Flatness and Surface Specs for Heat Sink Interfaces
Short answer: For most electronics cooling applications, specify heat sink base flatness at 0.05 mm per 25 mm of length (or 0.025 mm total for small bases), surface roughness between Ra 0.4 and Ra 1.6 µm, and a machined finish free of burrs, scratches, or anodize on the mating face. A flat, clean interface lets thermal interface material (TIM) compress to a 25–75 µm bond line instead of pooling in gaps. Tighter specs — down to 0.01 mm flatness and Ra 0.4 µm — are worth it only above roughly 50 W per device or when bond-line resistance dominates your thermal budget. BQUQ machines heat sink bases to ±0.005 mm on CNC equipment in Dongguan and returns quotes in 12 working hours.
Flatness is the quiet variable in every thermal stack-up. Engineers spend weeks simulating fin geometry and airflow, then lose 15% of the predicted performance because the base bows 0.15 mm in the middle and the TIM never fully wets the contact patch. This article covers how to call out flatness and surface finish on a heat sink drawing, what the numbers actually mean at the interface, and where over-specification quietly adds cost without adding cooling.
Why does heat sink base flatness matter more than fin design at the interface?
The fin stack determines how efficiently heat leaves the sink. The base-to-component interface determines how much heat gets into the sink in the first place. If the interface is poor, no amount of fin area compensates — the heat simply never arrives.
A real contact interface between two "flat" metal surfaces is a scatter of microscopic peaks touching. Air fills the valleys, and air conducts heat roughly 8,000 times worse than aluminum. TIM exists to displace that air. But TIM only works if the gap it has to bridge is thin and uniform. A warped base creates a wedge-shaped or domed gap, and the TIM flows to the thickest region, leaving the center starved.
The practical consequences:
- Higher bond-line resistance. Thermal resistance through a TIM layer scales linearly with thickness. Doubling bond line from 50 µm to 100 µm roughly doubles that layer's contribution.
- Hot spots. Uneven pressure means some die regions run hot while others stay cool, which matters for IGBT modules and multi-die LED arrays where junction temperature is set by the worst spot.
- Pump-out and dry-out. Thick TIM layers are more vulnerable to thermal cycling pump-out, where the paste migrates away from the hot zone over thousands of cycles.
That is why flatness is usually the first tolerance a thermal engineer tightens — and the first one a purchasing team should question when the cost delta appears.
How do you specify flatness correctly on a heat sink drawing?
Flatness is a form tolerance. It controls a surface relative to itself, with no datum reference required. That makes it cheap to inspect and unambiguous to machine — but only if you apply it to the right face.
Flatness vs. parallelism vs. profile
| Callout | Controls | Datum needed? | Typical heat sink use |
|---|---|---|---|
| Flatness | Surface deviation from a plane | No | Base mating face |
| Parallelism | Base face relative to fin tips or mounting face | Yes | Stack-up with a fixed clamp plane |
| Profile of a surface | Full 3D form against a theoretical surface | Yes | Complex cold plates, vapor chamber lids |
| Total runout | Combined form + location during rotation | Yes | Rare on heat sinks |
For a standard extruded or machined heat sink, flatness on the base is almost always the correct control. Parallelism is only needed when the fin tips or a secondary mounting surface must sit at a fixed distance from the base — for example, a heat sink that clamps between a PCB and a chassis rail.
Where to place the tolerance zone
Apply the flatness callout to the contact area, not the entire base. A 100 × 100 mm base with a 40 × 40 mm die footprint does not need the outer 30 mm ring held to the same tolerance. Restricting the tolerance zone to the functional contact patch typically cuts machining time and scrap.
Write it as: flatness 0.05, applied to the 45 × 45 mm central zone, with the remainder of the base held to a looser general tolerance.
Units and reporting
Metric drawings in millimeters are standard for thermal hardware. Require inspection reports in the same units, and specify the measurement method — a coordinate measuring machine (CMM) grid, a surface plate with a dial indicator, or optical profilometry. Different methods give different numbers on the same part, so agreeing on the method up front prevents disputes at incoming inspection.
