Jan 28, 2014 in Stuff
My mental design-year can be broken up into roughly two halves: Dragon*Con season, which spans from roughly March to August, and Motorama season, which is basically September through the next February. Of course, the observant would notice that these things seemingly coincide with robot tournaments.
But that’s kind of how it works. After each big event, I start thinking about what I want to do for the next. Robot fighting is a game which I do not foresee getting old for me for a very long time to come. And why should it? Of all the things I ever became involved in, besides being the first one over 12 years ago (…) , it’s the most freeform and unburdened by pages and pages of rules and procedures. Run what you brought, not what the organizers say you have to bring. It’s very much a high-energy (… literally…), destructive, and kind of redneck sport, and there’s no whitepaper to write at the end. I have literal dozens of sketched out designs that will likely never see the light of the day unless I drop everything and become a professional robot cockfighter, which I would love to do. Here’s just one.
Someone get on that right now. Build it so I can live vicariously through you!
But I digress. It’s been Motorama season for a while, and so my off-duty mental cycles have been devoted to engineering the latest entry in the Imperial Carolingian Robot Army’s Register. May I present CANDY PAINT & GOLD TEETH:
Wait a minute. Hold on, trap. That’s a giant spinning weapon, not a fancy multi-axis grabby thing. The hell is wrong with you?
I’ve been out of the high-energy spinner game for quite a few years. My last “large” kinetic weapon was Trial Bot, all the way back from 2005-2006. That design wasn’t very effective, so it was retired after only one event. I’ve built quite a few successful weaponed 3lbers such as Nuclear Kitten (and its earlier versions), plus Pop Quiz which was successful Back In The Day. Overall, I haven’t been enamored by the big spinning chunk of steel like many competitors have been, choosing instead to focus on drivetrains and manipulator type weapons like lifting and grabbing weapons.
With the flooding of the hobby R/C marketplace by Chinese components some time in 2004-2005, the spinner game began getting ridiculous, with each wave of new bots combining a larger brushless motor with a larger hardened steel weapon of some sort. The arms race began being lambasted as the “brushless penis race” as most of the competitors were young men, and most of the successful designs were the same in concept and execution, only differing by how big the spinning toothed drum or bar was. It got to the point where many matches were just a minute of robots hovering around each other, each aiming to deliver the match-ending hit, and finally a weapon-to-weapon hit that destroyed both robots or left them totally disabled to the point of only being to crawl around on one wheel.
It was, to borrow a previously used analogy of mine, like we were all dubstep groupies waiting for The Drop… or in this case, The One Hit. The only other bots that could survive were armored bricks which kept running into the spinner until something, one or the other, broke. I did not find this game very interesting, and ultimtely neither did quite a few builders, who began exiting the game after their last bots were exhausted, because it was getting expensive to rebuild over and over robots which were good for only a handful of matches, or even just one. I in fact retired Test Bot out of the 12lb class after it was “penised” at Motorama 2008. That year also saw the first Überclocker.
Several efforts sprung up to counter this, including the Sportsmans’ Class which I found as slightly reactive but ultimately aligned with my own goals of repopulating the ‘middle of the spectrum’. Back in the very old Battlebots days, the power to weight ratios of entries was so low that matches were more determined by driving skill, strategy, and reliability, than dancing around gyroscopically waiting for the right impact to land. This greatly contributed to the corpus of weird or interesting designs (check out the old Team Nightmare event pictures). In the current Sportsman’s class, flippers, hammers, and the odd clamp-and-lift or two (ahem) have all returned. The matches tend to go the full length and feature a lot of action and driving, with the added bonus that you could generally return home with something that did not require a dustpan to pick up.
It’s been a while, however. I miss the feeling of beating something with a chunk of steel. That’s where this design comes in.
There’s been relatively few successful overhead-bar spinners, known in the vernacular as “Hazard-style” after the multi-time Middleweight Battlebots champion. This type of design is hard to get ‘right’ because the blade tends to destabilize the robot if it’s too long or heavy, and then it’s more difficult to self-right if it does go over. Pop Quiz and Trial Bot are both this style of bot, so I have a historical attachment to this design too. Other examples of this design are Brutality
and Tornado Mer, which early versions of this design resembled the most.
Note that all these videos are from quite a long time ago.
I wanted to try making a bot that wasn’t square. Using a continuous round tube for the outer armor was appealing because it presents no obvious weak spots, unlike a square corner (or the backside of Brutality’s front wedge). I fixed the blade size at 2 foot span, 3″ wide, and 1/2″ thick, a relatively solid dimension that ended up being almost exactly 10 pounds in steel. This dimension has been my ‘default’ go-to for a weapon of this size. A chunk of S7 tool steel typically costs about $100 in this size. All things considered, this thing was basically a small Tornado Mer.
The design goals for this bot were basically:
I started with the “doodle assembly” in which I haphazardly make parts and outline sketches.
