going to rewind the weapon motor for 7.4v (2S) operation.

RazEr rEVolution

RazEr is officially all closed up and working again – the only thing the battery needed in the end was a charge and balance. God I love A123 DeWalt drill cells. What other kind of battery doesn’t mind getting zero-volted for a month straight? I don’t doubt that I have had some lifecycle and capacity loss because of the extended flattening, but RazEr is not a very high current system anyway.

This is what the battery looks like after I ripped off the layers of soda bottles insulating the pack. There are 2 balance leads, but no actual power connections associated with them – my balance-capable chargers can’t charge through the balance connector only. So this whole pack was kind of unserviceable from the start.

It also has bare balance wires passing directly over cells, something I recently learned was a very bad idea.

The resolution for this pack was pretty simple – add those power connections. The two Deans connectors represent the upper half and lower half of the cell. This enables it to be balance-changed by 4chan every once in a while to keep the cells level.

To insulate the balance cable better, I laid a layer of rubber cement underneath them. That will at least immobilize the wires and also keeps them, for the most part, out of potential electrical contact. Due to RazEr’s limited internal width, I couldn’t reroute the cables to the paper sides of the cell, which is what I would like to have done.

The final result, after coating with Real Giant Heatshrink!!! instead of more Mountain Dew bottles. Working with this was very refreshing – this is the massive thick rubbery PVC shrink often used for enormous power cable repair and the like. It actually doesn’t look like I threw it together in 5 minutes.

The Advanced Beast-it-troller

They’ve arrived.

From last update, the independent-input H-bridge version of the Beast-it-troller is now ready for assembly. Both of my Digikey orders for required parts, most crucially the IR2183 gate drivers, arrived the same day. Will it work?!

I also ordered (and received, too) a spare CIM motor to replace the toasted one in the left side drive. With luck, Land-Bear-Shark will be running (…again) for this upcoming Swapfest.

Alright, now that I’m settled in 900 miles from civilization out on the west side of the MIT campus (seriously, how do people live out here?), it’s time to comprehensively update everything. Dragon*Con is now in approximately 2 and a half weeks. So, the flying objects have been temporarily suspended in favor of two more immediate goals which I mentioned briefly before.

1. Getting the robots running again, since I ain’t going to no Dragon*Con Robot Battles without no robots.

2. Repairing some of the random vehicles, since they (along with the robots) will be exhibited at the Atlanta Mini Maker Faire hosted at Georgia Tech in a few weeks.

I’m just going to start laundry-listing everything. First, the most interesting of the upcoming builds, a rework of Pop Quiz 2.

This was the state of the robot as of two weeks ago:

Yeah. It’s kind of trashed.

PQ2 hasn’t seen any action since Robot Battles 2008, and it lasted about 20 seconds in the arena then. So how on earth did it get so destroyed? Mostly because I kept pulling parts from it and then stuffing them back without closing the bot up. I’m sure it’s also been dropped a few times, and I might have also landed Clocker on it a few times too. Pretty sad overall; while the frame is perfectly workable, I don’t like the way the frankenmotors worked out (and one of them has been missing from the start). The offset wheel axles mean that I can’t just directly swap in a micro gearmotor or similar without making the wheels much larger. Oh, and all 4 lithium cells are dead. It’s a little easier to just start over.

What few people know is that PQ2 has been up for revision several times before, each time I ended up pursuing something else instead.  I designed a new version of the robot to be built for the 2009 Robot Battles, but didn’t get around to finishing it.

This version was to have a waterjet-cut aluminum frame and replace the frankenmotors with Sanyo micro gearmotors. I even got around to cutting the frame:

It was made out of 1/2″ aluminum  with the intention of machining it down to 10mm (.39″). Now that I look at it again, I wasn’t sure how this was supposed to be done.

Actually, no – I do know. I was going to actually make a fixturing block for it and then meticulously machine it in sections, taking out a subset of the fixturing screws as needed. And then I was going to turn the head of the Bridgeport mill 45 degrees to machine and drill the holes in the front. Then I was going to flip it over and carefully use a boring head to make the inset in the center for the motor mount.

I remember being hardcore like that. But now, in 2011, there’s a better option.

YES! It’s another excuse for me to abuse Make-a-Bot. For a while, I’ve thought that ABS plastic should be durable enough to withstand 1lb and 3lb arena combat. So this is a thought experiment (turned physical) that will use a fully 3DP frame, with carbon fiber top and bottom plates as before. The drive motors are still Sanyo micro gearmotors, and the wheels will be machined plastic things with silicone tubber tubing tread. \

Here’s the entire frame being fired off at once. 5 x 5 inches is still with MaB’s build envelope. However, given the geometry of the part, I didn’t expect this to turn out well – MaB has no climate control, so large parts still tend to warp. Especially a 5 inch long solid bar.

I stopped the build after noticing that two corners were coming up. Because the robot is so short, a millimeter of curl would spoil the ground clearance and wheel contact.

The solution was to just split the frame into 4 quadrants, such that the longest continuous span was 2.5″. This relieves much of the stress from the parts not being in thermal equilibrium. In the best case scenario, I’d have MaB in a closed heated cabinet at about 120 degrees Celsius or more. Then I could also make cookies in it!

