I’ve been tinkering with implementing a capstan drive reduction system for “low” load stuff that can benefit from high precision position control. I think there are several potential applications (wrists, turrets, etc.).
I started with the test stand model in the above link, and have a hand cranked version so far. Initial results are really promising - there’s no discernable backlash. The reduction is around 8:1.
It’s grossly the ratio of the sheave diameter to the capstan.
OD
Driven Sheave
136
mm
Capstan Pinion
17
mm
Approximate Drive Ratio
8
In reality the non-planar motion of the rope makes the calculation more complex. I guess you need to account for the helical travel path. I don’t care to go down that math rabbit hole. I’m content to slap an encoder on the wrist or whatever and ignore the actual ratio.
The natural frequency is one of the characterizing parameters of 2nd-order linear ODES in standard form. It is equal to the square root of (stiffness / mass). It also is the stiffness a system will oscillate at if it had no damping.
In the applications of FRC, you want higher natural frequencies because it means your dynamics are faster, however, its not something I would account for when designing unless you are both familiar with system dynamics and classical control theory, where this type of content is usually taught and applied.
If you kick the mechanism, and measure the ongoing vibrations after, what’s the lowest frequency you see things bouncing at?
Mechanically, this is a measure of how stiff the mechanism is.
Controls-wise, this represents the limit on how fast you can vary the actuator with the output still moving in synch with the input… and, if it self oscillates.
No. Its more relevant than you think. Control theory is an extension of system dynamics that uses a lot of the same math, where all the control system is trying to do is change the system dynamics into the dynamics you want rather than what the system has by default. They are so related that many mechanical engineering programs will have controls as a required course, potentially more controls electives available. You are interested in designing a capstan for its nice system dynamics after all.
It is best (but highly uncommon in FRC) practice to design both the mechanical system and the control system together, as they are pretty much the same thing and good systems are designed to have nice dynamics to control. In fact, 971 almost certainly cares about natural frequencies for control reasons.
After coming across 1477’s offseason capstan elevator, I had to take a crack at making my own. So far I have a hand cranked version that seems to work pretty well. This style of capstan winch would lend itself well to a variety of linear tasks.
I’m playing with the sheave geometry. Currently I’m using combs like they did, but am experimenting with helical spool geometry as well.
I’ve drawn a couple of potential continuous elevator gearboxes. One is fixed geometry and one adds a daughter plate for the motors to accommodate different ratios. I’m not in love with it yet.
Interesting design. I think definitely workable. But it is not using the Capstan for gear reduction and the inclusion of the spur gears should introduce the backlash that it seems we are trying to rid ourselves of?
For the rotary actuator in the first post I totally agree. Here, I don’t think a single stage of spur gears would be too detrimental to an overall system. One degree of backlash in the first stage nets roughly 0.01" of difference in line pull.
I like this concept for a continuous elevator because there’s no spooling/unspooling to fool with, and there are no jackshafts across the whole robot.
I changed my mind about that. The way I wanted to rig this would have required ropes to transverse the robot anyway.
So I changed the design up a bit to keep the passthrough capability but simplify the rigging. The rope routing is very much inspired by 254 2018. The up ropes will each be tied off to a spring on the carriage. The down ropes will be connected and tensioned together at the carriage base. The actual rope route is non planar.
I think i can get enough wrap on a single capstan to make this work. If more friction is required, another capstan and a figure 8 can be added below the existing one.
I built a 3D printed small scale version of the above elevator. It works pretty well. I’m working on adding an additional stage. I’m really pleased with its performance so far. If 2025 needs an elevator, I’m taking a long look at this concept.
