I was on the red team with Corey (captain), Tom, Andrew, and Joe.
We had 4 days and $10,000.
As always, there are several rules that were not made clear in the show.
I don't really remember much about it. The focus was on using the same powerplant (a large motor) for all 3 events. This seemed particularly important because they hadn't yet told us we didn't need to jump the entire thing.
I was picked first, which is certainly nice. More importantly, we ended up with an exceptionally strong team because:
As always, we were able to get parts from anywhere we wanted. However, this challenge spanned a weekend, starting on Friday, and finishing on Monday. It also required a lot of components that we couldn't get from McMaster or the local metal shop. We got a bit lucky finding our chassis on craigslist and being able to meet the seller so quickly, but we had the most trouble sourcing motor speed controllers. My first choice had a 1-week lead time. My second choice was out of stock. We eventually found something we could order with next-day shipping (SyRen's). Then we realized that "next-day" meant next business day: the final day of our build. Then we tried to use what was available from a local hobby shop, but it was really meant for a different application and we ended up smoking them (discussed below) while attempting to use them for what we needed. Eventually, we decided to use the SyRens that were going to arrive on Monday. I'd never worked with that particular speed controller before, so all I could do was read the manual and hope everything worked on the first try when they finally showed up on the last day and we could drop them into the system. They did!
We pretty much addressed the three challenges with separate systems. They were really only tied together by scale. First, we decided on a pneumatic cannon for the throw and figured out the approximate weight of that. Based on that weight, we selected a kid racing go-cart for our chassis. We then estimated that total weight and divided by 3 to get our goal weight for the jumper.
Our chassis started out as a kid's racing go-cart. Then we used linear actuators to automate all the controls (steering, throttle, and brake). These were the same linear actuators used on the voodoo gun in challenge 2 (taken from the scrap heap). We'd have preferred to use faster actuators to get a quicker response, but while we had the budget to get better ones, we couldn't source any quick enough, especially with the weekend.
We also added a large (4.5 hp) DC motor (to which we gave the acronym BAM) in addition to the gasoline engine. This improved our acceleration though probably did little for our top speed. It also gave us the ability to move in reverse, which wasn't necessary, but was nice for driving around between challenges.
Keep in mind that the goal was 90.5 m, not as far as possible (see rules above). We therefore designed for consistency, not power.
We went with a pneumatic cannon for the throwing event. We also considered using a pitching-machine-like device, but felt that an air cannon would take less tweaking (which we had little time for) to work consistently. Also, an air cannon would allow us to leverage Andrew's fluid power expertise. As the fluid power expert, Andrew did most of the design. Tom did a great job fabricating the custom Aluminum accumulator. Tom also made plastic sabots to catch the air and push the javelin consistently. They had a thin flange in the back which was designed to flare outwards under pressure and seal against the wall of the barrel. You can see a dot go flying as well as the javelin during the competition; the dot is the sabot. We also had a linear actuator to control the pitch. This gave us fine control of our firing distance and also allowed us to lower it (most of the way) for the racing event to improve stability.
Based on the mass of our other systems, we knew we had to make the jumper at least 65 pounds (see rules above). We wanted as much of this to be in the jump system itself as possible, with the smallest, most pathetic, drivechain that could (just barely) move us up to the line. We were deciding between a pneumatic system and a giant crossbow pointed at the ground. I preferred the crossbow in part because I thought it was more scalable as we finished the other systems and found out our exact mass requirement, but the team decided on the pneumatic system because Andrew's expertise in that area allowed us to get it right on the first try.
The jump system was based on a 6" bore, 36" long tie-rod cylinder. In my episode 2 notes, I mentioned that off-the-shelf pistons do not move all that quickly, which is why we heavily modified this one. Rather than using the air ports the cylinder came with, we added four ports in the back plate for 1" hoses, each of which ran from a common accumulator through an electric valve. We also added big holes to the front plate to allow the air on that side to leave the cylinder quickly.
From a purely ballistic perspective, you want to jump at 45 degrees to get the most distance for a constant velocity. We were concerned about the foot slipping along the ground, so we jumped a little bit higher than was ballistically optimal, at something like 55 degrees. We also put rubber on the foot (a 4" square) to avoid slipping. In order to avoid wasting any energy in tumbling, we wanted the center of mass to be on the cylinder axis. Joe calculated the center of mass for several configurations and set the design to get the center of mass in the ballpark. To get the fine adjustment exact, I had the idea to hang the robot from a hook on the end of the cylinder. Then we adjusted the mount location for the cylinder onto the chassis until it hung perfectly vertical. The result was the eerily rotation-free jump you saw. I think we actually adjusted it a bit off from vertical so that it if it did tumble, it would go forward instead of backwards, but the flat plate of the foot was enough to make it jump perfectly straight.
The drive was hilariously half-assed. We had small plastic wheels direct-driven off of some harborfreight drills that we cut the handles off of. We added an omniwheel in the back to make it a tricycle. The smooth plastic drive wheels only barely had enough traction to move the robot. I had suggested using surgical tubing to improve the traction, but the team decided this was unnecessary. On the other hand, it also accomplished its job: move the robot a few yards up to the starting line, while being light weight and requiring little effort.
We were actually able to test most of this challenge, we played a large part in our success. We were able to test the driving in the alley behind the shop. We we afraid to take it up to full speed because it was narrow and littered with junk, but it was enough for Tom (our former race-car driver) to get some practice and for us to tweak the software a bit to give a better response.
We were able to test our air cannon on site on day 3 (shown here). Based on these tests, we knew what pressure we had to set it to to achieve 90.5 m. We're not exactly sure why the first shot on test day went so short. Leading theories include temperature change from being the first shot of the day (adiabatic expansion of a gas makes things cold) and the oil impregnated plastic in the sabots lubricating the barrel over multiple shots.
