Thermal Vias and Copper Area: Keeping Power Parts Cool

Thermal Vias and Copper Area: Keeping Power Parts Cool

Quick answer

Heat leaves a power device through a chain: die → package → thermal pad → copper → air. The copper and the vias are part of the thermal resistance, not an accessory. A 0.3 mm plated via through a 1.6 mm board is roughly 190 K/W on its own; a 3 × 3 array brings that to about 21 K/W. Copper area then spreads the heat into the board: a few square centimetres can move θJA from 60 K/W to 40 K/W, which is often the difference between a 110 °C and a 150 °C junction.

Executive takeaways

  • θJA is a board property, not a package constant. The same package can be 40 K/W on 4-layer copper and 80 K/W on a two-layer board with no pour.
  • Vias are resistors in parallel. Nine 0.3 mm vias are about 21 K/W; four are about 48 K/W. The count matters more than the diameter.
  • Copper area has diminishing returns. The first square centimetre does the most; beyond roughly 6–9 cm² the improvement is small and the board area is better spent elsewhere.
  • Solder voids are thermal resistance. A 20% void under a thermal pad can raise the junction temperature by 10–20 °C at 2 W.

1. The heat path and the model

Thermal design is a resistor network: power flows from the junction to the ambient through a series of thermal resistances. The junction-to-case resistance is in the datasheet; the case-to-board and board-to-ambient resistances depend on your layout. That is why θJA is quoted “on a JEDEC board” and why the number changes the moment you use a different copper area, airflow or enclosure; treat it as a starting point, never as a guarantee.

The model is simple enough to use on a napkin: TJ = TA + P × θJA. For a 2 W device in a 50 °C enclosure, θJA of 30 K/W gives 110 °C, and 60 K/W gives 170 °C — the difference between a working design and a failed one.

2. How much does a via carry?

A plated through-via is a copper tube. Its thermal resistance is R = L / (k × A), with L the board thickness, k the copper conductivity (about 385 W/m·K) and A the plating cross-section. For a 0.3 mm drill, 25 µm plating and a 1.6 mm board, the annulus is about 0.0216 mm², which gives roughly 190 K/W per via.

Via array (0.3 mm, 25 µm, 1.6 mm board) Approximate R Comment
1 via 190 K/W Barely useful alone
4 vias 48 K/W Small power pads
9 vias (3 × 3) 21 K/W Typical thermal pad
16 vias (4 × 4) 12 K/W High-power pads
Thermal pad with a 3 by 3 via array connecting to inner copper planes
Vias are parallel thermal resistors: nine of them turn 190 K/W into about 21 K/W, but only if they reach copper on both sides.

Placement matters as much as count. Put the vias directly under the thermal pad, with a pitch of about 1.0–1.2 mm; a ring of vias around the pad does far less than an array inside it. The vias should land on solid copper on the other side, ideally a ground plane, and that plane should extend beyond the pad so the heat can spread.

If the pad is large, a 4 × 4 array is a better starting point than nine vias spread over the same area: heat travels laterally through the pad, and the centre of the pad is the hottest point. Keep the outermost vias inside the pad edge by at least one drill diameter so the solder does not wick away.

Two details change the number. Plating thickness matters more than drill size: 25 µm is standard, 35 µm is better if the board house offers it. And the via must connect to copper on both sides; a via into a plane that is isolated by a thermal relief does much less than the table suggests.

3. Copper area and spreading

Once the heat reaches the board, copper spreads it laterally. The spreading resistance falls as area grows, but with diminishing returns: most of the benefit comes from the first few square centimetres. A practical table for natural convection on 1 oz copper (35 µm) looks like this:

Copper area (one side) Approximate θJA Notes
0.25 cm² 70–80 K/W Minimum pad, no pour
1 cm² 55–65 K/W Small pour
4 cm² 45–50 K/W Typical
9 cm² 38–42 K/W Diminishing returns
9 cm² + 9 vias to a plane 30–35 K/W Best simple case

Adding a ground plane on an inner layer is often more effective than adding copper on the surface: the plane spreads heat with less interruption from traces and components, and it is connected through the vias under the pad. On a four-layer board the plane is usually the second layer, directly under the component side.

