The Amplifier is a carriage on elevator featuring rollers to guide the Note in from the Feeder station or out to the Amp/Trap. The moving carriage with all its rollers, roller gearbox w/ motor, etc weighs only 8.2 lbs.
The carriage is comprised of 1/4" aluminum plates, pocketed only on the inside only partial depth (leave 0.030" thick web to be smooth surface for teflon).
Bearings on screws with spacers and locknuts provide the front-back guidance. 1/8" thick teflon stuck to the 1x1 tube upright and the carriage plate provide left-right guidance.
The carriage is lifted by a pair of 110T 5mm HTD timing belts, they are connected at the bottom by a jackshaft cleverly coaxial with the 3" tube redirect roller (roller is a deadaxle relative to this jackshaft). The âzombie axleâ (alive & dead at same time) design allowed us to get a jackshaft that didnât interfere with many Note paths in this compact space.
The elevator uprights are connected together to prevent spreading by a 1/4" plate attached to the tubes with a milled side-tapped plate/block that also holds the lift belt pulleys and a tensioner.
Comparing Carriageâs plate construction vs typical tube design:
The plates are much lighter and fewer parts, the plates can directly hold the rollers and the bearings, and are easily iterated if needed for new roller positions. For a robot-width carriage like this, having rigid crossbracing helps, but also having the jackshaft and driving from both sides ensured no binding. Thereâs a 1/4" thick, 2" tall plate attached with nut strips that helps as a crossbrace. The upper rollers being deadaxles also somewhat helps bracing, but not that much since the deadaxle donât go all the way through the rollers for saving weight (reminder, this robot weight 125.4 / 125.5 at the Championship.
We put a 1/4" thick plate with threaded holes inside the drivebase tubes so we can thread into those and they also serve as crushspacers for the swerve module screws going from bottom to top.
Blocks held up great, we used printed belt âclampsâ (they dont really even clamp on teeth) on the Laterator in 2023 and like those, these held up whole season because the load on the belt is relatively quite low. We are not climbing on this elevator motion. If doing higher loads like climbing, Iâd do a metal belt clamp, dyneema, chain, etc.
One of our most reliable and also easily serviceable subsystems. Can just untension the belt, slide it out of clamp, and pull whole carriage up out of the tubes. Most of our service was just refreshing the cattongue tape on the rollers. We even had some moments when we fell while climbing and the elevator uprights got bent and they were actually quite easy to replace, just some bolts.
I noticed that many of the elite teams are creating their own custom vision systems, and you guys have also developed some custom algorithms. With the those code you guys can easily create a better vision system. So is there a reason or advantage for using limelight?
Itâs funny you say this because after 2024 I had a bit of a minor revelation that our drawing boards usually have the right kind of concepts on them but we keep foot gunning ourselves because we keep choosing our robot direction recently based on our best prototype even if it has weaknesses to itâs long term success.
I made note of it to others in our team and weâre going to try not to do that again in the coming year, but itâs interesting to see some level of confirmation that choosing off the best prototype alone is not ideal decision making.
The advantage is we stick it on the robot and it mostly works, which frees up our (surprsingly limited) software people resources to go work on other things.
That being said, the cost of using a COTS item is not zero. There is always some work to handle, from understanding how to use it (and its interesting quirks and features), to integrating it into your larger system, to assessing failures and getting support from the vendor to remedy them.
We did enjoy using the photon vision simulation library this year though, and are experimenting in the off season with that DIY style vision system to see how we feel about using it.
In addition, we pressed in pins to prevent the print from dislocating between the layers. We typically use 1/16". The larger pin is around 3/16" (I forget what it is).
Auto-selection-from-a-large-library was one of my focuses at the end of last season.
Were you all happy with the solution you ended up at (hardcoded enumerated lists that detected if a final combo is legal or not)? Did it cause any pains adding new options as the season went on?
How undersized do you make your holes for pressed in pins on your markforged parts? Do you only do this with solid infill or does extra walls around the holes also work in lower load parts/if you are looking to add pins are we always looking at forces that would benefit from solid infill?
We tried doing a similar technique this year but instead of pins we used an undersized hole and self tapped a 10-32 though the piece with meh results. I know this is going to be a vague question/a it depends on xyz, but Iâm curious on what your thoughts (or anyone really) are on how this method would work vs a pin. One thought I had was if the threads on the screw created any weird forces or delamination in the hole possibly making the issue worse instead of better. Or if we shot ourselves in the foot by only doing extra walls and not doing a solid infill with this method.
