Report: Hyundai Will Offer 10-Speed Auto Gearbox from 2014

Can you count to 10? Well done, so does Hyundai, and in particular the Korean company’s engineering team responsible for car transmissions. According to a Bloomberg News report, Hyundai Motor Group president Park Seong Hyon said that 10-speed automatic transmissions will become available in the group’s luxury models from 2014. Only a few years ago, six-speed automatic gearboxes were available only in top-of-the-line cars. Then the gear count started to increase gradually: first at seven, which seemed like a novelty, and quite recently at eight, which most luxury carmakers rushed to adopt. Even though it seems like overkill, the reason for the increasing gear count is simple: CAFE, or Corporate Average Fuel Economy, regulations are becoming stricter every year. Downsizing engines is one solution to the problem, but it can only go so far, especially in large luxury models. Therefore, automakers employed another way to achieve low average consumption and CO2 emissions: gearboxes with as many gears as possible, with the top sets having a very long ratio. Porsche has even gone as far as introducing the world’s first seven-speed manual gearbox in the new 911, borrowing many components from the dual-clutch PDK semi-automatic. Now Hyundai is breaking new ground with the 10-speed auto, which will most likely find its way under the hood of the 2014MY Equus and Genesis saloons. Something tells us it’s only a matter of time before someone else announces an 11-speed auto, making Hyundai’s transmission old news… Story References: Bloomberg via Autonews

  • Mark Penrice

    Unfortunately no-one has yet found out how to make a CVT which can handle high torque and have a properly wide ratio range with low drag (ie high enough efficiency that it doesn’t end up ruining any advantage the infinite variability may have provided).

    In reality they originally looked to offer a good alternative to fixed ratio gears back when your typical options were a 3-speed auto or 4-speed manual, with an overall range (including torque converter multiplication) of about 4:1 between first and top. The general “width” of an engine’s most efficient range (if we consider the part of its hp-hour per gallon map, on rpm vs throttle load axes, where it’s within about 95% of the absolute peak), as well as that of its maximum power output, is narrow enough – and the rpm/load combination that’s best for efficient cruising at highway speeds typically far enough into the low rpm/high load corner – that such a limited choice and range of gears has no chance of keeping the engine within or reasonably close to the first region AND allowing use of the second, unless you have a quite low powered engine fitted to a very lightweight but unaerodynamic vehicle… which is why small capacity motorcycles have such frantic, tight gearing.

    So it was thought a torque converter plus a CVT with a relatively “long” differential ratio would be a good way of squaring that circle. It’d still have good launching power, have a wider overall range than a 3AT or 4MT and therefore have better cruising economy (especially with a lockup/bypass TC), with none of the problems associated with the huge gaps between each gear that would be necessary for the other types to achieve the same, or the need to fit a (heavy, more complicated, difficult to adapt for FWD transaxle use) separate overdrive unit. The control system merely modulates the selected ratio (and the *actual* throttle setting, in FBW systems) according to the sensed road load (ie amount of drag being experienced as a combination of rolling resistance, gradient and aerodynamics, thus the force needed to keep a steady speed), the current throttle pedal position and its rate/direction of change (IE driver’s input of desired speed/power changes), such to keep the rpm and load combination largely within the most efficient area of the engine map, or at least as close to it as possible should the power output demand exceed that boundary.

    (EG if you have an idealised smallish engine of that general period (say, about 1.5L) where the max torque is 80lbft at 3000rpm, falls off evenly and linearly on each side to 50% at 0 and a 6000rpm limiter, and the peak is around 75% also at 3000, with the 95% range being a rectangle around 2500-3500 and 65-85% – which isn’t actually that close to how most people drive! – and 90% maybe to 2000-4000 and 55-95%… you could get 34hp at the most efficient engine speed, which is about enough for a steady 78.5mph with something of contemporary aero and weight. Of course you don’t want that much when cruising around town, it’s not enough for maintaining speed up hill or accelerating rapidly, and even if the power is produced less efficiently you still use less gallons per mile by driving at a lower speed (and power output) because you use so much less power vs your rate of travel. Within the 95% box you can have between about 23 and 42hp, or 64 ~ 87mph on the flat; within the 90%, 14 to 48hp (about 50mph to low-90s mph)… outside of that the efficiency tends to drop off more rapidly. FWIW this setup peaks around 51-52hp in the mid-4000s, so it’s not very high tuned – like a lot of engines mated to automatic transmissions it’s focussed more on the torque. A real-world one would probably be able to rev at least a little harder and make maybe 60hp though, probably around 5000rpm, which is good enough for roughly 100mph – this actually matches the setup in some older watercooled VWs, for example, even though that’s not what I was consciously aiming for)

