"Automatic Transmission"
 

I'm sharing this with the community here because I need feedback on how to improve this design. Please, as you read this, keep in mind that it's not entirely complete -- but complete enough to be understandable. If I wait until I find the motivation to finish it before sharing it, it'll never get done. Also, please keep in mind that this is very much a prototype and I'm sure that it could be improved -- as I've already mentioned.

The two-speed shift-on-the-fly transmission has become a popular, almost required, feature of many robots. Similarly, coupling two or more motors together on a drivetrain has also become very popular and it offers many advantages over single motor drivetrains. Those benefits have been discussed in detail elsewhere.

There is, however, one large disadvantage to most of these motor coupling options. It is, of course, that they must obey the "law of averages" and we make compromises about the use of the motors. High torque motors sacrifice some of that available torque to achieve a higher speed. High speed motors are corraled a bit by the high torque motors. Both motors involved are often capable of providing a higher speed or higher torque individually -- but the compromises made by coupling them together allow for more torque at a given speed.

In this thread, Ken Wittlief of 578 talked about a gearbox design that takes advantage of the unique speed/torque characteristics of each motor. What follows is my interpretation of that idea.

Of course, the overall concept behind this is to create something that makes good use of the unique characteristics of two motors while sacrificing as little as possible of those characteristics.

For our purposes, we'll consider two motors.

• The Drill w/ Gearbox in Low -- this is our high torque motor with a free speed of 423 RPM.
• The Chiaphua -- this is our high speed motor with a free speed of 5,500 RPM.

When the robot's moving slowly -- probably because it's pushing something -- we'd like for it to have a lot of torque. For low-speed, high torque applications, the drill is ideal.

When there's nothing in the way, we want our robots to get where they're going as quickly as possible. For that, the high RPM of the Chiaphua motor is a real asset. That motors relatively high torque at high speeds is also a benefit.

So, the problem becomes finding a way of mechanically disconnecting one motor or the other from the drivetrain when it's not needed. Ken discussed this some in the thread mentioned above, but I'm going to repeat a lot of that for completeness.

Why do we need to disconnect the motors? Well, so our robot doesn't explode, of course! See, the drill just won't spin any faster than 423 RPM. So, if we're expecting to achieve a high speed, we need something that's going to spin a bit faster than that. (Unless, of course, we use 16" wheels or some such thing. My work was all based on 8.5" wheels.)

We're going to use the drill motor exclusively for low speed driving and the Chiaphua motor for exclusively high speed driving.

Car engines are always spinning -- even when the car isn't moving. There, a clutch is used to disengage the spinning parts (the engine) from the non-spinning parts (the wheels). Well, clutches are available off-the-shelf, but they're prohibitively expensive. So much so, in fact, that many I've seen exceed the $400 per part limit FIRST applies to teams. So, that means we need to come up with some other method of "disconnecting" our motors. In fact, we need two.

Ken talked about a one way clutch. Well, I scoured the internet looking for information on a serviceable one-way clutch or one-way bearing. I found next to nothing. There're a few things here or there for RC Helicopters, but not much more. I know this stuff exists, but I had to make do without it.

So, with that, we have my solution for the first "disconnect" -- the mechanical disconnect. I called it a "spiral clutch."



Missing from this image is the compression spring, mounted to one side, that would force the two halves together. That ensures good contact.

It works pretty much as you'd expect. When the right shaft spins counterclockwise, the spiral engages and the two sides act as a single shaft. When it spins clockwise, the two halves slip against one another and power isn't transferred to the left shaft.

What might be less obvious is that all that is required for this to work properly is "relative clockwise" motion; or, for that matter, what use it serves.

Imagine the right side is attached to the drill motor and the left is coupled to the Chiaphua motor as well as acting as an output to further gearing or straight to the ground. The drill starts at a standstill, spinning counterclockwise, and engaging the spiral clutch -- which acts as a solid shaft and drives the robot.

Well, during that time, so that it doesn't drag on the system, we also want the Chiaphua motor to be spinning -- but we don't want for it to see any load. To accomplish that, we have to employ our second "disconnect," but this time it's an electronic disconnect. By varying the voltage to the Chiaphua via PMW output, we can make it such that its free speed corresponds with the current drill motor speed while under load. This is our electronic clutch.

Once the drill has reached its highest RPM, the Chiaphua will still have plenty of its operating range left over. It can continue to accelerate to its top speed thanks to the clutch and "relative clockwise" motion.

Imagine that the left and right sides of the clutch each want to spin at 200 RPM thanks to their motors each outputting at 200 RPM. As I've mentioned, it'll act as a solid shaft because there's no clockwise motion. Imagine next that the left side speeds up to 1000 RPM while the right side remains at 200 RPM. There's a "relative clockwise" motion of 800 RPM. The clutch will slip.

The images that follow show the completed gearbox -- including the associated gearing for each motor. (Well, most of it. The drill output shaft is missing its 18 tooth, 1/4" pitch roller chain sprocket.)

The Chiaphua is geared down 15:1 over three stages when it meets the clutch shaft. The drill is geared down 2:1. I am probably going to revise these ratios -- I can already see where I've made an error. Can you? :) The top speed is 15 ft/s. The low speed torque output is ~57+Nm per side (depending on external gearing).

The gearboxes are big. Remember this is the first iteration. I sincerely hope I can make them 25-33% smaller. They measure approximately 10" x 10" x 4" each. They can be oriented vertically or horizontally.


Side view.


Front view.


Isometric view.

That iso view is the oldest of the views. I assure you that the end plate now has holes and bearings in it.

So, that's that. Feedback, comments, suggestions, etc. are encouraged. Please feel free to take any ideas from this thread and apply them to your own projects -- but please, tell me. I'd like to follow along and see how it turns out.

Edit: Sorry about blowing up the margins. I'll shrink those images and edit again as soon as I can.

Wow, that is so incredibly awesome, Ive been wondering what youve been designing for the last few months :). Very cool idea, like a CVT, but not as complex, and far more efficient, as well as being easier to build. You really explained the whole idea well also.

Congratulations on a great job!

Cory

Very interesting idea you have there.

Very interesting. Another method I've seen for coupling two motors whose speed needs to be independent of each other is the modified planetary used in the Toyota Prius. In the Prius, either the gas or the electric motor can run at any speed, and both will provide power to the wheels. "Gearing" is done by adjusting how much power comes from the lower RPM electric motor or the higher RPM gas motor.

The gearbox itself is almost identical to a standard planetary gearbox, except that the Sun and the Ring are both driven, one by the electric motor, and one by the gas motor. The Planet Carrier is then connected to the drive shaft. The advantage is that, if needed, both motors, each going at their optimum speed, could power the wheels at the same time.

Another idea that I came up with over the summer (while playing with mindstorms, of all things), is to use a standard differential, except that one wheel output is replaced by a second motor input. Mechanically, this would function identically to the Prius' system, with all the same advantages. However, building a dual power planetary gearbox or a differential (I'm not sure if they are available cheaply off the shelf) may be just as complex as building a CVT.

If I'm not mistaken, this is how the Thunderchicken's CCT works. There's a whitepaper on this site detailing it.

Quote:

Another idea that I came up with over the summer (while playing with mindstorms, of all things), is to use a standard differential, except that one wheel output is replaced by a second motor input. Mechanically, this would function identically to the Prius' system, with all the same advantages. However, building a dual power planetary gearbox or a differential (I'm not sure if they are available cheaply off the shelf) may be just as complex as building a CVT.

This sounds a lot like what Tytus has been working on. There are several threads devoted to debating the benefits of that design.

