CNC Milling Basics Course: What Engineers Need to Know
This is a working primer on what a CNC milling basics course should actually cover: how the cut removes metal, how a toolpath becomes a part, and where tolerances and setups decide the result. It is written for design engineers, process engineers, and buyers who review milling quotes and prints. Read it and you can judge whether a feature is millable, which parameters matter, and what to ask before a job starts.

How the Cut Removes Metal
Every milling operation is a controlled fracture. A flute edge enters the material, the material shears along a plane ahead of the edge, and a chip slides up the rake face. If the chip is too thin, the edge rubs instead of cutting and work hardens the surface. If it is too thick, the edge breaks down. The window between those two is the real subject of a CNC milling basics course.
Chip load per tooth is the number that decides which side of that window you land on. It depends on feed rate, spindle speed, and flute count. For 6061 aluminum on a 12 mm three-flute carbide end mill, 0.05–0.10 mm per tooth is a normal roughing range. Drop below 0.02 mm per tooth and the edge starts burnishing the wall.
Surface speed changes with the material, not the part. Aluminum runs fast, often 300–500 m/min. Stainless 316 runs at roughly a quarter of that. Titanium TC4 (Ti-6Al-4V) is slower still, and heat stays in the cut, so coolant delivery matters more than speed.
The radial and axial depth of cut set how much of the tool is engaged. A 50 percent radial stepover with a 1× diameter axial depth is a common roughing starting point. Full-width cuts at full depth push the tool into chatter unless the setup is very stiff.
- 1Chip load0.05–0.10 mm per tooth in aluminum, lower in steel
- 2Surface speedSet by material, not by part size
- 3EngagementRadial stepover and axial depth decide cutter load
From CAM Toolpath to Machine Motion
CAM software does not cut metal. It writes coordinates. The post-processor turns those coordinates into G-code for a specific control, and the machine turns G-code into axis motion. A mistake anywhere in that chain shows up on the part, not in the software.
G00 moves at rapid feed to a position. G01 moves at the programmed feed. G02 and G03 cut arcs clockwise and counterclockwise. Most of a milling program is those four codes plus tool changes and coolant commands. Reading a program is a useful skill even for engineers who never touch the control.
Cutting direction matters. Climb milling, where the tooth enters at maximum chip thickness, pushes the tool away from the finished wall and usually gives better surface finish and longer tool life on a rigid machine. Conventional milling does the opposite. On older machines with backlash, conventional milling is sometimes the safer choice.
Toolpath style decides cycle time as much as feed rate does. Trochoidal paths keep radial engagement low and let the tool run deeper and faster. Adaptive clearing does the same in CAM. On a part with deep pockets, that change alone can cut cycle time by a third.
- 1G00 vs G01Rapid position vs programmed feed move
- 2Climb millingBetter finish and tool life on rigid setups
- 3Trochoidal pathsLow radial engagement, higher feed and depth
Workholding and Setup Decide the Part
A perfect toolpath on a loose part produces scrap. Workholding has to resist cutting force in three directions and still let the tool reach the features. Vises are fine for blocky parts. For thin plates, vacuum chucks or adhesive fixturing spread the load and avoid distortion.
Zero point matters more than most people expect. A 0.02 mm error in the work offset shows up on every feature cut from that setup. Probing on the machine, rather than touching off by hand, removes most of that error and takes less time on a repeat job.
Thin walls are the classic failure mode. A 1 mm wall in aluminum will deflect under cutting force and spring back after the tool passes, leaving a tapered or wavy surface. Roughing both sides, leaving 0.3–0.5 mm for a finishing pass, and taking light finish cuts usually fixes it.
Multiple setups multiply error. Every re-clamp adds a new datum. On a part with tight relationships between faces, a five-axis machine or a mill-turn center can cut most features in one setup, which removes stacked tolerance. GreatLight runs 16 simultaneous 5-axis machining centers for exactly that reason.
- 1Vacuum and adhesiveBetter for thin plates than vise jaws
- 2On-machine probingRemoves hand touch-off error from the offset
- 3Wall deflectionRough both sides, leave 0.3–0.5 mm, finish light
Tolerance, Finish, and What Drives Cost
Tolerance is not a single number for the whole part. A bearing bore at ±0.005 mm and a mounting hole at ±0.1 mm can sit on the same print. Cost tracks the tightest requirement, not the average one, because the tight feature sets the machine, the inspection plan, and the scrap risk.
Surface finish is separate from dimensional tolerance. Ra 1.6–3.2 μm is a normal as-machined finish. Ra 0.8–1.6 μm needs a controlled finishing pass. Ra 0.2–0.8 μm usually means a fine finishing strategy, a sharp tool, and often a second operation. Calling out Ra 0.4 μm on a non-sealing face adds cost for no function.
Feature geometry sets the floor on cost too. A pocket 4× deeper than its width needs a long, thin tool that deflects. A sharp internal corner needs a tool radius no larger than the corner. Both are machinable, and both raise cycle time. Redesigning to a 2× depth-to-width ratio and a corner radius of 1.5–2× tool diameter usually cuts cost more than any parameter change.
Datum choice is part of the tolerance plan. If the print dimensions from a face that the machine cannot reach first, the shop has to add a setup or re-datum in CAM. Dimensioning from the same face that will be the first machining datum keeps the stack short and the cost down.
