CNC machining in London: a guide to precise engineering

• 1
  What removes metal, and what that means for accuracy
• 2
  3-axis, 4-axis, 5-axis and mill-turn: where each one fits
• 3
  Reading a tolerance band before you request it
• 4
  Workholding and tooling decide the result more often than the spindle
• 5
  How the part is measured sets what you can claim
• 6
  Which process fits the part in front of you

CNC machining is subtractive. A rotating tool with defined edges is fed through a workpiece along programmed paths, and material leaves as chips. The controller reads G-code, interpolates each axis, and keeps the tool tip on a path the CAM programmer defined. Everything downstream, including your tolerance, is a consequence of how stiffly that path is held.

Accuracy is therefore not a property of the machine alone. The part's own rigidity, the fixture, the tool overhang, the coolant, and the temperature of the room all enter the same error budget. A 3-axis mill in a stable shop can beat a 5-axis machine pushed past its reach.

What is controllable: position, feed, spindle speed and thermal drift. What is not: the material's internal stress, which moves a part after clamping is released. A 6061 plate roughed hot and finished cold will not land on the same dimension twice.

So the first question about any CNC machining in London quote is not which machine, but which sequence. Rough, stress-relieve, then finish keeps dimensions predictable on thin walls and long parts.

Three-axis machines cut from one direction. They are fast, cheap per hour, and the right answer for plates, brackets, housings and anything that can be reached from five faces with two setups. If a part needs three setups on a 3-axis machine, the setup error stacks three times.

A 4-axis mill adds a rotary table, usually Ø400 mm class, so the part indexes around one horizontal axis. This suits shafts, flanges, and parts with features on four sides. You lose nothing in rigidity if the table is well supported and the part is balanced.

Five-axis simultaneous machining tilts both the tool and the table. It reaches undercuts, keeps the tool normal to a curved surface, and shortens the tool so it deflects less. It also costs more per hour and needs a programmer who understands collision checking.

Mill-turn centers combine turning and milling in one spindle. For a part that is mostly round with cross-holes or flats, this removes a whole setup and the concentricity error that comes with it.

A drawing that says ±0.005 mm everywhere is a drawing nobody can quote honestly. Tight tolerances are local. Put the band on the features that mate, seal, rotate or locate, and leave the rest at general tolerances such as ISO 2768 medium.

The cost curve is not linear. General machining holds around ±0.1 mm on small parts without effort. Below ±0.02 mm you are fighting thermal growth, tool wear and fixture compliance. At ±0.005 mm, the shop must control the room, the spindle warm-up and the measurement loop.

Tolerance also depends on size. A ±0.005 mm callout on a 20 mm bore is routine on a good grinder or a jig borer, and on a 5-axis machine it is achievable with a boring head. The same number across a 1,000 mm length is a different problem, because thermal expansion of aluminum is roughly 23 μm per meter per degree Celsius.

Ask what the feature does before you tighten the number. A clearance hole at ±0.005 mm buys nothing and adds a second operation.

A part moves when you cut it. The softer the material and the thinner the wall, the more it moves. Soft jaws bored in place, vacuum plates, and low-melt fixturing all exist for this reason. The right choice depends on where the part is stiff and where it is not.

Tool overhang is the other half. Deflection scales with the cube of length, so a tool that sticks out 60 mm instead of 30 mm bends eight times as much under the same side load. On deep pockets, use the shortest tool that reaches, then a longer one only for the final passes.

Chip evacuation matters on aluminum and plastics. Recutting a chip doubles the cutting temperature and ruins the finish. Through-spindle coolant and high helix geometry are not luxuries on deep cavities.

For thin floors, leave stock, cool the part, and take a light finish pass. Roughing to final size on a flexible floor is the most common cause of scrapped parts we see.

A CMM in a temperature-controlled room reads differently from a caliper on a bench. Both are valid, but they cannot be compared directly. If your drawing calls a tight true position, the inspection method has to be named as well.

For most work, first article inspection plus in-process checks at the machine catch drift before the run is finished. Gauges for bores, micrometers for outside diameters, and a height gauge for step heights cover a large share of parts.

Reports matter for regulated work. Aerospace, medical and automotive programs usually want material certificates, dimensional reports and traceability back to the heat number. Ask for these before the run, not after.

We inspect 100% of parts before shipment, and reports are available on request. That does not replace your own incoming inspection, but it removes most surprises.