There are several ways of combining motors -- many of which compromise the motors involved in some ways. Continuously variable transmissions offer the same or greater benefit as the idea above, but I think they're a bit more complex. So, of course, it's a tradeoff based on what resources are available to you.

Would the spiral clutches wear out? It seams like repeated slippage and heavy spring tension wouldn't do them good.

I don't understand how the drill can power the robot in one direction until a certain speed and then the Chiaphua will take over propelling the robot in the same direction at a higher rate of speed.

I guess I'm just stuck on the idea of "relative clockwise" This is a very neat idea and I'm terribly frustrated that I don't understand it. Would someone put arrows of rotation on the drawings to explain how it works at what stage in the movements or give me an extremely detailed description?

Thanks. I would really like to understand this concept as I can see that the mechanics are quite simple.

This is very cool. Has this been used on a bot before? It seems that this would have been perfect for either the 02 or 03 games because of the need for speed followed by the pushing match for the goal position or king of the hill.

Sanddrag, I would explain it as I understand it, however I may just confuse you more. I'll leave it to the creator :).

M: What happens when you want to go backwards or turn? Would the clutch slip? Would you do this using only the CIM? Would the latter cause you to trip breakers?

Maybe I'm not understanding it totally. (Likely solution)

Great idea, though... It's making me think! =-]

George

What about a differential? Except a backwards one. Say, take a standard diff, two output shafts (the axle) and one input shaft (or gear, from the engine/transmission)

Except here, you power one of the output shafts with the drill, and the other with the chip.

So, as long as they spin the same direction, you get power out of both of them, and you don't need them to spin at the same speed or anything. you could even turn one off completely, and the thing would still turn.

I think this idea is worth looking into. you yessssss, I do indeed. The more I hink about this he more wonderful it sounds. I will post more after school (and after I've thought about it some more and maybe created some prelim drawings. )


This could be a winner, folks.


with this, you could get power from both motors, each going at their respective optimal speed . . . no need to use two motors only to use one at a time, you know?

tytus was(is?) working on a gearbox using a differential, and there are already a few threads dedicated to discussing its advantages and drawbacks, located here, here, and here.

Kevin A wrote:

Quote:

Would the spiral clutches wear out? It seams like repeated slippage and heavy spring tension wouldn't do them good.

I have no way of answering that, unfortunately. I'm hoping that I can have a set made and tested once I'm happy with the overall design. Admittedly, I'm really poor at choosing the right material for a job, so I'm not even going to venture a guess as to the best one for this application. In the drawings, the parts are shown as delrin -- mostly because of its low coefficient of friction. I'm unsure of the wear it'd see or if it'd have a significant effect. If the wear, even if it were bad, was consistent and even, the clutch should still work -- a bit like car brakes.

Sanddrag wrote:

Quote:

I don't understand how the drill can power the robot in one direction until a certain speed and then the Chiaphua will take over propelling the robot in the same direction at a higher rate of speed.

In physics, one of the first things you learn about is relative motion. I'm not sure if you've taken physics, so I'll briefly explain.

If you're standing beside a railroad track, not moving, and a train goes by at 50 mph -- well, it looks like the train is passing at 50 mph, right?
Now, if you're in another train, heading in the opposite direction of the first, but also at 50 mph -- well, then that first train looks like it's traveling at 100 mph.
Think of things the opposite way as well. If you're on a train that's running parallel to the first at 50 mph, it appears as if the other train isn't moving at all. If the first train speed up to 60 mph, it would appear to us as if it's going 10 mph. See?

Imagine both sides of the clutch are now spinning at 300 RPM. It appears as if they're not moving. Speed one side up to 1000 RPM and then it appears as if it's spinning 700 RPM when your frame of reference is moving at 300. I'm not sure if this is helping.

The real point is that "relative clockwise" motion never requires the motors to spin clockwise. It just requires that the left end spin faster than the right end such that the clutch slips. After a certain point, you could actually turn the drill off altogether.

generalbrando wrote:

Quote:

Has this been used on a bot before?

To my knowledge, no. My knowledge of things FIRST is fairly limited, though, as I've only been around for 6 years. If someone knows of a robot that's used something similar, please let me know. I would like to see their implementation. I know mine needs improvement.

George wrote:

Quote:

M: What happens when you want to go backwards or turn? Would the clutch slip? Would you do this using only the CIM? Would the latter cause you to trip breakers?

Without adding more complication, going in reverse would rely solely on the CIM motors. The error I mentioned in my first post, in fact, had to do with this very item. I need to double and triple check my gearing to make sure it doesn't strain the system too much while going in reverse. It is possible to make a design that uses both motors in forward and reverse, but it adds an additional level of complexity that may or may not be worth the hassle. In short, it'd involved two clutches and a shifting dog.

Turning is an interesting issue, I think, and something I'm still investigating. It seems that the gearboxes would behave very differently depending on how the robot turned. If it were to steer like a car, the gearboxes, together, could be used as a "differential" and the speed differences between the motors do not seem like they'd pose any real problem -- much like many robots. With a zero turning radius, however, things get a little stickier. For that, one side is going in reverse -- using the Chiaphua motor -- while the other goes forward using whichever motor is most appropriate for the speed at which it's turning. It may be that there's an easy solution, a programming solution, or that a zero turning radius is impossible to achieve. Again, I'm just not sure.

Frank wrote:

Quote:

What about a differential? Except a backwards one. Say, take a standard diff, two output shafts (the axle) and one input shaft (or gear, from the engine/transmission) Except here, you power one of the output shafts with the drill, and the other with the chip. ... with this, you could get power from both motors, each going at their respective optimal speed . . . no need to use two motors only to use one at a time, you know?

The differential idea, as I mentioned above, is Tytus' kingdom.

I agree that there are better solutions if you want to couple two motors together and run them all the time. You'll get more power at a given speed that way. Most of those solutions rely on averages -- as does the differential. This solution does not.
This design is explicitly for maintaining separation of your motors and taking advantage of the unique characteristics of each. The additional performance range possible via this solution means that your robot could simultaneously have the ability to output more torque and more speed than any other robot -- especially middle-of-the-road, average robots using a two motor drivetrain.

Again, thank you everyone for taking the time to read through that very long post and ask questions. You're raising issues that I haven't thought of yet or don't have satisfactory answers to and I really appreciate that. I understand that the design in its first iteration isn't close to perfect and so I'm seeking your feedback so I (and others) can make improvements before testing and using it -- if it's worth using at all.

M, this is a great design. You have any desire to come to school at U of A? Our team could use someone with your talents to teach our kids part design techniques ;)

That being said, I do have one question. You said that this works like a simplified version of a car clutch right? If so, does this mean that both motors are spinning constantly? The reason I ask is because I know several teams last year had battery drainage issues with high usage of both the drills and the chip. Let me know your thoughts :)

If I had a nickel for everytime someone told me to come join their team . . . :)

As for your second question -- the only motor that's spinning constantly is the Chiaphua motor. Even then, it's spinning at low voltage with a low current draw when the robots at the low end of its operating range. The drill can be safely turned off entirely once the robot's moved beyond the useful speed of the drill motor.

Great Thread
 

Threads like this should outnumber the pointless threads by 2:1"

M,

Great idea, but I will get back to you on comments I am proving some technical questions out on paper.

Ahecht & M,

The Thunderchicken CCT, while it does manipulate the advantages of a planetary gearset, is nothing like the Toyota Prius Mechanism. On howstuffworks.com it describes how the Prius' gearbox works, but it uses 2 electric motors and a gas engine. One electric motor acts as a generator. Toyota uses the planetary to couple the 2 electric motors and the engine together, but the main electric motor is attached directly to the driveshaft. There are several modes of operation, but the one that gets them fuel efficiency attaches the output of the engine to "help" the electric motor. The other electric motor spins to regulate the speed of the engine.