The print orientation of this part puts all the layers in shear. We dont add extra walls but that could be an option. I dont recall the exact dimension we used, but they were press fits. Almost every mark forge part is solid infill, there are very few parts with hollow infills such as the redirect wheels or the cats tongue shooter wheel. We are trying to prevent shearing on a layer line by having these pins help transfer loads to other layers and reducing propagation of a local failure.
As for the screw method, i think I would rather have many 1/16" pins rather than a few screws. We are very cautious with 3d prints. We had some pulleys 3d printed with circular pattern of pins, but we still replaced them with metal pulleys once we procured the correct parts. For the elevator part, we still added pins between the elevator plates and the 3d print part to avoid having screws as the only fasteners in shear.
As you say this is a very open ended question. I suppose we tend to overbuild 3d print parts. In practice itâs more like: quickly build stuff, panic when they break and then make it better.
Another question, I noticed that you switched from using custom swerve kinematics to using the CTRE swerveDrive class. Is there a significant advantage to using that?
I noticed that you use a variety of roller materials (aluminum, polycarbonate, 3D printed) and grip materials (silicone, cat tongue tape, black nitrile tread) throughout the build. What criteria do you use to determine exactly what combination of tube and grip material you will use in a particular location? I am assuming that it is some combination of geometry for compression/packaging, weight, serviceability, etc but am curious about your thought processes.
You are spot-on for many of the factors that need to be considered when designing a roller.
Grip material
In most cases you want the grippiest possible to get maximum energy transfer from roller into object
Requires prototyping, sometimes you want less grippy (this year we removed the cat-tongue on one of the Amp/Trap rollers to change the angle the disc entered the trap and make the Trap work more consistently)
Some materials are more susceptible to wear or damaging the game object. The treaded wheels on the Shooter stage 1 was to pick a material that wouldnât get clogged with dust from the skin of the Note and change its grip amount
Tapes like cat-tongue or grip tape are good because they can wrapped onto any size diameter or roller material
Grip material is sometimes coupled with roller diameter/material, can only put silicone on certain size tubes, canât put tread on something really small, etc
Grip material sometimes driven by need for compression, you need compression to generate friction and transfer energy somehow. Either
The roller compresses (flex wheels, thick rubber on roller)
The intake squeezes (floating/sprung arm enables moving rollers)
Or the game object compresses (object made of foam, inflated, or otherwise squishy)
Roller Diameter
Along with roller placement, driven by the architecture / implementation, compression amounts, no-dead-spots, etc
Whatâs even lighter than optimized rollers is fewer rollers total
Often coupled with grip material, see above
Larger is along stronger/stiffer (good for intake roller going outside robot) but larger is also hard to package, maybe heavier, and higher surface speed (more gearing-down required, so again more weight)
Axle implementation
Dead axle vs Live Axle usually about selecting for serviceability vs weight vs stiffness vs complexity
Dead axle often lighter weight and stronger (axle acts as a stiffening standoff, plates donât have big bearing holes in them)
Live axle sometimes necessary if shaft passes through a plate/tube (ex. cantilevered west-coast-drive wheels
Live axle enables smaller diameter roller/wheels, dead axle only works on rollers ~1" OD or larger
Roller Materials
Usually is driven by the above requirements which are more critical
Live axle? Choose between 3/8" or 1/2" hex pending stiffness and strength requirements
Dead axle? Pick a purchasable tube that is strong enough for loads
Polycarbonate is good for intakes outside the robot, can be lighter than aluminum if the same wall thickness
Aluminum is best for strength-weight ratio if you can get in very thin wall (0.030" is quite strong for internal rollers)
3D printed wheels can be ok if seeing low loads and can hollow infill to make real light.
Fewest number of parts and quick to assemble
Perhaps I should do a presentation on intake design and lessons weâve learned over the years.
How did you guys prevent a note from going back into the intake from your feeder and instead into your amplifier elevator system. Did you guys just reverse your intake rollers to pop the note up into the elevator or is there a physical barrier once the note is in the feeder that forces the note up to the elevator? How did you guys design that?