    (In a typical 3sp auto or 4sp manual, which are focussed more on being able to maintain progress up steep hills, accelerate moderately well and achieve top speed somewhere around or a little above peak power rpm, that means your highest ratio would be roughly 20mph per 1000rpm, and the lowest about 5 (including the multiplication effect of the torque converter; actual lowest mechanical gear in the auto would be more like 8mph/1000). Which is actually good news for highway cruising on the flat between 60 and 80mph, as you’re within the 90% range most of the time, and might pass over a corner of the 95+ region, generally moving more towards it than away when climbing moderate hills. However realworld engines tend to have rounded-off edges on their map regions, so actually that kind of setup would cover rather less of both, even with “normal” 5 or 6 speed manuals, and with further reference to VW one presentation slide I’ve seen addressing this shows how they adjusted the gear spread for their bluemotion and other more economy focussed models to raise the top ratios in a “natural” way, such that e.g. instead of a 5-speed version of that gearbox which merely lowered 1st a little and shunted an extra ratio in between it and top, 4th and 5th are instead adjusted so that they sit either side of the original 4th, with the load vs engine speed line for a steady speed on the flat being moved more into the economy zone for both medium to higher and the medium-low ranges, especially if having to push a bit harder up a gradient in the latter case… lower revs and higher load in the first case (eg, extending 5th to 23mph/1000, bringing 80mph just within the 95% range whilst just about hanging on to 60, and making 70 very close to optimal especially up a slight grade or into the wind, and keeping 50-90 within the 90% range, with top speed in that overdrive range somewhere in the 90s and low 4000s; 4th dropping to maybe 19mph/1000, meaning the lower 4 are still a little tighter or lower than the trad 4-speed, and 4th can be used for a little more urge when in the 50~70 (approx 95%) or 40~80 (approx 90%) range without it being terribly uneconomic – the mpg will still be lower as more power is being used at the same speed, but that power is produced more efficiently, so a car travelling the same route at the same speed will use less fuel overall than one that might have cruised at 20/1000 then dropped to a 15/1000 or lower 3rd, or opened the throttle fully (which usually also triggers an enriched fuel mixture) if it couldn’t quite make enough power to maintain speed… so long as the driver shifts back into 5th when the extra power is no longer needed. It can also reach the same, or very nearly the same top speed, just revving slightly harder, and won’t lose quite as much speed uphill if the driver is trying for as high an average as possible.)

    (Summarising that in a single two-line graph, they showed a couple of different cruise speeds/conditions in terms of constant-power lines across the rev range of the efficiency map, with the different ratios marked with X’s and arrows pointing from the original to the split ones showing how it moved the cruising condition from e.g. 91% to 96% at about 75mph, thus providing approx +5% efficiency just by changing the gearing in a way that didn’t particularly impact the driveability or the cost of production)

    Ahem, anyway – given the less forgiving nature of a real engine map, and the quite large jumps between gears (not often 2:1, but certainly 1.4:1 or more (3500/2500) for all but the upper ratios of 5 or 6sp’s – even my dad’s 4AT, 1.4L, 90hp Hyundai slushbox has approximately that much of a jump from 3rd to 4th, and at least as much 2nd to 3rd), it can mean you can’t even get within 10% of the most efficient engine regieme no matter how hard you try to be economic, and still have to make the choice between a transmission that’s good for quieter, more efficient cruising, and one that’s more responsive and delivers a sharper takeoff plus greater, more consistent power across the main range of actual road speeds… at the expense of either general lethargy (this is the approach Hyundai seem to have gone for – the car gives a very smooth, quiet ride, but has trouble keeping up with manuals that have two-thirds the power and hates hills), or thirstiness and lots of noise (VW’s original approach to making 5-speeds, which was very much “GTi” focussed, and what Renault still seem to do).

    The solution to these twin problems is either to employ the twist-and-go scooter solution of a CVT, which extends your range a little at each end (more than a 4MT/3AT, but not as much as a wider ratio type 5MT/4AT) and does away with all the ratio jumps in-between, so if you hold the pedal around 60% and the transmission/FBW system is smart enough it’ll modulate things to keep the revs near 3000 and the actual throttle near 75%, producing roughly 36hp in a very efficient way; or follow what bicycle manufacturers and truck builders have been doing for decades in order simply to get a usable amount of wheel torque from not very strong power sources with narrow peak output ranges – throw increasing numbers of gears at it, preferably via a multiplication and/or splitting system where a core set of 3~5 (truck) or (3, 4…) 5~9 (10?) (bike) gears are duplicated through 2 or 3 axle (or pedal crank) “ranges”, plus an intermediate, milder-overdrive “splitter” twin-range transfer box on the truck that effectively inserts the additional sprockets to its virtual dereilleur on the highest main range.

