Heat sink tolerance and base flatness explained: what extrusion can hold, what needs machining, and how flatness changes thermal resistance.
Heat Sink Tolerance and Flatness: What to Specify
Direct answer: As-extruded heat sink profiles typically hold +/-0.15 mm on fin thickness and +/-0.3 mm on overall dimensions. If you need tighter than that, or if you need base flatness better than 0.1 mm, the part must be machined after extrusion. Standard machined heat sink tolerance is +/-0.05 mm on critical features, with base flatness of 0.05 mm or 0.02 mm available at increasing cost.
What each process can actually hold
| Feature | As-extruded | Machined after extrusion | Notes |
|---|---|---|---|
| Overall width / height | +/-0.3 mm | +/-0.05 mm | Extrusion grows as the die wears, always in the plus direction |
| Fin thickness | +/-0.15 mm | Not applicable | Fins are formed by the die, not machined, unless skived |
| Fin pitch | +/-0.10 mm | Not applicable | Pitch error accumulates across a wide profile |
| Cut length | +/-0.5 mm | +/-0.1 mm | Saw cut vs milled end |
| Base flatness | 0.15-0.30 mm typical | 0.05 mm standard, 0.02 mm achievable | The single most cost-relevant callout on a heat sink |
| Base thickness | +/-0.25 mm | +/-0.05 mm | Thickness variation directly changes spreading resistance |
| Hole position | n/a | +/-0.05 mm | Drilled or milled after extrusion |
| Perpendicularity of base to fins | 0.3 mm typical | 0.05 mm | Matters when the heat sink is clamped against a chassis wall |
| Surface roughness on contact face | Ra 3.2 typical | Ra 1.6 standard, Ra 0.8 on request | Roughness affects thermal interface performance |
Why base flatness matters more than any other dimension
A heat sink transfers heat through a contact interface, and that interface is a function of flatness. A base that is 0.15 mm convex over 100 mm touches the component only near the centre. The air gap that remains is filled by thermal interface material, and interface material is 10-100 times less conductive than aluminium. The result is a contact resistance that can dominate the entire thermal path.
Practical effect: on a 50 W device with a 60 x 60 mm contact area, moving from 0.15 mm to 0.05 mm flatness typically removes the equivalent of 0.1-0.3 C/W of interface resistance. On a design with a total budget of 0.8 C/W, that is a meaningful fraction, and it costs far less than adding fin area.
The corollary is that specifying 0.02 mm flatness when your interface is a 0.5 mm gap pad at 3 W/mK is wasted money. Match the flatness callout to the interface you are actually using:
| Interface | Sensible flatness callout | Comment |
|---|---|---|
| Bare metal to bare metal, clamped | 0.02-0.05 mm | No gap filler, so flatness is the interface |
| Thin thermal grease, 0.05 mm bond line | 0.05 mm | Grease tolerates little gap variation |
| Gap pad 0.5-1.0 mm, 3-6 W/mK | 0.10-0.15 mm | The pad absorbs the non-flatness; tighter buys nothing |
| Phase change material | 0.05-0.10 mm | Thins under pressure, so flatness still matters |
| Potted or encapsulated assembly | 0.20 mm | Potting compound fills everything |
How to write the callout so it gets manufactured correctly
Flatness on a heat sink is only meaningful with a stated reference area. "Flatness 0.05 mm" without a zone is ambiguous; "flatness 0.05 mm over the 60 x 60 mm component contact zone" is a specification a machinist can hit and an inspector can verify.
Three further details make the difference between a callout that works and one that causes arguments:
- Say whether the measurement is before or after anodising. Anodising adds 10-25 micron per surface depending on class, and it is not perfectly uniform. A 0.05 mm flatness held before anodising can measure 0.07 mm after. If the contact face must remain bare, mask it and say so.
- Reference the datum to the mounting face, not the fin tips. Fins are not a stable datum. Measure from the face that touches the component or the chassis.
- State the inspection method. A surface plate and dial indicator, a CMM, or a flatness gauge all give different numbers on the same part. If you need a CMM report with the shipment, write it into the PO.
The cost of tightening tolerance
Tolerance is not free, and the cost is not linear. Going from as-extruded to machined base is a moderate step. Going from 0.05 mm to 0.02 mm flatness on a large base is a much bigger one, because it usually requires a second machining pass, stress relief between passes, or a different fixturing approach.
- As-extruded, no machining: lowest cost, suitable for low-power natural convection and for assemblies using thick gap pads.
- Machined base, 0.05 mm: the default for anything above roughly 30 W or anything using grease. This is where most designs should sit.
- Machined base, 0.02 mm plus Ra 0.8: for high power density, bare-metal contact, or cold plate interfaces. Expect a 15-30% price increase over the 0.05 mm version and a longer lead time.
- Lapped or ground contact face: used for direct-die or very high heat flux interfaces. Rare in extrusion-based heat sinks and usually a separate process step.
Why extrusion tolerance drifts and what to do about it
An extrusion die wears at the bearing, so the profile grows. A fin specified at 1.0 mm may come off the press at 1.0 mm when the die is new and 1.08 mm after two tonnes. Overall width drifts the same way. This is normal and it is why as-extruded tolerances are stated in tenths of a millimetre, not hundredths.
Two consequences for buyers:
First, if fin thickness is thermally critical, do not rely on the extrusion to hold it. Specify the thermal resistance you need and let the supplier manage fin thickness within the as-extruded band, or move to skiving where the fin is cut rather than formed.
Second, ask for a first-article report and a sample from the actual production run, not from the die trial. Die-trial samples are always on the tight side of tolerance and they are not representative of what you will receive in month eight.
Frequently asked questions
Can I get +/-0.02 mm on an extruded heat sink?
Not on the extruded surfaces. You can get +/-0.02 mm on machined features, and 0.02 mm flatness on a machined base, but fin thickness and pitch remain as-extruded characteristics. If you need tight fins, the process is skiving or forging, not extrusion.
Does anodising change the dimensions?
Yes, by roughly half the coating thickness per surface: a 15 micron anodic layer adds about 7-8 micron to each dimension. This is usually inside tolerance but it matters on tight callouts and on threads, which is why threaded holes are often masked or re-tapped after anodising.
What flatness should I specify for a CPU or IGBT interface?
0.05 mm over the contact zone is the working standard for grease or phase change material. Go to 0.02 mm only if you are running high heat flux or bare-metal contact with high clamping force.
How is flatness actually checked on a production line?
Usually a surface plate with a dial indicator across the contact zone, or a dedicated flatness gauge for high volume. A CMM gives a full map but is slower. If flatness is a critical characteristic for you, agree the gauge and the sampling frequency in writing.
Send your drawing and we will tell you which callouts are driving cost and which ones you can relax without losing thermal performance.
Heat sink sourcing guides
- Aluminum Heat Sink Price per kg: 2026 Cost Breakdown
- Custom Heat Sink MOQ and Lead Time by Process
- Heat Sink Extrusion Die Cost and Tooling Amortisation
- Heat Sink RFQ: What to Send for an Accurate Quote
- Heat Sink Surface Finish Options and Cost Impact
- Heat Sink RoHS, REACH and UL Compliance Explained
- Heat Sink Volume Pricing: How Breakpoints Really Work
- Heat Sink Prototype and Sample Service: What to Expect
- Heat Sink Packaging, Shipping and DDP Terms
- Heat Sink Supplier Audit and Second Sourcing
- Heat Sink Thermal Test Report: What the Data Package Contains
