This is a picture of the shifting mechanism of the friction clutch transmission that we did as a 2013 off season project. The hex broached part is moved similar to how a dog does, but with a 100lbf piston to give us sufficient friction force. The linkages connecting the selector to the shoes are at a 10 degree angle under engagement so we get around a 5:1 leverage in friction force off of the pistion force.
Our goal was to decrease our shift engagement time and be able to shift under load. The transmissions were used on one of our robots at the MadTown Throwdown and worked without issue.
It looks like with a drum on each second stage gear, that you would need air pressure to engage either gear set, regardless if it is low or high gear, is that correct? Seems like a big plus to that is you can shut off your air pressure after a match and the robot would roll freely with no motors engaged.
Based on the results of this off season project, do you see enough benefits to use this approach (instead of dog shifting) if the game requires a two speed drivetrain this year?
A roll pin connects the triangular central piece of the shifting mechanism to the shift shaft (that tiny one you can see in the photo). It went through the output hex shaft through a clearance slot, which isn’t shown in the photo. (EDIT: I’ve been told that it was a dowel pin, rather than a roll pin, which is stronger.)
I hope that this doesn’t get lost in the game hint chatter. This really is a huge technological leap, and a great experiment, similar to how a ball shifter was a few years ago. It’s just a pity it took 3 weeks to machine. If it didn’t, I’m sure 971 would be rocking it next year. I have to hand it to Travis, Austin, Brian, and co for pulling off such a innovative, different and cool idea.
This is half my opinion and half my recollection of conversations with Austin, so please don’t hold Austin accountable for my statements.
They had a clear problem they wanted to solve. An effectively utilized autoshift function with properly picked low/high ratios could potentially shave seconds off each match (4-10 maybe?). This time saved isn’t huge, but could amount to a 5-15% increase in scoring each season.
Both the dog and ball shifter do not shift instantly, and neither does this clutch method.However the clutch method does shift faster than the previous two, and with a smoother engagement; this better blends with their goal of smooth autoshifting.
You can debate whether or not their overall goal is valid or cost effective, but please don’t say that it’s a solution in search of a problem. Valid analysis and engineering went into this, and it wasn’t created for the sake of coolness.
If you are unsure as to the exact purpose of someone’s particular project (one that they’ve more than likely poured a tremendous amount of blood, sweat, tears and money into)- there are a number of different ways to approach that viewpoint that will not cutdown someone else’s hard effort.
Possibly begin a dialogue centering around why shortening shifting time is important to 971…
Also keep in mind that you do not necessarily need to optimize-for-efficiency every single action your team takes. Sometimes doing something the hard way, or doing something new, or even something for the sake of coolness can be a great exercise in teamwork and inspiration…
Travis, Austin and co., a job very well done. This is a really cool little clutch. Are you guys feeling good about trying to implement it this upcoming season? (Assuming shifting makes sense for the game).
I can see why taking seconds off of operation time would be helpful, and I concede that I shouldn’t judge the necessity of this mechanism without knowing the design process behind it.
I wonder, though, if time wouldn’t be better spent on low-hanging fruit – some mentors and I made a lively discussion of this a few months ago when talking about how 1678 could improve our competitiveness. I know 971’s bots always have high score ceilings, but often fail to execute to full potential. Would not it be more efficient to ensure basic functionality in each match, as opposed to some fantastic and some mediocre matches, rather than build a smoother autoshifter? (This isn’t to say that both of these improvements couldn’t happen simultaneously, of course.)
I’m not on 971 (actually on your team T^2 ;)), but you can’t make these assumptions without any understanding of team structure, goals, etc. I just spent the last 5 hours at 971’s shop, I can assure you that developing a friction clutch transmission is certainly NOT the only thing 971 is doing this off-season. Arguably, they’ve made larger strides this fall than we have.
Thanks! We think so as well. Brian, one of 971’s students, really took the idea and ran with it, making it what you see now.
As Roger said, the gearbox took a lot of time to machine. If we had sponsors lined up to help with it, we would seriously consider using it during the season. If we were to need to smoothly disengage power to something like a kicker in 2010 or 2008, we would consider using a similar design.
We can do all sorts of really cool things with the transmission that I would never try with a dog transmission. We were able to shift from high to low gear while traveling at top speed. The neat part about that was that you could hear the CIMs spin up to above their free speed and ‘engine brake’ the robot back down to low gear top speed.
That being said, this project taught us a lot about what is really important in a drive train and brought into question a lot of things we had assumed were known. We are now beginning to model things better and start to optimize for what we have just learned is important. Most of the algorithms developed for this transmission have already made their way back into the code for our dog shifters, and we will most likely pick different gear ratios next year. I consider this project a success just because of that.