    For a manual, this is a bit of a challenging proposition as you essentially have to do it the same way as with the truck boxes – similar to the traditional overdrive transfer case or differential on older models, which multiplied up 4th (and sometimes 3rd and even 2nd as well) to give a higher 5th (and 6th, and 3rd-and-a-half… or other such combinations) at the push of a button on the stick. Obviously, not quite as simple as just moving the stick into another position in the sequence, or having an auto box do that itself (which many could, with the “special” stick or button position being one that deactivated the O/D ability), and therefore a bit of a turnoff to those who don’t want to have to learn too much about a car’s operation before driving it. Worse still the 4x4s with low speed transfer boxes, or part-time 4wd (especially that which used clutch hubs), and the few “economy shift” efforts which actually had a whole other gearstick for range selection, much like earlier 3×2 and 4×3 truck efforts that sometimes saw you having to move two sticks in opposite directions for a single gearchange (modern ones just have a combination of a toggle switch for range selection and a button for splitting). Hence why the highest you usually get is 6-speed, Porsche’s 7-speed is a bit weird, and the most I’ve seen with a single range is a coach with 8 (presumably both the lower 2, or higher 2 forwards speeds, AND reverse need the driver to operate some kind of lockout mechanism to make it clear by feel which bit of the gate they’re using – thus the furthest you could go with it is either 9-speed with reverse on the same stick, or 10 if reverse is separately engaged using a button or pull-toggle of some kind after putting the main stick to neutral and coming to a halt).

    The only real way around that complication is to go the other way and simplify things as much as possible with a sequential setup that only allows direct selection of the next highest or lowest gear, by ratcheting a cam system forwards and backwards (often with neutral between 1st and 2nd, because of the common need to select it whilst braking without going through first, the certainty of having 1st when downshifting from neutral, and the minimal delay caused by the larger 1st-2nd shift time given that there’s also usually a larger rev gap; because of this, reverse gear, when it’s even fitted, is usually selected by a separate mechanism), or by signalling an electronic gear selector system (that, like the sequential, ultimately moves the more familiar selector rails internal to the box using indirect action) with a simple pair of buttons, which also allows additional functions like quick skipping up/down 2 or 3 gears at a time by double/triple clicking the button without fully releasing it, or immediately dropping into neutral by pushing both together for at least half a second (and then into reverse by doing the same, or alternatively pressing “shift down” afterwards), as the movement of the selectors isn’t locked to the pattern engraved on a rotating cam. Even so, oddly, these systems rarely have more than 7 gears… it seems to be the optimal number for the sportier vehicles those transmissions are bolted to.

    A traditional type automatic (or a more fly-by-wire type manual where the stick is not necessarily mechanically attached to the gearbox), can get around these problems thanks to its inherent nature as a sequential stack of epicyclic gears. Each sun-planet-ring triplet can only usually produce two (or in complex setups, three) forward ratios and one reverse, sometimes with an extra one if it employs a trick dual-cog-size / dual-carrier-radius planet system. Usually only two (or, with a dual planet, three) of those are useful, though, and one of them is a “straight through” 1:1 gear (with either one/two reductions, or one reduction and one multiplication if they can be made mild enough), so you tend to get at least two epicyclic combinations, next to each other and with the upstream one normally feeding the downstream (…about the only machines that have used a single epicyclic gear after about 1910 are the Ford Model T, with two very widely spaced forward gears and one reverse, some early torque-converter autos that relied heavily on the TC more as a full hydraulic drive than a clutch replacement, and various little centrifugally clutched, low speed offroad utility vehicles like golf carts and lawnmowers, early auto-shift scooters/mopeds, transfer cases in 4x4s and some trucks, etc).

    Interestingly – or annoyingly, if you prefer – this not only means you can take advantage of the two “reverse” ratios multiplying each other, but there are also about as many possible reverse gears as there are forward ones. The mathematics of how each planetary system bears on the next is extremely complicated – at least to my poor, addled brain – so it’s best if you just go look up some animated diagrams if you’re interested (there’s a lot of reciprocals, and most particularly negative reciprocals and weird fractional things going on thanks to the split-planet doodads). For the purpose of demonstration, just imagine that each one allows e.g. 1:1 straight through, -2.1:1 reverse and 1.4:1 forwards (don’t try working out the actual gearteeth needed for that 😉 …

    This means your possible gears with two *identical*, simple planetary systems are:
    -1.9 x -1.9 = 3.61 first gear
    1.4 x 1.4 = 1.96 second gear
    1.4 x 1.0 (or 1.0 x 1.4) = 1.40 third gear
    1.0 x 1.0 = 1.00 fourth gear
    -1.9 x 1.4 (or 1.4 x -1.9) = -2.66 reverse