The size of the bot was dictated primarily by what size giant tube I could easily find and purchase. Yarde Metals had a few promising candidates in the form of 15″ and 16″ tubing, so I started with 15. The rectangles represent the outlines of components I wanted to use. At this point, the choice of drive motor was a set of Banebots P60s with some 500-class DC motors. Fairly vanilla, and would give this bot an average drivetrain power for a heavy weaponed 30lber.
I experimented with several different possible heights, shooting for a 1.9″ blade height: 1.5″ wheels, 3/16″ ground clearance, and then a little over that to make sure I don’t whack myself. I’m used to sacrificing COTS parts for smaller packaging – Pop Quiz is still one of the lowest blade height antweights extant and for a while was the absolute lowest, and Trial Bot also had a blade height of 2.2″ using 2″ wheels.
The robot height decision was tied in with what I wanted to drive the weapon. Initially, I was inclined to make a super-wide custom direct-drive motor for the weapon – which would turn this bot into a giant Pop Quiz. One plan was to start from a chopped quadrotor motor on Hobbyking, except add external ring bearings (or metabearings – support rollers) to make it stiffer. I also took apart the 8-FUN motor again to measure the stator and considered using the motor inside as-is.
Further thinking and discussion led me to go for a brushless scooter style high-reduction indirect drive. I figured this bot was going to spend a lot of time upside-down and possibly crammed into a corner. A direct drive motor wouldn’t be as deterministic in such a scenario unless I also ran a sensored controller and used Hall sensors, which I didn’t want to do for reliability’s sake – one more part to jostle loose or break off its solder joints. The first image still shows a pretty insane and almost impossible (due to lack of tooth engagement or minimum pulley radii) and unnecessary 10:1 reduction. I put in a ~85mm motor diameter as a placeholder, since it was pretty clear that this had to be a custom job.
More components have been added here, including the outline of a Ragebridge as well as batteries. Height was the most important criterion when choosing batteries for this thing, since I would only have about 1.25″ of space to put them in. I elected to run an 8S power system with two 4S packs wired in series. My voltage philosophy is generally to run as high of a voltage as I can reasonably do so to minimize current draw and wire size; 8S was about as big as I could get without dangerously overvolting the average 12-18v RS550 motor. This high voltage would allow me maximum flexibility in choosing the weapon motor winding to best match it with the physical gear reduction.
I ended up deciding not to spend another $100-150 on batteries, but to use up more of my Nuclear Arsenal. I was going to split up two Thunder Power 7S 4.4Ah packs to make two 4S ones – two of them have dead cells, so they are not useful by themselves any more. 4 cells from those packs gives me a pack 30mm tall, or pretty much my maximum allowed height, as well as a great deal of battery energy for match length and number-of-spinups overhead.
The reduction is still shown as a slightly less ridiculous 8:1. I was also settling on what kind of power transmission to use:
I decided to give good ol’ V-belts a shot. V-belts often get written off as old technology, but they have favorable properties for this sort of thing – they allow some innate slippage due to the lack of teeth and do have tension members (they’re not just loops of rubber). The “L” series of belts (2L, 3L…) held promise for me since they’re designed to be small and flexible. A 3L belt seemed the best candidate here – 2L belts are tiny (1/4″ wide and 1/8″ thick!) and 5L is getting on up in huge. A 3L belt is 3/8″ wide and a little over 1/4″ thick, and the empirical smallest-pulley in use seems to be around 1.25″. Many larger bots have (and still do) use them for weapon drive, but they are not common in the little bots usually because of the minimum pulley diameter and thickness.
So naturally the next thing to get modeled is the Epic Drive Pulley. This shows the next weird thing on this robot that’s been getting me some stares. I’m electing to go for a live (rotating) axle suspended in dual tapered roller bearings rather than putting a hub and bearings on the blade, and only having a solid pole on the robot.
The latter is a mechanically simpler system, but in my opinion a live axle in this case saves both weight and height. With a hub and bearings that stick up above the blade, the center of impact force reaction is far up the shaft, which needs to be large in diameter and mounted solidly in the center of the bot to handle it. If I’m going to have a big solid block in the center, then I can put bearings in it instead, and move the center of impact much lower as a result. I also lower the blade height substantially doing so.
The big pulley is actually going to take up a substantial portion of the underbody real estate of the bot, since it’s on the other side of the bearings (to save, again, blade height). A lot of components need to fit under it, which ties into the 1.25″ of available workspace issue I raised earlier. This is what the blade spindle assembly looks like, in section; excuse the CAD-ahead going on:
Note the presence of Belleville washers on the top and bottom. These are often used to preload tapered roller bearings for zero-slop operation. In this case, I’m using the bottom washer for bearing preload (it pushes the bearings together, against the shoulder at the top) and the top big washer is only to keep the blade on. This keeps the bearing preload nominally separate from the preload of the blade. The idea is that the preload force has to be overcome before any of these parts even think of moving, and with (hopefully) most of the impact forces being side loads, it should prevent “blade wobble” from shaft compliance.
