Carbon fiber gives you weight reduction, high specific stiffness, and a woven surface that needs no paint. What it does not give you is complex snap fits, uniform thin walls, or antenna signal. Carbon fiber is conductive and shields electromagnetic radiation, so once an enclosure wraps around an antenna, efficiency drops. That has to be solved at the mechanical design stage, not after tooling. On cost, autoclave prepreg lands several times higher per part than injection moulding plus paint, and the tooling sits in a different bracket entirely.
There are only three things the designer actually has to get right: hold the wall thickness to a whole multiple of the ply stack, put the parting line where the weave can run continuous, and convert every snap fit and boss into an insert or an overmoulded part.
The short version: carbon fiber reduces weight, raises stiffness, and brings its own surface texture. It cannot do complex snap fits, it cannot do uniform thin walls, and it blocks antennas. Wall thickness, parting line and fastening method all have to be fixed during mechanical design.
1. Wall thickness is laid up, not drawn
A carbon fiber part is not a shape with a wall thickness pulled from it. The thickness is built ply by ply. How thick a shell ends up depends on how many plies, at what angles, in what fiber.
Production parts normally use 3K or 12K carbon cloth with epoxy, in one of three process bands. Dry cloth with vacuum bagging is the cheapest and the least attractive cosmetically. Prepreg in an autoclave gives the best surface and the best properties, at the highest cost. SMC compression moulding suits mid volume, with slightly lower weave fidelity. We pick by batch size: appearance parts in the tens to low hundreds go prepreg; above a thousand units we move to SMC or drop carbon fiber for something else.
Ply angle decides how stiffness is distributed. 0°/90° carries bending stiffness; ±45° carries torsion. A handheld device that mostly sees grip bending wants more 0/90. A long thin shell that twists easily needs a higher ±45 share. The stack also has to be symmetric, with an even number of plies. An asymmetric layup cures and cools with unequal shrinkage between plies, and the shell warps — it will sit on a surface plate with one corner lifted, and no amount of tool accuracy will fix it.
Practical thickness bands: vacuum bagging can technically produce 0.8 mm, but the cloth does not fill out at that thickness, resin distribution becomes uneven, and the cosmetic face develops blotching, so we do not go below 1.2 mm. A normal appearance shell runs 1.5 mm to 2.0 mm, which is 4 to 6 plies — enough to work with and easy for operators to handle. Past 3 mm, stacking plies stops paying: weight and cost both climb. That is where you switch to a sandwich — aluminium honeycomb or PMI foam — and gain thickness with far more stiffness per gram.
Numbers to carry: normal carbon fiber appearance shells run 1.5–2.0 mm in 4–6 plies; the practical floor is 1.2 mm with vacuum bagging; above 3 mm move to a sandwich core. Keep the layup symmetric and the ply count even, or the part warps.
2. Snap fits are where carbon fiber parts fail
A cantilever snap needs a thin root and repeated deflection. Carbon fiber plies are discontinuous at a sharp corner, and repeated load delaminates from there. A 0.8 mm snap arm is routine in moulded PA; in carbon fiber it is a delamination origin.
We handle it three ways: bond in aluminium inserts for threaded joints, overmould a PP or PA sub-component onto the shell to carry the snaps, or fit stainless steel springs through the wall. Which one depends on assembly direction and service requirements — parts that come apart repeatedly get springs, permanently fixed parts get inserts. Overmoulding suits cases with many snaps spread across the part, with moulded bosses on the carbon shell acting as the bond face.
There is a cost line that is easy to miss: carbon fiber in direct contact with aluminium corrodes electrochemically in humid conditions, so the interface needs an insulating gasket or coating. Threads cannot be tapped directly into the laminate either — you need inserts, or a blind hole with adhesive. Individually small, but the assembly time and consumables add up, and they belong in the quote.
3. Weave alignment is a design decision, not a factory one
Half of why customers want carbon fiber is the woven surface. The weave is set by three parameters: tow size (3K / 6K / 12K), weave pattern (plain, twill, satin), and lay direction. 3K twill is the default for consumer products — fine texture, strong weave character, and 45° diagonal lines read best on curved surfaces. 12K is coarser and suits large parts; on a small handheld it looks clumsy.
Where it most often goes wrong is the parting line. Fibers are cut at the parting line, so the weave necessarily breaks. A 45° twill break is particularly visible: a whole diagonal field stops dead at the line and restarts on the other side, obvious to the naked eye. This belongs in the drawing stage — put the parting line on a crease, a corner, or a visually quiet area, never across a large flat face. Draft angle should be 1° to 3°; too little and the part drags fibers on ejection.
Symmetric left and right parts need the alignment rule fixed in advance. Mirror alignment makes the two halves read as a balanced whole; parallel alignment makes the weave easier to continue across the joint. Either works, but it has to be decided at the ID stage and marked on the drawing, with a texture datum edge stated explicitly — "weave direction referenced to the bottom edge at 0°". Without it, the factory lays up in the most material-efficient direction and the two halves each go their own way, so the finished product looks like it was made from two different batches. We issue a dedicated weave-direction diagram alongside the parting-line drawing, so the mould shop and the layup team each work from their own reference. Handoff errors drop noticeably.
