When a discrete semiconductor hits its thermal limit, the silicon does not melt. The package does. The solder joint liquefies, the wire bonds lift off, and the device dies a slow, expensive death inside your enclosure. For power discrete components — MOSFETs, IGBTs, high-current diodes — the heatsink is not an accessory. It is an extension of the package itself.
The mechanical and thermal interface between the package tab and the heatsink determines whether the device survives 100,000 power cycles or fails after 1,000. This specification guide covers the structural rules that govern heatsink attachment for discrete semiconductors, from the mounting interface to the thermal compound selection.
The metal tab on a TO-220, TO-247, or D2PAK package is the primary heat extraction path. But it is also the mechanical anchor. The two functions fight each other constantly.
A tab that is perfectly flat transfers heat efficiently but creates zero mechanical clamping force when bolted down. A tab that is slightly domed clamps better but creates an air gap at the center that kills thermal performance. The specification sweet spot is a flatness tolerance of 0.05 mm to 0.1 mm across the mounting surface. Any deviation beyond that requires a thermal interface material thick enough to fill the gap, which adds thermal resistance.
For packages with isolated tabs (where the tab is electrically connected to the drain or collector), the mounting hardware must account for the voltage potential. If the heatsink is grounded, you need an insulator. If the heatsink floats, you need a dielectric washer rated for the full working voltage plus transients.
The hole pattern on the tab must match the heatsink mounting holes within 0.2 mm. Misalignment forces the bolt to bend, creating uneven clamping pressure. One corner gets crushed while the opposite corner lifts, creating a hot spot that accelerates die attach fatigue.
Most engineers treat the mica washer or silicone pad as a throwaway part. It is not. It is a precision thermal and electrical component.
Mica washers are the gold standard for high-voltage applications above 600 volts. They have a dielectric strength of 100 to 200 kV/mm and do not compress significantly under load, maintaining a consistent gap. The problem is that mica is brittle. If you over-torque the mounting screw, the washer shatters, creating a shard of conductive debris inside the package. This shard can bridge the tab to the heatsink and destroy the device instantly.
Silicone pads with ceramic filler are the modern alternative. They are compliant, which means they absorb mechanical stress and maintain contact even if the tab is slightly warped. But their dielectric strength is lower — typically 5 to 10 kV/mm. For high-voltage SiC devices, a silicone pad is risky unless it is specifically rated for the application.
The washer must cover the entire tab with at least 1 mm of margin on all sides. A washer that is too small concentrates clamping pressure on the tab edges, which can crack the internal ceramic insulator. The thickness tolerance should be plus or minus 0.05 mm. Thinner washers reduce thermal resistance but offer less electrical isolation. Thicker washers insulate better but add a thermal barrier.
The datasheet gives a torque value. Follow it. Exceeding that torque by even 20 percent can crack the package.
The internal structure of a TO-220 or TO-247 package includes a ceramic insulator sandwiched between the metal tab and the lead frame. This ceramic is strong in compression but weak in shear. When you over-torque the screw, you bend the lead frame slightly. This bending creates shear stress on the ceramic insulator. After a few thermal cycles, the ceramic cracks. The crack is invisible from the outside, but it electrically isolates the tab from the die. The device stops working, and you have no idea why.
The correct torque for a TO-220 is typically 0.5 to 1.0 Nm. For a TO-247, it is 1.0 to 2.0 Nm. These values are not arbitrary — they are calculated to provide enough clamping force to flatten the thermal interface material without exceeding the yield strength of the internal ceramic.
Use a calibrated torque driver. A hex key does not give you torque control — it gives you finger pressure control, which is useless.
In automotive or industrial applications, vibration loosens screws. A bolted heatsink that loosens by 0.1 mm loses 30 to 50 percent of its thermal contact. The device overheats, cycles, and fails.
The solution is a spring-loaded mounting system. A belleville washer or a dedicated spring clip maintains constant clamping force regardless of vibration. The spring compensates for thermal expansion and contraction, keeping the interface pressure stable from -40 to 175 degrees Celsius.
The spring force must be calibrated. Too little force and the interface gaps under vibration. Too much force and you crush the thermal pad. The target is a contact pressure of 1 to 2 MPa on the thermal interface material. This is enough to fill surface irregularities without squeezing all the compound out of the joint.
