We built a two-cavity tool for a 420 mm linear LED luminaire housing in PC/ABS, holding the 38 mm LED mounting channel to ±0.10 mm (Cpk 1.42) on a VDI 24 textured surface. The job took three steel corrections and 46 days to T1. Our verdict: on a textured part, texture depth is a tolerance variable, not a finish choice.
Why optical dimensions behave differently on a textured surface
A textured mold does not fill like a polished one, and it does not release like one either.
Acid-etched grain raises the cavity wall by 18–40 µm. Resin must flow over that roughness, so the effective flow length from gate to last-filled point grows by roughly 10–25%.
More importantly, texture changes the release condition. A textured wall grips the part. As the part contracts onto the core during packing and cooling, the core side and the cavity side stop behaving symmetrically.
On a smooth housing, a dimension measured across the part is basically the cavity dimension minus uniform shrinkage. On a textured housing, that same dimension also carries the pull direction of the textured wall.
For a cosmetic cover, this is acceptable. For a part that carries an LED board, it is not.
The challenge: five constraints at once
The customer is a European lighting OEM supplying retrofit linear fixtures into the German renovation market. Their existing supplier made the housing in a partner tool in Asia.
The part is a channel-section extrusion replacement, injection molded instead. Here is what the drawing required:
- Wall thickness: 2.2 mm nominal, thinning to 1.6 mm at the diffuser lip. The LED board sits on a 38 mm wide channel floor.
- Optical channel: 38.0 ±0.10 mm between the two inner rails that locate the LED strip. If the rails spread, the strip bows and produces hot spots on the diffuser.
- Surface: VDI 24 on all visible exterior faces — a matte, low-glare finish. The customer had rejected a glossier VDI 21 sample because of specular reflection off ceilings.
- Flatness: 0.15 mm over the 420 mm length, measured on the mounting flange with the part at rest.
- Volume and timing: 20,000 pieces per year, first shipment in 10 weeks. Tooling budget capped at $19,500.
The old tool had three specific failure modes, and the customer sent us the reject data: 6.1% of parts had LED strip bow of more than 0.4 mm, 3.4% showed visible sink on the textured exterior opposite the inner ribs, and 1.2% cracked at the diffuser lip during assembly.
Three defects, three different root causes, one tool.
What we did: three decisions that moved the dimension
1. Two cavities, not four, and a hot-drop gate per cavity
Twenty thousand pieces a year is a low volume for a 420 mm part. A four-cavity tool would have needed a 1,100 mm wide mold base and a press the customer’s molder did not have.
We quoted two cavities on a 696 × 496 mm base in 1.2738HH pre-hardened steel at 30–34 HRC. Each cavity is fed by a single Yudo hot drop, 1.8 mm diameter, landing on the hidden inner rib — not on the textured face.
That gate location was the first real decision. Gating on the visible textured wall would have left a vestige the customer would see on the ceiling. Gating on the inner rib means the vestige is invisible but the flow front has to travel around the channel.
Moldflow put the last-filled point 118 mm from the gate, against a 160 mm flow limit for 2.2 mm PC/ABS. That margin looked comfortable — and it was wrong, for a reason explained below.
2. Bake the texture into the steel before cutting the optical rails
This is the correction that saved the project, and it came from a mistake we made on the first iteration.
In the first machining pass we cut the whole cavity including the 38 mm optical rails, then sent the block out for acid etching. When the tool came back, the rails had moved. The textured mold measured 38.14 mm and 38.17 mm in the two cavities.
We had subtracted shrinkage from a polished-steel dimension and then let the etching process add material back.
The fix was a change in sequence. Now the cavity block is textured first, on a test coupon next to the part cavity, and the optical rails are finish-cut afterwards in the hardened block. The rails are then polished locally to SPI B1 — a narrow band, 3 mm wide, that is deliberately un-textured.
The customer accepted the polished rail band because it is hidden behind the diffuser. The visible faces keep their VDI 24 grain.
3. Split core and cavity temperature by 12 °C
The textured cavity side and the smooth core side do not want the same mold temperature.
A hot textured cavity fills better and releases cleaner. A cooler core pulls the part away from the cavity sooner, which reduces the drag that was bowing the optical rails.
We ran the cavity at 88 °C and the core at 76 °C, using a separate TCU circuit for each half and independent baffled circuits behind the two rails. Measured wall temperature variation across the cavity dropped from 9 °C in the first trial to 2.5 °C in the third.
Cycle time cost was real: 34 s at equal temperature, 39 s with the split. On a 20,000 piece annual volume, 5 s is roughly 27 machine hours a year. The customer accepted it because the alternative was a 6% reject rate.
Draft: the number the texture forced
VDI 24 corresponds to roughly 15–18 µm Ra on a mold surface. Rule of thumb: add 1° of draft for every 0.025 mm of texture depth on vertical walls.
The customer’s drawing specified 1° on the long side walls, which is correct for a smooth part and marginal for a textured one. On a 22 mm deep side wall, 1° gives 0.38 mm of clearance. That is not enough to overcome the grip of an etched wall on a contracting PC/ABS part.