There is a third route: mould the shell in ABS or PC and bond a carbon fiber skin onto it. Leave 0.5 mm to 1 mm of wrap on all four edges, close the edge with a wrap or a chamfer, and control adhesive thickness or the skin edge lifts. This costs far less than a full carbon part, looks different up close and nearly identical at a distance, and is worth considering where the weave is not critical. Hydrographic carbon-print film is cheaper still, but it cannot reproduce the depth of a twill — up close it reads flat, so it does not belong on a premium line.
Weave rules: keep the parting line off large flat faces, put it on creases or corners; mark the texture datum edge and the alignment method on the drawing; leave 0.5–1 mm of wrap on bonded skins.
4. The cost equation: tooling, unit price and yield
Carbon fiber cost is not a single number. It is three numbers stacked.
The first is tooling. A steel mould for prepreg compression has to be precision machined and polished to a mirror finish, costing several times a same-size injection mould. Vacuum bagging can use a simple aluminium or even fibreglass tool — cheaper, but short-lived, with surface accuracy drifting after a few hundred parts.
The second is unit price, and the third is yield. Prepreg costs an order of magnitude more per kilogram than engineering plastics, layup is manual work, and autoclave capacity per cycle is limited. On top of that come voids, dry spots, fiber misalignment and resin-rich surfaces. Any one of them scraps the part, and carbon fiber parts cannot be reworked — only remade.
So we use carbon fiber only when three conditions hold at once: batch size is small (tens to a few thousand), weight reduction is a hard requirement, and the surface texture carries a price premium. Missing one, look elsewhere. High-volume internal structural parts are fine in PC+GF30 or PA+GF — stiffness approaches the lower end of carbon fiber at a fraction of the cost. Appearance parts can go anodised aluminium CNC, IMD transfer, or PC with UV coating, and the texture fidelity is already convincing.
One more cost that gets missed: the joint between carbon fiber and metal. As above, carbon against aluminium corrodes electrochemically when humid, so the interface needs insulation. Threads need inserts or bonded blind holes rather than direct tapping. Small individually, material in aggregate — price it in rather than discovering it at cost review.
What designers can take away
Treat carbon fiber as a process decision, not a material upgrade. Fix the ply schedule and the parting line first, then talk about surface and cost. If a snap fit is in the concept, replace it before the drawing is released, not after the first mould is cut.
The takeaway for product teams: three conditions decide whether carbon fiber is the right answer — small batch, weight as a hard requirement, and texture that earns a premium. Wall thickness, parting line and fastening method are the three items that must be frozen at the mechanical design stage.
FAQ
Q1: Can you mould a cantilever snap directly into a carbon fiber shell?
Not recommended. The plies are discontinuous at the sharp root corner, and repeated deflection delaminates from there. The snap arm wall that moulded plastic takes for granted is a failure origin in a laminate. The reliable alternatives are bonded aluminium inserts for threaded joints, an overmoulded PP or PA sub-component carrying the snaps, or stainless steel springs fitted inside. Choose by how often the joint is serviced: springs for repeated opening, inserts for permanent fixing, overmoulding where snaps are numerous and spread apart.
Q2: What wall thickness should a carbon fiber enclosure use?
1.5 mm to 2.0 mm for a normal appearance shell, corresponding to 4 to 6 plies. Vacuum bagging below 1.2 mm is not advisable — the cloth will not fill out and resin distribution becomes uneven. Above 3 mm, plain ply stacking stops being economical; switch to aluminium honeycomb or PMI foam as a sandwich core. Thickness is also constrained by symmetry: use an even ply count, laid symmetric about the mid-plane, or differential shrinkage during cure will warp the shell visibly.
Q3: How should weave be aligned on symmetric left and right parts?
Both mirror and parallel alignment are viable. Mirror alignment gives the pair a mirrored weave and a balanced look across the assembly; parallel alignment keeps the direction consistent so the weave continues across the joint. Either way, the texture datum edge and the alignment method must be marked on the mechanical drawing. Otherwise the factory lays up in the most material-efficient direction and the two halves diverge. We issue a separate weave-direction diagram alongside the parting-line drawing, so the layup team and the mould shop each work to a clear reference.
Q4: When is carbon fiber the wrong choice?
When any one of three conditions fails: batch size runs above a few thousand, weight reduction is not a hard requirement, or the surface texture does not earn a premium. High-volume internal structural parts reach comparable stiffness in PC+GF30 or PA+GF at a fraction of the material cost. Appearance parts can use anodised aluminium CNC, IMD transfer or PC with UV coating with very high texture fidelity. Also watch the antenna problem caused by conductivity — if a wireless product is fully wrapped in carbon, antenna efficiency drops noticeably, and the structure needs an RF-transparent window or a local non-metallic section.