Now, in a good hit, everything will probably just munge together, but hey.
Back to the proper CAD order:
The internal frame rails were made using a “master sketch” that I reference all the solid models from. At this point, I didn’t know what any of the spacings or sizes needed to be, so I could just adjust the master sketch and the rails would resize to fit. The first time I did this, I made a sketch in the assembly; turns out Inventor doesn’t allow those kinds of references.
So I had to start over and make this sketch in a separate part and the insert the part into the assembly, letting other parts become Adaptive off of it. End random Autodesk Inventor tip.
Whoa, getting a little ahead of myself here.
After liking where the rectangles ended up, I began importing parts from previous robots and inserting them. The P60s are shown, as is the drive wheel solution: Banebots wheels on custom hex hubs, running on shoulder screw dead axles. I’ve never used Banebots wheels before, but they seem to be solid for a lot of other builders. The available of a hex bore was the swaying decision here, since it reduced the “hub” to a chunk of hex steel with a few retaining ring grooves cut into it.
To get power from the offset motors to the wheels, I have one stage of very short chain joining the motors to the rear wheels, and then another 1:1 stage from the rear wheels to the front wheels.
At this point, the protoform of the weapon motor has also been modeled for visual fitting.
This is the motor in a more complete form, though still showing the unrealistically sized 0.9″ pulley. Basically a one-sided hub motor with an integrated pulley, I’m going to blast it out from a solid round. No, not a two-piece welded assembly, nor a stack of waterjet-cut rings. I have a 3.5″ steel billet that’s been sitting around for far too long.
It sits on a 8mm shaft that rides in conventional 608 skate bearings. The very short load-to-bearing distance makes me confident that a good quality 8mm shaft (read: not made of the bullshit I make standoffs from, but real shaft steel) is sufficient for this motor; it’s also going to be hitting 12 to 15,000 RPM, and a bigger bearing would suffer.
A pizza appears.
The motor mount has a gratuitous number of slotted holes to let me adjust the belt tension if needed, but also maintain rigidity in the area.
This bot is of an ‘upside down’ construction. The top plate is rigid and integrated into the one-piece frame, and there’s only a bottom ‘dust cover’ which will be made from 1/16″ FR-4 (Garolite) laminate for the electronics. The batteries will have a bracket to retain them, but this is not shown yet. Basically, not a good idea in the old Battlebots days of floor-mounted hazards. I went this way because I didn’t want to have to remove the blade over and over to do maintenance. Everything in this bot drops in from the bottom and is bolted in place.
I also went to ‘duallie’ Banebots wheels because the single 0.4″ wide wheel seemed too fragile. These wheels have a thin web portion before they fatten up for the hub again. The chances of the bot landing on one side (or one corner, hence on the wheel) and breaking that off seemed fairly high. What I’ll likely do is put the two wheels side by side with filled epoxy in the middle to fuse them permanently into one.
A view from the top with the top plate made transparent. Buttonhead screws are shown, though for blade clearance issues, these must necessarily be flatheads in real life.
The design stood like this for a few days. I was satisfied with the roundness, but did not like how much wasted space there was inside. I needed to match the largely rectangular parts inside with a circle on the outside. I guess that’s just a trait of round robots.
After a while, though, I decided to refactor the design into one that was more practical. The roundness makes the bot more visually cohesive, but it is not very practical. I wouldn’t be able to push very well as a last-ditch backup, and being fully round is counterproductive for self-righting – a lot of it depends on luck and physics from jouncing around on-edge. A fully circular bot like Tornado Mer tends to “coin” around (though in that match it didn’t help that the weapon motor contactor locked on…)
I started playing with adding indentations or other features to the design:
Attempt number one: Cut off a chunk of the circle and append a polygonal wedge to it. Simple enough, and it would get the job done, but I didn’t like the fact that the two end corners were exposed. Sharp internal corners on this design would present an unnecessary vulnerability to opponent weaponry, since it would otherwise tend to bounce off the round sides or up the sloped wedge surface.
So I tried a ‘wraparound’ wedge instead. The wedge is not formed to the rounded surface at the corners, but is just tangential to it for a duration. This seam will be welded shut so it will resist peeling.
I liked this design a whole lot more, so I went with it:
Five front gussets support the wedge. I had to rebuild most of what was the rear end, including new cover plate shapes and motor locations. The battery cage has also been modeled – it comprises two 1/16″ aluminum folded sheet metal assemblies that each trap one modified 4S battery pack.