With all four frame quadrants made, I’m just waiting on carbon fiber sheets and spring steel to arrive.

I might also rewind the weapon motor to run on 7.4 volts instead of 14.8. Originally, I did it to avoid running high current in the robot, instead choosing to run a higher voltage for the same power. After building vehicles, on something this small, the difference is trivial. Using the lower voltage would mean the Sanyo motors will be much happier (they’re native 5 volts!) and I’ll have a chance to try out these VEX controller units, which are the cheapest small robot control option I’ve seen, ever.

überclocker

Suffering another loss last year due to my lack of care for detail and “baaaah, it will be fine”, ‘clocker is actually functional otherwise. The left side gearbox slipped my inadequate press fit on its output shaft early in a match, leaving the bot mostly handicapped through the tournament.

I took the bot apart to extract the gearbox, but also to clean everything up and tighten screws. It was missing a receiver (which after some digging, turns out I borrowed to use on the cute little prototype Coasterbot), and one of the Victor 883s was not responding to signal. I found out why it wasn’t receiving signal very quickly after extracting it and pulling the servo cable out: the entire internal header inside the Victor came out with it.

Oops. Solution: Make a short servo cable pigtail that passes through the cable shroud so I can still connect to it. After this, it worked fine.

The gearbox fix was quick and simple: press out the spun output shaft from the planetary carrier, throw it on a lathe, and carefully knurl it. The knurling increases the outer diameter of the shaft slightly by introducing ridges and valleys. Then shove it back in with a tanker-load of 609 Loctite, some of the stiffest retaining compound there is. Knurling also helps make threadlocker and retaining compound adhesion stronger due to the same reason – it seeps into the gaps and is therefore able to coat more surface area.

I have no pictures from the knurling process since it was done at another shop (MITERS not having a knurling attachment for the Old Mercedes)

Past putting the robot back together (and driving it into everything at full speed, repeatedly), Clocker will not have any modifications made. I might turn the clamp motor around so it has more travel available, however – right now, if I don’t pay attention to where the clamping arm is, it tends to wedge itself in the highest position and the motor is then unable to free itself.

Clocker and Pop Quiz are the only robots I intend on bringing this year – NK got banged up nicely last time, and I haven’t rebuilt it yet.

RazEr rEVolution

Ahh, RazEr. This thing has just been *working*, though I’ve only really brought it out for demo events because of its concrete-hard wheels. However, Maker Faire ATL is a demo, so it’s time to turn the power on and test it for functionality…

…

…wait, what do you mean I left it powered on?

Yup, RazEr’s battery is now completely flat after being left on for what must have been a month or more. I need to take it apart anyway – the way the pack is set up, it’s impossible to charge using my R/C multichargers. I split the pack into 2 strings of 6 A123 cells (but run in series for 12 cells), however I neglected to break out the middle connection so I can actually like… charge it as two packs. Or balance it, or do anything really. This should be a quick operation.

Segfault

Another one of my finished things that just works, Segfault has also been a demo piece for most of this year also. The breadboarded controller is becoming incrementally more flakey, however – probably because breadboard. Sudden direction changes or even slowing down/speeding up quickly tend to cause noise which appears as a jerk in the wheels. I can anticipate and compensate, but it’s enough for me to no longer let other people ride it.  I have a desire to put the controller on an actual printed circuit board that’s fully integrated so there’s less messy wiring, but that is unlikely to happen in the next week.

However, it does need some minor mechanical attention. The steering column potentiometer is heavily biased rightwards from when I replaced it. The left side gearbox for some reason tends to slip its output shaft (What’s with me and half-assed gearboxes?) which, fortunately, is not a Charles-induced manufacturing fault this time. Segfault uses 2 Banebots P80 type gearboxes, and the last stage planetary carrier is connected to the output shaft with a double-D flatted bore, which seems to be prone to axial misalignment. Usually I can kick the left wheel back in and it reseats fine – not a critical mechanical problem, but still annoying.

land-bear-shark

You know I couldn’t leave this thing alone for long. Last time, I said I wouldn’t touch it until Winter. However, seeing as how it just rained, it’s now chilly outside, so I declare Winter.

As usual, I will preface a LBS post with a new motor controller. Another new motor controller.

What’s different this time? Well, besides it being the densest through-hole board I’ve designed to date (because why not), it uses the IR21834 gate drive chips. Watch that number there – it’s the independent input version of the 21844 I am fond of. This means the high and low sides are switched independently, but it still has built-in deadtime.

This frees me from the implicit braking of synchronous rectification if I’m too lazy to implement current control (which I am). I’m more accustomed to freely coasting vehicles, which the implicit syncrec of the 21844 does not let me do unless there is a current sensor and current control loop (such that the motor controller output voltage matches the motor’s back EMF)…but that’s more software.

With the 2183, I should be able to drive the high side switches only, while otherwise keeping the low side switch from  turning on immediately afterwards. This would let a vehicle, say RazEr, coast freely. Regenerative braking can still occur if I command it.

A vehicle like LBS doesn’t coast at all due to the high friction of the tracks, but this controller will be a chance to test the 2183 before I put one in a more serious application like RazEr. It’s already been sent out to Retarded Circuits for fab and should be here later this week.

I’ll also need to perform battery and motor surgery as outlined previously, but this should be coming down with me too.