Abacus Drive is a Speed-to-Torque Game-Changer

Apart from the harmonic drive, the engineering community hasn’t really come up with any clever mechanisms for speed-to-torque conversion in the last few decades. However, recently a few folks at SRI have given us one more transmission to drool over: the Abacus Drive.

The Abacus Drive takes the standard concepts of a cycloidal drive, but takes the eccentric gear tooth pattern that we’re familiar with and converts it to two grooves in which an array of rolling spacers will ride. The benefit with this design is two-fold: it’s both constructed from entirely rigid components (unlike the harmonic drive), and it has a low-backdriving torque, enabling the application to more easily detect changes in load.

Achieving an affordable low-speed, high-torque transmission has been a holy grail among roboticists, where every motor-driven manipulator joint becomes an engineering design headache where the designers fight their application’s backlash, torque, and price constraints to get a functional robot arm. This problem stems from the fact that motors just don’t perform efficiently at low-speeds, where the near-stall conditions cause them to draw vastly larger amounts of torque compared to their full-speed conditions. While the Abacus Drive isn’t hitting the market anytime soon, we’ll let this idea stew in the community and hope to see some budget variants pop up in the near future.

[via IEEE Spectrum]

34 thoughts on “Abacus Drive is a Speed-to-Torque Game-Changer

  1. Interesting wear patterns already being shown on the top “cone” of those anodized “bearings balls”. To be fair, they should be hardened steel but how do you machine that shape cost effectively in quantity?

        1. Or they enclose the whole mechanism in grease, cover it with a rubber boot and seal it for life. Exactly like they do cv joints which also use ball bearings in odd curved tracks to transmit torque (lots of torque…).

  2. Also, an AC induction motor’s rotor gets to magnetic saturation at a certain (generally fixed) RPM, yes. But you can get around the motor spinning “too fast” issue and it has to do with motor physics of some types of motors but not all. They are painting this as a very broad problem and offering up a niche idea to solve it. It’s probably far better to solve it with a different motor type or other upstream solution.

    A Tesla or other electric vehicle has no clutch and a single fixed gear. This would be of no use in that application.

    Wonder why “gas efficient cars” are now cramming in 10 or even 12 gears into a small 4 cylinder vehicle? It’s to try to meet EPA standards for fuel economy. I mean, adding more gears is certainly one way to do it when you are “stuck” with a gasoline powered engine but it too has more costs and tradeoffs.

    If you care to read more on induction and DC brushless motors, Tesla wrote some about this back in 2007.

    As usual, there are several tradeoffs there too. Sound familiar yet?

    “Induction machines are more difficult to control. The control laws are more complex and difficult to understand. Achieving stability over the entire torque-speed range and over temperature is more difficult with induction than with DC brushless.”


    Seems like Tesla and others have dealt with the control side of things a heck of a lot better in the last decade than taking a fixed AC frequency motor and mounting it to a vehicle and hoping it works then realizing it doesn’t and then trying to apply something like this interesting speed to torque converter.

    Fix the problem higher up the chain, rather than trying to resolve it with this. At least in most applications. This might have some neat niche use but not for automobile use. Neat idea though.

    1. “Seems like Tesla and others have dealt with the control side of things a heck of a lot better”

      They haven’t. It’s still the same issues and inefficiencies in Tesla cars as everywhere else – they’re just talking about it in vague terms to give the impression that they’re doing something new, because the company is based exactly on the kind of off-the-shelf engineering as you describe – except without the part where they try to come up with clever workarounds.

      Tesla did aknowledge the problem of controlling an induction motor over a wide speed range very early on with the Roadster car, and they did try to engineer a gearbox into the vehicle, but they couldn’t hack it so they abandoned the attempt and simply added more batteries to compensate for the loss of efficiency. As a result, all Tesla drivetrains have all sorts of “gotchas” since they’re basically brute-forcing it and abusing the battery pack for more acceleration.

      1. Tesla actually went to one of the “top” transmission companies to get a 2 speed gearbox built for the Roadster, but they couldn’t handle the torque of the electric motor and kept breaking, so the first Roadsters shipped locked into 2nd gear and once their single speed (and other upgrades to enhance motor performance so a single speed was usable) were ready they replaced all of that under warranty.

        1. It was because they were trying to constantly back-drive the motor (no clutch) and work the regen through the gearbox at a high reduction ratio. The motor was already turning up to 13 kRPM on second gear, and the first gear made it even steeper.

          The situation was similiar to taking an ordinary car and bump-starting it on first gear. The gearbox is trying to convert the relatively slow turning wheels into very high RPM at the flywheel, and as you might guess that takes a whole bunch of torque. Any resistance, such as the motor acting as a generator, is amplified by the gear ratio when the gearbox is backdriven, so the forces required to do that, repeatedly over and over again going from acceleration to regen, will strip the teeth off the first gear.

          The irony is that using a second gear would save them about 15% energy and allow smaller batteries, which would reduce weight, which would reduce energy consumption further, possibly slashing the price of the car by $10k or more. The disadvantage is that they couldn’t offer “ludicurous” performance anymore because they’d have to mind not breaking the gearbox, and then it would be just a regular expensive car with the inconvenience of having to recharge a battery.

      2. Isn’t that what SpaceX is doing too? Take existing tech and apply it to space? Are they truly innovating or instead more bringing existing but disparate things together and throwing a ton of capital at it at the same time?

          1. Yah like generating fanboy’s by the bushel., never relaunching a single rocket, over priced terrible range cars for the elite, developing roof shingle solar panels no commoner can afford, building a battery production facility to satisfy non-existant demand They are just changing the world guy get with the program.

        1. Innovation means taking existing things and putting them together for a novel function. It also means taking an existing product and re-introducing it to the public as your own, like how Apple innovated the smartphone.

          It’s basically a bullshit word that means “we haven’t really done anything new, but our marketing department rocks.”

  3. There are a number of videos on Youtube showing cyclonic drives that have moving pins. It’s not clear why This would be better.

    This approach will only work for small loads because the bearing surface is an corner that that the bearing rides against on both the inside and the outside. There’s no good way to increase that surface area.

    if you want to detect if your arm bumps into something, waiting to notice the difference in energy requires to move the arm seems like a poor choice. Far better to have strain gauges built into the system instead.

    David Lang

    1. Look closer at the one with the conical bearings. You could tweak that to get wider bearing surfaces The guy above that mentioned the wear pattern misses the obvious fact that the bearings are made of soft brass and not hardened steel and the lack of lubrication.. These aren’t meant as bearings anyway. They’re meant for use as transmissions for things like robotic arms and such where space is an issue and just brute forcing with a larger motor is not an option. For all intents, its still the basic design of the strain wave , just eliminates the ring gear.

      1. I concede that they could certainly be brass, the rest of the mechanism appears to be black anodized aluminum. This looks more like a fancy “show how it works” appearance model. In practice, the bearing surfaces would presumably be a hardened material though, not brass or aluminum.

  4. I was pretty excited when I read the headline, thinking it was variable ratio, but it’s not. What we need is something like a CVT but with minimal mechanical losses for use in small systems like robots.

  5. They admit vibration is an issue just like a harmonic drive.
    They say it’s not for any kind of high power transmission application, just slow moving things like robotics.
    It’s the friction drive version of a harmonic drive, with all the negatives of both.
    It’s not a variable transmission, even though you see the tapered rollers. It has a fixed ratio.
    Efficiency will NOT be higher than traditional measures. The main benefit is the extreme reduction in backlash. When they compare efficiency to other transmissions, they conveniently make apples to orange comparisons… and even then it barely wins.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s