What flatness values are realistic for each manufacturing route?
Not every process can hit every tolerance. The table below shows what is achievable in volume production, not what a lab can demonstrate once.
| Process | Typical as-produced flatness | Achievable with secondary machining | Relative cost |
|---|---|---|---|
| Aluminum extrusion (as-cut) | 0.15–0.40 mm | 0.03–0.05 mm after face milling | Low |
| Die casting (as-cast) | 0.20–0.50 mm | 0.05 mm after milling | Low–medium |
| Forged aluminum | 0.10–0.25 mm | 0.02–0.05 mm | Medium |
| Skived fin (machined base) | 0.05–0.10 mm | 0.01–0.03 mm | Medium–high |
| CNC machined from billet | 0.02–0.05 mm | 0.005–0.01 mm | High |
| Vapor chamber (copper) | 0.05–0.10 mm | 0.03 mm typical | High |
Two things drive the spread. First, residual stress: extruded and cast parts move after machining as internal stresses relieve, so a base milled flat on Monday may bow by Tuesday. Second, fixturing: thin bases deflect under clamping pressure, so the flatness you measure on the machine is not the flatness in the free state.
For heat sinks produced on our extruded heat sink lines, we typically mill the base after extrusion and stress-relieve where the profile allows. For tighter work, CNC machined heat sinks from billet hold ±0.005 mm on the base face and are the right choice when the thermal budget is tight or the base carries mounting features.
When does 0.05 mm stop being enough?
Roughly, when the bond-line contribution approaches the spreading resistance. As a rule of thumb, target a maximum gap no greater than three times your intended bond line. If you want a 50 µm bond line, a 0.15 mm flatness deviation already exceeds that in the worst spot.
High-power IGBT modules, GPU cold plates, and dense LED arrays are the usual cases where 0.025 mm or tighter is justified. Consumer-grade LED lighting and low-power converters rarely benefit.
How does surface roughness interact with flatness and TIM?
Flatness controls the macro gap. Roughness controls the micro gap — the valleys between machining marks that TIM must also fill.
| Surface finish | Typical Ra | TIM wetting behavior | Best suited to |
|---|---|---|---|
| Fine ground | 0.2–0.4 µm | Excellent, minimal TIM needed | High-power, thin bond line |
| Precision milled | 0.4–0.8 µm | Very good | Most power electronics |
| Standard milled | 0.8–1.6 µm | Good with paste TIM | General electronics |
| As-extruded / cast | 2.0–6.0 µm | Requires thicker TIM or pad | Low-power, cost-driven |
| Anodized (matte) | 1.5–3.0 µm + oxide | Oxide adds resistance | Aesthetic or dielectric needs |
Two counterintuitive points matter here.
Smoother is not always better. Below roughly Ra 0.2 µm, some greases and phase-change materials struggle to wet the surface, and very smooth faces can promote pump-out because there is no texture to anchor the material. Ra 0.4–0.8 µm is a sweet spot for paste TIMs.
Anodizing is a thermal insulator. The aluminum oxide layer is electrically insulating and thermally resistive. If you anodize for corrosion resistance, mask the mating face or specify a machined finish afterward. Where dielectric isolation is genuinely required, a thin anodized layer can replace an insulating pad — but budget for the added resistance and say so on the drawing.
Burrs, chips, and contamination
A 0.05 mm flatness number means nothing if a 0.1 mm burr sits on the contact edge. Specify deburring on all machined edges, require cleaning after machining, and state that the mating face must be free of coolant residue, chips, and handling oils. Incoming inspection should check the face under magnification before thermal testing, not after.
What about mounting pressure and hardware?
Flatness is a specification; clamping is what actually closes the gap. A perfectly flat base with uneven screw loading will still produce a wedge-shaped bond line.
- Four-point mounting on a rectangular base distributes pressure better than two-point.
- Spring-loaded or shoulder screws with controlled compression beat fully torqued screws into a soft base.