    Which is, other than the unusually tall reverse gear, roughly similar to a slightly narrow 4-speed manual, and can be given effective ratios closer to a typical constant-mesh model (e.g. 3.20, 1.74, 1.24, 0.89), albeit with an even taller reverse (-2.36, still short compared to a typical 2nd gear), by reducing the differential reduction by a certain amount (in this case, 9/8ths or 1.125, so e.g. 3.88:1 falls to 3.45:1), with the additional torque multiplication and less damaging slippage needed to get moving from a stand (especially in reverse, though you usually use less throttle for that…) being provided by the converter and providing an effective range from 0mph to cruising speed closer to that of a 5MT. Indeed if the converter is effective enough, the differential ratio can be lowered still further to push top gear closer to a regular 5th (make it a flat 3.00:1, and top is effectively 0.77, and the other three 1.08, 1.52 and 2.79:1, hence how 2nd gear in a lot of autos is good for about 70mph on the redline and some drivers can go for a long time without realising they’ve accidentally locked out 4th gear or that there’s a fault in the box… and the launch through to 40-50mph region can be very “slushy” due to relying on the hydraulic multiplication for good starting torque and on its shock-cushioning nature to smooth out the otherwise pretty harsh rev drop, driveline shock and reduction in power at the 1st-2nd boundary, losing about 45% of the rpm at maybe 35mph).

    The simpler 3-speeds didn’t make use of the dual-reverse concept, because it is kind of difficult to synchronise going from both epicyclics in reverse to both in forward reduction, and just either used double, single, and no reduction (so here, maybe a deeper 1.5 second and 2.25 first, with 1.0 third/top) with either identical or slightly different gearing (one thing that the dual planet system was used for, in fact – reducing the part count and friction when doing that sort of thing, basically sharing the carrier between two otherwise separate epicyclics; in this case you might have a 1.675 reduction on one and 1.388 on the other, so your 2nd-3rd isn’t too steep, but you still get a decent gap between 1st and 2nd and a wider range overall – 2.32, 1.39, 1.00), or two different single-reductions with wildly different ratios on each epicyclic… so e.g. 1.5 x 1.0 and 2.4 x 1.0, but only giving 2.4, 1.5 and 1.0 overall. Though 2.4 x 1.5 to give 3.6 seems like an obvious combination for a 4-speed extension, it doesn’t seem to have been a common strategy. Presumably a mix of keeping things simple on the control side (more straightforward controller design, especially as it was originally a set of hydraulic lines and pressure valves controlled by the differences between the force coming from a pump on the output shaft, the two sides of the torque converter, a load sensor of some kind, plus the shift lever position; plus fewer brake-bands and intermediate clutches needing to be controlled, and potentially wearing out, jamming, etc) which both saved costs and made the transmissions more reliable, not making it seem too complex for the customer (the primary market was America, which saw even the retention of the gearstick a bit scary given that some models were push-button, and didn’t really want to have to consider anything more than “high” or “low” should the need for manual override occur… hence why a lot of even 4-speed boxes don’t allow any more advanced control than locking it in 2nd gear), and limiting the number of potentially uncomfortably rough changes when multiple, difficult-to-synchronise tranmission parts had to move in perfect harmony. They also didn’t need too many ratios, just enough that it would be able to screech the tyres a little (and therefore be able to climb any hill that the tyres could maintain grip on) and go up to a decent top speed without blowing the engine, with the slushverter papering over the rev gaps at shift time, because engines were big and torquey, allowing longer differential ratios, no-one cared much about fuel consumption, and even if you went much faster than 55mph, there wasn’t so much complaint about a noisy, racing engine at motorway speed, because it was just considered normal. If you really needed additional pulling power, or better economy and reduced NVH when travelling long distances at high speed, or both, you could always fit an overdrive differential, with either under or over corresponding to the normal setup (or each sitting a little above or below) depending on whether you wanted it to make towing easier uphill, or improve the speed, cost and comfort of doing cross country blasts, or a bit of both.

    And so it was with manuals too; 3 was OK, 4 was generally the norm after being popularised in the 2CV, Fiat 500, Mini etc, and 5 was a bit extravagant but not unheard of, at least on fancier cars. There wasn’t a need for much else and usually the provision fitted the general need of the vehicle and its intended usage, rather than the driver.

    Since we’ve started to move into the realm of electronic control systems, computer aided design, electromagnetic clutches / component brakes, iterative maths programs that can work through millions of potential combinations to find those that are most finely balanced, etc, it’s become easier to add further epicyclic gearsets inside a normal sized (or even a shrunk-down) transmission case without compromising strength, providing much greater opportunity for multiple reductions and reversals, multiplying up at one or more stages as well as just dividing (meaning less need for special differentials, because the output ratio can be lower than 1.00:1 – cost and ease of production is improved if that part can be shared with the manuals), and most particularly the shifts can be synchronised with pinpoint precision, all the multiple parts moving at exactly the same time, or in a millisecond-perfect cascade spreading out the driveline shock a little (and in concert with a momentary dip or pulse of the throttle depending on shift direction). Even with just three the potential sophistication is much greater, but some have four or even five, plus the split planets and all.

    …oh man, this has splurged. Can i split it in 2?

