Small parts are not just big parts scaled down. Halve the feature size and the physics change: the tool deflects more for its size, chips become proportionally huge, and a 5 µm error that is nothing on a 200 mm bracket is half the tolerance on a 0.5 mm pin. Anyone cutting sub-millimetre features for electronics or instrument hardware lives in this world, and most of the failure modes are not on the drawing — they are in the setup, the spindle, and the gauge plan. This is what actually decides whether a sub-millimetre feature holds.
1. Deflection and heat at micro scale
Tool deflection scales with the cube of length and inversely with the fourth power of diameter. A 0.3 mm end mill only 3 mm long deflects enough under a normal cut to miss a ±0.01 mm tolerance. The fix is not more spindle power — it is the shortest stick-out you can get away with, and a finish pass at near-zero radial depth so the lateral force collapses. We often take 0.02 mm radial depth on micro features and accept the extra passes, because the alternative is a springy cut that sizes nothing.
We also watch the holder. A collet that is not seated true adds runout that dwarfs the tool's own tolerance, so a 0.3 mm cutter measuring 0.32 mm on the far side is really a 20 µm eccentricity from the holder, not the tool. Shrink-fit and hydraulic holders with low runout are standard on our micro work, and we qualify each holder on a runout gauge before it touches a part. Rigidity is a chain, and the weakest link sets the number, no matter how good the spindle behind it.
At a 0.3 mm tool the cutting zone is tiny, but so is the mass that can absorb heat, so temperatures spike fast. We run micro-flood or mist rather than relying on air, because a dry micro-cut work-hardens stainless within a few passes. The coolant also helps evacuate the proportionally large chips — a 0.05 mm sliver is huge next to a 0.3 mm flute, and if it jams the flutes the next pass pecks into solid material and snaps the tool at the worst possible moment.
2. Spindle speed and chip load
Small tools need high rpm to reach a useful surface speed. A 0.3 mm tool at 60 m/min wants roughly 64,000 rpm — well above a standard 12,000 rpm VMC, which is why micro work runs on high-speed spindles or dedicated micro-machining centres. Feed per tooth drops to single-digit microns; below that the tool stops shearing and starts rubbing, which work-hardens the part and dulls the edge. There is a minimum chip thickness, and ignoring it is the fastest way to a burnt tool and a scrapped batch.
We set surface speed by material, not by habit. Brass and aluminum forgive a wide range; 316L and titanium want a tighter window or the edge smears. On a high-speed spindle we also manage acceleration — at 60,000 rpm a tool change in direction carries real gyroscopic load, so cornering speeds are limited even when the straight-line feed looks fine. The program's look-ahead and the spindle's ramp rate matter as much as the rpm number on the brochure, and we tune both before the first part runs.
3. Swiss-type support for tiny diameters
When the part is long and slender rather than flat, turning is the answer, and Swiss turning is what holds the tolerance. The guide bushing supports the stock right at the cut, so a 2 mm diameter pin has almost no exposed length to deflect. That is why micro shafts, bone screws, and catheter parts run on Swiss machines instead of a standard lathe, where the unsupported tip would wave under the tool and wander out of round.
We run sliding-headstock lathes with sub-micron servo resolution and live tooling for cross-drilling and milling in the same setup. A 0.8 mm pin with a 0.3 mm cross-hole and a 0.2 mm flat is a single operation, not three. The bushing also keeps the heat and the cutting force local, so the diameter a few millimetres away stays stable. For parts under roughly Ø32 mm and long relative to their diameter, Swiss is the process; everything else is a compromise we only accept when the geometry forces it.
4. Material behaviour at small scale
PEEK and medical-grade plastics micro-machine cleanly but can burr at exits, so deburring is a controlled pass rather than a hand touch. 316L and titanium micro-turn well on the Swiss-type because the guide bushing keeps the tiny diameter supported. The surprise for many engineers: at this scale, burrs and edge breaks often matter more than the body tolerance, because the part functions at the edge. A 0.05 mm burr on a 0.4 mm feature is a 12% error on the very surface that seals or slides, and it is invisible to a casual glance.
Work hardening is the quiet killer. Stainless and titanium that slow down mid-cut build a hard skin the next pass has to fight, so we hold a constant feed and keep the tool engaged rather than pecking in and out. Sharp edges are non-negotiable — a dull micro tool does not make a slightly worse part, it makes a burnt, burred, out-of-spec one. We track tool life in parts, not hours, because at 0.3 mm the edge is gone long before the diameter looks worn to the eye.
5. Measurement and a small example
You cannot confirm a 0.01 mm feature with a caliper, and a standard CMM probe ball (1–3 mm) is bigger than the feature. That is why micro work is measured with a vision (optical) measuring system for profile work, and a small stylus or an air/tactile method for bores on the methods that can reach them. Whatever the gauge, it must be settled before the first part — not after the cut is done and the customer is waiting.
In regulated or traceable work the gauge plan carries a documentation requirement as well. A part that passes the vision system but has no traceable measurement record may not be shippable, because the audit trail is part of the part. How long records are kept, and in what form, is something to agree in writing before the job starts.
Take a 0.8 mm × 6 mm locator with three 0.4 mm cross-holes and a 0.2 mm edge break on every sharp corner. What you can actually hold depends on the machine and the gauge plan: a high-speed micro-mill with vision-verified first-article checks and a controlled deburr pass is a different proposition from a standard VMC, where features like these tend to come back with visible burrs — technically "machined," functionally unusable on the assembly they were meant to seat into.
"On a 0.3 mm tool, rigidity beats horsepower. A short, sharp tool at the right speed will hold what a big aggressive cut cannot."
6. Process window and what to ask your shop
| Feature / tool | Spindle speed | Radial depth | Expected tolerance |
|---|---|---|---|
| 0.3 mm end mill, Al | 40,000–60,000 rpm | 0.02 mm | ±0.01 mm |
| 0.3 mm end mill, 316L | 50,000–64,000 rpm | 0.015 mm | ±0.012 mm |
| Ø2 mm Swiss pin | 8,000–12,000 rpm | n/a (turn) | ±0.005 mm |
| 0.4 mm cross-hole | 30,000–45,000 rpm | peck 0.2 mm | ±0.02 mm |
- Which machine? A standard VMC and a micro-mill are both "CNC" — ask which one will actually cut your feature, not which one is in the building.
- What gauge? Vision system, small stylus, or nothing? If they cannot measure it, they cannot hold it, and you will find out at incoming inspection.
- Holder runout? Under 3 µm matters more than the spindle's rpm rating at this scale, because runout is the first thing your tolerance pays.
- Tool life tracked in parts? Micro edges wear fast; an hours-based change interval will drift mid-batch before anyone notices the size walking.
7. When micro-machining is the wrong answer
Not every small part needs a micro-mill. If the critical features are larger than about 1 mm and the tolerances sit at ±0.05 mm or looser, a well-rigged standard VMC or a Swiss lathe will do it cheaper and faster. We tell customers this directly, because putting a part on a high-speed micro centre when it does not need one just adds cost with no benefit to the part. The skill is matching the process to the feature, not reaching for the most expensive machine on the floor and charging for the idle capability.
If your part has features under a millimetre, ask what machine and what gauge the shop will actually use. The right answer is a specific machine and a specific measurement method, stated before the quote, not a vague promise that CNC will sort it out. At this scale the difference between two shops is rarely the print — it is whether they understood the physics before they cut.