I decided to change out the motors entirely. Previously, I was planning on running a single-stage 5:1 P60 gearbox; however, their length and mass became issues. Plus, I’d have to buy them. Taking a page from 12 o’clocker’s “Angerboxes“, I modified the design to mount from the top down. The drill gearbox is a 6:1, yielding me a little reduction, so I could back down on the need to externally reduce from the intermediate chain stage. The simpler design saves a few ounces per side. Even though they’re plastic-cased, I think a well-supported motor (see the black rear mount) and use of the material in bulk will be sufficient.
The drivetrain is designed to hit 15mph, which is pretty zippy for a heavy weaponed bot.
A size comparison next to Überclocker, which it would bang up pretty badly. Clocker is a pretty good example of a design with too many pointy bits to survive in a big-weapons environment. At the very least, the clamp arm would have needed to be more vestigial to allow weight for a big armored plow in the back.
Physical progress-wise, I’ve accumulated the majority of parts for the bot, and more are on the way. First, a few weeks ago, I snagged these old weapon blades from a lightweight (60lb) bot which were reused in a 30lber years ago. They are…. 24″ long, 3″ wide, and 1/2″ thick, heat treated S7 tool steel with 1 inch bores. Exactly what I was designing! Well, that pretty much seals the blade decision.
The bottom one is solid, weighing 10.2 pounds exactly as it should, and the top has been weight-reduced to around 8 pounds.
I should be able to run the solid one; as-designed, the robot weight is 27.6 pounds, not counting some small hardware and wiring (which always adds up).
Most of the frame parts cut out of 1/8″ and 1/4″ aluminum. Caveat: I’ve only welded aluminum once and it didn’t go over too well. This will surely end well.
A few pieces are missing, so I can’t start just yet; I’m hoping to wander into the machine again this coming week with more parts for other bots.
However, I did sand and fit the rest of the pieces together. Pretend-o-bot #1 is complete!
I’m hoping to be able to roll the big outer hoop this week from barstock. I purchased two 1.75″ wide, 1/2″ thick 5 foot extrusions to make the hoop – only one is needed, but it was cheap and I’m most likely going to bang it up at least once.
I’ve been probing the local peer cloud to see if there are any skilled aluminum welders willing to take this job up. As much as I would like to learn on this bot, I also kind of don’t want to have it shatter on the first hit! If not, or if I’ll have to pay an absurd amount to get it done, well… here goes nothing in particular.
I’ve also gotten:
Parts that are already on-hand and just need to be modified or pressed into service:
I’ll still need to machine:
As for the giant output pulley, I did a make-or-buy study as soon as the design was finalized: I threw it on MFG.com, my go-to for hiring shady Chinese job shops to machine things for me (things made through them include all of the DeWut gear and my small run of hub motor parts). I will hopefully have the finished result back by the end of February – but if shit goes down, I’m going to cut out a circle from 1/2″ plate, stick it on a mandrel, and go to town.
In the next episode of Big Chuck’s Automotive Blog…
Happy new year and welcome to Big Chuck’s Automotive Blog! The mission of BCAB is to share and discuss all of our misadventures in being shadetree mechanics. Not only will I post all the questionably sound work on my own wreck, but every week, there will be one story submitted by you, the readers, about any aspect of your life pertaining to your own automotive project or rolling piles of garbage, whichever you would prefer.
The way my site visits and interesting search hit terms have been slippin’ lately, though, you’d figure I’d have gone full-time car blog. Luckily, that’s only partially true. It used to be that I got plenty of weird and interesting search hits, site referrals, and the like. I feel like I’m losing my touch there – these days it’s all full of “electric bike” or “electric go-kart” or “How to avoid electric shock installing I’m a hybrid battery” (sic) and stuff. Booooooooooooring. Perhaps I should be glad that I’ve been genericized to that point, such that my content has become more generally relevant. But I do miss the days of the Arduino powered butt massager.
This IAP, I’m watching over MASLAB which is using the IDC classroom and my shop space, while also ordering things and preparing for the next round of 2.00gokart in the spring. MASLAB is historically a ‘shopless’ activity… which means that students break into or ninja the use of whatever shops they can get into in order to finish their robots. This year, they faced difficulty getting their usual space in the EECS department, and several of their core team and students being my former students, I got pummeled with appeals for space. Now, it creates way more work for me (what amounts to an actual full term class’s worth of preparation and shop orientation sessions), but what better way to spoil even more undergraduates? Furthermore, I think it’s better for them that they have official access to much more resources that can be properly used (i.e. under my titanium fist rule) than students trying to steal and beg resources from any space they have access, or “get” access to; which in my mind is patently unfair to those who are also just starting out and don’t Know Somebody – MASLAB is often one of the first “Build a robot” things a lot of freshmen do.
Anyways, I went to Ikea:
I defy anyone to challenge me for the title of “Best Ikea Space-Filling Ratio”. Flat-pack furniture works best with a vehicle which can be 90% modeled in no more than 3 solid modeling features.
