COTS elevator to lift robot with single speed gearbox?

Technical Technical Discussion
,
1

There’s an active thread about COTS elevators, and I see the elevator in TTB’s video both lifting normally (i.e. to lift objects) and also doing a hook-climb lift of the robot.

I remember teams (e.g. 2910 in 2018 I think) using shifting gearboxes to maximize object lift speed in high gear and also accomplish the robot weight lift in low gear.

My question: Is that really necessary now? I seem to remember other teams being fine using single speed gearboxes, but I didn’t find a great thread on exactly how they optimized for that. In 2024, maybe we just need to use 2 of the more efficient brushless motors, use one of the engineering calculators to pick a goldilocks gear ratio, and that’ll just work with a COTS elevator with little fuss? Examples (using 1.75" radius to correspond with TTB’s 22t sprockets):
2 Vortexes
2 Krakens
@AriMB , are the above utilizing your mech ratio calculator correctly?

I haven’t tackled this problem before, but I think it’d be a good one to discuss with students. Any insights to share?

2

Do you mean single stage or single speed? Single stage would probably be rather difficult, but single speed doesn’t seem unreasonable. With 2 Krakens I was able to get 4 feet of travel distance in under a second. Getting 12:1 is probably going to be easier in multiple stages, especially since you’d likely want to gear it down some more to protect the motor, but it does seem like you probably don’t need shifting.

image
image871×519 14.1 KB

1 Like
3

Yes, I mean single speed - corrected the title. I also made some updates to the original post.

For this purpose, what’s the rule of thumb for max current? I see you’re at 40A, but I recall others going with more like 60A. Thanks!

4

I don’t really know. I’d assume it can be higher because it’s really only going to be at this load when climbing. I was just keeping it under 40 because I like to be pretty conservative with any numbers I post on here. In reality I’d probably just throw a couple 5:1 MaxPlanetary stages on this because it would still be moving as quickly as I’d really want it to and would just provide more protection.

1 Like
5

I use 20A for a lifter that activates once per match. This is because that’s ~3x-4x Factor of Safety (FOS) on a 40A breaker meant to operate for a few seconds. 3x FOS is pretty safe for FRC and makes PID tuning pretty easy as well. You can always increase the speed later if you find that you are held back by lower speeds. You can also add another motor and halve the gear ratio in the design stage if your calculations show that a 20A load is too slow.

In 2018, if you had a 2 stage elevator, you could often rely on the 1st stage being half the speed (and double the force) to lift yourself up, especially if you had a brake or ratchet to hold it after the initial lift. This would let you use a low FOS.

1 Like
6

We had a shifting elevator motor in 2018. The robot could actually lift itself in high gear, but could only lift a buddy in low gear. Optimizing a single ratio to lift both a ~20 lb elevator+cube, and a 300 lb robot+buddy would have led to sacrifices in cycle speed.

The other advantage is that in low gear the ratio was high enough that we didn’t need a separate brake to prevent our robot+buddy from falling quickly after the buzzer.

The GDC got rid of buddy climbing in 2022 though, so maybe this application no longer exists.

1 Like
7

You could throw more motors at it. 4 Krakens at a 12:1 on a ~1.7" spool will give you the power to do what you want while not drawing too much power.

1 Like
8

Only really if you aren’t buddy climbing. Even the older generation of brushless motors can climb and score on the same ratio if designed correctly

You can use cascade rigging to essentially get 2 different ratios in one. If you climb off the first stage, but put your manipulators on the 2nd or third stage, the climb gets the high torque while your manipulator gets an 1:N upduction compared to the climber.

On a 2 stage cascade, you can gear to 16:1 on a 2" diameter spool with 2 falcons to pull 38.5A per motor and climb at 3ft/s (too fast, but you can run the motors at lower power), while being able to lift gamepieces at 6.5 ft/s with plenty acceleration while pulling much less current.

image
image1193×662 70.5 KB

image
image1204×675 71.1 KB

However, in games like 2018, if you wanted to triple climb, you would need to gear 45:1 (with brushless motors that didn’t exist back then) to climb at 40a per motor at 1ft/sec. With a 3 stage cascade, this would have resulted in a 3.6ft/s speed for gamepieces which is slower than you’d want. There isn’t really a scenario in which single speed buddy climb works without major sacrifices.

I generally would pick 40a as the limit. You will get some factor of safety from the fact that breakers don’t pop instantly so the mechanism will still work fine if it ends up pulling more current than you calculated. In contrast, if you gear for higher currents like 60a, you can pop the breaker if you aren’t careful in making sure the mechanism won’t actually pull more than 60 or whatsoever amount of current. This is even more important if before the end of the match, you keep the robot up before the match ends by stalling the motors.

If you have a climb mechanism such that break mode is only sufficient for 5 or so seconds after letting go, if you climb early at all, you will have to pull 60a to keep the robot up for several seconds before the match ends which will likely pop the breakers.