The table assumes natural convection and a board in still air. Forced air or a metal enclosure changes the numbers; a sealed plastic box makes them worse. Use the table as a starting point and measure the real board.

4. The full resistance network

The junction sees more than the vias. A typical chain for a power device on a thermal pad looks like this: junction-to-case 5 K/W, case-to-solder 1 K/W, solder-to-board 2 K/W, board-to-ambient 20–40 K/W. The last term dominates, which is why the layout matters more than the package.

Element Typical value Controlled by
Junction to case 2–10 K/W Package and die
Case to solder 0.5–2 K/W Paste, voiding
Solder to board 1–3 K/W Pad, vias
Board to ambient 20–60 K/W Copper area, airflow, enclosure

The lesson is arithmetic: improving a 2 K/W term is nice, but improving a 40 K/W term is what changes the junction temperature. Spend the effort on copper and vias, not on a better thermal paste.

5. Worked example: a 2 W regulator in a 50 °C box

Take a switching regulator dissipating 2 W in a 50 °C enclosure, with a junction limit of 125 °C and a 15 °C margin target, so the design target is 110 °C.

  • Allowed θJA: (110 − 50) / 2 = 30 K/W.
  • Copper only: 9 cm² of 1 oz copper gives about 40 K/W — 130 °C junction, too hot.
  • Copper plus vias: 9 cm² with a 3 × 3 via array to a ground plane brings the board-to-ambient path down; the combined θJA lands near 30–35 K/W, which meets the target with margin.
  • If vias are impossible: the copper area must grow, or the power must fall, or the enclosure must be ventilated.

If the part is in a QFN instead of a DPAK, the same 2 W may need a 4 × 4 array and a larger plane, because the pad is the only path. If it is in a SOT-23, the leads carry most of the heat and the board area around the leads matters more than vias under the body.

The example shows the real decision: vias are usually cheaper than board area. A 3 × 3 array costs a few square millimetres; 4 cm² of extra copper costs board space and routing freedom.

6. Solder, voids and the interface

The thermal pad is only as good as its solder joint. Voids under a thermal pad are insulating pockets; at 2 W, a 20% void can add 10–20 °C to the junction. Two rules follow: use a stencil aperture that covers the pad without excessive paste, and check the reflow profile so the paste wets the pad instead of balling at the edges.

For large pads, a grid of paste deposits can reduce voiding compared with one solid block; the exact pattern depends on the stencil and the paste. The check is an X-ray of the first articles, not a guess.

7. Layout rules that actually move the number

  • Connect the thermal pad to a plane, not to an isolated pour. A plane on an inner layer spreads heat across the whole board.
  • Use solid connections for thermal pads; thermal reliefs are for hand soldering, not for power dissipation.
  • Place vias inside the pad with a pitch of about 1.0–1.2 mm; tent or plug them if solder wicking is a risk.
  • Use both sides: a copper pour on the bottom connected by vias roughly halves the spreading resistance.
  • Keep other heat sources away. Two 1 W parts next to each other share an ambient that is hotter than the box.

8. Who should not use thermal vias

  • High-voltage isolation barriers. Vias through the isolation gap reduce creepage and clearance; use an insulating interface instead.
  • Flex or thin boards. The via barrel and the copper area behave differently; thermal vias may crack under flexing.
  • Boards with wave solder. Open vias in a thermal pad wick solder and can starve the joint; plug or tent them.
  • Very low power. Below about 0.2 W the copper pad alone is enough; vias add cost without benefit.

9. Enclosure and airflow

Natural convection gives a heat-transfer coefficient of roughly 5–10 W/m²·K; forced air reaches 25–100 W/m²·K. That factor of five to ten is why the same board runs much cooler with a fan, and why a sealed plastic box is the worst case. A metal enclosure helps only if the board is thermally coupled to it; otherwise the air inside the box heats up and the ambient term rises.

Estimate the enclosure rise before choosing the copper. A 2 W board in a 0.5 L sealed plastic box can raise the internal air by 15–25 °C above room temperature; that rise is added to the ambient in every calculation. Ventilation slots, a metal chassis or a thermal pad to the case are the usual fixes.