It is much more complicated than the ThunderChicken CCT and has 2 separate patents filed: 5,914,575 and 6,067,801. We had to reference these patents to make sure we did not patent the same idea.

-Paul

bzzt... wrong answer.

The differential idea (aka Planetary gearset) was the Thunderchicken's invention (and patented) long before Tytus even thought of it.

Such an idea would not work unless the stall torques of each motor were the same. The only way to overcome this fact is to put a anti-backdrive mechanism on the weaker motor, hence the worm gear on the 2002 CCT.


I've considered many similar designs to your clutch. Also, I would like to note that a simple search at McMaster Carr would yeld clutch bearings (aka ratchet bearings, one-way bearings). I would know because I've looked at them before for this purpose.

The bad side is: propulsion in reverse. You could only do this using the high-speed, low torque motor. Everybody knows that robots often get pinned. A high-torque reverse is a nice thing to have...

In any case, my point is that the differential design, whoever it is developed by, is different and is the subject of its own threads and discussion. Why it does or does not work isn't relevant to this thread. Thanks.

Quote:

I've considered many similar designs to your clutch. Also, I would like to note that a simple search at McMaster Carr would yeld clutch bearings (aka ratchet bearings, one-way bearings). I would know because I've looked at them before for this purpose.

I've looked at what's available through McMaster-Carr in the past and either ruled it out or flat out forgot about them when I began this design last week. I'll look again and see if those products are applicable to this concept. If they are, that's a great thing because it greatly reduces the machining complexity required and brings the design within reach of most teams.

Quote:

The bad side is: propulsion in reverse. You could only do this using the high-speed, low torque motor. Everybody knows that robots often get pinned. A high-torque reverse is a nice thing to have...

Agreed. But, remember, this is just a gearbox. There's nothing stopping anyone from attaching it to something fancy like a swerve drive or something. Additionally, if the products available through McMaster-Carr are viable (I'm concerned about the torque loads they're designed for), creating a similar system that works in forward and reverse becomes a bit less prohibitive. It's still possible in either case as I already mentioned, however. This is just the starting point and there are many improvements that could be made upon this concept.

Paul -- I am very interested in what you come up with and your suggestions.

I'll admit that my understanding of motors and gearboxes and things is a bit rudimentary. There could be large swaths of important information that I've failed to account for.

Also, forgive me for my bad comparison of your CCT to the Prius' arrangement. Where planetary gearsets are concerned, the moment someone talks about locking the ring gear in place, my mind shuts down. I've never been able to sufficiently imagine how that's accomplished. I'll go read through your whitepaper again. :)

Edit:
It seems like part #2489K14 from McMaster-Carr (www.mcmaster.com) would do the trick. The only mitigating factor might be how high above 40A the current can spike before the breaker trips -- if it's too high, above about 70A -- it'll exceed the specifications of the clutch bearing. This all assumes my drill gearing of 1:2 as well. Anything lower than that would also exceed the specifications of the clutch bearing.

Additionally, it seems like you can reverse the orientation of the clutch bearing on one side -- eliminating the usual need of having to run your drill motors in opposite directions and compensate for the difference in programming.

I'll work on a second iteration that includes this clutch bearing. The first problem I see is finding a way of mating the clutch bearing to a gear -- either by welding it (they seem to be plastic, partly) or by press fit. I really have no idea at all if a press fit is capable of transmitting 25-35Nm of torque.

Has anyone done some rough calculations on the power output of this transmission. I know from the past post that the main advantages were the lack of having to shift.

It would seem to me that you would have higher power available to you in both high and low at the output shaft using a shifter with the motors individually geared to match free speeds and all driving the output. From my first breeze over this I think everyone is saying that the Chip is overdriving the drill to go faster than the drill can when in high and the chip is essentially ideling in low. In the 45 shifter both motors are chipping in in both low and high. Shifting with the 45 design gives all of the power you could want in low (Can spin the treads!) and it is fast in high, and the gearbox is tiny by comparison.

I may be missing the whole point though??? Wouldn't be the first time.......

One Last Time....
 

I'll try to explain it one more time:

The Thunderchicken CCT is nothing like Tytus' differential idea. Please read the thread discussing the Tytus gearbox for the explanation. Tytus is using a differential with the motors connected where the wheels (on a car) would normally be. The output is where the input (on a car) would normally be. The Thunderchicken CCT uses a planetary (not a differential) and places one motor on the input and one on the ring gear.

JNadke,

The motors will not overrun each other on the CCT when both motors are powered. The reason we put the worm gear was to not overrun a motor with no power to it. During low speed mode, only one motor was running (much like what M. Krass is trying to do) and the ring gear motor was turned off. We decided to use a worm gear to HELP not backdrive the unpowered motor. In the end, we had to pulse the motor 'on' every once in a while in low speed mode in order to stop the occasional back driving. Both motors were turned on in high speed mode with the SAME torque available as high torque mode, but much more power consumed (aka battery life).

Matt R.,

I agree with your initial thoughts. I am doing some equations to show why combining motors at a compromised speed is still better than one motor optimized.

I hope this clears it up,

Paul

Like Swerve^2 :rolleyes:

Krass has a really neat design, but without doing any analysis to the design, I agree with Matt. I am looking forward to what Paul determines with regard to analysis.

Innovations like the Krass design and the Gerrish design make us all think, and that is a good thing. This is how concepting a new idea happens... we bat around ideas for a while until someone hits a winner.

Andy B.

Okay, I did some of my own very special, hackneyed "I think I know what I'm doing" math.

This is the conclusion I came to when comparing this design to 45's shift-on-the-fly transmission. I assumed different gearing for this comparison such that the output RPM in high and low matched those of the Technokat's design. The gearing I have now is far slower, but correspondingly more powerful.

In low, assuming the drill gearbox is in *high* rather than low (to achieve 375 RPM), with a 4:1 reduction, the output is 50.8 Nm. 9% more powerful.

In high, with the Chiaphua geared down to 1500 RPM via a 3.6:1 reduction, the output 9 Nm. 23% less powerful. The transition between motors would also overload the breaker on the Chiaphua motor.

In high, with the Chiaphua geared to 1000 RPM (67% of the Technokats top speed), the output is 13.5 Nm. While more powerful than 45's extreme end, it's still 23% less powerful at its own top speed.

All of these numbers are skewed because 45's whitepaper only gives the RPM under load conditions. My RPMs are assumed for no load. So, take the numbers with a grain of salt and assume that, for all speeds, 45's shift-on-the-fly transmission is more powerful by as much as 30%, I'd guess. Andy, can you give an idea of what the no load RPM of high and low is? I'm interested in seeing how far off I am.

So, that poses questions about the viability of the concept. Teams already have access to the Technokat's whitepaper for a more powerful gearbox and drivetrain.

My immediate conclusion is yes, it's still viable. Mostly, it's because I spent some time fleshing this out :). But, also, I think there are still two distinct advantages.

• It's still an automatic transmission. It will react to increased load on its own, it doesn't require pneumatics or additional motors, and, though less powerful, the moments saved could be crucial. It also seems to reduce the risk of popping breakers because it'll automatically distribute the load to the right motor.
  
• It's easier to manufacture. Given Jeremy's mention of the clutch bearings available at McMaster-Carr, I think that manufacturing this, after it's redesigned, will be within reach of most teams with access to a manual mill. There's really no complex machining work left once the spiral clutch is eliminated.

It has its advantages and disadvantages, I suppose. While not as powerful or fast as I'd hoped, it is a simple solution that seems like it could, at the very least, be competitive with the Technokat's and other's transmissions.