4 Likes
9

Wouldn’t rule it out for future games, though. 2-robot or 3-robot climbs have occasionally cropped up over the years even if single-robot climbs are the norm.

2 Likes
10

If we all acknowledge that we’re living in a brushless future - then there’s no real reason to shift to do a dual purpose lifting elevator.

If memory serves me, 2018 254 had a high speed scoring ratio, then a lower lifting ratio. I believe this was common among the top tier in 2019 as well. Most of those robots were using 775’s in the elevator - so they’d typically add an additional motor (or two) and then add shifting if things got iffy.

After seeing what teams have done with 1 or 2 Brushless motors - Neo, Falcon, Pick - there’s no need for a shifting ratio, the mechanism just needs to be designed right. If the breaker curves can be trusted, there’s about 150% of rated current available through a given channel indefinitely, and all FRC BLDC motors (excluding NEO 550s) have been proven to survive at least 60A, for any plausible period of time.

So essentially you could design around pulling 60A through a single motor, knowing that it needs to be done once, maybe twice, for less than 3 seconds, each match.*

Would that be the literal edge of sane design? Probably, but if you dialed it back to 50A, it all still works.

*there’s a caveat here, especially if you’re doing everything based on current limiting, current limiting on any “high inertia mechanism” needs to be based on the highest transient current and not the highest load current. In the case of an arm or elevator, your highest current will be during acceleration from a stop, especially when moving against gravity. This often will spike to something well in excess of the breaker capacity, but it’ll do so for, milliseconds. Lifting is a bit different, since it’s a sustained load, but you’ll still see similar transient behavior, so the loaded current will be much higher, and so will the initial spike. (That said, you won’t pop a 40A breaker if you can get the work done in less than a second or two)

1 Like
11

Thanks for the responses. We have quite a number of Neos on hand, and we’ll be looking for ways to use them up because the recently announced motors seem like the future, and we want to keep up. We do swerve (assuming game allows) now, and we also have gearboxes in inventory that are more oriented to doing tank, including some AM Evo SS and I’m sure some toughbox minis.

It seems like a decent way to go would be to just use 4 Neos + a couple 2-motor gearboxes, hacked onto a TTB elevator. Even the default toughbox mini gear ratio of 12.75:1 seems to be safe enough to try. Any gotchas with that concept? (Tryin’ to be thrifty)

12

I may have missed something here, but that elevator would either be grossly overpowered or best disguised as an industrial forklift.

Unless you were trying to lift 3 full weight robots at once, and even then, it’s still a bit overkill. One of the big things that jumps out to me, personally, is the diameter of your pulley. 1.75” OD isn’t big, but it’s also not small. Decreasing the pulley OD will require less ratio to get your torque up, similar to how a smaller wheel requires less ratio.

(4) Neos is A LOT of power to throw at anything. Like full blown, let’s hit things with a heavy hammer sort of approach to the problem. It’ll work, that’s for sure. It also may break itself because you have enough power to destroy anything that is worth working with.

(Unless you’re lifting 3 robots, in that cause it gets fun, and keep doing what you’re doing. You’ll want to watch the ratings of your power transmission members, and be very specific about the output stage support and how you pull down on the lift. #35 chain would technically be a “no brainer” but you need to support it from the center, two 25 chains that are (Actually Heavy Duty) would work, so would (2) 15mm Belts, but you’d need to load them symmetrically.)

TLDR is that if 4 NEOs are really your answer and you gear at 12:1(ish) and you stick with a 1.75” pulley, you need to treat your power transmission more like a drivetrain than an elevator. If you look at the math radiometrically, 1.75 OD at 12 is roughly the same as 4” OD at 5.XX. Play out the loads through the power transmission, and you’ll see that it breaks a lot of things without proper planning. (Your 500+ lb load case will break most light weight FRC stuff, especially considering how quickly 4 neos can generate that torque)

1 Like
13

When using rope, large diameter spools have the advantage of being better spools. They use up less horizontal width for a given amount of rope, meaning you have much smaller fleet angles which makes rigging significantly easier and doesn’t require as good of a tensioning system. Especially when doing continuous rigging, perfect spooling eliminates most of the rigging problems. In most cases, I would prefer a larger diameter spool with the extra gear reduction stage rather than the smaller (and slightly lighter) spool because it just eliminates most rigging problems.

14

The TTB elevator power transmission includes two # 25 chains on two 22t sprockets (which are 1.75") on a hex shaft, so that’s where the 1.75" number came from. Agree it’s a lot of power, and that was on purpose… I was thinking of it as a way to do a SS, use up Neos, & still have a really fast elevator that could climb. Software and current limiting are tools that could be used along with it. That said, if you think it’s at risk of breaking, good to know… Maybe won’t do that then. Your feedback & the discussion overall’s what I was looking for on this, because I don’t have a strong intuition from experience on this one.

