Thermal Paste Selection for Tablet Thermal Management
Thermal Paste in a Fanless Tablet: What a 0.1 mm Layer Does to a 6 W Budget
A tablet removes the one component that normally makes thermal design forgiving: the fan. A 6 W system-on-chip on a 70 mm2 die produces roughly 8.5 W/cm2, modest next to a notebook, but there is no airflow to carry it away and no heatsink volume to absorb it. The only path is the die into a shield can, the shield into the midframe, the midframe into a graphite spreader, and the back cover into the hand of the user. In that chain the interface contributes less than a kelvin in absolute terms, yet it sits inside a junction-to-ambient budget of about 55 K at full load, so a thermal paste layer nobody measures can consume 3 to 4 percent of the entire thermal headroom. This is where tablet thermal management is actually decided.
The stack has three hops, and only one is a paste joint
The mistake most often made in a thin device is to treat the die contact as the whole thermal problem. In practice heat crosses four boundaries before it reaches ambient air, and each of them behaves differently. Only the die interface needs a thermal interface material of paste-like thickness; the others are spreading and mechanical contact problems.
| Boundary | Flux | Method | What governs it |
|---|---|---|---|
| SoC die to shield can or midframe | 6–12 W/cm2 | Paste, 20–40 µm | Bond line thickness and pressure |
| Shield can to midframe | 3–6 W/cm2 | Paste or thin gap filler | Metal-to-metal contact area, clamping |
| Midframe to graphite or vapour layer | 1–3 W/cm2 | Adhesive or pad | In-plane spreading, not interface resistance |
| Back cover to air and hand | < 0.5 W/cm2 | No interface | Surface area, emissivity, natural convection |
Once heat has left the die the limiting resistance is spreading: 6 W over a 100 x 150 mm back cover is still a natural-convection problem that no interface material fixes. A good thermal interface material protects the first two hops only, and should be specified for exactly that role.
Why thermal paste and not a 0.5 mm pad
The thickness budget decides, and so does preload. A tablet midframe is typically 0.4 to 0.8 mm thick and the display stack sits directly above the board, so there is no room for a compressible pad of 0.5 mm; forcing one in preloads the glass and, on the next assembly revision, cracks it. The numbers also rule the pad out on performance grounds.
| Option | Joint thickness | Rise at 8.5 W/cm2 | Fit in a tablet |
|---|---|---|---|
| Thermal paste, 0.03 °C·cm2/W | 20–40 µm | 0.3 K | Preferred on the die |
| Thermal paste, 0.080 °C·cm2/W | 30–50 µm | 0.7 K | Preferred on shields and frames |
| Gap filler, 0.5 mm at 5 W/m·K | 500 µm | Approx. 8.5 K | Exceeds the thickness budget |
| Graphite sheet alone | 25–50 µm | Spreads, does not bond | Correct above the midframe, not on the die |
Five constraints that a datasheet does not mention
Low and uneven clamping. Tablet thermal management is limited by how the device is closed: screws around a midframe and adhesive around a display, so die pressure is often below 10 psi and varies across the die. The bond line measured at 30 psi in the laboratory is therefore optimistic; the design should be qualified at the pressure the product really applies.
Die warpage on a small package. A 70 mm2 package warps more per unit area than a large one, and a 20 µm joint cannot absorb a 40 µm bow. Either the shield can is coined to follow the die, or the joint is specified thicker and the penalty accepted.
Long, shallow power cycling. Unlike a gaming notebook, a tablet spends most of its life at 1 to 3 W, so the thermal interface material ages slowly rather than being pumped out quickly; creep and oil separation over three to five years still show up as a device that throttles earlier than it did when new.
Migration into optics and antennas. In a fanless tablet a camera module, a microphone port and antenna windows share the same thin volume. Low-molecular siloxane from a silicone-based thermal paste can fog a lens or deposit on an antenna carrier, so volatility is a design constraint in a 6 mm chassis, not a detail.
Service exposure. Battery replacement and screen repair expose the joint to opening and re-closing. A non-curing thermal paste tolerates one rework; a cured adhesive joint does not, and a pad that has been compressed twice is worse than a fresh print.