Again -- thank you everyone for offering your insight into this. Threads like this are the best kind here on CD because we can all learn from one another's ideas, successes and failures.

Actually, I like your idea of using a coupling as a clutch. It allows you to transmit more torque than a clutch bearing. I was merely trying to point out that there is more than one way of doing what you wanted to do.
Really, I've never ordered those, so I'm not sure if they could be adapted to work. You might be able to weld a gear around the clutch bearing, but it would probabaly damage the internal workings. I'm not really sure.

I understand that it's the first design, but you'd probabaly want to make the coupling smaller.

I'm sure it'd probabaly be the loudest robot on the field though with the constant clanking. :)

I thought about designing a belt-driven CVT, but the physics involved would be very complex, too much stuff that needs to be very precise. Basically, a belt-driven CVT works off a very similar principle as your clutch. You have a spring that applies a known force to half of a v-belt pulley. The other half is mounted permanently. Basically, as the robot goes faster, the half begins to separate (applies a centripetal force to the spring), thereby decreasing the radius. With that said, you can achieve anywhere between a 1:2 and 2:1 gear ratio.

I'm not sure if such an idea has yet been implemented on a FIRST robot, but I'd be willing to work through the math and find the correct parts if I had someone to draft it up for me :rolleyes:

From what I understand of the Prius, the CVT in the Prius is used to split the torque between the wheels and the generator. Their are three modes of the prius:
In the cruising mode, the CVT splits the engine torque between the drive shaft and the generator (generators induce a negative torque while they generate electricity). The generator, in turn, charges the batteries of the vehicle.
In high-acceleration mode, the CVT splits the engine toruqe between the drive shaft and the generator as before, but the electricity generated is sent back to the drive shaft where it powers a motor.
In low-speed and reverse, the vehicle is powered by the motor only.

I could be wrong, but I read a lot about it, and this is what most websites say. From a Physics point of view, this way would make sense.

I agree with the above posts. It is definitely a cool idea, and something that has never been done before. But, I also agree with what Matt said. I see no reason why building one of these is more beneficial than building a simple shifter. The Technokat style dog shifter is great, but somewhat difficult to make. But, anyone with a mill and a lathe can make a "pin style" shifter similar to what we made this year. *shrug*

It's certainly a cool idea though.
Keep 'em comin'!;) :D

John

M,

Your numbers are similar to what I would have thought. However, one of the biggest advantages to this design is be it's efficiency converting electrical power to mechanical power compared to the "averaging" design.

Do you have any way of figuring that out, without actually building one?

I don't know much of anything about anything when it comes to gearboxes, but I do know that when I was on RAGE last year we looked into constructing a "belt transmission" because it would allow us to achieve a _really_ high speed in a short period of time, with a near seemless journey there. So, my question.. shouldn't a belt transmission be geared towards robots that need to be really fast more than to robots needing to be "strong?" Throw something at me here, so I know what to think.

I don't think those are belts, the pulleys are drawn too narrow. Those are chains & sprockets.

Depending on the belt, chains and belts are nearly identical in the concept and function. The difference is that chains are typically more durable and can transfer more torque. If you use v-belts, they usually can transmit a certain amount of torque before they slip (slide without rotating the pulley), which would be a good thing if you have a very high-geared robot and you don't want the breakers tripping.

Other than being lighter and easier to assemble/dis-assemble, belts have no significant advantage over chains & sprockets. Of course there's the whole slip thing, which is good when you're dealing with combustion engines (stall = bad). On robots, however, we seem to like to stall our motors with all the pushing (since nobody has designed a true CVT, yet), so I guess this could be considered a bad thing.

Actually, 190 had a toroidal CVT in 2002, and was planning to use it again in 2003, but it had to be left out due to weight and time considerations. Also, I heard that 222 had a CVT designed and built this past year, but their robot design didn't need it. The hard part with a CVT is not the actual design and build, it's the programming. There is a fine balance between not reacting fast enough and occilating wildly that is very hard to overcome.

There is a good description of toroidal CVTs at http://www.barloworld-cvt.com/varibo...based_cvt.html

Interesting... didn't know a true CVT had been built before. I thought about a toroidial CVT, but then you have the problem that you brought up (control of gear ratios). Also, you have the issue of torque transfer between two non-interlocking metal surfaces. How did you guys solve this problem? I suppose you could use a high COF material and use acid-etching to make the metal surface porous. Or am I making this problem larger than it is?

If I remember correctly from the explanation given to me at 2002 Nationals, the toroids are polished metal, and it works flawlessly. Very high Mu, they told us :D

Ahh Cool im getting mentioned now as an inventor

I Personaly Can,t stand automatics Its All Manual For me, But on an atonomus robot i see the advantages And i Didn't steel my idea from the thunderchickens I have only known about it for a month now when Brandon said a similar device already exists, ive been ctudying the CCT and it uses the same mechnical princapls as the "Gerrish Gearbox" But It Is Very Diffrent!. I went with a reverce diff thinking its how Dual-Engine Helicopters merge their power. Ive yet to draw the Srag chutches to prevent backdrive in my models, its gonna get in there tho

BTY: I NEED AUTOCAD!

Nobody ever really said you stole any ideas. We were just saying that the concept has been done before.

Both from a physics standpoint and from a design standpoint a bevel gearbox "differential" is the same as a planetary gearset, only re-arranged. In both cases, the stall torques need to be the same on both motors in order to prevent backdrive (unless a clutch mechanism is used), and in both cases the overall speed of the gearbox is the average of both motors.

More discussion has already been done here. The differential you have designed is merely a planetary gearset with the carrier fixed.

Not quite...

Tytus' tranny is not "merely a planetary gearset with the carrier fixed." Think about it. He has two inputs and one output. Holding the carrier fixed in a planetary gear set leaves you with only one input and one output. As I see it, the design is more like a special (and physically impossible) case of a planetary set where one input bevel is the sun and the other is the ring:

The output speed of Tytus' differential will ALWAYS be the average of the speeds of the two inputs. The output speed of a planetary gear set (i.e. the rotational velocity of the planet carrier) will only be equal to the average of the sun gear speed and the ring gear speed if the sun gear and the ring gear happen to be spinning at the same speed (when this happens the planet gears stop spinning and the sun, ring, and carrier all have the same speed).

Looking at the equations that relate the speeds of the sun, ring, planets, and carrier to each other, the only way that the carrier speed is the average of the ring speed and the sun speed is if the sun and ring gears have the same radii – which is not possible. So, like Paul C. has said before, the bevel differential is a close cousin of the planetary gear set, but it is not the same thing. As far as his claim that the CCT is not a differential, I’m not so sure, but I think that is more a matter of semantics than physics.

P.J.

Not Semantics
 

PJ,

No. A differential and a planetary set (like the one is the CCT) are different. The differential, by definition, has 2 outputs (or inputs); for the car example it is the 2 side gears. The CCT, and planetary gears in general, only have one sun, one carrier, and one ring gear. The differential for a car (and Tytyus' implementation) has 2 suns, 1 carrier, and one ring gear. Unlike the differential, the planetary gearset with 2 motors has independent control of each motor's Torque/Speed profile (i.e. one will NOT override the other if they are both supplied with power.

-Paul

Differential/Planetary analogy:
 

The analogy between the differential and the planetary set is as follows:

one of the input bevels is the sun

the other input bevel is the ring

the intermediate bevels are the planets

the ring gear is the carrier.

There are not two suns, it's just that the sun and the ring are the same and are actually interchangeable.

In a standard planetary set, if the planet gears are made infinitely small, the sun gear and the ring gear approach eachother in size and the behavior is almost identical to the differential design proposed by Tytus.

P.J.

