Join 3,512 readers in helping fund MetaFilter (Hide)


Steve Durnin's D-Drive
May 15, 2010 8:31 PM   Subscribe

Steve Durnin's D-Drive is a fascinating new infinitely-variable transmission that doesn't use friction components or a clutch of any kind. Video of a prototype with detailed explanations is included.
posted by odinsdream (44 comments total) 7 users marked this as a favorite

 
Planetary gearing used very cleverly. I'm not sure how the speed of the upper and lower shafts are controlled without some sort of friction.
posted by Jumpin Jack Flash at 8:44 PM on May 15, 2010


My D:-Drive has a malware-infested copy of Vista that barely runs which I'll probably have to delete one of these days.
posted by koeselitz at 8:47 PM on May 15, 2010 [3 favorites]


Eh, not to sound dumb, but:

In order to control the spinning speeds of the upper and lower shafts, you have to input a certain amount of energy - for instance, to put the D-Drive transmission into neutral, you have to spin the bottom shaft around at a speed that equals the speed of the driven top shaft.

So, it's a kind of a differential?
posted by AzraelBrown at 8:51 PM on May 15, 2010 [1 favorite]


After reading the comments on the site it's either a clever shell game recreation, or something only slightly novel. The comments also led me to this awesome video of a John Deere transmission. Mmmm...transmissions and techno.
posted by odinsdream at 8:54 PM on May 15, 2010 [1 favorite]


I'm not sure how the speed of the upper and lower shafts are controlled without some sort of friction.

One of the shafts is connected to a second (apparently speed-controlled) electric motor. That is, part of the power input into the transmission must be in a form that can be controlled without requiring its own transmission. Hence his talk of electric motors and hydraulic possibilities.

They're claiming that the regulation power is just a fraction of the main drive power. Maybe that's true, but I don't understand how that could be. Work is always force times distance. But then I don't understand all that eccentric stuff going on at the input side at all.
posted by Western Infidels at 9:04 PM on May 15, 2010


Also, I like stories like this, I'm grateful for the post, and I hope it's as promising as they make out. I don't mean to come off like a grouchy cynic, it's just a natural talent of mine.
posted by Western Infidels at 9:15 PM on May 15, 2010


Here's an old video on differentials from a FPP: http://youtube.com/watch?v=yYAw79386WI#t=3m22s

The theory is that the second shaft doesn't really have much load on it, so it would take very little power to make it do its thing. Now they just have to test to see if the efficiency gains over other types of transmissions are greater than the cost of powering that shaft.
posted by xorry at 9:26 PM on May 15, 2010


Technically, I suppose it should be an FPP and not a FPP.
posted by xorry at 9:27 PM on May 15, 2010 [1 favorite]


Yeah, I'm pretty skeptical. How much force is there going to be on the inner shaft by the main motor? It seems like there would be as much force on the inner shaft from the main motor as their would be on the crank.

So, I think you would need a motor that was just as powerful. But maybe you could use it to slow something down temporarily, using a powerful motor. But then, that does just seem pretty similar to a differential.
posted by delmoi at 9:38 PM on May 15, 2010


This falls into the category of things that excite me but I completely fail at comprehending. I even asked the very smart and technically / mechanically oriented folks over at MeFightClub about this thing. A buddy of mine is an ME and emailed me about this thing 4 times over 3 hours.
posted by lazaruslong at 10:46 PM on May 15, 2010


Another continuously variable drive, already in production: NuVinci.
posted by werkzeuger at 11:57 PM on May 15, 2010


werk,

I'm not a mechanical engineer, but how can a smooth sphere provide anything approximating pushback?
posted by effugas at 12:26 AM on May 16, 2010


The video
posted by dibblda at 12:47 AM on May 16, 2010


Well, they aren't perfectly smooth.
posted by Soupisgoodfood at 12:53 AM on May 16, 2010


Soup,

Compared to a gear? With teeth?
posted by effugas at 1:10 AM on May 16, 2010


Apply enough pressure and you can get a lot of traction from a smooth metal surface. Depending on how much power you're putting through, such a device may be more than adequate.
posted by Soupisgoodfood at 1:49 AM on May 16, 2010


This is a pretty clever little device, even if it isn't efficient or has some other flaw that reduces its utility. Thanks for posting this!
posted by Blazecock Pileon at 1:56 AM on May 16, 2010


Well, I spent a fair few years searching transmission patents, and geared continuously variable transmissions based on differential gearings (which is what this is) aren't anything new. At all.

