Why an Impact Driver Wobbles by Design

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Because the quarter inch hex interface is built around a clearance, and the clearance is not a tolerance failure, it is the thing that makes the tool work. A new driver has it, every driver has it, and it produces two separate motions that people lump together as wobble.

One is a tilt, about 0.43 degrees, which shows at the tip as a circle strictly proportional to how far the bit sticks out: 0.45 mm on a standard bit, 1.95 mm on a long one. The other is a rotation, 7 to 12 degrees of free play before the flats even touch, and that one is the interesting number, because it only falls as the square root of the fit. Tighten the socket by a factor of three and you remove 42 percent of it. There is no machining your way out.

The clearance has to be there

An impact driver is a rotary hammer in miniature. A spinning mass winds a spring, lets go, and strikes an anvil, three to four thousand times a minute. The anvil is the bit holder. For that to work the bit has to be held by something that can be loaded and unloaded fifty to sixty seven times a second without being a press fit, and that something is a hex socket with a spring loaded ball sitting in a groove.

Both halves of that need room. The ball needs radial travel to ride into and out of the groove, and the hex needs enough clearance that a bit pushes in with a thumb. One of our own records publishes the dimension this all hangs on: a pack of magnetic bit holder rings describes sliding onto "a standard 6.35mm quarter inch hex shank bit", and 6.35 mm is the nominal across flats figure. Shanks are made at or slightly under it and sockets slightly over, which leaves a diametral clearance in the region of a tenth to a quarter of a millimetre.

That is the whole input. Everything below is geometry on top of 6.35 mm and 0.15 mm.

Motion one: the tilt, which is linear in bit length

A bit sitting in a socket that is 0.15 mm wider than it can lie over to one side. The pivot is the deep end of the part of the socket that actually bears on the bit, which after the entry chamfer and the detent groove is about 20 mm. So the angle is 0.15 divided by 20, which is 0.0075 radians, or 0.43 degrees.

That sounds like nothing and it is nothing, until you multiply it by a lever. The circle the tip traces is twice the protrusion times the angle, so:

A standard 50 mm bit stands about 30 mm out of the socket and traces 0.45 mm. A 150 mm installer bit stands 130 mm out and traces 1.95 mm, which is wobble you can see from across the room, and which is also exactly 4.3 times the short bit for no reason other than its length.

Which puts a number on reach. A 10 piece diamond tip S2 screwdriver bit set is sold partly on being longer than usual, its record noting each bit runs "2-9/16 inches long for deeper reach than standard 2-inch bits". That is 65 mm instead of 50, and it is a genuine advantage in a recess. The unstated half is that 15 mm of extra reach is 50 percent more tip circle, from 0.45 mm to 0.675. Not a defect, and not a reason to avoid it. Just a trade nobody writes down, and the right response is to own both lengths and use the short one by default.

Motion two: the rotation, and why it cannot be fixed

The same clearance lets the socket turn relative to the bit before anything touches, and this is the motion people feel when they wiggle a bit by hand. The geometry is exact. A hexagon of across flats dimension A, sitting in a socket of A plus c, can rotate by the inverse cosine of A over A plus c, because a hexagon's width in a given direction grows as one over the cosine of its rotation.

Work that through. At 0.05 mm of clearance the free rotation is 7.17 degrees. At 0.10 mm it is 10.10. At 0.15 mm it is 12.33. At 0.25 mm it is 15.82.

Look at the shape of those numbers. Clearance went up fivefold and the angle barely doubled, because for small clearances the inverse cosine is very close to the square root of twice the clearance over the across flats dimension. It is a square root law. Which means a maker who machines sockets three times tighter, at three times the difficulty, delivers a bit that still has 7 degrees of rotational play instead of 12. Forty two percent of the problem, for an enormous change in manufacturing.

That is the real answer to the question. The play is not proportional to the quality of the fit, it is proportional to its square root, so every quarter inch hex tool in the world feels much the same and a better driver does not help. The interface gives that clearance away deliberately, because what it buys back is a bit you can change with one hand.

And then the tool slams the bit across that gap. At 3,000 to 4,000 impacts a minute, the anvil accelerates through 7 to 12 degrees of nothing before it meets the flat, fifty to sixty seven times every second. That is the hammering that rounds the corners off a hex shank, and it is why impact rated bits are made from tougher steel with a deliberately weakened torsion section: not because the torque is higher, but because the loading arrives as a rapid series of impacts into a joint with free travel in it.

Where our own page gets the diagnosis wrong

We have a page on the accessory most affected by this, and one sentence in it needs correcting. Our guide on using a drill chuck adapter with an impact driver says, of the shank: "A loose or wrong shank invites wobble and lost bits."