I am wrong
 

P.J. is right. The differential idea proposed by Tytus is exactly like the CCT with a 1:1 ratio between the sun and ring gear. Hey, even us unsung FIRST heroes are wrong from time to time.

That being said, only one motor contributes to the output torque of the system. The other motor acts like a speed increaser at the same output torque (if balanced correctly). All the equations derived in my whitepaper for the CCT apply to Tytus' design.

-Paul

Re: I am wrong
 

Yes, I was going to mention that, but ran out of time. The bevel/differential gearset can can do the one thing that the planetary gearset can't: achieve a 1:1 gear ratio between the sun and ring gears. In order for this to happen, the sun gear would have to be large enough such that it approaches the size of the ring gear. Of course, this can't happen, but calculus says the limit exists. As the sun gear size increases (and therefore the planet gear size decreases), the gear ratio decreases (assuming that the ring gear is the input and the sun gear is the output and the planet carrier is fixed).

Of course, luckily for us, the stall torques of FIRST motors are never the same. So a more compact dual-motor gearbox can be made by simply using a planetary gearset, taking advantage of the difference in gear ratios.

Actually, both motors will contribute to the output torque of the system. However, in order to avoid backdrive, both motors must be configured to have the same stall torque. So basically, you would end up with twice the torque of a single motor. However, the overall top speed will be the average of the two individual motors.

This is complete opposite to the standard spur gear method of combining motors, where the free speed must be configured the same on both motors. The overall torque will be the sum of the two motors, but they will share the load unequally.

I relize this thread is getting off topic, but I just have to ask.

Why not just use 2 identical motors on a differential type gearbox? As in, use the two drill motors on the left side, and the two Atwood motors on the right and then just gear them to give you the same speed. With two motors of the same stall torque you avoid the main detraction of a diff gear box, which is the lower torque motor being over powered when stalled.

I'm still not completely clear on what happens next, when the motors actually start to show their torque differences. Would the side powered by the more powerful motor 'overpower' the other side in a pushing match, and pull the 'bot to one side? Am I correct to think that the drills would be geared higher, and the Atwood's geared lower? Or does the Diff mix this all up??>

Like I said, I don't really understand how it all works out- I've yet to receive any formal mechanical education. I know that 121 has used two drills on one side and two Atwoods on the other (this past year I believe?). From what I remember, they used a fairly traditional transmission, except the two were not identical. My team has never used multi motor gearboxes, so I'm a little in the fog here.

Am I just missing something painfully obvious here? I get the feeling I am. Again, sorry for pulling this thread further off topic, but my curiosity is killing me.

-Andy A.

You wouldn't notice any difference, except that the drill motors are slightly more powerful than the CIM motors, so one side will accelerate faster than the other. Although you won't really notice it in pushing matches, or when cruising at top speed, you will notice it everywhere in between. So, when accelerating from top to full speed your robot would travel in a slight arc.

Also, it's extremely hard to match free speeds exactly. Although, in general, the numbers will be so close it won't make a whole lot of a difference. Especially if you have a 4-wheel drive robot. If you have a robot that's 2 wheel drive with casters you will definately notice it, other than that.

Next year, the situation could change though. This year you could have gotten away with it, because the power of both motors are approximately the same (the drill is 50W more powerful, I think, which is 8-10%).

I'm not sure what type of gearbox they used, but i'm pretty sure that 121 used both drills on hte left, and both chips on the right last year.... someone from their team might be able to give more detail

Tom

that is definitely what 121 has done for the past couple years. I have talked to members of that team, and they say it works great.

Yes, you are. The toroids were made of machined cast iron (only because FIRST didn't allow steel thicker than 1/4" that year), and were very smooth. We ran CVT at fairly high speed, and did further gear reduction afterwards, so that there was never that much torque going through it. We never had any trouble with the it slipping.

After seeing 190's CVT I was a little research on something else and found to my surprise that the coefficient of friction for iron on steel was actually fairly high. BUT the rolling resisitance is very low and the contact area is small.

Which explains why locomotive wheels are still cast iron. Actually it might have been in a railroad book that I found that little gem. Unfortunately I don't remember the reference.

Locomotives also drop sand on the rails at times to increase their traction.

Actually the coefficient of friction of Cast Iron on Steel is 0.4. This is compared to the coeficient of friction of Steel on Steel which is 0.8. By comparison, some rubbers (neoprene, gum) can achieve a COF of 3.0 or more on most surfaces.

The real reason cast iron is used is that it won't deform (compared to steel). Steel can deform as much as 35% before it fractures, whereas cast iron will only deform less than 1%. This is a good thing when you want the wheel to stay round under all that weight. Also, cast iron as better vibration dampening than steel. Keeps everything from rattling as it goes over the many joints of the railroad tracks.

in any case, differentials can be heavy and at most, i see only a few teams using them as power (not to mention the input motors must have equal max-torques or else you loose major energy)

The major advantage of a differential (or planetary) link system is that you don't loose as much energy to motors fighting each other as you would with a fixed gear setup . . .


On topic: Personally, I would avoid using that type of clutch system where one motor does low speed and one high speed because at low speeds (as I understand your design) the chip would be nearly stalled (and therefore you have a possible breaker popper) and at high speeds and reverse your only power come from the chip. The main advantage is that it is automatic and you can forget about it. Still, If I'm going to put two motors in a gearbox I want them both to pull their weight. What would work instead could be a wheel-switching mechanism (one wheel geared highs than the other, lower it using pneumatics). You could even give the computer control over that and turn it into an automatic.

Using the original post's design, you would have a robot that is quite energetic until it hits 5 fps or so and then is fast but a little sluggish, and in reverse it would have a hard time moving at all (much less skid-turning)

if you could somehow make the drill take over lowspeed power in reverse also (maybe accomplish this with a clutch controlled by a governor?) , then you would have a great box.

Team 980 did this very successfully last year. It was amazing they got it all in there along with all that other stuff they have. Truly amazing.

Well, personally, I've only conquered that beast and am on V2.0. This was an attempt at something different.

For me, anyway, switching drivetrains is old hat.

True. I enjoy new things myself. This year, I am pushing for (and will probably get, seeing as I'm design lead) pretty standard drivetrain, cept its a four-wheel killough-style.

I personally think the future of this competition is in omnidirectional drivetrains. I mean, when it gets down to it, its all about speed and manuverability, and, to a lesser degree, power. nothing beats killough in manuverability. four wheel styl lets us distribute the motors more evenly. and make the math easier.

Team 190 did this in 2001 and 2003 (although the two sets of wheels in 2k3 were at 90 degree angles to each other). The only problem with it is that is is much bulkier and heavier than a SotF transmission (which 190 had in 2000, IIRC). However, as we learned this past year, it can be really nice to have a backup driveline in case you pop a chain on your primary one.

Exactly. The only reason 190 didn't use steel on steel in 2002 was that it wasn't allowed. The CVT we started in 2003 (but never used) was designed to be steel on steel.

Torques do not Combine
 

jNadke,

OK, I was wrong about the differential not being like the CCT, but I know I am right about the torques NOT combining. The calculations are done in my whitepaper. Conservation of energy dictates that the torques DO NOT combine.

Put simply,

The second motor acts like the housing on the drill transmission: all it does is REACT the load of the input motor. To be exact, its reaction torque is (Torque of output carrier - torque of sun motor). The backdriving phenomenon you are describing is as if the housing breaks. If the ring gear motor does not have enough torque, then it can't react the load and is backdriven or put to stall and pops breakers.