Indeed, if you have a look in Durnin's published international patent application, and specifically into its Search Report, you'll see that the examiner at the Australian Intellectual Property Office is citing no less than four prior art documents in the X category ("particularly pertinent") for all 18 claims.
posted by Skeptic at 3:35 AM on May 16, 2010 [4 favorites]


Also worth noting, speaking of that international patent application, it is that it is claiming a priority date of September 2007. When you file an international patent application you have 30/31 months (depending on the countries) to enter the "national phase": getting the patent application examined in each of the individual countries or regions in which you want a patent to be granted. This is very costly: the applicant may have to disburse several tens of thousands of dollars at this moment.

In short, it is just the moment when a cash-strapped individual inventor may wish to attract venture capital. Now, I don't want to piss on Mr. Durnin's parade, and this may turn out to be a patentable and useful invention, but this breathless, uncritical, not-particularly-informative report strikes me as a bit of a marketing exercise.
posted by Skeptic at 3:48 AM on May 16, 2010 [4 favorites]


It's just a differential. And it's exactly the same as the Prius drive, not just in function (input and output configuration) but also in physical layout (planetary gears)!.

Here's why it's not useful: to get it into neutral, the "control input" must sink ALL of the engine's power and to get it to top speed, the control input must source as much power as the engine. If you want to use it as a CVT, the "control input" must itself be driven via a CVT mechanism - in other words, this just moves the CVT outside of this mechanism to somewhere else. It is not a CVT, it is a differential. And it looks like a pretty inefficient diff, at that.
posted by polyglot at 3:49 AM on May 16, 2010 [2 favorites]


I like his claim that infinite torque is possible since it's a planetary geared mechanism. Apparently he's never read the specs for any bike's three-speed hub, which always recommend minimum gear ratios: If the chainring is too small and the cog too large, an average human's weight on the pedal will apply sufficient leverage to strip the hub's hardened steel drivetrain.
posted by ardgedee at 6:09 AM on May 16, 2010


It's a very cool prototype, but after watching dibblda's video link, I have to agree with polyglot that the crucial element to this transmission working as a CVT is a continuously variable input. In the demo, that's supplied by an electric motor.

I'm not an engineer, so I could be mis-interpreting what I'm seeing. But they said that the output speed is determined by the relative speeds of the input and output shafts. Which means the input shaft's speed has to be continuously varied.

I'll get excited when I see the input shaft driven by something that isn't varying it's speed or torque, e.g.: an engine running at a fixed RPM. Do that and show me the perfectly smooth transition from reverse-neutral-topspeed. That would be a CVT.
posted by cyclopticgaze at 6:11 AM on May 16, 2010


As I understand it, it's got two inputs; the non-variable main drive and a much smaller control drive, which is variable, with a single output. In the prototype, he's using electric motors for both, but you could use a conventional IC engine for the fixed input and an electric motor or bleed off some main power and dump it through a conventional CVT for the control input.

I don't claim to understand how the planetary gear system actually does what he claims upon first blush, but his claim is that the control drive alters the planetary gear system, such that the ratio between the two inputs (i.e. non variable main power and variable control) is what determines the mechanical gear ratio, and thus the RPM of the output. But it appears to be reasonable.

In fact, it appears to be a type of Infinitely Variable Transmission, (IVT) which is usually a CVT + planetary gear system. To quote wikipedia:

"Most IVTs result from the combination of a CVT with a planetary gear system (which is also known as an epicyclic gear system) which enforces an IVT output shaft rotation speed which is equal to the difference between two other speeds within the IVT. This IVT configuration uses its CVT as a continuously variable regulator (CVR) of the rotation speed of any one of the three rotators of the planetary gear system (PGS). If two of the PGS rotator speeds are the input and output of the CVR, there is a setting of the CVR that results in the IVT output speed of zero. The maximum output/input ratio can be chosen from infinite practical possibilities through selection of additional input or output gear, pulley or sprocket sizes without affecting the zero output or the continuity of the whole system. The IVT is always engaged, even during its zero output adjustment."