We probably could have tested our jumper outside, but we were afraid to. We didn't think it would survive the jump, especially without the sand pit. To our surprise, it seemed to survive the actual test jump, only pushing the gasket at the bottom of the cylinder out, which could have easily been pushed back in.
We knew we were going to lose the race as soon as we saw the red team's design. They were light and small and based off of a very powerful RC car. However, this same design choice, which helped them so much in the race handicapped them enormously for the other two events. The RC car, while extremely fast, had a very limiting payload capacity.
Anway, they had way more acceleration than us. Even with the "BAM" (large electric motor, which we added to the existing gasoline engine), we didn't have nearly the torque to get our 260 lb-ish robot moving quickly. I think we had about the same top-end speed as them once we got going, but that hardly matters in a 100 m sprint. They beat the human record, finishing in 7 seconds; we finished much later at 11 seconds. We could have shaved about 1 second off of that time, but since it was clear we were losing anyway, Tom braked early. We had never tested the brakes at full speed and were nervous about how they would perform. Turns out we needn't have worried, but there was no point in destroying our robot to lose by less. I think the other team did end up damaging either their suspension or the gears in the drivechain when they stopped at the end of the race, though I'm not certain.
As mentioned above, the goal was to throw as close as possible to 90.5 m, not "as far as possible" as implied by the show. We knew we had this one when somebody on the other team asked us how far ours shot since the relevant question to us was accuracy, not distance. Our first shot was very short (about 2/3 of the desired distance). Our second shot came up 7 m short of the target, and our third was 4 m long. The red team had insufficient power and could not release their trigger under load. Trouble with triggers seems to be a persistent problem for teams on this show and was one of the reasons why we went with a design that only required triggering electric valves.
We had never tested this for fear of breaking it. We did all the calculations and knew what was supposed to happen, but it was easy to imagine something going wrong when pushing the pneumatic cylinder so far beyond what it was designed for. When it jumped, we were ecstatic! Finally, somebody had completed a challenge.
I helped out with a bunch of details here and there, but my primary responsibility on this one was electronics. This was somewhat ... interesting because while I was the most qualified person on the team, I'm not exactly qualified. I'm a mechanical engineer who writes software. I've done 2 significant electroncs projects in my life and I know the basics, but I was way out of my comfort zone.
The jumper was pretty simple electronically. We used a standard hobby RC setup and ran 3 relays (left wheel, right wheel, jumping valves) directly off of the RC receiver. Due to issues sourcing speed controllers, we were forced to use relays for the drive, meaning we only had on-off control. We didn't even have enough relays to give it reverse capability.
We also had an intermittent issue with cross-talk. Sometimes, when throwing one relay, one or more of the others would throw for a brief instant. This wasn't a major concern for the drive, but it meant the robot might jump early when turning the drive on. I went to my go-to solution for all problems electrical: I added capacitors. Everywhere. It helped.
The main robot was a bit more complicated. We needed software position control for the linear actuators, which meant we couldn't control them directly from the RC receiver. Instead, we ran the control through a small computer (Arduino), which listened to the RC receiver. On/off control wasn't good enough, so we needed speed controllers, which we had a hell of a time sourcing over the weekend (described above). The linear actuator for controlling javelin launcher pitch didn't require high speed, so we could run that one at lower voltage, meaning we could use an Arduino motor control shield.
I thought the Arduino motor control shield used a separate power supply for the power electronics. To my surprise, it connects its power input to the power regulator on the Arduino it attaches to. I connected my 12 V battery to the shield input and a 9 V battery to the Arduino power regulator. The result was that the 12 V and the 9 V were wired in parallel with the 12 V basically trying to charge the 9 V. The result was a 9 V battery that was bulged outwards and a wet patch on the plywood electronics board. Remarkably, it still held a charge. Wish I had a picture.
As a non-expert at electronics forced to develop systems in a rush, I sometimes (often) make mistakes that result in releasing the magic smoke.
As I've mentioned a few times, we had a hell of a time finding speed controllers that we could get over the weekend. We went to the hobby shop and the only thing they had were these marine speed controllers designed for use with a bipolar power supply (positive and negative power). The instructions only said how to wire it up for the bipolar supply, and I could not find any information about how to use it with a single power supply. If I knew how it was wired internally, I could have figured that out myself, but I couldn't find that information either. After much searching, it became time to guess. Turned out it was internally wired differently from how I guessed and I ended up shorting the battery. Based on that test, I thought I knew how to do it right if we tried again, but we decided to just wait for the SyRen's to come in on Monday and hope I could get the system running with them in a few hours.
We were finished about an hour early and were just sitting around, charging our batteries. We were simultaneously charging the large battery and swapping the small battery for the computer. With so much stuff loose, it became possible for things that weren't supposed to touch to touch. Of course they did and there was a puff of smoke. It looked like just the battery connector itself was damaged but we had a tense few minutes of testing because there was no time to fix anything else that was broken. Amazingly, there was actually a spare (empty) clip for a 9V battery in the system due to a design change. We had accidentally incinerated the only redundant component in the system. It's tempting to claim I did this on purpose, but it was just luck.
We named our robot Thurbo. I think this was just after we saw Wreck-It Ralph (we finished testing build 3 early, and got a rare night out). So we named it Turbo, with an extra "h" in reference to the conservation of h's incident in Episode 2. It was Thurbotastic!
Including calibration and all the testing we wanted to do, we finished about an hour ahead of time. We also had about $1,500 of our $10,000 left over. We found a hot tub on craigslist and filled out an order form to get it for the house. Unfortunately, this order worked its way up the chain of command until the Executive Producer vetoed it, possibly for insurance reasons.
Back (Eric's TBBT Home)