10. Package pads: what each one gives you

The package decides how much of the heat can reach the board. A SOT-23 has no thermal pad and most of its heat leaves through the leads; a SOIC-8 with an exposed pad, a DPAK or a QFN has a dedicated pad that can be soldered to copper and vias. That is why the same die in a QFN can run 20–30 °C cooler than in a small leaded package.

For leaded packages without a pad, the lead frame is the path: wide, short traces from the leads to copper help, and vias under the body do nothing. For QFNs, the pad is the path and the via array under it is the design. Match the effort to the package, not to a generic checklist.

11. Measuring the real junction temperature

A thermocouple glued to the package case gives the case temperature, not the junction. Convert with the datasheet θJC: TJ ≈ TC + P × θJC. A thermal camera gives the same case number but faster, and it shows whether the copper is spreading heat evenly.

Run the board at full load until the temperature stabilises — five to ten minutes is typical for a small board — and measure at the maximum ambient you expect. A measurement at 25 °C in open air tells you very little about the 50 °C enclosure.

12. Five mistakes

1. Using the datasheet θJA as if it were your board. The JEDEC board is not your layout.

2. One or two vias under a power pad. At 190 K/W each, that is not a thermal path.

3. Vias into an isolated pour. The heat has nowhere to spread.

4. Thermal reliefs on a thermal pad. The spokes are a bottleneck.

5. Ignoring voiding. The first prototype works; the production board with a different stencil does not. X-ray the first articles and check the void percentage, not just the presence of solder.

13. Checklist

  • Power dissipation and maximum ambient, including enclosure rise.
  • Junction limit and the margin you need.
  • Allowed θJA from TJ = TA + P × θJA.
  • Copper area on both sides and connection to a plane.
  • Via count, drill, plating and pitch.
  • Via connection to planes on inner layers.
  • Solder paste pattern and expected voiding.
  • Neighbouring heat sources.
  • Isolation, flex or wave-solder constraints.
  • Verification: thermocouple or thermal camera on the real board at full load.

FAQ

How do I know the copper is doing anything?

Measure the case temperature with and without the pour, or use a thermal camera to see the isotherms. If the copper is not spreading heat, the gradient will be steep around the package.

Do I need vias if the pad is on the bottom copper?

No, if the device is mounted on the bottom and the copper is exposed to air. The via array is for moving heat from a top-side pad to inner or bottom copper.

Is a thermal camera accurate enough?

For relative comparison and for finding hot spots, yes. For absolute junction temperature, use the case temperature and θJC, and account for the camera's emissivity setting.

Should I use thermal paste or a pad between the board and the enclosure?

If the board touches the enclosure, a gap filler or pad helps; if it does not, the interface does nothing. Mechanical contact and pressure matter as much as the material.

How many vias do I need under a thermal pad?

A 3 × 3 array is a common starting point for a few watts; 4 × 4 for higher power. The count matters more than the diameter, and the vias must reach copper planes on both sides.

Is 2 oz copper worth it?

For high current or high dissipation, yes: the spreading resistance falls roughly with thickness. For a 2 W part with a good via array, 1 oz is usually enough.

Can I use the pad as a heatsink without vias?

Only for low power. Without vias the heat stays on the top copper; the board's other layers do not help.

Why is my junction temperature higher than calculated?

The usual causes are voids, an isolated pour, thermal reliefs, or a hotter local ambient than assumed. Measure the board, not the datasheet.

Do thermal vias reduce electrical isolation?

Yes, if they cross an isolation barrier. Keep vias out of the barrier or use an insulating interface.

Glossary

θJA
Junction-to-ambient thermal resistance, in K/W.
Thermal via
A plated through-hole used to conduct heat between copper layers.
Spreading resistance
The resistance to heat flowing sideways through copper.
Void
An air pocket in a solder joint; a thermal insulator.
Thermal relief
Spokes that slow heat flow, used to make hand soldering easier.

Sources

Designing a power stage that has to stay cool?

Browse our power management and MOSFET lines, or send the dissipation and enclosure conditions for a recommendation.

Browse In-Stock Components  ·  Submit Your BOM

Check stock by part number

Turn this insight into a sourcing decision.

Need pricing or a drop-in alternative? Submit a BOM / RFQ or talk to our team.