We drafted the long walls at 2.0° and the short end walls at 2.5°, and added a 0.4 mm radius to the texture boundary at the flange so the grain does not terminate on a sharp edge. The customer approved the geometry change in week three.
The result: three trials and the numbers
Trial 1 ran the tool as first cut. Trial 2 added the temperature split and the rail polish band. Trial 3 was a steel correction: we removed 0.06 mm from the core side of each rail to compensate for measured post-mold contraction, which was larger than Moldflow predicted because the textured surface restricted contraction on the cavity side.
All dimensions below are from CMM reports on 30 parts per cavity, measured 24 hours after molding.
| Parameter | Trial 1 | Trial 3 (released) | Drawing limit |
|---|---|---|---|
| Optical channel, cavity A | 38.14 mm | 38.02 mm | 38.00 ±0.10 |
| Optical channel, cavity B | 38.17 mm | 38.04 mm | 38.00 ±0.10 |
| Cavity-to-cavity difference | 0.03 mm | 0.02 mm | — |
| Cpk on channel width | 0.71 | 1.42 | ≥1.33 |
| Flatness, 420 mm length | 0.31 mm | 0.11 mm | 0.15 mm |
| LED strip bow after assembly | 0.62 mm | 0.18 mm | 0.40 mm |
| Texture depth achieved | 21 µm | 17 µm | 15–18 µm |
| Sink visible on exterior | 3.4% of parts | 0.2% of parts | — |
| Cracked at diffuser lip | 1.2% of parts | 0.0% in 600 parts | — |
| Cycle time | 34 s | 39 s | — |
| Draft, long wall | 1.0° | 2.0° | — |
The sink at 3.4% had a separate cause we have not described yet. The inner ribs backing the LED floor were drawn at 62% of nominal wall, which is standard. But on a textured exterior, a 62% rib leaves a shrinkage void that the texture then amplifies visually, because the grain scatters light differently over a depression.
We cut the ribs to 55% and added a 0.15 mm core-side relief under the rib roots. Sink dropped to 0.2%, and the visible defect disappeared entirely — the remaining 0.2% is a dimensional deviation, not a visible mark.
The diffuser lip cracking was a radius issue. A 0.5 mm radius on a 1.6 mm lip left a sharp stress riser where the textured surface met the polished edge. We opened it to 0.9 mm and the crack rate went to zero in 600 parts.
Tooling landed at $18,900 against the $19,500 cap. T1 samples shipped on day 46, which includes the three days lost to the re-texturing sequence change. Approval came on day 53, and production started in week 10 as planned.
What carries to your project
If your next part has a textured surface and any dimension that other components locate against, these five points are the short version.
- Decide the texture before you decide the tolerance. A VDI number is a fixed offset on every textured wall. If the drawing says 38.0 ±0.10 mm and VDI 24, ask how the toolmaker plans to protect that dimension from the etch.
- Do not cut critical features before texturing. Either texture a test coupon first and adjust, or texture the block and finish-cut the critical features afterwards. Cutting first and etching second is how 38.00 mm becomes 38.15 mm.
- Split cavity and core temperature when the part is textured and dimensionally critical. A 10–14 °C split costs 4–6 s of cycle and can halve the bow.
- Draft is a function of texture depth, not of part size. 1° is fine on a polished wall at 20 mm depth. At VDI 24, go to 2° minimum and add radius at the grain boundary.
- Rib thickness on a textured exterior is 55%, not 62%. The sink is the same depth, but texture makes it more visible, so the safety margin has to be larger.
We have run this sequence on textured housings across appliance, automotive interior and lighting programs. It is not faster than cutting everything in one pass. It is, however, the difference between one steel correction and three.
Our answer on textured parts with optical features
Our answer: on any part that carries an optical component and a visible texture, treat the textured surface and the locating geometry as two separate manufacturing problems. Texture the block, cut the optics after, polish the locator bands, and split your mold temperatures. RCH Plastic reference — this housing was designed and built at our Dongguan tool shop: mold design in-house, molding through vetted partners, which is why we can focus entirely on steel quality and dimension rather than on protecting press time.
The mold ran on a two-cavity 1.2738HH base with Yudo hot drops, 2.0°/2.5° draft, and an 88/76 °C temperature split. Every dimension above came from a CMM report shipped with the T1 samples, along with the trial photos, video and full process parameter sheet.
If you have a textured housing with a tolerance on it, the useful first step is a drawing review, not a quote. Our engineering team returns a written DFM report in three working days at no cost, and it will tell you which of your dimensions the texture is going to move before any steel is cut.
Related guides in this series
- Mold design and DFM guide — the hub for everything below
- Texture and VDI 3400 guide — how to choose a grain grade
- Injection mold surface finish guide — SPI grades and what they cost
- Case study: appliance fascia, VDI texture and fingerprint resistance — the gloss-side companion to this project
- Draft angle guide — how much draft each feature needs
- Injection molding tolerances guide — achievable accuracy by material
- Cooling system design — balancing core and cavity circuits
- Mold steel comparison — P20 vs NAK80 vs S136
- Defects and quality guide — sink, warpage and stress cracking
Send your 3D file and texture specification to our engineering team and we will tell you where the grain is going to move your dimensions.