Now, none of this is actually mine, since my own life is containerized into a number of typical milk crates, and I wouldn’t touch anything Ikea produces with the most bargain of Harbor Freight allen wrenches. But while on the tour in the most perfectly structured consumerism experience, I naturally gravitated to their tool section. The selection was naturally all custom-commissioned products geared towards assembling only their shit – again, part of the most perfectly structured consumerism experience this side of Buy & Large.
(To quickly skip to the other sections, here’s…
I spent a little while looking at the FIXA (I keep wanting to say Fixya) power tool series – they have things as interesting as a 14.4v hammer drill and a standard two-speed drill. Ikea being an entity that nominally prides itself on inexpensive low-key quality (as opposed to, say, Harbor Freight, which prides itself on Fuck You), I did expect that these tools would have worked just fine in their intended household lives. It’s like a domesticated goose – all you really need is a guarantee that it will poop everywhere, perhaps not with the flamboyance of a wild Canada goose.
I found their 7.2v drill/driver interesting. This is because it evoked the shape and function of the classic Handiworks mini-drill found at Walmarts back in the Early Noughties. This little thing fueled the rise of the 12lb weight class. For a while, Harbor Freight carried a 7.2v variant which made it into the 2nd and 3rd iterations of my own Test Bot. That was about 2005-2006. Those drills disappeared with an increasing RMB to USD trading ratio, as did most of the low-v0ltage (9.6v, 14.4v) drill/drivers from Harbor Freight.
An overwhelming sense of curiosity and nostalgia drove me to pick up one of these units. I’ll say right away that for $24.99, it may not be worth it in general, even if it were identical to the old Handiworks. However, the package ended up being more compact and a higher ratio – it definitely could be robot-applicable for somebody. So thus begins the Beyond Unboxing of the FIXA 7.2v drill/driver.
The casing is shed with a few Phillips-head screws from one side. No hidden screws here. The first thing I found is that it really IS lithium ion! There are two cells, 1500mAh each, size 18650, of lithium cobalt or lithium manganese chemistry (not LiFePO4). These 1500mAh cells contrast with the modern generation of laptop and other device cells which are typically 2400mAh, likely because they are “power” cells made for industrial use – wider temperature ranges and higher allowed burst currents – than “energy” cells which simply try to provide the longest runtime.
It has a cute little BMS board attached to it that handles both charging and discharge protection. The large FET at the top is connected to a current sense circuit that actually causes the drill to shut off if it’s near stall or suddenly locks up. This manifests itself as suddenly losing power, but it resets once the trigger is let go of. A nice protection to have if you sell your tools to total rubes for sure.
This current sense circuit depends on a sense resistor, which, like the Jasontrollers, can be easily chopped to a lower resistance if somehow you are compelled to do so, God help ye.
Four more screws and the gearbox comes apart. The gearbox is unlike the standard 36:1 or 24:1 drill gearbox. Rather, the gears are somewhat smaller in pitch, 0.6 module by my closest guess (about halfway between 32 and 48 pitch, which is what they look like). What was surprising is that the first stage of the geartrain is all metal. Usually, the first corner to be cut on these is to replace the first stage with nylon gears, ostensibly for noise reduction but we all know really why.
The gearbox is 3 stages of 16:14:45, resulting in a total ratio per stage of 3.8125 and a final ratio of 55.41:1. The final stage has 5mm thick gears, compared to the 4mm thick in the rest of the thing, to handle high torque demands.
The ratio is a little high for my tastes for a robot drivetrain, but for those not aiming to hit 15-20mph, perhaps just a slightly larger wheel will suffice. Remember that I’m clouded by a decade of smashing robots into each other; very few parts which are generally useful make it into the top echelons of the battle-tested.
I wasn’t quite curious enough to take off the chuck, since the left-handed locking screw was better installed than most Harbor Freight drills and I wasn’t in the shop at the time. I suspect that the traditional drill gearbox bellhousing, albeit in a smaller size, is on this one. The drill shaft is also most likely a 3/8″-24 thread like normal, but I won’t speculate more unless I have it taken apart. It has a nominal rating of 400RPM – which, through the gearbox, yields a motor speed of about 22,000 RPM, in line with the typical small drill motor. The motor in question is a 7.2v Mabuchi RS380 knockoff, unlabeled.
Remember the Landbearshark video? Well, after that and the additional snowstorm a week ago…
Whoops. I guess I went a little too hard. I noticed something was wrong after the batteries never recovered above 16 volts even after a day of sitting. Both battery packs had cell groups which were either at 0 volts completely, or at severely damaged levels like the 1.38v group above. This was the batteries which caught fire once and also survived months of tumbling in the original Landbearshark, finally having been done in because the rest of the thing worked too well.
Damn. Well, with the potential for more weather in the next few months since this winter has really been making it
rain snow, I had to replace the damaged batteries before LBS could work again.
