Liquid Cold Plates: Channel Design and Materials
Short answer: a liquid cold plate is a metal heat exchanger with internal channels carrying coolant, and a well-built copper plate for a 100×100 mm IGBT or CPU-class load typically lands at 0.02–0.05 °C/W at 1–2 L/min of water — roughly five to twenty times better than a good air-cooled heat sink. The two decisions that set performance are channel architecture (straight, serpentine or microchannel) and material (copper versus aluminum). Copper wins on conductivity at roughly 3–4× the material cost; aluminum wins on price and weight but needs corrosion control. Plan for a pressure and leak test at 5–10 bar on every production plate.
When air cooling runs out of headroom, the answer is usually not a bigger fan but a liquid loop. Cold plates appear inside inverters, laser drivers, EV battery systems, medical equipment and high-density computing, always doing the same job: collect heat from one or more components into a coolant stream that carries it to a remote radiator. Because the liquid side is so much better at carrying heat than air, the cold plate does not need heroic surface area — it needs well-designed channels, the right metal, and honest manufacturing that does not leak.
The Core Trade: Channel Geometry Sets Both Performance and Pressure Drop
Channel design is a straight exchange: more surface and more turbulence give better heat transfer but cost pressure drop, which your pump must pay for. Three architectures cover most designs. Straight parallel channels give the lowest pressure drop and the most even flow when the source covers the whole plate, which suits large battery or busbar surfaces. Serpentine channels force the coolant to cross the hot zone repeatedly in one long path, giving high heat transfer under a concentrated source with a moderate pressure penalty. Pin-fin and microchannel arrays give the highest heat transfer per area but the largest pressure drop, and microchannels below about 0.5 mm width demand clean coolant and filtration to avoid clogging.
| Channel type | Heat transfer | Pressure drop | Best fit | Typical coolant path |
|---|---|---|---|---|
| Straight parallel | Medium | Low | Large-area sources, batteries | Short, even |
| Serpentine tube | Medium-high | Medium | Single concentrated die or module | Long, single path |
| Serpentine milled channel | High | Medium-high | IGBT modules, laser diodes | Long, wide channel |
| Pin-fin array | High | High | High-flux spots, compact plates | Short, turbulent |
| Microchannel | Highest | Highest | Research and very high flux | Very narrow, filtered |
There is no universal best geometry, only a best fit for your heat map. If the heat comes from one 20×20 mm die in the middle of the plate, a serpentine path crossing that zone is effective and easy to manufacture. If heat is spread evenly across a large surface, parallel channels deliver more total cooling for the same pump power. If two different devices sit on one plate at different powers, the flow can be split with channel widths sized per zone — a trick machined plates do naturally and tube plates cannot.
Copper Versus Aluminum: The Material Decision
Copper carries roughly 385–401 W/m·K against roughly 167–180 W/m·K for 6061 or 6063 aluminum, so a copper plate spreads heat faster and runs a smaller temperature gradient between the device and the coolant. That conductivity advantage is worth most when the heat source is concentrated, because spreading resistance dominates. Aluminum becomes attractive when the source is spread out, weight matters, or cost dominates, and when the loop uses a corrosion-inhibited coolant such as the ethylene-glycol mixtures common in automotive and industrial systems.
The corrosion question is the one that trips up new designs. Copper and aluminum should not be mixed in one wetted loop without corrosion control: aluminum is anodic to copper and will corrode preferentially in plain water. Automotive systems therefore standardize on aluminum everywhere with inhibited coolant, while laboratory, medical and many industrial loops standardize on copper or stainless. Decide the whole loop material family before you design the plate, not after.
| Material | Thermal conductivity | Typical cost vs aluminum | Weight | Loop compatibility |
|---|---|---|---|---|
| Copper (C1100/C1020) | ~385–401 W/m·K | 3–4× | Heavy | Copper or stainless loops |
| Aluminum 6061 | ~167 W/m·K | Baseline | Light | All-aluminum loops with inhibited coolant |
| Aluminum 6063 | ~200 W/m·K | Baseline | Light | All-aluminum loops, easier extrusion |
| Stainless 304 | ~15 W/m·K | 2–3× | Heavy | Corrosive fluids, but poor conductor — usually tubing only |
Cost figures are indicative material multipliers, not finished-plate prices. Stainless appears in the table mostly as a caution: it is corrosion-proof but such a poor conductor that solid stainless plates only work with microchannels or thin wall sections, which is why stainless appears as tubing and fittings in practice rather than as the plate body.
How Cold Plates Are Actually Manufactured
Manufacturing method is a design decision, because it decides what channel shapes are possible and what the leak risk is. A machined plate is the workhorse: a CNC mill cuts the channel field into a solid block, then a flat cover is bonded or brazed over the top. This route handles serpentine, parallel and pin-fin shapes, holds tight channel dimensions, and suits low-to-mid volumes. A stamped plate starts from sheet metal: half-shells with the flow pattern formed by a progressive stamping die are joined into a sealed plate, which is how many automotive battery cooling plates are made at high volume and low per-part cost. Tube-in-plate is the oldest route: a bent copper or aluminum tube is embedded in a cast or machined body, simple and leak-tolerant but limited in heat transfer because the coolant never touches the component side directly.