The first IVT does date from the 1930's, and used friction rollers - a similar design is still in production today.

IVTs can theoretically be more efficient than CVTs, given most of the power passes through the primary planentary gears rather than the control system (same idea as a transistor), and torque transmission is supposedly better. Which makes me wonder why IVTs aren't already in further use; perhaps the cost and complexity of having a 2nd rotary control input outweigh the efficiency gains, at least when using an IC engine?

Maybe IVTs in general will prove much more useful with electric vehicles, especially in a design like this (i.e. gearteeth rather than friction rollers or friction clutch plates). It certain seems interesting enough to be worth a 2nd look, though I'm not sure how revolutionary it is, especially - it looks more like reinvention of an existing idea, to be honest.
posted by ArkhanJG at 6:58 AM on May 16, 2010 [2 favorites]


Like everyone else says, this relies on having a continuously variable-speed input to the 'control' shaft. It's not *just* a differential although it is basically a differential. The big question is how torquey the input to the control shaft has to be during acceleration or deceleration I think the transistor analogy is probably apt: if this thing does indeed work it's because the planet/sun arrangement allows the system to use a much smaller input for ratio control than the main power input. Essentially it's just offloading the CVT to the control input.

(It's nowhere near as hard to understand as the clip says; it's freakin' simple).
posted by unSane at 7:12 AM on May 16, 2010


Oh, and an IVT shouldn't be confused with a power sharing transmission (PST), which is what is in the Prius (also known as Electric Variable Transmission) - there you're combining power from two inputs (electric motor and ICE) for a single output, altering the ratio to change how much power you draw from each input; or splitting one input into multiple outputs, i.e. by running in a different way, you choose how much mechanical power you use for powering the vehicle, and how much you route back into the electrical system to charge the storage batteries.

In an IVT, you're using a small input to control a large input to get a variable output by varying gear ratio. Both use a CVT and a planetary gear system, but for different purposes.
posted by ArkhanJG at 7:19 AM on May 16, 2010


The transistor analogy is a good one. The point is that this transmission (purportedly) requires negligible input to its control "circuit" to effect a large change in its output speed. It is, in effect, a valve for torque.

It isn't a differential, because the control is different from the output. If you load the output, it doesn't change its output speed. (Until the power engine can't keep up any more.)

However, I didn't see how it works for speed multiplication- presumably, it works just fine, but its hard to tell from the explanation. Further, an efficiency problem I see is the need for the control motor to be spinning at all times when the transmission isn't at full speed.

Also, it would appear that it would require computer controls to be suited for automobile usage. The computer would have to read the accelerator input and vary the speed of the engine and the control circuit.

Lastly, although it eliminates clutches and whatnot, gears do have friction. Square-cut gears are noisy, and angle cut gears add friction.

It's a fantastic idea, but the inventor is completely correct- they need to build a prototype and see how it actually performs compared to current solutions.
posted by gjc at 8:59 AM on May 16, 2010


I think most people are not really getting it. The speed of the upper shaft in the D-Drive is controlled by a smaller D-Drive using power from the first D-Drive. What controls the speed of the smaller D-Drive? D-Drives all the way down, of course.
posted by digsrus at 10:06 AM on May 16, 2010 [1 favorite]


Yep. Transistor analogy holds true.

What folks are missing about the control input, with regards to efficiency, is that it requires basically no torque! It only requires enough torque to move the geartrain. It's not torquing against the load, nor is it torquing against the input torque.

They're claiming that the regulation power is just a fraction of the main drive power. Maybe that's true, but I don't understand how that could be. Work is always force times distance. But then I don't understand all that eccentric stuff going on at the input side at all.

That's the point. While distance may be quite high (infinitely spinning doohickey), force is very, very low.

It only requires controllable speed. A three-phase AC motor speed controller is quite an efficient device. And not especially expensive, either.

Also, it would appear that it would require computer controls to be suited for automobile usage. The computer would have to read the accelerator input and vary the speed of the engine and the control circuit.

Modern cars are fly-by-wire anyway for the throttle. Really, this is just a few extra lines of code and a fist-sized electric motor.