  • Mark Penrice

    …yes, I can 😀

    Some of those where the ratio spread is given can go from somewhere north of 4.4:1, down through 1:1 as like 6th gear, and then out to 0.63:1 as 9th, with reasonably small gaps between each. The torque converter can be simpler, lighter, more efficient, and completely bypassed out of the circuit fairly early on as it doesn’t need to act as a shock absorber any more, even for those where there are multiple *reverse* gears (basically keeping the engine at idle throughout to limit maximum speed, but still allowing a fairly strong push from a standstill), though it still helps contribute to the total multiplication which can go up over 6.2:1 … total spread of about 7:1 without considering the converter bonus, and nearly 10:1 with it, meaning our model engine has a 14:1 speed range from the bottom of the 95% zone when operating through the converter in 1st, to the top of it when in 9th gear lockup (or, say, 6.5 to 91mph; without the TC, approx 10:1 from ~9 to ~90), and 20:1 (or 14:1 again) going from the bottom to top of the 90% range (so essentially 5 or 7mph through to 100, and able to creep at a natural 2 or 2.8mph at 800rpm with no riding of the brake). With the average ratio gap being about 1.275:1 (7 taken to the 8th root, as there are 8 shifts), or closer than 3375/2625, and equivalent to the typical 4th-5th gap of a small diesel. Given that it would be entirely acceptable for 1st-2nd and 2nd-3rd, if no others, to be a little wider (1.4 and 1.333, together 1.867 or about the same as a normal 1-2, leaves 3.75 to cover with 6 more shifts, average of 1.246; 3-4 being 1.293 leaves 2.9 with 5, average of 1.237; 4-5 as 1.26 leaves 2.3 with 4, average of 1.231… so far we’re down to 1.447, for 7th to be 1.00 then that’s 1.203 average, then the last two are 1.26 again… ok, this needs some refinement, but already that could actually work, even if it seems a little lopsided and might actually work best without a 1.00:1 gear at all — e.g, continuing to slowly drift down to about 1.22 each time, which is approximately equal to shifting from 3300 to 2700 between 8th and 9th (similar to a typical manual 4th-5th or 5th-6th gap in a reasonably close ratio box), and in this idealised system maybe increasing throttle from 67.5 to 82.5% to maintain the same output power. It’s not quite the same as sitting right on top of the ideal peak with a CVT, but it comes close enough to it (less than 5% away a good deal of the time, and less than 10% almost all of the time except at the very lowest or highest speeds and loads where that’s simply not possible with any transmission anyway, and the 9-speed still provides a reasonably close facsimile of the CVT – the biggest rev drop at full power being e.g. 5600 down to 4000 1st-2nd, some of which would get smoothed away anyhow so that the shift might start before 5400 in 1st and not complete until after 4200 in 2nd, and 3rd is anyway as far away from 1st as 2nd would be in a regular box, and 2nd is itself about where 1st normally would be; and if creeping along, where essentially you want to keep the revs as low as possible so that the throttle is opened more than the extremely inefficient minimum at least some of the time, it would shift up with a post-shift aim of 800rpm at no higher than 1120rpm, or 5mph in 1st, then 6.7, 8.6, 10.9…19.9mph, when a typical 4AT might not bother with the first shift until about 10mph and the last until more than 30, because of the large jumps that would make a very stark difference in available power and the bigger torque converter’s higher “stall” rpm) that the additional improvement wouldn’t be worth significant investment, and the differences between a stepped or stepless automatic drive would end up swamped by the actual inherent differences in efficiency between the available physical implementations. It also means you don’t get the same kind of kickdown lag as with most CVTs (aka the “rubber band effect”), and there’s still an audible reminder of rising or falling speed from the revs gradually building then falling back near-instantly, which is something CVTs either have to emulate to give the driver an idea of changing speed, indicate in a different way (eg having the rpm setpoint vs throttle position gradually increase with rising road speed, though less rapidly than with a fixed ratio, so the pedal sort of determines the rough gear range and if held in place the revs gradually build to give a cue of gathering pace, even though the gear ratio continues to lengthen and the residual acceleration once the increasing drag is subtracted gets ever less until the car reaches a stable speed in a natural fashion), or just flat out ignore and hope the driver was tone-deaf anyway and always used the speedometer as their sole source of information. It would, however, still be about as flexible – though it would still lose the slight edge (and gain it back through less internal friction and slippage) from being able to sit at a certain rpm or very narrow range of them, having a road speed vs rpm-and-loading conversion ability with a typical granularity of 1.3:1 / 1:0.77 or better does allow you to stay a lot closer to the mark than when 1.3 isn’t much wider than your smallest gap, with all the others being greater. EG climbing a long, moderately steep hill with a full car, trying to maintain 60 or so, the box may be able to choose between about 3200, 4000 or 5000rpm if there’s three gears that are 1.25x each other (having revved up already from 2500ish), and drop one further gear if they go below about 4400~4500, up to 5500~5600. +/- 1000rpm when calling up a wedge of hill climbing power at an arbitrary speed isn’t too shabby. The regular 4-speed might have to go from 2667 to 3750 in one jump, and then straight to 5250 from there, with no intermediates anywhere. On a long run, that can really eat into both your performance and your economy. And if you’re just trying to keep roughly the same output power in the face of small variations in gradient or headwind that have the road speed (and thus rpm) move up and down slightly, with 75% at 4000 being generally alright, then there’s more latitude to increase the FBW throttle opening first before going for a downshift (where the equal-power setting might otherwise be below 60%, so doubly less efficient), at least up to maybe 85-90% (more like 68-72% post downshift; wider throttles at high rpm aren’t actually *that* terrible, it’s just that they make lots of power which generally means you’re either accelerating, climbing a hill, or going fast, so the resulting mpg is low… what’s bad is high rpm and low throttle…). Similarly if the gradient slackens a little the “real” throttle can be backed off some before shifting up again, allowing the engine to maintain road speed at lower rpm without having to jack it fully open (shifting up from 75% would need nearly 94% of the maximum available power… letting it ease off more towards, again, 68% or 64% would mean a less extreme opening of 80-85% afterwards, and so leaving a little latitude to raise it again temporarily without hunting up and down between the gears, and rather less engine noise too, without a huge difference in effective economy. Then when the load at 3200 again falls off enough that looks to be about to go off the cliff below 55%, it can shift up again and go straight from 60% at 3200 to 75% at 2560, and even lower should the road level off completely, with a difference of almost 2.5x between the amount of power output on the flat and climbing the steepest grade without having to make a third gearshift but no more than a 10% difference in efficiency between any of the steady states. Whereas the 4sp may not have to shift out of top quite so early, but it’s a lower top gear to start with, the throttle will have to go almost to maximum (certainly 90%+) before the downshift triggers (leaving you around 64% in 3rd, at higher revs than would otherwise probably be needed), and again it’ll be able to wait quite a bit longer before shifting again, but only because it’s almost at the rpm the 9-speed would have reached after two shifts. And when it does, it ends up at higher rpm and lower throttle again (with a second 90 to 64% jump) than the 9-speed would have after *three*. Of course, it also takes longer to upshift, as you have to ease off enough that the same power is still available with a little in immediate reserve after quite a big drop in rpm. The CVT seems to do better than both as it can smoothly follow the line of best gearing for road speed vs power demand, right through the efficiency peak at 34hp, and hit both the upper right and lower left corners if needs be without ever burning more fuel than is needed to make a particular amount of hp at the crankshaft. Unfortunately, the transmission itself may well consume so much more that the advantage disappears vs the 9 speed, and in extreme cases even the 4-speed. The opponents aren’t going to be in the positions where the biggest gap is displayed between the stepped versions and the CVT after all, and in certain cases one or the other will be an exact match for it, or close enough that the difference is hard to measure. The statistical assumption would be that it averages about half the peak difference. Therefore if it’s potentially up to 10% different in the most pathological case (and really, I doubt it’s actually more than 5%), the CVT only needs to be inherently 5% less efficient (or indeed 2.5%) before it’s not possible to tell between them on the economy OR performance front (what consumes extra fuel by sapping power also slows acceleration and reduces top speed), and the comparison switches to other features… and whist CVT is generally a lot smoother, people often don’t like how its power delivery progresses, and the sounds it makes. It can also be more prone to wearing out and strange faults unique to itself (e.g. a slack drive belt, which I’ve experienced on a rental scooter and made the thing almost unusable on hilly stretches of road because it had a lot of trouble increasing the reduction ratio below that of somewhere between a normal 2nd and 3rd, and so max rpm for accelerating up to the speed where it would otherwise have used that position anyway fell to the take-up point of the centrifugal clutch, which is fine for the first second or two where it’s taking some energy from the flywheel anyway, but not grinding along for a minute or more at a time on a 1-in-10 hill after coming out of a tight hairpin, eventually getting up to 40km/h or whatever and having the engine finally rev up just before having to brake again – now imagine that in a vehicle that you can’t hop off and give a helping push if the slope becomes too much with yourself on board… regular autos, even multispeed ones, can at least be locked in a reduced selection of gears if something goes wrong with one of the subcomponents), and a lot of mechanics refuse to touch them because adjustments and repair seem like a black art compared to the more familiar systems…