Sealing and testing decide whether any of these survive in service. Covers are typically vacuum-brazed for automotive aluminum, soldered or bonded for copper, and the finished plate must pass a pressure and leak test — commonly 5–10 bar air under water, pressure-decay testing, or helium leak detection for automotive and medical grades. Ask any supplier three questions: what leak test every plate receives, what burst pressure the design targets, and what the flow rate versus pressure-drop curve looks like at your operating point. A plate that passes visual inspection but skips leak testing will find its defect in your customer's rack.
| Manufacturing route | Channel flexibility | Typical volume | Leak risk | Indicative relative cost |
|---|---|---|---|---|
| CNC machined + brazed cover | Highest | 1–5,000 pcs | Low with testing | High per part |
| Tube-in-plate (bent tube) | Low | 100–50,000 pcs | Lowest | Medium |
| Stamped half-shells joined | Medium | 10,000+ pcs | Medium | Low per part |
| Vacuum-brazed aluminum | Medium-high | 10,000+ pcs | Low | Medium-high |
Volumes and costs are indicative and overlap heavily in practice. The practical takeaway: for prototypes and production runs up to a few thousand pieces, a CNC-machined plate with a brazed or bonded cover is the fastest route to a validated design, and a source factory that runs both CNC machining and stamping lines can move you between routes as volume grows instead of forcing a tooling bet on day one.
Sizing a Cold Plate: What the Numbers Mean
The performance number buyers should quote and compare is thermal resistance in °C/W from the component mounting surface to the coolant. A rough budget for a 100×100 mm class plate is 0.02–0.05 °C/W with water at 1–2 L/min, which means a 500 W IGBT module runs roughly 10–25 °C above coolant temperature. Compare that with the 0.1–1.0 °C/W territory of air-cooled sinks and you can see why liquid loops exist: they move the same heat for a fraction of the temperature rise, at the price of a pump, radiator, fittings and the leak risk.
Resistance splits into three parts you can attack separately: the spreading resistance inside the plate material between the device footprint and the channels (cut by using copper or a thicker base), the convective resistance between channel wall and coolant (cut by more channel surface or turbulence), and the bulk coolant temperature rise along the flow path (cut by higher flow rate). When a cold plate underperforms, diagnose which of the three dominates before changing geometry. A quick flow test with a thermometer at inlet and outlet tells you the coolant temperature rise directly — if outlet temperature is already close to the device temperature, the loop flow is the problem, not the plate.
Frequently Asked Questions
Q: What is a realistic thermal resistance for a liquid cold plate?
A: A well-made copper plate in the 100×100 mm class with water at 1–2 L/min typically achieves 0.02–0.05 °C/W from mounting surface to coolant, which is roughly five to twenty times better than a good air-cooled heat sink. Exact numbers depend on source size, channel design and flow rate, so validate with a test.
Q: Copper or aluminum for a cold plate — which should I choose?
A: Choose copper when the heat source is concentrated and you need minimum temperature, at roughly 3–4× the material cost of aluminum. Choose aluminum when weight, cost or an all-aluminum automotive loop matters more. Never mix copper and aluminum in one wetted loop without corrosion-inhibited coolant.
Q: What pressure and leak test should I specify for cold plates?
A: Specify a 100% production leak test, commonly air-under-water or pressure-decay at 5–10 bar, with helium leak detection for automotive or medical grades. Also request the flow-versus-pressure-drop curve and a burst-pressure figure so the plate matches your pump and your safety margin.
Q: Can cold plates be made in low volume without expensive tooling?
A: Yes. CNC-machined plates with brazed or bonded covers need no dedicated tooling and suit runs from one prototype to a few thousand pieces. Stamped half-shell plates only become economical with dedicated dies, typically at 10,000+ pieces, which is when you migrate the design.
Q: How do I know if I need a cold plate at all?
A: Work the thermal budget first: if the junction temperature stays acceptable with an air-cooled heat sink at your ambient and reliability targets, stay with air. Liquid earns its complexity above roughly 300–500 W in a compact area, in sealed enclosures where air cannot flow, or where noise and air temperature rules block forced convection.
Related Resources
- Air cooling vs liquid cooling in electronics — the loop-level comparison that decides whether a cold plate is worth it.
- Aluminum vs copper heat sinks — the same material trade applied to conventional finned sinks.
- About BQUQ: an ISO9001 source factory in Dongguan machining and stamping cold plate components under one roof.
- Contact us: send your thermal requirements and drawings for a quote within 12 working hours.
Authored by the BQUQ Engineering Team. BQUQ is an ISO9001-certified source factory in Dongguan, China, running CNC machining, metal stamping, custom springs, heat sink and collet lines under one roof. Send drawings to sc@bquq.com or WhatsApp +86 13713157787 for a quote within 12 working hours. www.bquq.com