I'd really like one of these on my all-electric roadster in a few years. I don't want a friction-based CVT because I want the massive torque an electric motor can produce. I don't want a standard transmission because the torque makes clutching/flywheeling essentially impossible for a human--you'd need to rev match on up and down shifts, and just one hill start would smoke your clutch. And don't even talk to me about an automatic transmission. Torque converters are a laughable, inefficient solution to the problem of bloody lazy drivers.
posted by Netzapper at 4:24 PM on May 16, 2010


I think most people are not really getting it. The speed of the upper shaft in the D-Drive is controlled by a smaller D-Drive using power from the first D-Drive. What controls the speed of the smaller D-Drive? D-Drives all the way down, of course.

Actually, this could work. I know you're joking, but with enough D-drives, you could control the speed of your final control shaft with a tiny DC motor connected to the first control D-drive. Like, the one that vibrates your cellphone.

I don't know how many steps you'd need to go through, but you could bleed power off your main engine with a fixed gear ratio to the torque input of each control D-drive. Then connect the output of each control D-drive to the control input of each subsequent D-drive.

[I tried to make an ASCII graphic explaining how this would work. But, I gave up.]
posted by Netzapper at 4:33 PM on May 16, 2010


I'd really like one of these on my all-electric roadster in a few years. I don't want a friction-based CVT because I want the massive torque an electric motor can produce. I don't want a standard transmission because the torque makes clutching/flywheeling essentially impossible for a human--you'd need to rev match on up and down shifts, and just one hill start would smoke your clutch. And don't even talk to me about an automatic transmission. Torque converters are a laughable, inefficient solution to the problem of bloody lazy drivers.

If you're talking all electric, why do you want a variable transmission at all?
posted by Soupisgoodfood at 6:53 PM on May 16, 2010


If you're talking all electric, why do you want a variable transmission at all?

Because even an electric motor has a peak efficiency RPM or peak torque RPM and you want it to operating there.
posted by unSane at 7:14 PM on May 16, 2010


Without reading the article or even these comments, I already know this is bullshit. No real engineer, working to solve a real problem, tries to invent a mechanical infinitely-variable transmission. Hello, just use electricity, FFS.

And by the way: Welcome you to the 21st century. It's awesome!
posted by ryanrs at 7:56 PM on May 16, 2010


On preview: electricity, yay! See also: AC motors & inverters.
posted by ryanrs at 7:59 PM on May 16, 2010


Because even an electric motor has a peak efficiency RPM or peak torque RPM and you want it to operating there.

But is the difference enough to warrant all the extra mechanics? Can you give an example of an electric car or train that uses variable gearing during normal operation? I've only ever seen it on home-built jobs.
posted by Soupisgoodfood at 8:53 PM on May 16, 2010


But is the difference enough to warrant all the extra mechanics? Can you give an example of an electric car or train that uses variable gearing during normal operation? I've only ever seen it on home-built jobs.

The difference is speed for cost. You can use a lower-torque, higher-speed motor, then gear it down for launching torque and gear it back up for higher top speed. Or, you can use a lower-torque, lower-speed, lower cost motor.

The Tesla Roadster originally had a two-gear drivetrain. They dropped it in the production version because it was unreliable and kept breaking, not because it was a bad performer.

I'll grant you that an ideal AC motor would be better than a variable geartrain. But, a real-world AC motor, available at production-car prices, can very well benefit from one. With a $2000 motor, $1000 controller, and a $1000 D-drive, you can easily achieve what it would take a $50,000 motor to replicate with fixed gearing or direct drive.

Anyway, I'm specifically thinking about home-built solutions. I really, really want to electrify my Impreza. And I don't want a high-efficiency granny car. I want a sleeper Tesla.
posted by Netzapper at 11:48 PM on May 16, 2010


I'm pretty sure you're mistaken there, Netzapper. A transistor is a nonlinear device and this contraption is a differential - it has no nonlinear mechanism at all, just a linear combination of speeds and torques. It's not a valve, though the inventor would like you to believe that it is. It's worth noting that you cannot implement a switch or valve using purely linear terms; there must be nonlinearity and this design has none.