I went digging in my lithium nuclear arsenal, which I obtained after the MBARC class ended and I confiscated all the lipos (with exception of those taken by R/C airplane experienced students). Most of the packs were in the 5S and 6S range, which was good for LBS, but they did not have built-in battery management boards and I didn’t want to add a big balance harness to LBS. However, there were also these:
One of the teams went commercial/industrial and picked up these from Batteryface. These are sold with a “PCM” module built-in, so they don’t need to be externally balance changed. I’ve used these boards a handful of times before in not-my-own applications, and they do work just fine, but I find them a little too wimpy on the discharge: for most high burst current or other high power apps, I prefer running straight battery, because the management board usually introduces more resistance or has built-in current limits.
But LBS is not particularly high power. I could also fit four of them in the space left by the 6S6P A123 pack, netting me much higher energy density: 22.2v 40Ah instead of 19.2v and 26.4Ah. I’d trade the unneeded brute force for ease of use and built-in protection.
Sounds like what these were made for! So in they go.
To get four packs in the space of two, I had to put Y-harnesses on my Y-harnesses. I chopped the discharge leads off my old battery, which had a type of 6mm bullet connector I no longer had on hand, and spliced them to two Deans plugs each. The students added quick disconnect terminals to their batteries, which I cut off and replaced with Deans.
Installing the batteries was a fun game of OH GOD DON’T TOUCH THE FRAME RAIL WITH THE EXPOSED PARALLELED CONTACTS.
The batteries are mounted to the electronics box with strips of Velcro. Their height is just under that of the box itself, so they shouldn’t be going anywhere.
Suddenly, the wiring looked less nest-like than before. Not because I made it better, but now all the excess runs were the correct length to tuck next to each other! Science.
LBS has yet to make it back outside since the weather has been… “nice”? Test riding around the building showed me that it was very much more responsive. Not only because the voltage has jumped a few from the A123s, but that the batteries must have been damaged for a while and have been sagging more for the same current draw. Hopefully the next bout of winter commuting will put these to the test.
Rewinding before the new year once again, I’ve officially commenced…
It could refer to several things. First, the old magnetic disc drives that used straight iron oxide (rust) to store information; the earliest kinds that went into the “refrigerator” hard drives. Next, the fact that you can’t quite remember something. Finally, all of the really shitty bodywork I’m about to do to prevent more problems down the line.
I’ve been leery of doing bodywork for a while, despite a slow buildup of arms in the interest of doing so. The past has shown me that I have no patience for making smooth and clean lines or blending paint. However, the recent pressure of winter and its associated wet salt slush has caused me to examine some of the spots in more detail. I’ve determined that there’s some areas where I’m getting close to now-or-never, because the underside and “hidden” rust. Remember these boarding step holes? They’ve gotten bigger:
Soon, they will soon break the outside body lines… and hell if you’re getting me to rebuild external lines. Other trouble spots include the majority of the left underside for some reason – the right side is pretty clean, but the left is all sorts of beat up.
Before tacking the more complex curvature of the step, I decided to practice more on a less visible spot – the left rear corner. Here’s what it looked like in May:
A complex confluence of edges in the corner with quite a few holes and thin areas to patch up. The plan I formulated was to cut away as much of the bad areas near the holes as I could get, then grind or wire brush off the rest. About two weeks before starting on this, I thoroughly coated the interior of the bodywork in the area with that “rust converter” compound and let do its job for a while. Hopefully this will help prevent the interior sheet metal from being a problem in the near future.
Let’s get started. I once again dibbed the corner of the garage for a weekend, though I didn’t need the lift. What I did need was a spot that wasn’t -30 degrees out, so things could actually cure.
When I was using the lift before, I had noticed that the arms block the area I need to work on, regardless of orientation. So I had to use a whole trade of jackstands (the proper collective noun for jacks is a trade) in that area. Since I’ll be violently thrashing on this area for a while, I used not only a stand on the frame, but on the corner of the rear suspension also, kept the floorjack a little pressurized under the differential, and chocked both right wheels in both directions. A little paranoid? Perhaps, but I also prefer to have thickness.
This is what that region has devolved into since that time. The holes have grown a bit, and much of the weaker rust has fallen off. The treatment compound is seen in green.
The excise begins by gently hammering at the panels to loosen up more internal rust. This is item #2 on the list of 3 things Mikuvan does very well: dropping little flakes of rust everywhere. The other two, of course, are emitting black mucuses of various viscosities, and raining bearings.
Maybe I should have done this before spraying the converting compound…
Next up is imprecise angle grinder cutoff wheel excise. The biggest trouble spots went first.
About 1/3rd way through the process. When the angle grinder became too unwieldy to maneuver, I switched to a Dremel with a small cutoff wheel. My goal was to eliminate as much of the obviously rusted metal while retaining features that will help rebuild the area. I cut off a piece of the wheelwell (the right angle upside-down-L cut is center in the picture) to gain more maneuvering space for cleaning the area behind it. After the cutting, liberal application of wire wheels knocked out the rest of the surface rust in the surrounding area.
