Why a Twist Drill Cannot Enlarge a Hole

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

The short answer is that the best way to enlarge a hole with a cordless drill is to not use a twist drill for it, and the reason is more interesting than the advice. A twist drill is not guided by its point. It is guided by two narrow strips behind the cutting edges, and in this particular job those strips never touch the work at all.

Once you see where a drill's guidance actually comes from, two surprises fall out of the same piece of arithmetic: why enlarging a hole produces a lobed, oversized mess, and why the same bit misbehaves in thin sheet for exactly the same reason.

The margins are the bearing, and they arrive late

Look at the business end of a twist drill. There is a cone, there are two cutting lips, and behind the lips, running up the length of each flute, there is a narrow land at the full diameter of the bit. Those lands are called margins, and they are the only part of a twist drill that is a bearing surface. Once they are inside the hole, the bit is a shaft in a journal and it goes where the hole goes.

The point does something different and smaller: it stops the bit skating at the instant of contact. After that it contributes almost nothing to position, because a cone sitting in a cone it cut itself cannot resist being pushed sideways.

Now the timing. The margins start where the lips end, which is at the top of the cone, so they reach the surface only after the whole cone is below it. Our page on why a drilled hole has two depths works out that distance exactly, and this page uses its numbers rather than re-deriving them so the two agree: the cone is half the diameter divided by the tangent of half the included angle, which is 0.300 of the diameter at the ordinary 118 degree grind and 0.207 at 135 degrees.

So the sentence worth carrying is this. A twist drill is unguided from first contact until it is 0.300 of its own diameter deep. On a 20 millimetre bit that is 6.0 millimetres of unguided cutting, every single time, with nothing but the operator's hands deciding where the axis is.

The first consequence: thin stock never gets guided at all

Turn that round. The bit only becomes guided in material thicker than its own point depth, so the largest diameter that gets guided at all in a sheet of thickness t is t divided by 0.300 at the ordinary grind.

In 1.5 millimetre sheet that is 5.0 millimetres. In 3 millimetres it is 10. In 6 millimetres it is 20. Go above that and the hole is finished before the margins arrive, which is precisely the familiar experience of a large twist drill in sheet metal: it grabs, it walks, it leaves a triangular opening, and no amount of care fixes it, because the feature that would have fixed it is still in the air.

And this is where a published point angle turns into a real number. A 19 piece cobalt grade set of stepped point drill bits with 135 degree split tips publishes exactly the grind this arithmetic needs. At 0.207 instead of 0.300, the largest guided diameter in any given sheet rises by 0.300 over 0.207, which is 1.449, about 45 percent. In the same 1.5 millimetre sheet it goes from 5.0 to 7.25 millimetres. That is not marketing. It is the only mechanical benefit of a point angle that can be computed from a listing, and it follows from one published number.

The split grind does a second thing worth separating from the first. On an ordinary point the web at the centre is not a cutting edge at all, because the cutting speed at the axis is exactly zero and no rotation rate changes that; it indents rather than cuts, which is why a centre punch helps and why that region takes a disproportionate share of the thrust. Grinding the web back so the edges reach the axis is what the record means by "self-centers and penetrates with less pressure", and it is true.

Two honest notes on that record. Its own drawbacks field says the listing "mixes M35 and M2 grade references", which is a genuine contradiction between a title and a description and the record is right to flag it, so treat the steel grade as maker stated. And its "four times sharper" claim is a comparative with no quantity attached, which its own cons also say. Neither affects the point angle. Full record in our review of the 19 piece stepped point set, and the heat side of drilling metal, which is a separate failure mode entirely, is in why your drill bits go blunt in metal.

The second consequence: enlarging a hole removes both of the drill's locating features at once

Here is the question as asked. A drill of diameter D opening an existing hole of diameter d engages each lip only over the band from d over 2 out to D over 2. The rest of the lip, the inner part, is passing over a hole. So the engaged fraction of each lip is one minus d over D.

Open an 18 millimetre hole to 20 and that fraction is 0.10. Ten percent of each cutting edge is doing anything at all.

The mean cutting radius moves outward at the same time, from D over 4 to D plus d over 4, which is 5.0 millimetres to 9.5, a factor of 1.90. Torque goes as the engaged length times the mean radius, so it falls to 0.10 times 1.90, which is 19 percent of drilling from solid, and thrust, which does not have the radius in it, falls to 10 percent.

Every one of those numbers says the job is easy, and the job is not easy. The reason is the term nobody computes.

A production twist drill has a lip height error: one lip stands slightly proud of the other, because they were ground separately. Call it five hundredths of a millimetre, which is ordinary. From solid, that error is 0.5 percent of a 10 millimetre engaged lip, and the resulting side force is trivial. Opening 18 to 20, it is 5 percent of a 1 millimetre engaged lip.

The same grinding error is amplified by exactly D divided by D minus d, which here is a factor of ten. And it is amplified at the one moment when both of the things that would resist the resulting side force are absent: the chisel edge at the centre is hanging in air over the old hole, and the margins are above the surface and, because the hole is already there, will never arrive. A two edged tool free to move sideways in a hole it is cutting produces a lobed hole, which is why the result is reliably oversize and out of round rather than merely rough.

So the whole difficulty of this job is in one ratio, and the fix is not technique. It is bringing your own bearing surface.