Thermal compound selection is not about picking the highest conductivity number. It is about matching the material to the assembly process and the operating environment.
Thermal paste has the highest conductivity — typically 3 to 8 W/mK. But it is messy, hard to control in volume, and prone to pump-out under thermal cycling. The pump-out effect occurs when the compound gets squeezed out of the joint during expansion and is not sucked back in during contraction. After 500 cycles, the joint is dry.
Thermal pads are pre-cut, clean, and consistent. Their conductivity is lower — 1 to 4 W/mK — but they do not pump out. For high-reliability applications where maintenance is impossible, pads are the safer choice. The thickness must be selected so that under full clamping force, the pad compresses to 50 to 60 percent of its original thickness. This ensures maximum contact area without bottoming out.
Phase change materials (PCM) sit between paste and pad. They are solid at room temperature and melt at operating temperature, flowing into surface voids. They offer the conductivity of paste with the handling of a pad. The downside is cost and long-term stability. PCM can dry out over years of storage before the device is even used.
For high-voltage discrete devices, the thermal compound must be electrically insulating. Standard thermal pastes use metallic silver or zinc oxide fillers, which are electrically conductive. Using a conductive paste on a live tab will short the device to the heatsink.
The specification requires a dielectric strength of at least 5 kV/mm for devices rated above 400 volts. For 1200-volt SiC devices, specify 10 kV/mm or higher. Check the datasheet. Do not assume.
The heatsink itself must be designed for the specific airflow conditions of the application. A heatsink that works in free convection will fail in forced airflow if the fin spacing is wrong.
For natural convection, fin spacing should be 6 to 10 mm. Wider spacing allows hot air to rise freely between fins. For forced airflow, fin spacing can be tighter — 2 to 4 mm — because the fan pushes air through the gaps. But if the spacing is too tight, the airflow creates turbulence and noise, and the pressure drop across the heatsink increases, reducing fan efficiency.
The fin thickness should be 1.5 to 2.0 mm for aluminum extrusions. Thinner fins bend under vibration. Thicker fins add weight without proportional thermal benefit. The base thickness under the mounting tab should be at least 3 mm to prevent bowing when the screw is tightened.
Anodized aluminum heatsinks look nice, but the anodized layer has a thermal conductivity of only 1 to 2 W/mK compared to 200 W/mK for raw aluminum. That anodized layer acts as a thermal insulator.
For high-power discrete devices, specify a raw aluminum finish or a nickel-plated finish. If corrosion resistance is required, use a thin nickel plate (5 to 10 micrometers) rather than anodizing. The nickel plate protects the aluminum without adding significant thermal resistance.
The surface roughness should be 1.6 to 3.2 micrometers Ra. A rough surface improves mechanical grip on the thermal pad, but a surface that is too rough creates air pockets that reduce contact area. A smooth, flat surface with a thin layer of high-quality thermal compound gives the best results.
Qualification testing must be done with the heatsink attached, not just the bare package. The thermal mass of the heatsink changes the cycling profile dramatically. A bare device might cycle from 25 to 150 degrees in 10 seconds. The same device on a large heatsink might take 60 seconds to reach 150 degrees.
The slower ramp rate changes the stress on the solder joint. Longer dwell times at high temperature accelerate intermetallic growth at the die attach interface. The test plan must account for the actual thermal profile the device will see in the application, not just the datasheet limits.
Vibration testing must be performed with the heatsink mounted and torqued to spec. Loose heatsinks resonate at different frequencies than tight ones. A loose heatsink can amplify vibration at the mounting point, creating a fatigue failure at the tab hole that would never occur with proper clamping.
Run random vibration from 20 Hz to 2000 Hz at 10 to 20 G for 30 minutes per axis. After testing, measure the thermal resistance. If it has increased by more than 10 percent, the interface has degraded. Cross-section the joint to check for pump-out or washer cracking.
The heatsink is not a separate component. It is part of the semiconductor package. Design it, mount it, and test it as one system.
Contact: Joanna
Phone: Info@addcomponents.hk
Tel: 852 5334 3091
Email: info@addcomponents.hk
Add: FLAT/RM C -13/F HARVARD ,COMMERCIAL BUILDING 105-111 THOMSON ROAD,WAN CHAI HK