The CVT: why it feels wrong, and whether that means anything is wrong

The corpus describes transmission slip as engine speed rising without a matching increase in road speed, and says to treat it as a fault.

On a CVT that is the normal, designed behaviour, and it is the single most reported non-fault in motoring. A driver used to anything else presses the accelerator, the engine goes to 4,000 rpm and stays there while the car gathers speed underneath it, and every instinct says the clutch is slipping.

Nothing is slipping. So the first job of this article is to explain why it does that, and the second is to say how to tell it apart from the times when something genuinely is wrong — because a CVT does slip, and when it does the repair is usually the whole unit.

What is actually in there

Two pulleys and a belt, and that is nearly all.

Each pulley is a pair of cones facing each other on a shaft, and they can slide closer together or further apart. The belt runs in the V between them. Push the cones together and the belt is forced outward to a larger running diameter; let them apart and it sinks to a smaller one. Do that to both pulleys at once, in opposite directions, and the ratio between input and output changes smoothly, through every value in between, with nothing to engage and nothing to release.

That is the whole idea, and it is why there is no shift to feel. The original describes this part well.

The belt is the interesting bit, and it works backwards from what you would guess. The original says it is "a special belt of metal or plastic". Plastic belongs to scooters, not cars, and the metal one is not really a belt at all: it is a stack of several hundred trapezoidal steel elements threaded onto two bundles of thin steel bands. Clamped between the cones under enormous pressure — measured in tonnes — the elements are pushed along their loop by the driving pulley rather than pulled. It transmits torque by compression, like shoving a column of washers along a rod. Hence its name, the push-belt, and hence why it can be made of steel at all.

Two corrections to the history

The original opens by saying the first CVT was patented in 1986, and then, within a few hundred words, credits a Leonardo da Vinci sketch of 1490 and says the Dutch maker DAF was fitting them to cars in the late 1950s. Those cannot all be true, and 1986 is almost certainly 1886 with a digit lost — which is the plausible date, from the earliest years of the motor car.

The rest of its history is sound and worth keeping: Hub van Doorne of DAF designed the push-belt that is still the basis of most car CVTs, and most manufacturers still buy their belts from the company he founded.

One phrase to soften. The original says a CVT gives infinite ratios. It gives a continuous range between a lowest and a highest — which is the useful property, and it is not unlimited. The spread between those two limits is a real specification and it is what decides whether the car can both pull away steeply and cruise quietly.

Why it feels wrong

A conventional gearbox couples engine speed to road speed in fixed steps, so acceleration comes with a rising engine note that you learn to read. A CVT is free to break that link, and the most efficient thing it can do is hold the engine at the speed where it makes the power you asked for and let road speed catch up.

So you get the two complaints, and neither is a fault:

• The rubber-band effect. Revs jump and then sit still while the car accelerates. It is the ratio changing continuously underneath a constant engine speed.
• The drone. Constant revs plus constant load is a constant note, and a constant note is more tiring than a changing one.

Manufacturers agree it feels wrong, which is why most modern CVTs now fake it: the control software steps the ratio deliberately so the engine note rises and falls as if there were gears, and some add paddle-operated pretend ratios. That is a comfort feature and it costs a little efficiency, which tells you how strong the objection is.

The test that separates feel from fault is whether the relationship is stable. Hold a steady throttle on a level road. Engine speed and road speed should each be steady, and the ratio between them should change smoothly and predictably when you change the throttle. That is a CVT working.

When it genuinely is wrong

The original stops before this and it is the part an owner needs.

The reason slip matters so much more here than in an ordinary automatic is that a CVT has one friction interface and it is also the thing that sets the ratio. A slipping clutch pack in a conventional box wears a friction plate you can replace. A slipping belt polishes and then scores the hardened cone faces it runs against, and once the pulleys are scored the belt cannot grip them at any pressure. The repair at that point is the transmission, which is why the flare in the first row is worth acting on the day you notice it.

The fluid, and it is not negotiable

Every article in this group ends up here and this one most of all.

CVT fluid is not automatic transmission fluid, and it is not interchangeable between makers. Its whole job is opposite to normal: a conventional automatic's fluid must let clutch plates slide smoothly into engagement, while a CVT's must maximise the friction between a steel belt and a steel pulley while still lubricating the bearings and carrying the hydraulic pressure that clamps them. Filling a CVT with ordinary ATF gives exactly the belt slip described above, quickly, and the damage is the expensive kind.

So: use the fluid the manufacturer specifies, by name, and change it on the interval in the book — which for most CVTs is a real interval and a short one, often somewhere around 60,000 km, and sometimes half that for city or towing use. See transmission fluid for why fluids are specified rather than graded, and the AL4 for what happens when a manufacturer tells everybody the fluid never needs changing.

Three things not to do

Do not tow it with the driven wheels on the ground. The pump that provides the clamping pressure is driven by the engine, so with the engine off there is no pressure holding the belt against the pulleys while the wheels are turning them. Check the handbook; most say a flatbed, and the ones that allow towing set a strict speed and distance limit.

Do not coast in neutral. Same reason, in miniature, and it saves nothing.

Do not treat it like a torque converter automatic on a launch. CVTs are built to a torque limit — it is why they appear mostly in smaller and mid-sized cars — and repeated hard launches with the belt at maximum clamping load are exactly the duty cycle it is least suited to.

Whether to buy one

The honest summary the original does not give. A CVT in a small car, driven normally, with its fluid changed, is efficient, smooth and long-lived; the engine spends its life at the revs it is most economical at, which is a real advantage and the reason they exist.

The risks are that the fluid is easy to neglect, the failure mode is unforgiving rather than gradual, and the repair is a unit rather than a part. So the question to ask about a used one is exactly the question to ask about an AL4: when was the fluid last changed, and with what.

Video guides

Video, for the parts of this that are easier watched than read.