Radial Chip Thinning in Milling, Explained

By Bradley Taylor · August 2026

Here is a thing that catches almost everyone at some point. You take a light finishing pass, maybe a ten percent stepover, at the same chip load that worked fine when you were roughing. The cut sounds wrong, the tool wears fast, and the light pass that was supposed to baby the endmill ends up killing it. The tool did not fail because the cut was too heavy. It failed because the cut was too light, and the reason is radial chip thinning.

What the programmed chip load really is

When you program a feed, what you are really choosing is feed per tooth, the distance the table moves between one tooth cutting and the next tooth cutting. The tooling catalog gives you a recommended chip load because the cutting edge is designed to work at a certain chip thickness. Thick enough that the edge actually bites into the material and shears it, thin enough that the edge does not overload. But feed per tooth is only the actual chip thickness under one specific condition, and that condition is when the radial width of cut is at least half the cutter diameter. Slot the tool or bury it halfway and the chip really does reach the thickness you programmed. Go lighter than half the diameter and it never does.

The geometry in plain words

Picture one tooth swinging around the centerline of the tool. The chip it takes is not the same thickness everywhere along the arc. The chip is fattest at the one spot where the tooth sits directly in line with the centerline along the feed direction, and it tapers to nothing at the point where the tooth crosses the machined surface. A detail worth getting right, because most explanations fumble it: at that fattest spot the tooth's cutting motion is actually running square across the feed direction, not along it. Think of slotting. The fattest chip is at the far end of the slot, dead ahead of the spindle, where the edge sweeps sideways across the line of travel.

Which side of the arc your tooth uses is what climb versus conventional means here. In a climb cut the tooth enters the material at the thick end and the chip tapers down to zero as it exits at the machined surface. In a conventional cut it is the reverse, entering at zero and thickening to the exit. Either way, as long as the stepover is half the diameter or less, the chip changes thickness in one direction only through the cut. The tooth never passes the fattest point while in the material unless the stepover goes past half the diameter, and only a full slot uses the whole rise and fall.

Now shrink the stepover. With a light radial cut the tooth only clips the edge of the material, using just the skinny end of the arc near the machined surface. It never gets anywhere near the point where the chip would be fattest. You programmed a chip load of 0.003 but the steel never saw a chip anywhere near 0.003 thick. The math is one line. Maximum chip thickness equals feed per tooth times two times the square root of the stepover times diameter minus stepover squared, all divided by the diameter. As an equation, h = fz × 2√(ae·D − ae²) / D, and the compensation factor is just the inverse of that ratio. At a ten percent stepover the actual maximum chip thickness is only about 60 percent of the programmed feed per tooth. At five percent it is down around 44 percent. The lighter you go, the bigger the lie between what you programmed and what the edge experiences. One more honest footnote: this is the story for a square shouldered endmill cutting on its side. The number the formula gives is the peak thickness, and the average chip keeps shrinking with stepover even above the halfway point, which is why power and torque predictions still move up there.

Why a thin chip is a problem

Every cutting edge has a real radius on it, even a sharp one. For the edge to cut, the chip has to be thick enough to climb up over that radius and shear. When the chip gets thinner than the edge can bite, the tool stops cutting and starts plowing and rubbing. Rubbing generates heat, and worse, the heat has nowhere good to go. In a proper cut most of the heat leaves in the chip. In a rubbing cut the chip is too thin to carry it away, so the heat soaks into the tool and the workpiece instead. The edge softens and dulls, and a dull edge rubs even harder, which is a loop that only ends one way.

In work hardening materials it gets uglier. A tooth that rubs 304 or 316 stainless instead of cutting it hardens the surface, and then the next tooth has to cut that hardened skin, so it dulls faster and rubs more, and now you are machining a material you did not quote. This is why stainless has a reputation for punishing timid cuts. The gentle pass is not gentle. It is the exact condition that kills tools at the light cuts that were supposed to be easy on them, and the instinct to lighten up when a tool struggles only makes it worse.

The fix is more feed, not less

The correction is simple once you accept it. If the geometry is thinning your chip to 60 percent of programmed, you multiply the programmed feed by the inverse of that factor to put the real chip back where the catalog wanted it. For the ten percent stepover example that means feeding about 1.67 times your normal number. It feels wrong the first time. You are taking a delicate finishing pass and the fix is to shove the tool considerably faster. But the edge does not care about your feed rate readout, it cares about chip thickness, and the faster feed is what restores a proper chip. You get the heat back into the chip where it belongs, the edge stays cutting instead of rubbing, and tool life usually gets better even though the tool is doing more work per minute. The speeds and feeds calculator on this site computes the thinning factor from your tool diameter and stepover and shows the corrected feed automatically, so you do not have to carry the formula in your head.

Where this matters most

Two places, mainly. The first is light finishing passes, where stepovers of five or ten percent are normal and the thinning is severe. The second is high efficiency milling toolpaths, the trochoidal and adaptive style paths that CAM systems generate. Those strategies deliberately hold a small radial engagement, often around ten percent of diameter, and run deep axially. The entire approach depends on chip thinning compensation. The aggressive feeds those toolpaths use are not the software being reckless, they are the software correcting for thinning so the chip comes out right. Run an adaptive path at uncorrected catalog chip load and you get all the rubbing and none of the productivity.

When you can ignore it

At half the diameter of engagement or more, there is no radial thinning to correct. The tooth reaches full chip thickness during its arc, the factor is one, and your programmed feed is honest. Slotting, heavy roughing at 50 or 60 percent stepover, facing with most of the cutter buried, all of that runs at catalog chip load. Heavy engagement brings its own separate concerns, harsher entry and exit angles and more teeth in the cut at once, but the chip thickness math is not one of them. The correction only starts mattering as you drop under half the diameter, and it grows quickly from there.

The same idea shows up in the axial direction with lead angle tools. A face mill or chamfer style cutter with a 45 degree lead spreads the cut across a longer edge and thins the chip the same way, and the fix is the same, feed up to compensate. Different axis, identical logic.

The ball nose version, where everyone gets burned

There is one more thinning trap, and in 3D milling it is the most common one of all: the tool is not cutting on its full diameter. A ball nose endmill taking a shallow axial cut is only working on a small circle near the tip, and both your surface speed and your chip math need to use that effective diameter instead of the catalog size. The formula is effective diameter equals two times the square root of the depth of cut times the quantity diameter minus depth of cut, or Deff = 2√(ap(D − ap)). A quarter inch ball at ten thousandths deep is cutting on about a tenth of an inch of effective diameter, so at the rpm you picked for a quarter inch tool it is really running at forty percent of your intended surface speed, rubbing along at the tip where the speed is zero. Bull nose tools with a corner radius do the same thing on a smaller scale whenever the depth is inside the corner radius. Spin the rpm up to match the effective diameter, or the lightest, most delicate finishing pass becomes the slowest and rubbiest cut on the whole part. The speeds and feeds calculator computes the ball nose effective diameter in milling mode and bases the rpm on it.

The short version is this. Chip load is a promise about chip thickness, and below half the diameter the geometry breaks that promise. Feed up by the thinning factor and keep the tool cutting instead of rubbing. Check your numbers in the calculator before you blame the endmill.

As always, this is general practice, not a spec. Tooling manufacturer recommendations, prints, and customer requirements win every argument.