Three tools, three different bearings, each one stated in its own record

A leading step that rides the old hole. This is what the stepped point set above is for, and its record says so directly: "Enlarging existing holes is a stated use as well as fresh drilling." A step small enough to enter the existing hole turns the hole you already have into the bushing for the cut you are about to make, which restores the margin function at the one moment a plain twist drill loses it. Its record publishes no step diameter and no land length, so that mechanism is described here from geometry rather than from the listing, and that is the honest limit on it.

A side cutting flank, with no point at all. A diamond coated finger bit of 1 and 3 eighths inches for enlarging tile holes publishes 35 millimetres of diameter, a 50 millimetre work length, 65 millimetres overall, a 5/8-11 thread for an angle grinder with an adapter for a 3/8 inch hex drill, wet or dry use, and 6,000 to 11,000 revolutions a minute. Its own drawbacks field is the clearest statement of role in this catalogue: "Only for predrilled holes", and the maker "recommends using it only to enlarge, shape or round predrilled holes, not to start fresh ones."

That tool has no lips and no point, so none of the arithmetic above applies to it. It is a mill, and a mill in a hole is located by its own flank bearing against the hole wall. Divide its two published lengths: 50 over 35 is 1.43 diameters of bearing surface, against a twist drill's zero in the identical job. That is the entire difference, and it is a published ratio. More in our review of the diamond finger bit, with the wider job in how to cut a large faucet hole in tile or countertop.

A pilot that hands over to the kerf. A 2 and 1 eighth inch diamond tile hole saw with a removable pilot, in a two pack publishes 55 millimetres of diameter, a 3/8 inch diamond height, a 1 and 3 eighths inch cutting depth, 2 and a half inches overall, and the sequence that makes it work: start on the guiding pilot, then "a hex wrench lets you take the pilot out and finish with the diamond rim alone."

Read that as a bearing problem and it is elegant. The pilot provides guidance for exactly as long as there is no kerf; once a groove exists, the groove is the bushing and the pilot becomes a liability in the middle of the cut. Five millimetres into a 55 millimetre saw that is 0.091 diameters of bearing, which is modest and is emphatically not zero, and it is the reason the pilot is removable rather than fixed. Its own honest limit is thermal rather than geometric: "Water cooling is needed to protect the edge." Details in our review of the diamond tile hole saw, and the family view is how to choose a diamond core drill bit for tile and stone.

What this says about two other pages on this site

How to use center drill bits for accurate metal holes says a centre drill "makes a precise starting point, not a finished hole" and that its cone "keeps larger bits from wandering". Both correct, and the arithmetic above says what that is worth and where it stops: a conical seat fixes the start, and the following drill is then unguided for 0.300 of its diameter anyway, because a centre drill cannot bring a margin with it. A deep, narrow seat buys you the first instant. The rest is still on your hands.

Why an impact driver wobbles by design reaches the same conclusion from the other end of the tool and says the right thing in four words: "let the work locate the bit rather than the tool." That is exactly what the three products above do, each with a different surface.

And on the one case where none of this is needed: in wood, a brad point or a Forstner bit has a rim that cuts the circumference before the centre removes anything, which means its bearing is the outside of its own kerf from the first millimetre. The whole margin problem is a steel and masonry problem because the twist drill's geometry is a steel geometry.

The honest limit, and three things not computed

What is exact: the engaged fraction one minus d over D, the mean radius shift from D over 4 to D plus d over 4, their product at 0.190, the amplification factor D over D minus d, and the 1.449 ratio between the two point angles. Those are algebra and they hold for any pair of diameters.

What is reused rather than derived, deliberately: the 0.300 and 0.207 point depth multipliers come from our own point depth page, used unchanged so the two pages cannot disagree. That page also flags the real softness underneath them, and it still applies here: the 118 degree grind is "the one figure no bit listing we carry states anywhere", so the entire first column of the thickness table rests on a convention rather than a specification.

What is representative rather than published. The 0.05 millimetre lip height error is a typical production figure and the amplification result scales directly with it, though the factor of ten does not depend on it at all. The 20 and 18 millimetre example diameters are illustrative. And the torque estimate assumes a uniform chip load along the engaged lip, which is not true: the load falls off near the centre, so drilling from solid is slightly less torque-hungry than the comparison implies, which makes the 19 percent a mild underestimate and the instability argument unaffected.

Three things this page deliberately does not compute. It does not compute the number of lobes or the roundness error of the resulting hole, because that needs the bit's radial stiffness and the workpiece's clamping, and the second of those is your bench rather than a published quantity. It does not compute a cutting force in newtons, because no bit listing in this catalogue publishes a feed rate and a force without a feed is a guess wearing units. And it does not recommend a technique for making a twist drill behave in this job, because the honest result of the arithmetic is that the missing thing is a bearing surface, and no amount of hand pressure is one.

Three numbers to carry. A twist drill is unguided until it is 0.300 of its diameter deep, so in 1.5 millimetre sheet nothing above 5 millimetres is ever guided. Opening a hole leaves one minus d over D of each lip cutting, ten percent on an 18 to 20 millimetre step, which drops the torque to 19 percent and makes the job feel easy. And the same step amplifies the bit's own grinding error by D over D minus d, a factor of ten, at the exact moment both of its locating features are gone.