The latter is exactly what happened to us at Great Lakes in 2002. I, like you, thought the motors combined torques and that got us in a lot of trouble. We popped breakers constantly. When we got home and I actualy sat down and did the calculations required (in the whitepaper), I discovered that the torques did not combine and the ring gear motor was merely a speed increaser. This is the biggest disadvantage of the CCT: add a motor and get speed, not torque.
However, the cool part is that you get more speed with the same output torque. In 2002 that was a big advantage, because we could move both goals at high speed.

If you are not convinced, you can e-mail me and I will discuss the details with you.

-Paul

Ahecht mentioned that 190 has done this, before, but I thought I'd add some pictures from 258's venture into this kind of system. This robot is from the 2002 season. The huge bolt through the 1x2" box aluminum acted as the pivot point, and the 2" bore, 4" (?) stroke pneumatic actuator did the shifting. If it's implemented properly, it can be a great method of shifting, but otherwise, it can be pretty bad. besides the driven wheels, we only used casters, and I must say, the casters didn't help. With a little off-the-cuff math ((550rpm/60s)*7in*pi), I calculate that we were going at, or a little under 16fps, in high-gear. Straight wheelchair wheels would have provided lots more control, I believe.

Also, speaking of a hybrid manual / automatic (for lack of a better way of putting it) shifting system, it sounds nicely challenging to develop a system (mechanical, electrical, and programming) which monitors current draw in the drive motors, and shifts accordingly. (I believe that 190 had this kind of system for their CVT, in 2001, but please correct me if I'm mistaken) Id est, starting in high-gear, get into a pushing match with someone, the system recognizes that you're drawing a lot of current on the drive motors, and shifts into low gear. You'd want to build in some hysteresis, so that it wouldn't go crazy at whatever you determined to be the threshold.

Just a couple thoughts, not sure if they're really applicable.

Actually if you don't cut power to the chip motor at low RPM it will stall and will subsequently over heat and literally fry the motor.

If you do cut the power to the chip it will act like a magnetic brake because it will be generating electricity; due to the conductor, magnetic field and relative motion between the two. That's assuming that the chip motor has permanent magnets.

Motors only act as brakes if dynamic breaking is enabled on the Victor. It is selected by the position of a jumper between 3 pins on the base of the Victor it's self.

Check the Victor docs for more on that.

-Andy A.

Re: "Automatic Transmission"
 

Getting in on the action of this thread, albeit a week or so late, here are my thoughts....

First, I think someone mentioned this before, but I'll say it again: this monster's going to be loud when it's slipping/drill motor turned off to save power. I also can't help wonder how all this slippage and slamming of the plates will hold up. Mechanical things aren't my forte, but it seems that, especially at the CIM's higher RPMs, this is going to cause a lot of wear and breakage, not to mention heat. Rather than 'turning off' the drill motor when going fast, I'd use some more power and put the thing in low so the relative motion decreases and there's less slamming of the plates.


Now for my real problem with this design: from a programmer's point of view, this is a complete nightmare.

Something like this cannot easily be determined from motor curves alone. This is going to require testing, testing, and even more testing. You can't wait three, four weeks to build this gearbox and possibly expect the programmers to have the e-clutch figured out in a few days. I've begun thinking of several ways on how to calibrate this, but regardless, the programmers need a lot of trial-and-error time to get the e-clutch working at max efficiency. Another thing, it seems to me the 'golden values' would be different at 12 volts than at 10, 9volts (seeing as more voltage causes the motor to spin faster). Once your voltage begins dropping in those last 60, 45 seconds of the match, it seems to me like the efficiency of the e-clutch will be going down and down.

M, it's an interesting idea, but as I understand it, if anyone implements it, that person better hope he has very talented/patient programmers with plenty of time to find the magic values. Someone correct my ideas if I'm wrong, though - like I said, I'm an electrical guy, not a mechanical one...

[edit]
Thinking about it even more, there seem to be just so many variables to take into account to get it working at max efficiency - the voltage of your battery, the drill motor load rpm vs. the cim no-load rpm (is the relationship linear, or is it complicated?), the actual amount of load or resistance to motion being expierienced, the wear (and so, the changing coefficient of friction) on the clutch plates, differences in individual motors (you'll have to calibrate each side seperately), and probably more to be discovered. If even one of those changes, the programming is going to have to account for that. This idea is going to be like jamming an iron rod through your programmer's face.

How about a custom circuit feedback sensing system?
 

Extremely! (whack whack whack... ;) ) I was wondering about the wear and heating effects, too. That's a LOT of hammering while clutching. (I'd be tempted to add a pneumatic clutch just to reduce the noise and wear on the joint! :D )

Well, I'm not so sure. IMO, the solution to that will be in adding "appropriate feedback sensing" to make it a closed loop system.

The question then becomes: What data do you need to collect to determine the case, and appropriate control action? In some way, you'll need to know a loading factor per motor, and to sense an imminent motor stall condition.

Off the cuff, current monitoring of the motors may yield enough data to determine the case and close the loop, but you may also the true speed of the motor with either encoders or "PWM off phase generator speed sensing" as well.

Note that to prevent backdriving, you can always include some kind of a mechanical antibackdrive system in each motor's train (like worm gearing, or backdrive pins). Another option would be to wire the coast/brake jumper to switch the Victor into braking mode under program control. I'm just curious though how much back drive the Schottky kickback diodes within the Victor's MOSFETs can safely take before popping! :)

I agree with you, Dan. Dropping this one onto the programmers in week 5 (or 6!) without even a proof of concept prototype, and somehow expecting them to "figure it out" would NOT be a good idea (and possibly suicidal). This is definitely a case where "off season prototyping" experiments are in order before you even consider choosing it.

- Keith Mc.

Re: How about a custom circuit feedback sensing system?
 

You drop this on the programmers that late the only whack whack whacking you'll hear is them beating you with something blunt.

Growing up in an auto shop I have learned some things about how transmissions work, and how gear ratios are figured, and how mechanics in general works. I think that the inital idea of having a sort of screw that connects the powered shaft from one motor is a good idea, but... Let's think how a real automatic transmission works, you have a ring gear, planet gears, and 1 or 2 sun gears. In an automatic transmission in a car there are 2 of these systems connected together to give the car multiple gear ratios. Now you can make a level of the gears, such as the planets stop moving, this gives you a different output gear and a different ratio. What if you could construct a planetary gearset that was somewhat like the one inside of a car. That when you lock one section of the the gearset you get a different ratio, but you can also allow another section of the gearset to spin freely. Thus allowing a drill to be powering the wheels via sun gear, and the whole apparatus rotating around the sun gear from the atwood, and then at a specific point, the atwood powers up, and a clutch of sorts is released and another engaged, and the drill motor is stopped, while the whole shabam is being powered off of the atwood. I'll get a copy of this idea up as soon as I can, I've been working on this concept for some time now.
But that's my two cents on the whole automatic transmission idea.
Ivey

I agree that there are different -- probably better -- ways of accomplishing the same task. This was just an idea I tossed out there for criticism and improvement.

What you described sounds a lot like 217's CCT (There's a whitepaper describing its design and function on this site), though, admittedly, I don't know much about how the CCT works -- so I could be far off base.

Thanks for taking the time to read through this thread -- especially since it had been dormant for so long.