Both halves of that are true and the implication is not. It reads as though wobble is a symptom of something being loose or wrong, which invites the reader to go looking for a fault, tighten something, or replace a worn bit. There is nothing to find. A perfectly correct, fully seated, brand new quarter inch hex shank in a brand new anvil tilts 0.43 degrees and rotates 12 degrees, because both numbers live in the standard rather than in the fit. The right test is not whether the shank is loose, it is whether the wobble you are seeing is bigger than the numbers above, and for that you need the numbers, which is what that page was missing.

Stacking, which is where it stops being academic

Each thing you add between the anvil and the work adds a stage, and the stages add up rather than averaging out.

A quarter inch hex shank drill chuck adapter set is the clearest case, and its own record is admirably straight about it: "An adapter on an impact driver will not match a true drill for runout or control." Correct, and here is the figure it does not give. The adapter's own shank tilts by the same 0.0075 radians, but the tip of a bit held in its jaws now sits around 95 mm from the anvil face rather than 30, so the angular contribution alone is 1.43 mm of tip circle. On top of that a small three jaw chuck has its own centring error, typically one to three tenths of a millimetre, which adds directly. Call the total about 1.6 mm.

For a hole, that 1.6 mm is a ceiling on how oversize the mouth can start. A 5 mm bit can open an entry up to 6.6 mm across before the walls of the hole begin to hold the bit straight, which is a third oversize at the surface and is why these holes look ragged at the top and fine further down. The adapter's stated clamping range tops out around 8.0 mm, and at the small end of that range the runout is a larger fraction of the hole than the hole is.

Push the idea further and a flexible 360 degree socket extension bar set abandons the constraint altogether. It is spring steel with a ball detent, made to bend, and bending is the entire function: it exists to reach bolts that are not in line with the tool. Angular error is not a tolerance on a tool like that, it is the feature. Its own drawbacks say the honest thing, that it is "A reach tool with sensible torque limits" and "Not for very high-torque or seized bolts", and the same sentence should be read as applying to accuracy. Use it where access is the problem and nothing else is, and never for a hole.

The accessory that adds nothing, which is not the one you would guess

Worth noting because it inverts the expectation. The magnetic rings above, and the five pack of magnetic bit holder rings in particular, look like the token accessory in this category and they are the only accessory here that adds zero runout. They slide over the outside of a bit that is already in the socket. They are not in the load path, they do not add a hex joint, and they do not lengthen the protrusion. Their own records are unusually careful about what they are, saying plainly: "A magnetizing ring, not a locking bit holder."

So the hierarchy is not about how much a thing is sold for. It is about how many hex joints you have put in series and how far from the anvil the cutting edge has ended up. A ring adds neither. A longer bit adds reach. An adapter adds a joint and a chuck. A flexible shaft adds an unbounded angle.

The honest limit

Two numbers do all the work here and neither is published anywhere on this site. The 0.15 mm of diametral clearance is the middle of the general tolerance band for quarter inch hex shanks and sockets, and the 20 mm of effective bearing is a typical one inch socket depth less its chamfer and detent groove. Not one impact driver, bit, holder or adapter listing we carry states a socket tolerance, an engagement depth, or a runout figure of any kind.

That matters less than it might, because of how the two laws behave. Tip circle is linear in both, so if your socket is actually at 0.08 mm, halve every figure in the middle panel and the ranking is unchanged. Rotational backlash is a square root, so it barely moves at all: even a very good 0.05 mm fit still gives 7.17 degrees. The conclusion survives any plausible value, which is the only reason it is worth stating without a measurement.

What to do, and none of it is a purchase

First, stop treating it as a fault. Our impact driver coverage already answers the practical question correctly in its own FAQ, saying an impact driver "can [drill] with hex shank drill bits, but it is not ideal for precise holes". That is exactly right, and now it has a number attached: somewhere between half a millimetre and 1.6 at the cutting edge, depending on what you have stacked.

Second, notice that the wobble is irrelevant for the job the tool is actually for. When you drive a screw, the screw head locates the bit. The recess is a socket too, and once the bit is engaged the fastener is holding the tip concentric whatever the anvil is doing. Wobble does nothing to you while driving, which is why nobody notices it until they try to drill.

Third, use the shortest bit that reaches. The tilt is linear in protrusion, so swapping a 150 mm bit for a 50 mm one takes 77 percent of the tip circle away for no hardware at all. It is the single largest free improvement available here.

And fourth, let the work locate the bit rather than the tool. A centre punch mark gives the tip somewhere to sit for the first revolution, which is the one where a wandering bit decides where the hole goes. Better still, the technique from our page on why a drilled hole has two depths: clamp a block of scrap on top and drill through it. A hole in a rigid block is a bearing with nearly no clearance at all, which is a better constraint than any chuck on this page, and the block is free.