It has two inputs ("motor" and "control", though they're functionally identical), call them A and B. One output, Z. In terms of rotation, Z=0.5(A+B) and in terms of torque, z=a+b; those two equations are all you need to describe (excluding losses) the operation of the device and it doesn't matter how the components are arranged internally, i.e. the use of planetary gears is irrelevant. Power input is A*a+B*b, power output is Z*z.

With stationary output and the control input spinning, Z=0, B=-A, a=b=z=0. You could just leave B disconnected and it would spin freely. Due to the lack of torque on any input or output, there is no power transfer anywhere. Similarly if you lock the control input you get B=0, Z=0.5A, b=a and z=2a, in other words it's a direct drive with some ratio... so far so good because the control input power is zero (B*b=0 because either B=0 or b=0).

The problem is that when you try to transition between these two distinct ratios, the control input power becomes significant. Say you want O=0.1A, which requires B=0.9A. If you want ANY torque to appear on the output then you must apply a torque to the control input, otherwise the control input will spin freely. Now you have non-zero torque and non-zero speed on the control input, and that means non-zero power - the exact balance is up to you, but the device becomes a power splitter exactly like the Prius drive depending on how much torque and speed you apply to the control input. You can either put power into the control (and therefore get it from the output), or the control can sink power and thereby reduce the torque available at the output.

Further proof: conservation of energy in conjunction with no control power either in or out would mean Z*z=A*a and B*b=0. If you expand that out, the only solutions are where b=0 or B=0, which means it's not continuously variable unless you have a powerful control input.
posted by polyglot at 11:58 PM on May 16, 2010


oops, Z=0.1A in the 4th para. I changed nomenclature and failed to fix all the instances.
posted by polyglot at 11:59 PM on May 16, 2010


grar and B=-0.8A to get Z=0.1A. And disregard my z=a+b statement because it's actually z=2(Aa+Bb)/(A+B) which you can see from conservation of energy.

The point being though that B*b != 0 while you're varying speeds.

To make an analogy, it's just like a clutch wastes no energy when fully engaged or fully disengaged. While you transition though, the clutch is slipping and absorbing power which is transformed to heat.

If you take the Prius case, the presence of powerful electric motors means that this arrangement can act as an efficient clutch without requiring friction plates. During the transition phase, the energy is transferred into the battery instead of to heat. But then, that's not particularly new if Toyota has been doing it for most of a decade...
posted by polyglot at 12:26 AM on May 17, 2010


I get what your saying. I would have thought that if there was enough of an advantage, that it would be more common, but haven't seen much data on the subject.

I've also read about people breaking transmissions with such conversions, let alone high-end prototypes, so I hope your Impreza's up to the job! Be sure to make a post about it if you start the project.
posted by Soupisgoodfood at 12:46 AM on May 17, 2010


polyglot, your analysis is depressingly convincing.

However, it appears to me that the control input acts essentially as an idler gear of variable speed. That is, the main load of torque passes through it without affecting it so long as it's allowed (driven in this case) to rotate at the appropriate speed.

Are you saying that during the transition itself from one ratio to another that high torque is needed on the control input, but that control torque drops after the new ratio is achieved? Or, are you saying that the torque on the control input is always proportional to the desired output torque?

If it's the former, this thing is still useful, but radically less so than the inventor indicates. If the latter, then, yeah, this thing is pretty bloody useless.

Also, how does the eccentric (elliptical) gear factor into this? Is it just acting as a substitute for a crank?

[I'm not an ME. I'm a software guy. I apologize that I lack the skills necessary to follow your analysis more closely.]
posted by Netzapper at 2:08 AM on May 17, 2010


I'm saying that with any ratio other than 0 or 1 then the B input needs to source or sink power. If you want a stationary output, you can let B spin freely (b=0) or if you want a driven output, you can lock B stationary (B=0) and no power flows through B. Any other speed ratio requires that B either provides or absorbs power.

On the matter of transistors, they make an extremely good analogy to a clutch. With the transistor off (no collector current, voltage can be anything), there is no power flow and no power dissipated by the transistor - just like a disengaged clutch. With the transistor on (collector current can be anything, voltage drop is ~0), there is power flow and very little dissipation by the transistor. Anywhere in-between those two states, there is both a voltage across the transistor and current through it, and it dissipates power as heat, just like a clutch does while slipping. The control inputs (clutch actuator pushing plates together / base current into transistor) have tiny but non-zero amounts of power.