- Specify a torque range, not a single value, and note whether the base is stiff enough to avoid bowing under that torque.
- For thin bases, add stiffening ribs or increase base thickness rather than chasing tighter flatness — a 3 mm base that deflects 0.1 mm under clamp load will not hold 0.02 mm in service.
This is where a thermal spec sheet pays for itself. Documenting flatness, roughness, mounting torque, and TIM type together lets a supplier optimize the whole interface rather than one number in isolation.
How does machining route affect achievable specs and cost?
The cheapest way to hit a flatness number is to start with a process that is already close. If an extrusion comes off the line at 0.25 mm, you are paying to remove 0.2 mm of material across the whole base. If a forged or billet part starts at 0.05 mm, you are paying for a finishing pass only.
Our CNC milling guide for heat sinks covers the practical side: fixture design, face-milling parameters, and how to avoid the classic mistake of measuring flatness while the part is still clamped. The short version is that face milling with a fine feed and a sharp insert produces both the flatness and the surface finish in one operation, and that trying to save time by skipping a finishing pass usually costs more in scrap than it saves in cycle time.
Where cost pressure is real, the honest conversation is about relaxing flatness on parts that do not need it. A cost reduction review on a mature heat sink design usually finds two or three tolerances that were copied from an earlier, higher-power product and never revisited.
Frequently Asked Questions
Q: What flatness should I specify for a standard aluminum heat sink?
A: For most electronics cooling, 0.05 mm over the contact zone is a sound default — tight enough for a 50 µm bond line with paste TIM, loose enough to machine economically in volume. Go to 0.025 mm or tighter only for high-power devices where bond-line resistance is a meaningful share of the total thermal budget, and relax toward 0.10 mm for low-power applications using thermal pads.
Q: Does a smoother surface always improve thermal performance?
A: No. Below roughly Ra 0.2 µm, many thermal greases and phase-change materials wet poorly and become more prone to pump-out because there is no surface texture to anchor them. Ra 0.4–0.8 µm is the practical sweet spot for paste TIMs. Going smoother adds machining cost and can actually degrade long-term stability under thermal cycling.
Q: Can I anodize a heat sink and still get good thermal contact?
A: Only if you mask the mating face or machine it after anodizing. Aluminum oxide is thermally resistive and adds a measurable layer to the interface. Black anodizing on fins is fine and often desirable for radiation. On the base, specify a machined finish and note it explicitly on the drawing so the finishing vendor does not coat the contact zone.
Q: How do I inspect flatness on incoming heat sinks?
A: Agree on the method before production. A surface plate with a dial indicator works for coarse checks; a CMM grid or optical profilometer is needed below 0.05 mm. Measure the part in its free, unclamped state, at controlled temperature, and check the contact zone rather than the full base. Record the method on the drawing so supplier and buyer measure the same way.
Q: Does flatness matter for heat sinks bonded with thermal adhesive?
A: Yes, but less. Adhesives and gap fillers tolerate thicker bond lines — often 100–200 µm — so a 0.10 mm flatness deviation is usually acceptable. The trade-off is that adhesive bond lines add more thermal resistance than a thin paste layer. If you are bonding, specify flatness around 0.10 mm and focus instead on adhesive thickness control and cure schedule.
Related Resources
- About BQUQ — ISO9001 factory in Dongguan with four production lines under one roof
- Heat sinks — extruded, skived, bonded-fin, and CNC machined options
- CNC machined heat sinks — ±0.005 mm base flatness and machined interfaces
- Extruded heat sinks — cost-effective profiles with milled bases
- Technical articles — thermal, machining, and stamping engineering guides
- Industry trends — what is changing in electronics cooling sourcing
- Contact — send drawings for a quote in 12 working hours
Authored by the BQUQ Engineering Team. BQUQ (Dongguan) runs CNC machining (±0.005 mm), metal stamping, custom springs, and heat sink production in one ISO9001 factory. Source-direct from Dongguan, China — quote in 12 hours: sc@bquq.com | WhatsApp +86 13713157787 | www.bquq.com