15

While none of 254’s elevators are COTS, they may give insight into power and reductions we’ve gotten away with in the past for fast climbs.

image
image3141×653 194 KB

2018 we were lifting both ourselves and a partner (combined 300 lbs) for ~3 seconds, so we wanted the current per motor to be below 40A, with 4 775 pros the only way to achieve this was with a 2-speed, so we had a dog shifter in the gearbox.

2019 the elevator climber just had to lift itself (150 lb load) so we had just a single speed gearbox with 3 775 Pros, again geared to below 40A.

2022 we uniquely had the Truss React climber which had a peak load of (we calculated with CG vs lever arms) 380 lbs during the truss reaction. It was also our first year using the new REV PDB and its new breaker. For CVR we originally geared kinda standard, but then we experimented and found that for the brief 1sec during the truss react we could actually pull over 90A without tripping any breakers, so we swapped gears and went much faster for SVR/Champs.

7 Likes
16

Reject modernity return to PTO

3 Likes
17

The additional context here makes your math make much more sense. If you’re planning around 25 chain, then the 22T sprocket will lessen the tooth load as you should have 10-11 teeth engaged at any moment.

You’ll still be dancing around the breaking strength of the “wrong” 25 chain - HD chain will be better, but you’ll need to make sure that you keep the current limits low enough to keep the maximum output within the tensile strength of the chain.

The above being said, you can reference Torrance’s post about the logic behind the Poof Elevator. In their case, they threw motors at the problem to decrease the current draw under worst case loading / endgame loading. This is arguably ideal, since it gives you some design headroom should a battery ever be “bad/not good” and starts seeing high 11V at rest in end game, in theory, you can still pull enough power out to get the robot up without anything too difficult, especially if you’re only lifting one robot.

That said, your gearing is still probably on the aggressive side, and you may not be able to catch a “spike” in current before you snap the chain. You could either drop the gearing down a smidge to reduce the torque and raw lifting power available, or do dynamic current limiting on the NEO (basically you’d push a new limit to the motor when assuming a climbing pose that is designed to ensure that the power transmission stays under safe operating load)

1 Like
18

Hey Andrew, thanks for sharing those screen shots! I hope folks noticed those as they were surfing around yesterday because it’s really helpful to connect them with your write-up for each year. The closest year to the hypothetical scenario I’m looking at is 2019.

I’m now going to dive down a rabbit hole, looking at the linear motion scenario in the different calculators available…

I used @dydx’s calculator (available here) that is based on JVN to replicate Andrew’s results:

image
image886×537 26.2 KB

BUT when I try to replicate the scenario using @AriMB 's ambcalc.com, I get very different results for the projected current draw (78A instead of 32A). I see that the ambcalc displays a low motor efficiency of 28.8% based on where the scenario sits on the motor power/efficiency/torque/speed graph. Maybe that’s most of the difference? Anyway, since I like Ari’s calculator pretty well, I have been using it, and that is mainly what led me to go in the direction of 4 Neos instead of 2 or 3… I wanted to lower the current per motor and increase the efficiency, while retaining high speed.

Since we sort of count on these calculators to help us identify safe useful starting points for our designs and because the variance between the first two calculators was so large, I decided to also plug the numbers into reca.lc by @jtrv . I learned the linear motion calculator here is trying to help with different things (really cool things actually) and doesn’t provide projected current draw as an output. BUT if I plug 40A in as the current limit, the projected time to move the travel distance of 31.5 inches is 2.5 seconds rather than less than a second projected by JVN. So my assumption is this calculator, like the AMB, is calculating a higher current draw (if not limited) will be required than what the JVN/dydx calculator projects.

If the current draw projections were in the same ballpark, I would just chalk this up to acceptable variance, but because the results are so different, I’m curious what folks think is going on here. I assume Ari and Justin are including some considerations in their calculators that JVN/dydx are not in theirs. On the other hand, 254 in 2019 happened, and their elevator was really fast, and they geared it as Andrew stated. The climb was much faster than 2.5 seconds… but I wonder about the actual current draw since I assume the motors weren’t current limited (or the limit was high if they were). @Torrance, was the actual current per motor logged back then to confirm it was really in the ballpark of 30A as projected by JVN/dydx?

Interested in what people think about all this. The obvious point is these calculators are for identifying reasonable starting points and teams have to test from there to arrive at a final solution.

1 Like
19

In the ambcalc you linked, you had the radius set to 1.75" when it should be 0.875", I think that’s the majority of the difference in that calculator at least.

I’m no software guy, but more than likely, yes, the supply side and stator current limits we higher than 40A. So for brief periods like the initial acceleration where the load is higher than the static gravity load the robot will draw more amps and still move fast.

2 Likes
20

I think 4 motors is merited if the calculations demand it. We ran 4 775pro in both 2018 and 2019 on 1072’s elevator. It’s a fairly good use of a couple pounds.