M.Krass

I finally found this thread - took a long time to read through all of it (i wish you would have PM'd me :^) ]

first off, I love what you have done with the idea - you've really taken the ball and run with it here.

second - I believe the idea is sound and very clean as you have implemented it. Ive taken note of some of the objections in the thread, but I dont think any of them are serious problems

third - I have only one suggestion to make the system better. In your first drawing you show the mating surfaces, the part that will catch and transfer power when the clutch locks, at 90° relative to the clutch face. Id recommend you put an angle in there, so that, when the clutch engages, the two halves pull themselves together - a little bit of the spiral, but in the opposite direction - then as soon as it starts to catch, it will pull itself together tight, not depending on the pressure of the spring to engage fully

and that brings us to one of the repeated objections - noise? clatter? I dont think so. When you ride a ten speed bike, and stop pedalling, is the clatter from the one way clutch in the back wheel deafening? No! does that clutch inside the gear assembly wear out frequently? have you EVER had one wear out on you?

no

With that little bit of back angle on the catching surfaces, the spring that pushes them together can be very soft - remember, this is only going to engage at very low rpms, and once the edge catches, they will 'screw' themselves together (if you angle that surface a little like I mentioned above).

As for the SW control problem, thats simple. You dont want to lean on the high speed motor when?

when the clutch is engaged

when is the clutch engaged?

when its pulled together!

it would be simple to ride a contact switch on the back side of the moving part of the clutch, and when its disengaged, the clutch is pushed back a little, the switch closed

and SW knows its time to allow serious power to be applied to the high speed motor, and if you want, to power off the low speed motor.

I think some people are interpreting the concept behind this wrong - dont think of it as a traditional automatic transmission, with each 'gear' covering half the operation range.

Design the high speed motor and gear ratio so the bot will drive the way you want, as if it were the only motor - dont gear it up so high that the bot cannot start without popping the 40A breakers

then design the low speed transmission geared down so low, that you can spin your tires if you want to - it might have a top speed of 1mph

which is what you want in a shoving match - kinetic friction has no knowledge of how fast a wheel is slipping on the carpet - once it starts to slip, the force you are able to apply is the same, no matter how fast you spin the wheels

so you could have a bot that drives 'normally' in both directions, is pretty quick on the playfield

but when you want to push, you have the all the torque you need (ie, once you start spinning your wheels, thats all you need)

Nicely done! I really like it. :c)

Another easy solution to the SW control issue.

The driver will know when he is in a shoving match - you could have a button or use the trigger on the joystick to powerdown the high speed motor (give it some power, but not alot)

this then becomes an automatic transmission most of the time - when the bot is not pushing an obstical, you can apply full power to both motors from a standstill, and the bot will acellerate out of the low gear range quickly, and transisition into 'high gear' smoothly

and then when the driver wants to really load up the high torque motor, he can squeeze the trigger or button, to keep the high speed motor from over heating

kinda like pulling your auto tranny into low when you are going down a hill, or plowing through deep snow.

Another possible way to implement this

if you want 4 wheel drive - put the drill motors, geared way down, with the one way clutch, on one set of wheels (lets say the back wheels)

and the high speed motors on the other set

when you are moving around quickly you dont really need 4 wheel drive

and when you are pushing someone, most of your weight is shifted to the back wheels

might save on some of the complexity of the mechanics.

Yeah, I'm glad I found this tread, it has a lot of good information, but I have worked on a design and have finished it today! But there are a few problems, I haven't been able to decide a way to make the planet gears to stop when I need them to, but I have a pretty good idea of how it should work, and it also relies a good amount on making the ring gear rotate by friction, because at some points you must let the planets rotate freely and such. The biggest problem is that one motor must overcome the resistance put forth from it's opposing sun gear, this could cause problems in the electrical system. But the two sun gears are where the power comes from, and then transfered to the planet gears, and then to the ring gear. The ring gear is the gear that will be attached to the rest of the drive system. I'll post the link as so as I get it uploaded.
Ivey

Think Power...
 

There is a ton of good thinking in this message thread...

...but... there is another ton of sloppy thoughts in this message thread.



Too much to deal with on a case by case basis, so I will leave it as an exercise for the student to sort it all out.

but I do have a thought I two I would like to share.

#1 No one discussed the fact that the drill transmissions themselves have DOUBLE ACTING one way clutches in them.

The motor can drive the output, but the output cannot drive the input. Using such devices on two inputs to a planetary gearset (or its equivalent -- any of a family of devices with equations of the form w1 + (R2*w2) + (R3*w3) =0 where R2 & R3 are allowed to be negative), you can avoid the some of the problems that Paul has had with his Thunder Chicken Tranny -- you don't need to use a worm gear drive for example.

#2 People who design robots really should think in terms of power. The speed and torque of a motor are more or less just an accident of birth. Given the right amount of power, you can always use a ratio to trade one for the other.

Thinking in terms of power you can see the advantages of using 2 motors to drive a differential even though the input torques to the two motors have to balance -- the output speed doubles at any given torque -- this is another way of saying that the POWER doubles -- if you pick your ratios right, you can do twice as much work per second. This is often a good thing.

#3 To my mind, the main reason for multiple motor drives is to increase power. If you want to a very different speed/torque curve (more speed now, more torque later) I think you will almost always be better off with a shifting transmissionrather than motors that engage and disengage, especially if you want to use the drill transmission as your shifter.

Relatively robust systems can be made tolerably easliy if you are willing to live without shift on the fly. Using multiple motors makes using the standard drill tranmission a bit more tricky but not as difficult as rolling your own shifter (imho). CVT's and other more complicated systems quickly reach the point of diminishing returns.

I have more to say, but that will have to do for now.

Joe J.

Hey Joe, I really want to hear more about your thoughts on drive trains. The members who were a driving force in design last year graduated, and now we have people that are creative and innovative, but we need some of the skills, our team sponsor/teacher is a genius when it comes to this stuff, but Chainsaw is only one man. By the way here is the link to my design, planetary gearset.
Ivey

Complicated issue...
 

Ivey,

I am really sorry that I have more or less give up monitoring forums, mainly watching from a distance and only once in a while writing a reponse.

I love these forums and would like to be more involved but between my family, my church, my work and FIRST, something had to give in my life.

Now to your point. I can't really go into the detail that is needed, but I will say that robust drivetrains are the most important part of a robot. They are really worth all the discussion this thread and others have put into them.

That said, I have really become a believer is simple simple simple drive trains (and this from a former 4 wheel drive, 4 wheel INDEPENENTLY driven swerver advocate -- to be honest, I have a 4 wheel swerver, 4 wheel drive SHIFT ON THE FLY design sketched up that I had been noodling with as a future Chief Delphi drive train). The K-3 rule and the push-a-thons that FIRST has been giving us lately have really forced me back to basics.

For my money, a simple geartrain with a chain drive is a real winner.

As to chain, if you are careful, you can usually use #25 chain, but, I have become a big believer in #35 chain of late (based on our robot last year and based on 3 years of "OCCCRA" robots). It is SO easy to work with and it lets you get away with SO many mistakes and SO much abuse, that I wouldn't even consider #25 chain for my drive system.

As to the number of wheels and orientation of them, this is very game dependent. All other things being equal, I like all my surfaces that touch the ground to be powered. But, I will say right now, I am not a big fan of treads, though people of good will can disagree on this point (for me, I think there is almost always a wheeled solution that is simpler and more robust -- but I have already admitted my bias on this point). Back on driven wheels vs. casters, depending on the game (and especially the floor), casters can be an excellent solution (espcially if there is not a ramp).

As to 8 wheel drive (4 on a side) or 6 wheel drive (3 per side) or 4 wheel drive. I confess to be totally in love with 6 wheel drive with the middle wheels lowered by 1/4 to 1/2 inch (the machine rocks). This system is a good compromise in my view between 4WD with the wheels in the corner (which can lead to a very difficult machine to turn -- especially if you have grippy tires -- one year we had a 4WD machine that would not turn at all unless one wheel was powered in reverse, full power to one side with the other side off STILL drove the machine in a straight line) and 8WD or NWD (N>6). The ratio of distance between the sides and the distance between the 4 wheels that are touching the ground is very important for determining how much scrubbing occurs when your machine turns (this can be managed with omni-wheels but that is another topic). Bigger is better. This is why I advocate lowering the middle set of wheels on a 6WD system -- it effectively doubles this ratio over a similar 4WD machine.

