What a CNC Gloucester Expert Actually Controls

• 1
  Key takeaways
• 2
  Where the Last 0.005 mm Comes From
• 3
  Heat Is the First Thing That Moves
• 4
  Why Five Axes Change the Tolerance Map
• 5
  Finish Callouts and What Tools Deliver Them
• 6
  How the Number Gets Verified Before Shipment
• 7
  Which Machine and Setup Fits the Tolerance

A tolerance callout of ±0.005 mm looks like a single number on a drawing. On the shop floor it is a budget. Every step in the process spends part of it: the machine's own positioning error, the fixture holding the part, the cutter's runout, and the material moving as it gets warm. A CNC Gloucester expert works by adding those errors up and keeping the total under the callout, not by trusting one instrument.

Start with the machine. A three-axis mill with box ways and a calibrated ball screw holds roughly ±0.005 mm positioning, but only within its thermal steady state. A simultaneous five-axis machining center with a Ø400 mm rotary table adds two more rotary axes, and each one contributes angular error that becomes linear error at the part. That is why we measure the rotary centerline at setup and re-check it after long runs, not once a year.

The fixture is the next consumer of the budget. A part clamped on three points can bow under cutting force. A thin wall pushed 0.02 mm by a vise jaw springs back after unclamping, and the final measurement looks wrong for no obvious reason. Soft jaws bored to the part profile, or a vacuum plate for flat thin parts, keep clamping force low and repeatable.

Finally, the cutter. Radial runout of 0.01 mm on a 6 mm end mill cuts a slot that is wider on one side, and no amount of machine compensation fixes a wobbly tool. We check runout at the spindle taper with a dial indicator before finishing passes, and we keep separate holders for roughing and finishing so a worn collet never touches a final cut.

• 1
  Positioning errorBall screw and scale error, typically 0.003–0.005 mm on a healthy machine.
• 2
  Fixture errorClamping deflection, often the largest single term on thin parts.
• 3
  Tool runout0.005–0.015 mm on a worn holder, visible as a taper on walls.
• 4
  Thermal driftGrows with part length and spindle run time; small parts cool fast.

Metal expands before the tool touches it. Aluminum 6061 grows about 23 × 10⁻⁶ per °C, so a 300 mm part that warms 1 °C from a cold shop floor to a running spindle grows roughly 0.007 mm. That single degree already eats the entire ±0.005 mm budget. A CNC Gloucester expert deals with this by letting the machine idle to steady state, and by measuring the part after it has cooled back to room temperature, not while it is warm.

The spindle is the second heat source. A spindle running at 12,000 rpm for an hour can climb 5–10 °C above ambient, and the spindle nose grows with it. On long finishing passes this shows up as a slow drift in Z depth. We break long passes into shorter segments and re-reference between them, so the depth error stays inside the band instead of accumulating in one direction.

Material choice changes the arithmetic. Steel 1045 expands at about 11 × 10⁻⁶ per °C, less than half of aluminum, so a steel part is more forgiving of a warm shop. Titanium TC4 (Ti-6Al-4V) is closer to 9 × 10⁻⁶, but it cuts hot and conducts heat poorly, so the tool and the local surface heat up while the bulk stays cool. That local difference is what causes chatter and poor finish on titanium, not the overall part temperature.

Coolant matters as much as the machine. Flood coolant keeps the tool and the chip cool but can shock a thin part and move it. For thin aluminum plates we often run air blast with a mist instead, accept a slower feed, and measure twice: once at temperature and once after the part rests on a granite plate for 30 minutes.

• 1
  Aluminum 606123 × 10⁻⁶ per °C. Fast to cut, least thermally stable.
• 2
  Steel 104511 × 10⁻⁶ per °C. Drifts about half as much as aluminum.
• 3
  Titanium TC49 × 10⁻⁶ per °C, but heat stays at the cut, so local growth rules.

Every time a part is unclamped and re-clamped, the origin shifts. On a simple plate that shift might be 0.01 mm. On a part with five setups it compounds, and the feature you care about may sit on the last setup. Simultaneous five-axis machining removes those setups by tilting the tool and the table together, so features on five faces come off one origin. That is the real reason to pay for five axes on a complex part, not the ability to cut a fancy angle.

The trade-off is that five-axis motion introduces its own error. The rotary table has a centerline position and a tilt error, and the two combine into a linear error that grows with distance from the table center. A feature 200 mm from the rotary center can pick up 0.015 mm of error from only 15 arc-seconds of tilt error. We map the rotary centerline with a ball and a dial indicator, then post the values into the CAM setup so the toolpath compensates.