    IE a well sorted 9-speed, and particularly 10-speed, should be able to emulate a CVT quite closely in terms of performance and economy even if both of them are as inherently efficient as each other. As a fully geared transmission, even one with 3+ sequentially coupled gearsets increasing, reducing, or reversing the input rotation is generally at least a few percent less wasteful internally vs a CVT built with current technology, they are at worst on a level playing field, before we consider the other advantages and disadvantages of each. Essentially, we only really tried CVT because we didn’t have the engineering (and possibly mathematical / designing) sophistication to make a practical multi-speed automatic (or even manual, without resorting to heavy truck tactics) for private vehicle use, nor was there much motivation to attempt it for the particular applications where CVT seemed a natural fit at first. However the geared models can now easily exceed the belt-and-cone jobs in terms of ratio range, and the gaps between the middle and upper gears where keeping a relatively constant, midrange rpm is most important are narrow enough that the energy converted into motion by the engine from each gallon of fuel only differs by a few percent, instead of potentially 10, 15% or more, and the closer steps in how much throttle is then needed to keep the same amount of power going to the wheels if their speed or the engine rpm tips up or down just enough to trigger a shift are also smaller, allowing quite a bit of overlap between the useful, efficient range of each gear, reducing the reliance on preset shift points and instead using a bit of fuzzy logic that takes into account previous driving conditions and likely future ones, etc. Thus we instead have to consider them on other merits, and in terms of weight, engine note / audio cues, responsiveness, power handling etc, the CVT ends up losing.