As to types of wheels, I am a fan of pneumatic wheels now that they are legal. 8 inch "Mountain board" wheels worked pretty well for us and a number of other teams. They typically come with a 2 piece hub that has a bolt pattern that makes mounting a sprocket pretty much a walk in the park (the hub itself is your hole template -- how hard is that?)

As to transmissions, I am pretty much agnostic on the whole roll your own vs. use the drill tranmission controversy. Good arguments can be made on both sides. For me it comes down to putting your resources where they have the biggest impact. If you simply cannot live with the package space taken up by the drill transmissions OR you are convinced shift on the fly is your ticket to Atlanta this year, by all means roll your own transmission. BUT AGAIN, only make it as complex as you need it to be.

I don't think I speak too braggingly to say that I have designed some pretty sweet FIRST gearboxes in my day and I have never had to use anything other than a 2 or at most 3 flat plates to hold the bearings for my shafts. I pretty much use straight spur gear transmission. Most of the complexity of my gearboxes comes from acknowledging that I am not smart enough to always get my ratios right (sometimes, I the game is just not what I thought it was so I need more torque or more speed or both -- if I need both I need a shifter OR more likely I need another motor). The complexity this drives is that I have to think hard and have BERG, Small Parts, Stock Drive Products & Mcmaster catalogs open so that I can make sure that I can do 2 things: buy my gears off the shelf (thus the catalogs) AND change ratio without having to change my gearbox plates (this is the hard thinking part). If I can, I try to put a relatively large set of gears (not wide, but large) no more than one stage after the motor. I attach these gears the shafts using trantorques. I know this will shock a lot of folks, but I love them for 2 reasons #1 The only modifications I have to do to off the shelf gears is to put in the right size hole for the trantorque and #2 I can change my gear ration in a few minutes.

If you can use the drill transmissions, I think that it would be hard to justify not using them, especially since they went to the 1/2 inch drill output.

One final point is that before you decide you just HAVE to have that multi motor drive system, make sure you are not using the cross axis helical gear system FIRST provides or that you are using some other inefficient stage in your gearbox (if you have bevel gears or worm gears or -- to a lesser extent -- a home brew planetary gearbox or differential in your drive system THIS MEANS YOU). The best you are going to get out a cross axis drive like the ones in the kit is something like 70% effeciency. If you can get that stage out of you drive train it is almost like a 50% increase in output power to the wheels. This often makes the difference between willing a pushing match and constantly popping breakers.

Well... ...this message has definitely gotten out of hand.

I will end it now but again, not because I don't have more to say but because time is short.


Joe J

I thougth of one limitation of the transmission idea presented in the first post.

When you take a 10 speed bike and roll it backwards, the one way clutch engaged, and drags the pedals around backwards.

so what this means is, not only would the low rpm motor not be able to apply a force in reverse direction, it will also limit that max reverse speed

because the low rpm motor will have to spin backwards (powered) to allow the clutch to rotate backwards

so if your low speed motor is geared down to have a max speed of 1mph, then that is the fastest you will be able to go in reverse - even though it is only the high speed motor that can move the robot backwards - the low speed motor will have to be spun backwards to 'get out of the way'

this could easily be fixed by holding the clutch open mechanically or pnuematically - but then you have implemented a shifter

and if you start down that path, you have to do a tradeoff study against a normal (shifted) two speed transmission.

Re: Complicated issue...
 

OK Joe, NOW you got my attention... I love trantorques, too. I used them at Huron in the CDI two years ago. The idea of using them in a gearbox for ratio changes is intriguing.

But I'm missing something here. HOW do you change a gear ratio in a two plate gearbox without having to drill new holes in the plate??? Changing a gear to a different tooth count changes it's diameter, which affects the spacing between the shafts and changes everything. Can you please describe (or show an image of) your "floating gear" mounting to allow diameter changes to still engage its neighboring gears properly without a new plate set?

BTW, when you make a gearbox, do you make different versions for the two robot sides so that clockwise on a motor always drives the robot in the same direction instead of having to reverse power to one side? We've been noticing "arcs" in our robots from gearbox 180 degree rotations because forward and reverse motor behaviors are different for the same power level. If you ARE making different versions per side, how are you implementing it: via a jackshaft, an extra gear to reverse one side, a double sided output shaft, or what?

Thanks!

- Keith

I appreciate the continuing discussion in this thread. :)

As for changing ratios using a trantorque -- wouldn't that simply mean that you can change to any ratio that maintains the same center-to-center distance of your initial gearset? Maybe there's a better way, but that's what I interpreted that to mean.

Trantorques and Biased motors...
 

As to keeping the center distance, M. Krass is right, all you have to do is keep the gear set such that the gears add up to the same number of teeth.

I often refer to it as "flipping a tooth" from on gear to another. For example, I may use 13:47 and 14:46 on the same center distance.

The problem with the above scenario is it is not much of a ratio shift in this case and more importantly, they are gears that are almost certainly not available from PIC Design, W.M. Berg, McMaster, Stock Drive Products, or Small Parts Inc. (In addition it is a bad example because unless the module is huge -- meaning the D.P. is small -- a 13 Tooth gear is not likely to fit on a Trantorque).

While I am thinking about it, I always add a bit (5-10 thousandths of an inch) to the theoretical center distance. Backlash will almost never hurt you. Gear tooth interference from out of round gears is a killer.

Back to the changeable gear ratio gearbox, it is not as simple as making sure that Trantorques fit in the center of the gear and that off the shelf gears are available. You have to really layout the gearbox to make sure all the gears actually package in your gearbox. Sometimes I have to move boltholes between the plates so that they miss the gears in extreme cases. It is easy to put the holes in place while I am making the plates. The key is to plan them from the start not to have to add them after the fact. You also sometimes have to reduce shaft diameter or move a motor or whatever in order to make sure all your gear combinations can be accommodated.

Now for this whole, motor bias issue. Essentially this is when a motor performs differently in one direction vs. the other. When you have a robot drive system that has the right and left sides mirror images of each other, the motors drive in opposite directions when the robot goes forward or backward. This leads to a robot that wants to turn (especially at the start).

Yes I agree that motor bias is real, particularly on drill motors. Where I disagree with folks is where the fix should be. Many people feel very strongly that it should be fixed by either putting another gear stage in one of the motors to flip it's direction or doing something else to make sure the motors turn in the same direction.

My take on this is that if this is easy, be my guest, but if it is driving ANY complexity at all, it is not worth the bother. Given a competent driver and drive time, it is very easy for a driver to learn how to compensate for this problem. Alternatively, in 2 lines of code, you can have the computer compensate somewhat for this effect. Many folks even go so far as to use the Yaw Rate Sensor to allow the computer to fix this problem.

Last year, for whatever reason, we had one side that was a bit less efficient than the other side; this made the machine want to go in an arc. WE LEFT IT ALONE. Our driver was comfortable making the corrections himself and we were comfortable working on other, more pressing problems.

I am sure that I will get a flood of zealots on the other side of this issue calling me a heretic, but... ...so be it. I have to call 'em how I see 'em.

Joe J.

we used the yaw rate sensor last year to close the loop on steering - it worked so well I cant believe we never tried it before

and it only takes two lines of code to implement.

You take the difference between the joysitck X input, and the yaw rate sensor output (the difference between what the driver is telling the bot to do and what it is actually doing)

mulitiply the differnce slightly if you want tighter response (we used 2X)

then use this difference signal in place of the joystick X axis command - it works SO nice - the robot responds like a servo.