What I do not have is a picture of the completed surgery, since much of this process was mentally streamed. More of the steel on the inner wall to the left was removed, as was the area with the perforations in the upper left, extending about 4″ towards the front (where the wirebrushed paint starts).
I retained my tactic of using 3 layers of fiberglass cloth (I’m not sure of the weight, but it is pretty heavy) that were nipped from Solar Car.
I decided to split this work into two sessions to make sure I didn’t have to hold onto too many things at once. I patched the outside first and let the glass cure overnight.
The next day, I worked on the inside. To cut the cloth to shape, I just mashed the fabric against the repair area and used a marker to get the rough outline, then cleaned and simplified the marker scratches to a cut pattern. The pattern was used as a template to make two other pieces, each very slightly smaller. The marker dissolving into the fiberglass resin is the cause of the blue outline.
This area looks pretty gnarly because of the untrimmed glass and the fact that I didn’t try to rebuild the down-facing curvature of the original body section.
The day after was cleanup, filling, sanding, and painting. The tattered glass edges were trimmed flat with a Dremel and cutoff wheel first, then the whole area manually sanded with a sanding sponge and then some fine regular sandpaper. I used a small amount of Bondo to smooth the transition between the glass layers and the remaining bodywork, but as the masked area shows, did not attempt to resmooth the surface from where I wire brushed off the paint.
Paint was the same procedure of primer, color, and clear I used on the rear hatch. This took several hours by itself, then I let everything dry overnight once again.
Once the outside was dry enough to put some masking tape on, I sprayed a few coats of underbody coating compound on the inside repair to seal it as well.
Here’s what it looked like as of a day or two ago – it’s gotten a little dirty since:
I make no claims to ever passing auto body school.
Based on my research, a real auto body guy would have removed far, far more metal than I did, and also have remade at least some of that inside corner box section in steel, if not straight up remake the entire sheet metal of the wheelwell area. When I can afford this service, I suppose I’ll have that done…
I’ve learned since that they make this stuff called “spot putty” which helps fill in the very small resin bubbles that are visible; plus that I’m not spamming enough resin onto the top ply to start with, a phenomenon also visible in the rear hatch work. These lessons will hopefully be put to use in repairing the boarding step hole soon, since that is a more visible location (with the door open, anyway).
A quick break from inhaling styrene and toluene led me to try and figure out exactly what the deal was with the “subwoofer-like device” that I touched upon previously. I thought it was barely working, but it turned out to be sympathetic vibration transmitted through the front sheet metal and dashboard components. It was in fact totally out.
I’m sure a normal person would have replaced this with a set of 12″ subs in the back, but I dunno, it’s already there and most likely working anyway. What if it was as simple as some dumb fucker not connecting one of the wires? Wouldn’t I feel foolish for not trying to make use of it at all!?
Besides, the 12″ subs come after the electric drive conversion, as do the tacky underglows and stancing.
It was 20 degrees out, in the middle of winter, in Massachusetts. And here I am, outside, with nothing but flashlights, using an oscillosope and soldering iron to probe the paths that the signals took in an attempt to debug the amplifier board. Consider the frightening possibilities if I had put this much effort into actually studying something.
I ran into a slight metaproblem – it was so cold that my small cheap soldering iron, which travels in the robot service toolbox normally for use in the field at events, froze its power cord off. Literally. It probably deplasticized in the cold and in the process of me unfurling the cord, it broke off.
I borrowed a Weller station from MITERS in the mean time, which seems to use a plasticizer that didn’t also grow up in the South like me.
So if you’re ever stuck debugging the subwoofer amplifier circuit of a generation 3 Mitsubishi Delica, here’s what it is. The whole thing is OEM’d by Matsushita (a.k.a Panasonic). There’s 7 wires leading to the board – three of them are the ground, 12v, and “power on” lines shown, the others are two channels of signals and their return lines.
What the frontend of this amplifier does is add the two stereo channels together, then severely low-pass filters it before sending it to the amplifier power IC. This is all done actively, with op-amps. In fact, the circuit is eerily reminiscent of this generic mono amplifier circuit.
The ENABLE line controls the coil of a little relay that is in between 12 volts and the amplifier chip. Guess which wire was open circuit?
Naw, couldn’t be that someone forgot to wire it up.
(Alternative explanation: The new head unit that came Free With Purchase did not have an external amplifier enable output, so this was left unwired, but that doesn’t explain why someone took the speaker totally out…)
I took the cheap and dumb way out: Jumping the enable pin directly to 12 volts. When I turned the ignition key, I heard the faintest click of a relay and a little pop from the speaker.
Scoping the speaker’s terminals shows this nice waveform coming out. The cutoff frequency does appear to be around 150Hz.