Machine size sets the practical limit too. Our largest travel is 4,000 × 400 × 150 mm for long parts, and our medium platforms run 750 × 1,150 × 550 mm and 600 × 600 × 600 mm. A part that fits a 600 mm cube usually holds tighter tolerances than a 3 m part on the same tolerance callout, because the machine has less travel to correct and the part has less room to move. When a drawing calls ±0.005 mm on a 2 m part, we check whether the tolerance applies to the whole length or only to a local feature. It almost always applies locally.

For parts that do not need five-axis motion, a four-axis mill with a rotary table or a mill-turn center is often the better choice. Mill-turn keeps a turned diameter and a milled flat on one setup, which removes the concentricity error that comes from moving a part between a lathe and a mill. On shafts with a bearing seat and a keyway, that single-setup advantage is worth more than any tolerance number on the drawing.

• 1
  One setup, one originFewer re-clamps means fewer position errors stacked in the final size.
• 2
  Rotary error grows with radiusCheck the tolerance at the feature's distance from the table center, not at the center.
• 3
  Long parts drift moreTravel and thermal length both work against a 2 m part holding 0.005 mm end to end.

Surface finish and tolerance are linked, not separate line items. A Ra 0.8–1.6 μm finish is the normal result of a clean finishing pass on aluminum or steel with a sharp carbide tool. Getting to Ra 0.2–0.8 μm needs a smaller step-over, a higher spindle speed and a tool that is fresh, and it usually adds a pass. It also makes the part more sensitive to chatter, because a lighter cut removes less material and any vibration shows up in the surface.

As-machined surfaces at Ra 1.6–3.2 μm are fine for brackets, housings and most internal features that nobody sees. Specifying a fine finish where it does not matter adds cost and time without changing function. The engineering question is whether the surface is a sealing face, a bearing surface or a sliding surface. If it is none of those, the as-machined finish is usually correct.

Tool choice follows the feature. A 6 mm end mill with a corner radius handles a pocket floor better than a square-corner tool because the radius spreads cutting force and reduces chatter at the corner. Deep pockets need a tool with enough flute length to clear chips, because recutting a chip is the fastest way to break a small tool and lose the part. For slots narrower than 3 mm, we slow the feed and use a high-pressure coolant or air blast to clear chips.

Hard materials change the tool material. Inconel and 17-4PH stainless work-hardened surfaces wear carbide quickly, so we use coated carbide with a lower cutting speed and accept a shorter tool life. Titanium TC4 needs sharp edges and generous coolant because it conducts heat poorly and the cutter edge reaches high temperature. These are process choices, not preferences. They decide whether the finishing pass comes out at Ra 0.8 μm or tears the surface.

• 1
  Ra 0.2–0.8 μmSealing faces, bearing seats, optical or sliding surfaces.
• 2
  Ra 0.8–1.6 μmStandard finishing pass on most aluminum, steel and stainless parts.
• 3
  Ra 1.6–3.2 μmAs-machined brackets, housings and hidden internal features.

A tolerance you cannot measure is not a tolerance. We check the drawing first to see which dimensions are functional and which are reference, because measuring every dimension on a complex part costs more than the part. The functional dimensions get measured with the right instrument: a micrometer for a shaft diameter, a bore gauge for an internal diameter, a height gauge on a granite plate for a step, and a CMM for position and profile tolerances.

Material comes in before the first cut. We check the certificate against the drawing and, where the drawing calls it out, run a hardness or composition check. A part made from the wrong heat of steel can machine perfectly and still fail in service. In-process monitoring catches drift: the operator measures the first part, then a part every few pieces, and adjusts the offset when the trend moves toward the tolerance limit.

Final inspection is 100% before shipment. That means every part gets checked against the critical dimensions, not a sample. Reports are available on request, and for regulated work we can document the measurement method and the instrument used. The ISO 9001:2015, IATF 16949:2016, ISO 13485:2016 and ISO 27001:2022 systems cover the process side, from traceability to how drawings are stored.

The qualification rate on our line runs 99.99%. That figure matters most on tight-tolerance parts, because a single out-of-tolerance feature can scrap a part that already has hours of five-axis time in it. When a dimension is difficult, we say so during the DFM review and suggest a change to the design or the callout. A small geometry change often turns an unstable tolerance into a stable one.

• 1
  Functional dimensions firstMeasure what the part does, not every line on the drawing.
• 2
  In-process offsetsAdjust the tool offset when the trend moves, before parts go out of tolerance.
• 3
  100% final checkEvery part checked against critical dimensions before it ships.