To consider another (mechanical) way of thinking about why this box cannot do what is promised. Consider a black-box, rigid[1] D-Drive with motor & control inputs and a drive output. What happens when you want to start up the output, i.e. apply a torque to the stationary output shaft? Clearly the output power is zero (shaft not moving yet) but the input power (engine is running!) is non-zero. Either the power must be dissipated through some lossy mechanism (doesn't exist in this case; a friction-clutch is the usual solution), be stored (no storage mechanisms are present) or it must be transmitted out of the control port. Anything else is a violation of the law of conservation of energy.

You can get sprung clutches for cars - they're a spring in series with a clutch plate. When engaged, some power is lost to friction/heat and some power is stored in the spring and then released once the friction plates engage. That's a combination of the storage and dissipation solution.

Prius differential drive system simulator linked in the gizmag comments is educational.

The planetary gear arrangement is just a standard way of building a differential. A common feature of all diffs (I think) is that they have three gears; the two differential ports (A,B) are connected to the two outer ones and the middle gear can spin freely but motion of the common port (Z) implies lateral motion of the centre gear. In a basic car diff, the centre gear (Z) is the cage of the diff with the crown wheel and its central pinion gear; the two free pinion gears are A and B. With an epicyclic diff like the D-Drive, the sun and outer ring are A and B and Z is connected to the shafts of the planets. Epicyclics are a bit messier though because you have different ratios on the sun and outer ring and therefore a different torque balance, but the same conservation of energy applies. It's functionally like having a symmetric diff and then gearing one of the outputs down a bit.

Yes, B is an idler. But when you try start up the output, B is spinning like mad. Unless you apply some torque (b>0) to B, then you will get no torque at the output and the motor will just continue to make B spin.

On second thoughts, I suspect my original z=a+b guess was maybe correct. You need all three equations (summed speed, summed torque, conservation of energy) to completely constrain the system. I'm not sure though - radars & software are my usual stomping ground and it's more than a decade since I did mechanics as an undergrad.

[1] rigid: has no friction losses and no energy-storage mechanisms like springs.
posted by polyglot at 5:51 AM on May 17, 2010


ok, this is bugging me now. z=a+b is clearly wrong and I now think 1/z=1/a+1/b is more like it.
posted by polyglot at 6:21 AM on May 17, 2010


Yo Dawg, I heard you like CVTs, so I built a CVT that needs another CVT so you can vary your ratios while you vary your ratios.
posted by pjaust at 6:21 AM on May 17, 2010 [1 favorite]


polyglot- I think the device is meant to be a torque converter. I don't have the ME skills to analyze it from the video.

But I *think* it is is a sort of windlass with an adjustable ratio. By feeding the motion back onto itself, you get the ability to have a zero output speed with very little input power. I think the idea is that the prime mover is throttled to deliver only the torque necessary to maintain the RPM setpoint. If it's an electrical motor, it only draws as many amps as necessary to keep spinning.

Abstractly, it's a Hybrid Synergy Drive with another order of magnitude added onto it. He added another differential onto it so that the control motor doesn't add or subtract power, it just "keeps up".

I don't think it's meant to be magic, it's just meant to be a better solution than friction or viscous clutches.

In a car, I could see this replacing the torque converter rather than the transmission. Hang a small, lightweight three speed "automatic manual" off of it and I think there would be a good efficiency gain. The d-drive can handle torque management so that the transmission only has to deal with changing gears and doesn't have to have friction plates or bands or complicated hydraulic circuits. The d-drive just rev-matches and off it goes. Or, at least, if a friction clutch is easier, it can rev-match so that there is almost no time the thing is losing energy to friction.

(To get really esoteric, you could call this a PWM amplifier rather than a linear amplifier. Each gear tooth pushing from the prime mover to the output is a pulse- the control just changes the rate at which these happen.)

((Again, in theory. I'm not defending it, just explaining how I see it, and how I see it to be fundamentally different from a differential or a HSD.))
posted by gjc at 8:01 AM on May 17, 2010


« Older Go watch some baby sloth's inaction....  |  Late Night with Jimmy Fallon c... Newer »


This thread has been archived and is closed to new comments