About now in most of my projects is when things start breaking, not fitting, or not working. Let me be the first to say…

AAAAAAAAAHHHHHHHHHH FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK FUCK

There. The short story: The bot doesn’t work.

The details: The arm jams, the Spektrum Rx doesn’t communicate with the Victors, and the SyRen25 arm controller current limits really fast. And the wedges still aren’t done.

Anyways, some quick pics from the last few days…

Here’s another gratuitous bot-shot.Several things have been done to the arm structure, namely the addition of ginormous inter-beam standoffs and the channel cuts in the beams themselves. The standoffs keep the arm links an exact distance apart and also gives the structure more integrity. The channeling was to reduce weight.

Preliminary wiring gives the green (okay, orange and blue) light for the controllers. Yes, they work.

A botcessory that I cooked up during the design process was to create a charging port. I wanted to find some way to charge the battery with balancers without taking the whole bot apart to get to the connector, so everything was routed through this DB9 connector. The charger will have its own integrated plug. Shove into top of bot, done.

This is Ghettobooster. It’s a 5v regulator strapped to some signal transistors and the PWM cables. It sits between the Spektrum receiver and the Victors and amplifies the signal such that the Victors actually notice the receiver.

It worked great the first time, but now nothing works again. And I’m out of small transistors.

So, onto the problems that I need to solve before Wednesday:

1. Everything bends.

UHMW especially loves bending. The torque from the motor combined with the momentum of the whole arm assembly when it’s traveling at .1C causes the external spur gear stage to bend away from eachother until there is so much force between them that even the motor can’t get things back down. This is a direct consequence of having an overhung gear shaft combined with a UHMW motor mount.

Unfortunately, there is no easy way around it at this point. The simplest solution is to cook up some sort of external support for the shaft at the gear, but since said gear is trapped between two arm links, that’s rather tough.

The other solution is to get rid of the pins holding the arm at the upper position such that the links can keep going further. This gives me more response time, but doesn’t solve the jamming issue.

I can also try some sort of software limit using a “signal interceptor” microcontroller to read limit switches, some sort of optical interrupt, or a hall effect switch with a magnet strapped to the arm. I have some AVRs. I have a parallel port ghetto-burner. This is a possibility.

2. Victors are the Cadillacs of ESCs

And they take some serious power to drive. Victor 884s are opto-coupled controllers that apparently have opto-couplers the size of a capybara. Not all receivers can output enough current to drive them, which is why Victors frequently need buffers. So I made Ghettobuffer earlier, which worked for a few minutes and at least allowed the bot to putz around the ground, but now… it doesn’t work again? What, do I really suck that bad?

I’ll need to figure out what’s wrong or get some commercial inline R/C signal amplifiers ASAP.

3. Overprotective ESCs are like overprotective mothers and overprotective governments

The SyRen25 has regenerative braking, temperature and current limiting, and the ability to be controlled by smoke signal. It’s almost unkillable by conventional means, but that current limiting also means it quickly clamps down on the power that the lifter motor can produce. The 4-bar arm in TB is very close to the bottom toggle point due to the shape of the bot and it takes a significant burst of current for the motor to generate enough torque to lift another 12lb object. The SyRen detects this and thinks I’m shorting the outputs into eachother. Result? No lift.

At least I can blow up the Victors when I want to. Not that I want to. Oh snap, please don’t let that happen.

The bottom line is, if the bot is not running by Wednesday night, I am not going to Motorama. That’s right, I’m skipping the event. It’s not worth missing two days of classes, including one lab, if the bot is going to be flakey and dysfunctional.

We’ll see where this goes. SP1 is almost there.

So now that classes have begun, I’ll have slightly less time to work on the bot. But it’s getting close. It’s getting Real Closeâ„¢. Here’s the conglomerate update over the past three days or four days.

The arm ESC in its “riser card” whose profile fits into a slot in the left side EBay. I actually think this is a great arrangement for single-board controllers, and might use this layout in future projects.

One of the standoffs is really fucked up. Wow, finally an off-center machine product that I wasn’t responsible for!

It also says something about the quality control at whichever company it was that made this. Oh well, can’t catch ’em all.

Both assembled EBays. These pieces also got some rough handling by the waterjet. As a result, I had to move some of the standoff locations. The empty holes are visible in all the plates.

Hey, a fitted arm link. This is the rear driven link. The gear will have additional holes drilled to reduce weight. Or, if I can get ahold of a rotary table, it will have some bits milled out of the core. The latter method retains more integrity under torque.

The front link assembly. In retrospect, I’m not sure why I insisted on putting the little linklets on the ends. Symmetry, I suppose. And, at least for the front link, to distribute loads over a wider area.

All are attached together with some counterbored 4-40 screws.

A quick toss-together to check for fitting…

And some movement for visualization and to check for interference.

Test fitting the EBays in their final locations. They actually add a great deal of rigidity to the chassis in terms of side-to-side deflection since they take up the entire length and width of their cavities. In fact, mounting them is a very light press fit against the chassis rails.

And now some adventures with garolite delamination. My method of choice ended up being using water-thin CA glue and wicking it between the layers such that they filled part of the interstitial bubbles. I locked the sheets in a vise after each bubble fill to let it set with the layers pressed together. This worked great for the most part, but some of the large bubbles actually had abrasive grit stuck in them and could not be closed all the way. No matter.

The small hole-area delaminations were filled after I countersunk the holes such that I could wiggle the CA nozzle into them.

Holes all countersunk. I went a little shallower than what the screw head called for in order to retain the strength of the material around the holes some more. Hence, the screws stick up very slightly above and below the bot. Not enough to cause trouble.

I also went over each of the countersunk areas with CA glue to seal the joint and hopefully prevent “transcendental garolite syndrome” which has been seen on TB since build 4.0 in 2006 – where the countersunk screw head magically passes through the hole without (visible) damage. It’s the strangest thing ever and I have no explanation for it besides the break occuring so quickly and cleanly that it snaps back together and I don’t see the separation.

Fitting the chassis along with top and bottom plates together. Yep, it fits.

Components in their final positions, but not yet mounted. The battery is slightly too tall to fit between the plates, but they have foam spacers which can be compressed somewhat. Otherwise, everything fits as designed. I love 3D design.

After the semi-successful planing operation described yesterday, it’s time for a fitted powered test! Things were a bit rough and I didn’t have real 1/4″ pins, and so no heavy lifting was done. However, the mechanism moves as planned.

Well, mostly. The arm does have a “hyperextension” mode where the driven link goes past the “straight point” pictured here. Past TB arms have had built-in hard stops at this point since they were C-channel designs. However, with this two-beam system, there is nothing preventing the link from traveling past it. I did include hopes to put in pins that act as hard stops, but I suspect the hyperextension might come in handy in some matches – the arm goes almost completely vertical.

However, if I don’t stop there, the linkage has enough degrees of freedom to actually swing back around the underside and lift the bot off the ground. This underside position is also one that cannot be recovered from using motor power, since it is a toggle position. Uh oh.

Is it time for some software limits?!

Or better linkage design?!

Bot on. Like, seriously. Moto is a week and a half away.