    (in the lower gears, below whatever is equivalent to a typical 5MT 3rd, or “2.5” of a typical 4AT, speed and load generally varies so much that no automatic can easily track it and keep the rpm steady without wavering up and down continually, and power demand may rapidly go back and forth between strong acceleration and very low or nothing at all, with normal steady-speed power demand being low enough that it may as well be covered by the electric drive of a hybrid system anyway, so close-set gears aren’t so much of a requirement, and the 1.4:1 & 1.33:1 estimated above may well seem needlessly close if the system ends up repeatedly bouncing between 1st, 2nd and 3rd in quick succession when a regular auto staying in 1st would likely have not been noticeably more economic)

    (I also mention 10sp as well as just 9sp, as having done the maths on my model setup here, it tops out somewhere around 24.5mph/1000rpm, and even considering the engine’s low output, it could stand to go a little higher, either through a taller differential or an additional internal gear both expanding the total range and making the gaps slightly smaller – after all the estimated cruise speed for the most efficient combination of throttle and rpm (again, not most efficient *speed*, just least wasteful conversion of fuel energy into energy used solely to push the vehicle along rather than being turned into noise and waste heat) is indicative of something closer to 26mph/1000. With an extra gear shoved in we can easily increase the top gear by that 6-7%, maybe drop the bottom one a little, and tighten up all the others, and it’s coming close to a fairly optimal setup where any other changes get into the realm of diminishing returns and we should instead look at making it lighter/smaller, simpler, more robust and reliable, inherently more efficient in terms of its own drag on the engine/wheels, cheaper to manufacture, etc.
    That additional bit of range is also even more important for more modern, more powerful engines (even those which are smaller, capacity-wise) vs this old crock I’ve used for an example (chosen because its maximum capabilities are fairly modest and thus we can realistically hold it at a large fraction of full power without ending up at 100+ mph). Their power is useful for good acceleration and towing/hill climbing without thrashing the bolts off it all the time, but also means that our less demanding cruising – where the power input is likely somewhere in the 10 to 40hp range depending on speed (approx 40 to 85mph) – is a much lower fraction of the overall potential, even at lower rpm (case in point: my own car has a 1.2L engine which can happily stick out more than 60hp at 4000rpm, and the power keeps on building up to 6000…). Thus this means smaller throttle openings with the same rpm, and ultimately lower efficiency in most cases other than where the terrain or cargo becomes challenging. It might not even be possible to keep it in the most economic range at lower speeds, but what we can do is push the revs/load combination as close towards optimal as possible and hope that it’s enough, given that it’s using a modest amount of power in the first place. At motorway speed it should still be possible to get within the 90% or even 95% envelope, if the engine isn’t entirely excessive, or has some onboard tech like cylinder deactivation, variable compression, parallel hybridisation where the battery is recharged off the engine in periods of sustained low load in order to provide extra power without extra high revs later on, or for full-EV silent running in city centres, making the best possible use of the potential generation capacity.
    This of course means raising the top gear ratio… for a really OTT engine this might mean pulling it down to 2000 or less even at 80+mph … which some older 6-speed boxes fitted to unashamed muscle cars like the Viper and Lotus Carlton offered, to try and pull some kind of sensible between-pumps range out of them when not being used to turn rubber into vapour and money into rictus grins. If we want to increase the top gear towards that from where it is right now, that’s a fair jump; if we essentially turn 1st gear into 2nd, multiplying everything by an otherwise fairly appreciable 1.4x, even that only gets about 34mph/1000 not 40… though it’s not too far off, stacking up an 68mph cruise at 2000rpm, being easily into three figures before passing 3000, and therefore potentially capable of 200+… And, well, normally losing 1st gear through a much longer differential ratio isn’t the best of ideas… you need a stronger clutch/TC and more starting torque from the engine (not an issue here of course), it’ll idle/creep faster so traffic and parking needs much more brake riding, and certainly at lower speeds the between-gear gaps are effectively a bit wider too… not settling into the reasonably close sequence until the car is travelling 40% faster. So things may generally get less efficient even with higher gearing.
    So here we can insert that extra gear, but put it all at the top end, and make the top 2 or 3 shifts slightly wider, and the 2 or 3 below them a little tighter, readjusting where the larger and smaller gaps are in relation to how each gear will likely be used and the rather different engine characteristics. The increase in range might only be about 1.2 or so, but that covers half of the increase otherwise produced by lengthening the differential ratio… if that is instead itself stretched by 1.2 not 1.4, the overall effect is 1.44x … giving us just over 35mph/1000, which is quite long in most people’s books. It’s enough to essentially *literally* idle up to about 30mph from a standstill in lethargic city situations, cruise with just a sniff of power and hardly any engine sound other than a quite burble (the revs may be low but the throttle might still open fairly wide) whilst off the motorway (never really exceeding 1800rpm unless there’s a sharp rise in the road or you want to get past someone), and able to cruise at 70 with a modest and fairly efficient 2000rpm, or just kiss 88mph at 2500, which isn’t exactly screaming for a modern engine… and if you avoid flooring it too hard, it could run all the way up to 140mph without exceeding 4000, thus the fuel efficiency not counting external drag forces barely wavers from a medium jogging pace all the way up to supersaloon territory. And, well, the shift points are still moved up somewhat, so the effective shift gap at any particular speed is about 20% wider, but they’re narrow enough, and the main run of close gears still starts early enough (especially after rearrangement) that it doesn’t really make any practical difference. And in fact with the necessarily larger torque converter it might be OK to pull top just a mph or two per thousand higher. Plus of course it can just sit +/- about 10 to 15% of max power rpm (certainly never more than +/- 20%, once it’s actually launched and revved to that zone, at all of about 20mph) all the way from the point where the traction control stops cutting in to keep the tyres gripping, up to wherever Vmax is encountered, if you need to go fast… or whatever point is best between the various extremes. Moving the revs (and/or throttle setting) up and down by 20 or 25% at a time is a pretty minor change by most measures, unless you’re already up into what would usually be considered the overdrive range, and making it granualar enough that it might change by 0.1% at a time only offers a relatively minor improvement over changing by 30 to 50%.

