Speeds and feeds, with the machine's limit in the arithmetic

The bottom, on any setup you have not run before. The band is wide because it is covering things the table cannot see — how far the tool sticks out of the holder, whether the cut is interrupted, how rigid the workholding is, and how long you need the edge to last. Speed is the variable that consumes tool life fastest, so the cheap experiment is to start low, confirm the chip looks right and the sound is steady, and then raise the speed in steps of ten percent. Production work runs at the top of a band because somebody already paid for the experiment; a one-off part in a vise does not have that history behind it.

Find the class it behaves like and then take the bottom of that band, not the middle. The seven here are chosen because they bracket the ordinary shop: a 12L14 free-machining bar runs faster than the mild steel row, a prehardened P20 mold plate runs at half the alloy steel row, and Inconel is slower than the titanium row by a wide margin. When you have genuinely no idea, the honest move is to look up the alloy's hardness and work from that instead — hardness tracks cutting speed far better than an alloy name does, which is why the pages here print the condition and the Brinell range beside every band.

Sometimes, and only the coating supplier can tell you by how much. TiAlN and AlTiN raise the temperature the edge tolerates, which is why they buy speed in steel and cast iron; in aluminum they buy nothing and can actively hurt, because the failure there is built-up edge rather than heat and a coated flute is rougher than a polished one. The carbide column in the table is written for uncoated and coated grades together for exactly this reason: the spread between two coated grades from two makers is bigger than the step from uncoated to coated, so a single multiplier would be fiction.

Whichever one is turning. On a mill or a drill the tool spins, so the surface speed is at the cutter's outside diameter; on a lathe the workpiece spins and the tool is stationary, so it is the bar's diameter that goes in and it falls as the cut goes deeper. That is why constant surface speed exists on a lathe control: facing a 6 in disc to center at a fixed rev/min means the cutting speed falls to zero at the middle, which is exactly where the finish goes bad and the insert rubs. G96 asks the spindle to hold the surface speed and let the rev/min climb, up to whatever G50 caps it at.

The speeds do; the feeds usually do not, and the reason is rigidity rather than arithmetic. A benchtop machine deflects under a cut that a 40-taper machining center does not notice, so the chip load the table gives is a ceiling you approach rather than a starting point — halve it, take a shallower axial depth, and let the extra passes make up the metal. The other constraint on a router is the opposite of a mill's: the spindle turns fast and has very little torque, so a 1/2 in cutter in steel is out of reach even though the recommended rev/min is well inside its range.

In steel and stainless, modestly; in titanium it is not optional; in cast iron and often in aluminum it makes things worse. Coolant does two jobs, cooling and chip evacuation, and only the second one is always wanted. On an interrupted cut in cast iron a jet of coolant thermally shocks the carbide every time the edge leaves the work, and the edge chips instead of wearing. Gray iron is normally cut dry because its own graphite lubricates the cut. In aluminum an air blast that clears the chips will often beat flood, because recutting chips is what welds material to the flute. Titanium is the exception in the other direction: it puts nearly all the heat into the tool, so flood coolant is part of the setup rather than an improvement to it.

Because the chip load is what the edge is being asked to cut and the feed rate is just how fast the table has to move to deliver it. Feed in inches per minute is rev/min times flutes times chip load, so halving the speed halves the feed and leaves the chip exactly the same thickness. This is the single most useful thing to understand about the pair: rev/min and feed rate are not two independent dials, they are one decision — the chip — expressed in the units a control accepts. Change the speed on the machine without changing the feed and you have silently changed the chip load, which is how a cutter that was fine at 3,000 rev/min snaps at 1,500.