Selecting a grade for each hop
The TSAD Series from TOUSEN is a solvent-free, non-curing, thixotropic gray paste covering 1.2–6.5 W/m·K, with measured impedance from 0.201 down to 0.03 °C·cm2/W at 30 psi, specific gravity 2.5–2.9 g/cm3 and a 12-month shelf life. Because thickness runs out in every direction, tablet thermal management is specified hop by hop: the thermal interface material is chosen by the flux of that joint and by the bond line the assembly can hold, not by the highest conductivity in the catalogue.
| Grade | W/m·K | Impedance @ 30 psi (°C·cm2/W) | Density (g/cm3) | Typical position in a tablet |
|---|---|---|---|---|
| TSAD60 | 1.2 | 0.201 | 2.5 | Shield can to bracket, low-flux filler |
| TSAD66 | 3.5 | 0.080 | 2.7 | Shield to midframe, camera and speaker modules |
| TSAD66-P | 4.0 | 0.070 | 2.6 | Rougher or warped midframe contact |
| TSAD83 PLUS | 5.2 | 0.050 | 2.6 | Mainstream SoC at 4–6 W |
| TSAD88 PLUS | 5.2 | 0.030 | 2.9 | Thin printed joint on a hot bare die |
| TSAD90 | 6.5 | 0.030 | 2.7 | Premium SoC at 6–8 W, high burst power |
| TSAD100 | 6.5 | 0.030 | 2.8 | Highest flux with the tightest joint target |
Mass, not a small dot
In tablet thermal management the dispense is a process, not an art: convert the bond line into mass and the guesswork disappears. At 2.7 g/cm3 a 30 µm joint needs about 8 mg per cm2, so a 70 mm2 die takes roughly 5.5 mg and a 100 mm2 shield contact about 8 mg. On a device this small, one extra milligram is a measurable extra bond line, and a balance is cheaper than a rejected lot.
Print rather than dispense where the geometry is flat
For shield cans and midframe contacts with a defined footprint, stencil printing at 0.1 mm with 80–90 percent coverage gives a repeatable joint in a few seconds and avoids the operator-dependent dot. Dispensing remains the right choice on the bare die, where the area is small and the placement must follow the die outline.
Where a pad or a gap filler is still correct
Above the midframe, between the battery and the back cover, and around camera modules, a thin silicone pad or a gap filler is the right part: those joints exist to absorb tolerance and to spread, not to conduct at high flux. Using paste there creates a joint that can migrate and offers no mechanical benefit.
Qualification matched to fanless usage
Run a 30-minute steady-state load at the product's sustained power and log both junction and skin temperature, because in tablet thermal management the customer experiences the second value. Then run 1,000 shallow power cycles between idle and boost and re-measure impedance on a sacrificed unit. Add 500 hours at 60 °C and 90 % RH, a drop sequence on each edge, and a gravimetric oil separation check at 150 °C. Where a camera or antenna shares the volume, add an outgassing test. Shelf life is 12 months, so a cartridge held over from a previous build should be re-qualified rather than topped up.
Assembly and service notes
- Clean the die and the shield with isopropyl alcohol; residue under a 30 µm joint is a hotspot that no spreader can hide.
- Torque the midframe screws to the drawing value in a diagonal sequence; over-torque loads the glass, under-torque thickens the joint.
- Keep the print area inside the die outline; paste outside the die adds no conduction and creates a migration path.
- On service, remove the old film completely and reprint the thermal interface material; two layers read worse than one fresh layer.
- Store the cartridge closed at room temperature away from flame; the product is non-hazardous but not indefinitely stable.
Conclusion
Tablet thermal management is a passive problem: 6 W must leave through a midframe and a back cover, and every kelvin removed from the die buys sustained clock for the user, which is why the thickness of the paste layer matters far more than its position on a datasheet. Thermal paste cannot fix the spreading resistance of the chassis, but it is the only layer the assembly process still controls after the industrial design is frozen. Specify it per hop, by flux and by the bond line the clamp actually produces, size the dispense in milligrams, and validate on an assembled unit at the real pressure. Between a 30 µm joint and a 100 µm one at 8.5 W/cm2 there are more than half a kelvin, and in a fanless device half a kelvin is the difference between holding a frame rate and dropping it.
Impedance data, conductivity grades and packaging options for the thermal paste described here are listed on the TSAD Series thermal paste product page, where samples can be requested for print trials on your own midframe.
Thermal Paste Selection for Laptop Thermal Management
Related Article