    All those additional ratios that you can choose from, both above and below the normal range, and interspersed within it, simply allow you to use the engine’s power a lot better for both speed and range, without having to essentially feed it through the not-much-more-efficient mechanical equivalent of a series hybrid starter-generator and its remote motors. And the changes required to produce them all are actually fairly minor. It’s something we could have done basically a century ago, if we’d been bothered, and could have made things to sufficient tolerance (or simply just *thought* about it enough… I’m sure a simpler implementation of electrical control could have been produced just with centrifugal governors, valve tubes, comparators etc, and it wouldn’t be TOO much worse than a hydraulic or transistor-and-microchip controlled one… or even a combination hydraulic and electric, using each for the part where they’re most suited…) … in some cases it’s nothing more than reprogramming of the control system and not even adding any extra parts, just using the existing ones in an alternative way.

    Funnily enough it’s rather like how bike gears went from 2 to 5 speed cylinder type (epicyclic, btw!), then switched to the more easily expandable, cusomisable and maintainable dereilleur, and grew from 1×5 through 2×6 to the modern 3×10 types … the overall gear range of a 30-speed isn’t 6x that of a 5-speed, nor is a 5-speed cylinder 2.5x that of a 2-speed. They’re more like manually operable versions of a multispeed automatic. There’s various overlapping ranges, some duplicate gears, some which, mechanically speaking, it’s not a good idea to try, etc. When you bring it down to the useful combinations, then maybe there’s six rear gears that can be sensibly used with each chainring, so that’s 18, though some are more or less duplicates of each other, and that can be proven by changing up a ring whilst dropping 2 or 3 rear gears, only to find that the pedal vs wheel speed is about the same. So maybe overall it’s more like 10 to 12 really useful ones, though the duplicates are handy to avoid having to muck about with the rings too much. And vs the older ones with just five gears, your lowest and highest ratios are much extended, making both insanely steep hills easier and nice long gentle descents with a slight tailwind a lot faster. On the way from one to the other, the change in ratio, and in how hard you have to push and at what cadence, is kept to a nicely tolerable minimum. Not eliminated, but small enough that it’s barely noticeable, and may as well be part of a sequential series… in fact, there are now cylinder gears with 8, 10, 12, 14 distinct speeds, set up inside much like a 10-speed automatic, which are operated using a sequential click-shifter… with about the same range and step width as that 30-speed dereilleur.
    (Even the 10-speed auto has plenty more potential combinations available, I’d wager; if you can get an average of 2.5 useful forwards ratios out of each gearset – or, say, alternate between 2 and 3 – then that’s 2 for one set, 6 for two, 12 for three (already there’s two spare or duplicated), 36 for four… and even if it’s only 2 for each, you’re still up to 16. There just comes a point where the extra options aren’t actually of any additional use, no matter how carefully you engineer each gearset to given entirely unique combinations that aren’t equal to any other single speed produced by a neighbour – or two, or three of them multiplied together. Particularly as so many of them will actually be various flavours of reverse, anyway.)

    So there you are :p