- 1Tightest feature sets costNot the average tolerance on the print
- 2Finish is a separate calloutRa 0.4 μm where it is not needed adds cost
- 3Depth-to-width ratioPast 4×, tool deflection becomes the limit
Materials and Their Machining Behavior
Material choice changes the cut more than any other decision. Aluminum 6061 machines cleanly at high speed and holds a good finish. 7075 is stronger but more prone to distortion after machining because of residual stress in the plate. Rough it, stress-relieve if the geometry is thin, then finish.
Stainless steels split into two groups. 303 and 304 cut reasonably well with sharp tools and steady feed. 316 and 316L work harden quickly, so a light finishing pass with a dull tool makes the next pass harder. 17-4PH in the solution-treated condition is gummy; in the aged condition it machines better but the heat treat has to come before the finish cut.
Titanium TC4 (Ti-6Al-4V) and Inconel sit at the hard end. They hold heat in the cut, so the tool edge softens and wears fast. Low surface speed, high coolant pressure, and a rigid setup are not optional. Expect shorter tool life and plan the process around it.
Plastics behave differently again. POM and ABS cut cleanly but melt if the chip cannot clear. PEEK needs sharp tools and slower speeds. Carbon fiber is abrasive and wears carbide quickly, so coated tools or diamond-coated edges are worth the cost.
- 16061 vs 70757075 distorts more; rough then finish
- 2316 and 17-4PHWork hardening punishes light dull-tool passes
- 3TC4 and InconelHeat stays in the cut; coolant pressure matters
Process Choices by Part Requirement
Match the part to the process before quoting
| Requirement | Best fit | Why | Watch out for |
|---|---|---|---|
| Prismatic part, 3 faces | 3-axis mill | Simple setup, low cost | Extra setups add datum error |
| Features on 5 sides | 5-axis machining | One setup, fewer datums | Higher hourly rate |
| Deep pocket, 4× width | Trochoidal roughing | Low radial engagement | Needs CAM support |
| Thin wall under 1 mm | Vacuum or adhesive fixture | Even load, less distortion | Slower load and unload |
| Round part with flats | Mill-turn center | Turning and milling in one setup | Limited to smaller diameters |
| ±0.005 mm bore | Finish boring on 5-axis | On-machine correction | Inspection time rises |
| Prototype, 1 piece | No-MOQ milling | No tooling cost | Cycle time per part is higher |
| Ra 0.2–0.8 μm face | Fine finish pass | Controlled stepover and feed | Second operation may be needed |
The One Rule That Saves the Most Money
If a feature is not functional, loosen its tolerance and finish before you shop for a cheaper shop. If it is functional, spend the tight tolerance on a single setup, not on more inspection. A part with one datum and honest callouts will always beat a part with five datums and one tight bore.
Questions Engineers Ask After the Basics
How deep can a small end mill cut without chatter?
As a rule, keep the axial depth at or below 1× tool diameter for a roughing pass and the radial stepover at 50 percent or less. A 6 mm end mill in aluminum can run 6 mm deep at 3 mm stepover on a rigid setup. Push past that and the tool deflects; the wall comes out tapered and the finish degrades.
For deeper pockets, use a smaller stepover with a longer reach, or step down in multiple passes. Long reach tools need a lower chip load, not a higher one.
What tolerance can milling hold on a production run?
Standard milling holds ±0.05 mm on most features without special effort. Tight features can reach ±0.005 mm when the process is controlled and the feature allows it. That is not a blanket number for every dimension on the print.
The controlling factors are machine condition, thermal stability, fixture stiffness, and whether the feature is cut in the same setup as its datum. A tight tolerance across two setups is much harder than a tight tolerance inside one.
Does climb milling always give a better finish?
On a rigid machine with a preloaded ball screw, yes, in most cases. The tooth enters at maximum chip thickness, cutting force pushes the tool away from the finished wall, and tool life is usually longer.
On an older machine with backlash in the axis, climb milling can pull the tool into the wall and leave marks. Conventional milling is sometimes the better choice there. Check the machine before you commit to a strategy.
How do I know whether to use 3-axis or 5-axis?
Count the faces that need machined features. If they all face one direction, 3-axis is cheaper and faster. If features sit on four or five faces, or if the print has tight relationships between angled faces, 5-axis removes setups and the stacked tolerance that comes with them.
The trade is hourly rate against setup count and accuracy. On low quantities, 5-axis often wins on total cost once you count the extra fixtures.
What should be in a milling quote to make it comparable?
Ask for the material specification, the machine type, the number of setups, the inspection method, and the finish callout. Two quotes that only list a price are not comparable.
A useful quote also flags any feature that drives cost. If a shop points out that a 5× depth-to-width pocket is the reason for the cycle time, the conversation gets productive fast.
Can a prototype and a production run use the same process?
Often they should not. A prototype may be milled from billet to prove geometry. The production part may be cast or forged and then milled on the critical faces. That changes the stock allowance and sometimes the datum scheme.
Decide the production process early if quantity is likely to grow. Designing the prototype around the production stock saves a redesign later.
Check Your Design Before You Cut Metal
Send a print or a STEP file and we will return a quote with a DFM review within 12 hours. No minimum order quantity, from one prototype to 10,000+ parts.
12-hour quote±0.005 mmNo MOQ100% inspection