You are correct that you need pressure to engage the gear sets and our initial thinking was exactly along the same lines as you outline, however we found that this didn’t turn out to be the case. When you drain the pressure on the robot, the transmission stayes engaged in gear to the point where the motors spin as you push it and it wasn’t possible to dis-engage it by hand. We took appart one of the transmissions to check wear, and found that the shoes were wedged into the drum pretty well. We are not sure of the exact mechanism of this or how strong the engagement is, as we got everything together at the last minute and have been so busy with other projects since that we haven’t gotten a chance to dig into this further. The nice thing about the linkage approach rather than say a taper lock is that whatever lock we have dissengages pretty smoothly because the shoes pull away from the face as you pull on the selector.
We are working on that too, but a lot of that is very game specific, so it’s hard to address with specific projects in the off-season. We’ve had a very detailed discussion of the robot issues last year and how to avoid them again. I will also say that there has been a conscious choice during build to go for riskier designs with higher potential the past few seasons.
This was one of several off-season projects, one of the others being a 3rd robot, HP-load only for Ultimate Ascent that we ran at Madtown. That was designed and built almost entirely by freshman and sophomores, with some help from the upperclassmen. The friction clutch is a lot cooler, and I don’t know of anyone else in FRC who has tried to make one as a replacement for the dog shifter, so we thought it would be more interesting to CD.
I’d argue that this project has a lot larger benefit than just smoother shifting. Dog and ball shifting were both adapted for FRC by teams using them in projects like this. And then brought to a larger market by companies like vex pro and Andymark. I could definitely see something like this catching on in the larger FRC community once to difficulty to manufacture comes down. I could think of a lot of situations where this would be more desirable than a dog shifting mechanism. And I’d argue that just the exercise of attempting something complicated and gaining engineering knowledge from it is kind of the point of FRC. Some of the most rewarding projects I’ve worked on have been ones where the mentors and students were both learning from them.
So my hats off to all those involved at 971. On another note I was wondering what ratios you guys geared this box for? Also what material were the clutch pads made of?
This is the closest we have of what you are asking. You can see the large gear and steel drum that is in this picture, and then the opposing steel drum and smaller gear. Unfortunately all of the inner workings get hidden once you assemble it.
Low was geared at 14.3:1 and high 4.9:1 with a 3.5" wheel. (5.8 and 17 free speeds). In hindsight, we would have made our low a bit higher. It took 0.5 seconds to get to top speed in low, which made the shift point a little too early. One of the things we learned from this gearbox is that the optimal speeds for accelerating to high gear assuming shifting are different than speeds chosen for pushing. We have a student who is currently doing the simulation effort to see what the best speed pairings are for an auto shifting robot. That and how fast we can make our low gear before we start blowing breakers in low gear also.
The clutch pads are made of a machinable brass. The drums are something like a 4140 steel. We really didn’t know what the best friction pairing to use would be, so we ended up settling for brass and steel because it has a decent coefficient of friction and because we have seen this pairing used in other friction applications. We ended up designing the clutch based on a friction factor assuming that it was greased brass on steel because we figured we couldn’t garuntee that we wouldn’t have grease get in there, and we figured that it might help with wear if we preemptively greased it. So far we have been pleased with the clutch material chosen, however we have in no means run it through the paces of an entire FRC competition season. If anyone knows of a better clutch material pairining, please let us know!
Part of why we built the transmissions was because the theory behind the design was solid, but there were enough uncertain factors like this that we really couldn’t tell for certain if it would work without building it.
On the note of clutch material, did you ever try any of mcmasters clutch lining? It seems like using something like clutch lining would let you use a much weaker cylinder and save some weight.
If you haven’t already, I would suggest looking into moped and scooter clutches. They use very small pieces of clutch lining glued to the shoes and operate with very little force between the shoes and drum.
Thanks for posting these pictures and fantastic job on the transmission! you’ve definetely given me some food for thought.
33 made this point a few times when talking about optimal acceleration (here’s a link). I second Adrian’s suggestion of trying McMaster’s brake/clutch lining. (Their thinnest is 1/8".) On another note, are you worried about deformation of the steel drum? It looks like your wall is around .050, and with the amount of force you’re applying with the cylinder, I could imagine that the wall would get bowed out over the course of a season.
Something (semi) interesting to note, we did some extra testing of this drum transmission today at 254’s lab. When we removed all air pressure, and ran the robot against the wall, we could slip the tires without slipping the clutch. There seemed to be more enough friction in the clutch, even without normal force from the piston, for the clutch to remain fully engaged.
Further testing at low but non-zero air pressures failed because… I broke their robot removing the breaker from the compressor spike.