I packaged everything back up after this fairly simple hack, and immediately ran back inside to defrost. Let’s be honest here – this little thing didn’t add that much to the experience; finally some noticeable low end now, but it seems that it saturates (clips) relatively easily. Not that I blame it at all. It was another box ticked off on the checklist of completion.
Item #4 on the list of things Mikuvan is good at: leaving small droplets of oil wherever it goes.
It’s done that ever since the first start. I’ve always attributed it to a crank seal problem, but recently I started suspecting otherwise, because the symptoms didn’t really line up with just a crank seal issue. If I had a leaky rear crank seal like I suspected, then the oil drops would be coming from very specific, concentrated locations. Same goes for the front seals. I’d at least see a consistent, concentrated ‘shot pattern’ from the two locations in my parking spot… which I assure you is terrifyingly disgusting.
Instead, it just seems like it’s been shitting everywhere. Since I’ve been getting under it more recently, I’ve also been keeping track of the cleanliness of the underside: Every time I look under it after cleaning up all the oil and grime, there’s more of it everywhere. There was no one consistent spot at all – the whole underside near the engine would be wet all the time and spots would appear almost at random. It was less oil leak and more Self-Applying Undercoat.
As the weather got colder, it just started getting ridiculous, and once again I was faced with a now-or-never scenario. I was beginning to suspect the oil pan gasket a few weeks ago when I first began noticing that it was always wet on the outside. Hey, shouldn’t a gasket keep the leaky stuff on the inside?
During the suspension work on the lift, I gave that area a very fine look-over.
This is the forward left side of the oil pan. First, that gasket is pushed out completely and ripped. Second, it’s disgusting.
I figured, once again, that even if it was not the main problem, it could be a contributor or aggravating factor, and that I should at least inspect it. I braced myself for yet another Yak Shaving Session where I end up having to remanufacture the entire assembly. How bad could it be!?
(Always Famous Last Van Words)
I looked at the service manual for a bit and then began disassembling the oil pan screws.
First to come off is the oil level sensor. I have no idea how this is supposed to actually work – and it just barely does. Usually, if I park on a non-flat area, it’ll throw an oil level light; and not knowing how leaky the thing actually is, I check every time, only to find the majority of the time it’s totally fine.
I have no pictures of the pan removal process, since my hands were well-covered in oil, and the whole thing just sort of fell off after I undid the last screws and put a little pressure on it with a scraper. Well, that’s certainly a bad sign. From my Youtube instructional video surfing, you’re almost supposed to use said scraper to cut the whole thing off.
Oil pan removed! This is the first time I’ve ever physically seen the inside of an engine, from the bottom end. Who the fuck thought this was a better idea than a brushless motor?
The fact that oil is everywhere on the alleged gasket sealing surfaces is, again, not a good sign.
So here’s the deal with the gasket. First, on top, there’s a layer of silicone. Not, say, specially formulated gasketing compound, but I swear it was just clear RTV used for bathroom tiles.
Next, there’s a paper/felt sort of gasket, the type that you would buy specifically to fit a model of engine.
And finally, there was another layer of silicone.
Silicone-on-paper-on-silicone didn’t exactly strike me as a professional repair. I suspect, again, that this was like 5 different dudes’ repair hacks and I am the 6th.
Unlike bodywork, I considered this a blasphemy against the mechanical gods. I rage-cleaned and stripped the entire pan, paying special attention to the gasket seal surface. I also cleaned up the bottom of the pan some. Luckily, there were no metal particles to be found, but there was a sizeable amount of brown and black sludge; likely from before I was also meticulously keeping track of oil condition.
Here’s a shot down the line of crankpins and big ends. Once again… who thought this was a better idea than twirling a magnet (or a blob of copper and steel) on a stick?
Here’s a picture of a 3-floor building sized engine’s crankcase while we’re at it. It’s only a little bigger.
I didn’t get any pictures of the re-gasketing process, but it entailed borrowing a small amount of this RTV material designed for gaskets and laying it out in a roughly 1/8″ wide bead in the pan’s top groove, around the outside perimeter, and in a circle around the bolt holes. I then let this cure overnight under the influence of a halogen lamp, and retorqued the screws according to specification the next day.
After a week and a half of this, I’ve only seen 3 new oil drops after having placed a white spill mat on the concrete parking spot. They were concentrated around this spot:
I didn’t notice this little vent in the bottom of the transmission bell housing until I was under there looking at it. Under the cover is the torque converter and its crankshaft adapter plate. If I had a crank seal leak, I would have seen the majority of the oil drops originate near here. It might still be leaky; I have not confirmed its health in either sense. For now, however, the oil-shitting problem seems to have been resolved in the majority.
This concludes the latest Big Chuck’s Automotive Blog entry. Make sure to check back next week as I make even more mechanics and auto body technicians cry!