CNC Machining of Gyroscopes: How Geometry Controls Drift

• 1
  What actually sets gyroscope drift
• 2
  Material choice changes the error, not just the weight
• 3
  Why 5-axis work fits gyroscope housings
• 4
  When CNC machining is the wrong process
• 5
  Which process holds which gyroscope feature

A rate gyroscope measures angular velocity. The sensing element, whether a vibrating ring, a tuning fork, a spinning rotor or a fiber coil, converts motion into a signal through a geometric relationship. Change the geometry and the scale factor changes with it. That is why the machined housing matters more than most people expect.

Take a ring gyroscope with a 30 mm resonator diameter. The drive and sense modes sit a few hertz apart. If the ring bore is oval by 5 μm, the stiffness around the circumference is no longer uniform, the two modes split further, and the zero-rate output walks with temperature. The electronics can null some of this. They cannot null a geometry error that changes with load.

The same logic applies to spinning-mass designs. A rotor that is 2 g·mm out of balance at 24,000 rpm produces a synchronous force that the bearings must absorb. That force shows up as vibration in the output. Balancing a rotor is easier when the bore and the outer diameter are concentric to within a few micrometres to begin with.

So the error budget of a gyroscope is split between electronics, assembly and machining. Machining owns the part you cannot fix later. Once the housing is cut, the bore position, face squareness and surface finish are locked in.

• 1
  Bore roundnessControls mode splitting in ring and cup resonators.
• 2
  Face squarenessSets the mounting axis relative to the sense axis.
• 3
  Hole positionDetermines where the sensing element sits after assembly.
• 4
  BalanceRotor mass distribution drives synchronous vibration.

Thermal expansion is the quiet error source. A housing in 6061 aluminium grows about 23 μm per metre per degree Celsius. Over a 60 °C swing that is roughly 1.4 μm on a 1 mm feature, and more on a long bore. If the resonator is a different material, the gap between them moves. Matching the coefficient of thermal expansion between housing and sensing element removes that term.

Titanium TC4 (Ti-6Al-4V) is common for rotor housings because it holds strength at temperature and has a lower expansion coefficient than aluminium. It also machines slowly. Cutting speeds drop, tool wear rises, and thin walls move under clamping force. A 1.5 mm titanium wall needs light radial passes and a rigid setup, not a heavy one.

Stainless 17-4PH (SUS630) is the usual choice for bearing journals and shafts. It takes a good finish, resists corrosion, and can be heat treated after roughing so the final dimensions survive the age-hardening shift. Beryllium copper appears in some flexure and spring elements because it combines stiffness with electrical conductivity.

Plastics and carbon fibre composites show up in lab and low-cost units. PEEK holds dimension better than POM under heat, but neither matches metal for stiffness. If you are chasing bias stability below 1 °/h, the housing is metal.

A gyroscope housing is rarely a simple cube. It carries a precision bore on one axis, a mounting flange on another, and often a set of sensor pockets at angles in between. On a 3-axis machine these features need multiple setups. Each setup adds a datuming error and a re-clamping error. On a housing with a 20 μm true position callout, two extra setups can eat the whole tolerance.

Five-axis machining cuts the bore, the flange face and the angled pockets in one setup. The rotary table positions the part, the spindle reaches in, and every feature is measured from the same datum. On a Ø400 mm rotary table we can hold ±0.005 mm on bore position and keep the mounting face perpendicular to the bore axis within that band.

Thermal drift during the cycle is the other reason for single-setup work. A long run on a 3-axis machine warms the spindle and the part, and the last operation is cut at a different temperature than the first. On a compact 5-axis cell with a short cycle, the part stays closer to ambient and the dimensions stay closer to nominal.

Finish is part of the process, not a separate step. Boring with a sharp insert and a light finishing pass reaches Ra 0.8–1.6 μm. Where a bearing or a seal rides, we push to Ra 0.2–0.8 μm with a wiper insert or a fine boring head.

• 1
  One setupBore, face and angled pockets from a single datum.
• 2
  Short cycleLess thermal drift between first and last cut.
• 3
  In-process probingCatches a shift before the part is finished.

Machining is not always the answer, and saying so saves money. If the part is a thin flexure with a 0.1 mm web, milling will deflect it and the spring rate will scatter. Photo-chemical etching or laser cutting holds that geometry better. If the part is a complex 3D lattice or a hollow shell with internal channels, additive manufacturing wins, and then the critical faces get machined after.

For a resonator that must have matched modes, machining alone may not reach the required symmetry. In that case the housing is machined to a near-net shape and the resonator is tuned after assembly by removing a small amount of material. The machined part sets the starting point; the tuning step finishes the job.

Volume also decides. One prototype housing in titanium is a 5-axis job. Ten thousand identical aluminium housings may be die cast and then machined only on the bore and the mounting face. We run both routes, so the choice is based on the drawing, not on what we prefer to sell.

The test is simple. If the tolerance is tighter than the process can hold after heat treat and assembly, pick a different process for that feature. Machine what needs to be precise. Cast, print or etch what does not.