971's chassis

This year I think 971 has built one of the most effective and innovative chassis, and consequently I want to emulate what they have done.

I noticed from the Flikr account that this year 971 departed from from what I presume the “original” three piece design. Could the octagonal shape not be achieved in three pieces? (image here)
That brings the questions: do the ribs provide significant rigidity?
What’s the advantage of one piece modules over the traditional railed approach? (image )
What is the advantage of PEM nuts, is there a closer image of them in use, how do they hold up to the beatings and vibration? (image)
Finally, what are those giant copper type plugs?

Those Copper plugs are called cleco’s and they are basically a temporary rivet. You place the cleco in the holes you want clamped together and it keeps them aligned with each other. It provides an easy way to fab lots of sheet parts with less risk of assembling as you go that using rivets to start might cause.

We ran the “original” 3 piece design in 2012 and 2013, and it has served us very well.

We couldn’t find a way to make an octagon with a 3 piece chassis, which means it probably isn’t possible or easily to machine. We wanted an unbroken flange to help form a backbone to tie everything to and to tie everything together. The best way to do that that we found was to set the chassis up as you see this year. We started by figuring out where to put the wheels and pulleys, and then built the frame to hold everything in place. This resulted in the jog on the inner face, which resulted in an internal corner which couldn’t be bent without breaking things up into more pieces. We also needed a parallel face to bolt the wheel tensioners to, which drove the inner frame rail shape that you see. You can see the set of 8 #6 holes that define the tensioner location in one of the pictures.

In theory, they should provide more torsional rigidity and prevent the tube from collapsing. We could probably leave them out, if we really wanted to, but I’m a fan of overkill when it doesn’t hurt you… We also choose to bolt our superstructure down either to them, or right next to them, which spreads the load a lot better by spreading it through the rib.


What’s the advantage of one piece modules over the traditional railed approach? (image *)

By railed, do you mean a classic WCD chassis? We have lots of sheet metal sponsors, and a large amount of experience working with sheet metal. We build robots that play to our strengths, and this chassis does. By building a big tube instead of 2 rails like the old AM chassis, we get a lot more structural rigidity. It is quite hard to flex one of our chassis. Our wheel tensioners are designed such that the act of tensioning the wheels locks the shaft into the frame, so it really isn’t very hard to get the wheels out. The wider wheels and wider body tube lets us flip the CIMs over the wheels, which opens the belly pan up significantly.

PEM nuts give you the strength of steel threads in aluminum. When used properly, they are quite strong and have no problems. When used improperly, they fall out and are a pain to use. When using PEM nuts, make sure that you never let any load be put into them that would try to press them out of the material. On our drive base, we only used them to mount speed controllers to our belly pan. In our superstructure, we really only used them in a couple places to let us remove parts quickly.[/quote]

I meant something like this, but I think you answered my question. I don’t quite understand how the tensioners work; can you post an image of them in action? Also why not exact center to center belt?

I would also be very interested in learning how these tensioners work. I saw the picture you guys have, but how do you make sure that both sides of the shaft are aligned so that the wheel is straight?

I have a couple more questions:

How hard is it to replace the timing belt with this setup?
Sheet-metal wise, what tolerance does your manufacturer provide?
What thickness and alloy of sheet metal do you use?
Do you guys press in bearings into the sheetmetal? If not, what else do you add to keep them in there, and if so have you had any problems with them falling out?


What is the advantage of having the center drive wheels offset right/left, compared to other traditional 6 wheel drive set-ups?

This is a pretty good picture of how our tensioner blocks work. The block in the upper left is positioned such that the side facing up would be bolted to the sheet metal.

The inner block slides relative to the outer block, the bolt comes from the front and provides a pull tensioner. The outer block has a lip that combined with the sheet metal fully captures the block. The springs return the block to neutral when you de-tension the wheel.

When the wheel is fully de-tensioned, the dead axle lines up with a hole in the frame. This allows us to push out the dead axle. Our wheels then drop out as one module (there should also be pictures of our wheels in the picassa albums). Once the wheels are tensioned, the dead axle no longer lines up with the hole and is captured by the frame rails. We find this to make wheel changes very easy. Also, with belts, it is critical that you are able to replace a broken belt quickly, and this does a good job of enabling this.

We align the wheel angle by eye. It turns out that the human eye is pretty good at judging this, we typically sight if the wheel is parallel to the lightening hole. We have never had an issue of having a miss-aligned wheel. We ended up going to a separated tensioner design because we really wanted a pull tensioner for various reasons (2012 was push and had lots of problems), and we couldn’t come up with a way that was compact enough while maintaining accurate wheel angle.

We don’t do exact center to center because it is hard to install them this way given our frame, but more importantly because you do not get the full strength performance of the belts if they are not tensioned properly. Our experience is that it is not reasonable to do this reliably with FRC tollerances. The drive belts are particularly important to get every last bit of performance out of because they are running out of spec. We do exact center to center in a lot of other mechanisms to save design and manufacturing time when the loads are low (IE the belt tying our two intake wheel sets together and the first stage motor reduction on our tusks).

