A practical reference for engineers picking pads for power supplies, EV inverters, telecom radios and dense PCB stacks. We cover the spec numbers that actually matter — and the ones that quietly cause field returns.
A gap pad seems like a simple thing. You stuff a soft conformable sheet between a hot chip and a heatsink, and the heat is supposed to flow. In practice we see engineers get caught out by the small stuff — picking a pad too thick for the closure force their fasteners can actually deliver, or specifying a number from a datasheet that was measured at compression rates you'll never see in real production.
Every customer enquiry starts with the same five variables: pad thickness, real-world compression, true thermal conductivity, failure modes nobody puts on the datasheet, and a short checklist to paste straight into your design review. Get these wrong and the pad is the wrong pad, doesn't matter how good its spec sheet looks.
Get the thickness wrong and nothing else matters. Pads come in standard thicknesses from about 0.5 mm up to 5 mm, sometimes thicker for special builds. Pick a pad that matches your nominal air gap exactly and you are already in trouble — the pad needs to compress to seat properly, and to make low-resistance contact with both the chip lid and the heatsink.
A reasonable starting point is 20-50% compression on a typical silicone-based pad. So if your physical gap is 1.5 mm, a 2.0 mm pad compressed to ~25% works well. If your gap varies (it usually does, once you stack-up min and max tolerances) pick the pad off the largest worst-case gap. The smaller-gap regions just compress further, no harm done.
| Nominal Thickness | Compression Range | Typical Application |
|---|---|---|
| 0.5 – 1.0 mm | 10–30% | SoC, FPGA, small power IC under tight-tolerance lids |
| 1.5 – 2.0 mm | 20–40% | Telecom modules, EV onboard chargers, motor drivers |
| 3.0 – 5.0 mm | 25–45% | High-voltage power conversion, IGBT modules, large gaps |
Pad thickness tolerance is typically ±0.1 mm or ±10%, whichever is larger. PCB warp, lid co-planarity, and component height variation can easily stack to ±0.3 mm on dense boards. With 30+ thermal contacts on a single PCB, the tolerance budget matters way more then most teams think it does.
This is where most engineers underweigh the spec. Compression force vs deflection curves are non-linear. A pad rated "soft, 10 psi at 25% compression" might need 40 psi to reach 50% — and 100 psi+ to push to 70%. Your screws and the stiffness of the heatsink plate decide where on the curve you end up, not the datasheet.
Why does this matter so much? Because the bond-line thickness (BLT) — the actual pad thickness after compression — drives thermal resistance more than the W/mK number in most assemblies. A 1.0 W/mK pad sitting at 0.5 mm BLT can outperform a 3.0 W/mK pad sitting at 2.0 mm BLT. The expensive material is not always the right answer.
Datasheet "soft" pads with low Shore 00 hardness feel mushy in hand but can still need real force to fully seat. Pads with a carrier (mesh or fiberglass) don't compress past the carrier thickness, period.
Compression set — the pad relaxing under load — eats into your contact pressure over 1000+ hours. We've measured 8–15% set on common silicone pads after 18 months in always-on telecom.
Our rule of thumb: compress soft pads to 20–40%. Enough to wet the surfaces, not so much that the carrier limits BLT or you put stress on solder joints. Anything past 60% and the failure risks shift from thermal to mechanical.
Pads are sold by their bulk thermal conductivity — W/mK at some reference temperature. What that number does not tell you is contact resistance. A pad with a stickier surface conforms better to micro-roughness on the chip and the heatsink, so it has lower contact resistance. Two pads with the same W/mK can perform very different in a real assembly, just because of how the surface wets.
Lab measurements use polished plates with even pressure that you don't have in the field. Always test in your actual stack, with your actual screws torqued to spec, before you commit to a part number. We've seen 1.5 W/mK pads beat 3.0 W/mK pads in the same assembly because the cheaper pad wet the surface better.
Pad gets squeezed out from under the chip after repeated thermal cycling. Mostly seen with very soft pads in big delta-T environments (-40 to 105°C).
Silicone pads can release oil over time, contaminating optical sensors or RF connectors nearby. Pick low-bleed grades when that matters.
Pad doesn't recover after long compression, contact pressure drops, thermal resistance climbs. We see this most on units running >18 months always-on.
Operators puncture or tear the pad while seating the heatsink, specially around through-screws and standoffs. Specify reinforced pads or use install jigs.
If the pad has a PSA on one side, hot-side debonding under high temperature is real. Make sure the adhesive is rated for your continuous use temp.
Over-tightening fasteners to "fully compress" the pad can fatigue BGA solder balls. We've seen 0.5 mm of unnecessary compression cause failures 12 months in.
Get these answered before you ask a vendor for samples. With the first four nailed down, we can usually pull together a credible shortlist same day. Send the rest along with your inquiry and we'll have samples out within 48 hrs in most cases. No NDA needed.
ITAR registered, ISO 9001:2015 certified. Engineering reply is usually within one business day.
Note: Test in your actual stack with actual fasteners torqued to spec. Lab bench numbers and field numbers are rarely the same thing.
Send your gap, dissipation, and force budget. We'll send the most likely thicknesses so you can test on a real assembly, not a polished lab block. Engineering reply usually within one business day.
Request Sample Kit Talk to an Engineer