-Replacing timing belts is pretty easy. The center wheel is on a dead axle also, so it can drop out without us having to disassemble any of our transmission (the problem of belts in a traditional WCD).

-We always find that our sponsors are better at the tolerances than they are willing to commit to. They will give us their standard bend tolerance of /pm .015", but our experience is that they are much better than this. I think we spaced a .010" gap between anything in the drive train that was a bent part that needed to fit inside of another bent part. We didn’t have any fit problems in assembly.

-We use mostly 060 on the robot, however the drivebase was half 060 and half 090. Any sheet metal that could be probably impacted by other robots (bellypan, outer rail, and maybe one or two others that I can’t remember) were 090. The rest was 060 to save weight. This year was really brutal and both chassis still looks like they are new (probably helps that the bumpers also held up very well).

-We do press a lot of bearings in, however we are re-evaluating that for the future based on some bad experiences this year. We were disappointed to find that the vex pro bearings have a very large undercut by the flange that is about as thick as a piece of 090. Their bearings (especially the hex bearings) have a very poor press in sheet metal. We will likely be sourcing different bearings for next season.

So the axle is held in place by the block and friction against the frame rail? Couldn’t it just be cantilevered?

Not if they wanted to make it easily drop-outable, one of the primary advantages of dead axles. The way it makes sense to make their frame, the belt is right next to the wheel anyway, so they might as well avoid the complexity of making that shaft live. The whole retaining/dropping out is really smart, and I’ve never actually noticed that specifically, despite having looked at their drivetrain more times than I can count.

By the way, good to see you come out of your CD hibernation Travis and Austin.

No, the axle is sandwiched between the frame, there are two tensioner blocks per wheel. This picture shows the tensioner in de-tensioned state. Once the axle is installed through the hole, the inner block (with the bushing) will move to the left and thus the axle will be captured. Not sure what you are envisioning with cantilevering…

I see; I meant something similar to how they cantilever west coast drive axles.

Thanks for the quick reply: some follow up questions.

  • I’m assuming you guys are using 5052. Confirm/deny?
  • Are the tolerances on the laser cutter and the turret punch the same?
  • How has your experience been in bending **over **holes? Do the holes deform and you only use them for access sake? Or is it reliable to allow critical parts to interface with those holes?

Thanks a lot for answering these questions; very very helpful.

5052-H32 for the parts that are bent, and we use some 6061-T6 when things aren’t bent and need to be strong.

No. I’d trust a laser to ± 0.001 or better, and a punch to ± 0.005. Sponsors tend to be pessimistic about what their machines can do.

We don’t do it to the extent that we will redesign the part to avoid doing that, or only use them for access. We keep all holes about 3/8" away from any bends, otherwise the holes will blow out and the k-factor will be affected. If there is a hole too close to a bend, we will make it into a slot and break the bend. Some machine shops will leave the holes out and put them back with a mill after bending it, which is a waste of everyone’s time.

We do use 5052 for its ability to bend.

I do not believe that the laser and the turrent punch have the same tolerances. The laser can hit holes within .001 and I believe that the turret pouch will get within ~.005, and Travis and or Austin can correct if I am wrong. We will usually spec out holes on transmissions and whatnot that need to be within laser tolerance and we try to make everything else “punchable”.

As for bending near holes, you can deform the hole when it is located near a bend, but in a somewhat predictable way. For holes that don’t require major accuracy like our tensioner allen key clearance holes, we are okay placing them near bends, but we try to avoid it when possible. You can also see in the siderail of our drivetrain we put a hole on the inside flange where the transmission goes. While we don’t like breaking up the flange, the hole is necessary for our second reduction from out transmission.

slots tend not to deform much when the bend is in the middle of the slot. There have been a few side projects where we do some interesting things with this, but I cannot find any pictures right now…

Hope I have answered your questions

Edit: Darn I was too late

The students have been doing a great job answering all questions on CD, so there hasn’t really been anything to add.

Based on the pictures it looks like you’re adding a countersink before putting in your rivets. This seems rather time consuming, so I’m guessing there is a good reason. Care to elaborate?


Is the CAD model for this available somewhere? I’d like to take a closer look at how you worked around the internal bends on that inner face.

If we had to do the countersinks ourselves, yes it would be time consuming, but it is relatively easy for our sponsors to quickly do the countersinks (and to do them right for the rivets). This then reduces the number of protrusions that we have to avoid interfering with or that might catch on things.

We use countersunk rivets. There are punches on the CNC turret punch which will make a .129 hole that is countersunk to 120 degrees, all in 1 hit. This makes it so the bottom, front, and sides of our bot are perfectly smooth and can’t catch on anything.


The CAD model isn’t available, and we have no plans to release it. We subscribe to the 254 CAD release philosophy, and choose to share pictures and explain the why about what we do rather than directly share the model. We have a team meeting tonight though, and I’ll (try to) have one of the students put together an exploded view that should answer your question.