Thermal Profiling for Thick Boards and Heavy Assemblies

A thick board or a heavily populated assembly does not simply run slower through the oven; it runs differently. The laminate acts as a heat sink, the copper planes spread heat laterally, and the components on the top side shield the board beneath them from the convection above. A profile that produces a textbook curve on a 1.0 mm test coupon can leave the centre of a 3.2 mm backplane several tens of degrees cooler, and the joints that fail are the ones in the middle rather than the ones at the edge. Thermal profiling that ignores this documents the oven rather than the assembly.

Why Thick Boards Behave Differently

The thermal mass of a board scales with its volume, and the FR-4 laminate has a specific heat roughly twice that of an equivalent volume of water on a per-gram basis, while its thermal conductivity is around 0.3 W per metre-kelvin, which is very low. Heat therefore enters through the surfaces and must conduct inwards, and the interior lags the surface by an amount that grows with thickness and with the number of copper planes.

The copper changes the picture. A large ground plane conducts heat laterally far faster than the laminate conducts it vertically and horizontally, so a board with heavy planes reaches a more uniform temperature than a thick board without them, but it also takes longer to heat because of the total mass. The practical consequence is that two boards of the same thickness can need different profiles, and the number and weight of planes is as important as the thickness. Moisture taken up by the laminate adds to the same problem, so the baking schedule should be checked before a thick board is profiled.

Thermocouple Attachment and Placement

A thermocouple reads the temperature of its junction, so it has to be attached so that the junction is in thermal contact with what is being measured, and so that the wire does not conduct heat away. High-temperature tape alone is not enough: the tape holds the bead against the surface but the bead also sees the air, and it reads the air temperature rather than the joint temperature. A small dot of thermally conductive adhesive, or a bead captured under a solder joint made for the purpose, gives a reading closer to the real value.

Placement should cover the extremes rather than the average: a large component body, a small chip beside a heavy plane, the corner of the board furthest from the conveyor centre, and a point in the middle of a large copper area. Four to six thermocouples on a thick board is normal, and the profile is judged by the worst case rather than by the trace that looks best. The attachment should be checked before the run, because a bead that has come loose during the pass produces a trace that looks like a successful profile and means nothing.

Reading the Profile on the Board, Not the Oven

The oven display shows the air temperature in each zone, which on a thick board can be 40 °C away from the board temperature during the ramp. Setting the profile from the display rather than from a measurement leads to an oven that is hotter than intended, which shortens the life of the heating elements and the blowers and can overheat the components on the top side while the interior is still cold.

An increase in board thickness therefore usually means a longer oven or a slower belt, not a higher setpoint. Raising the setpoint to force heat into a thick board also raises the surface temperature, and the surface is where the components are. The compromise is a slower profile with a longer soak, which brings the interior up without overheating the exterior. This is one of the cases where the belt speed is the correct lever and the setpoint is not.

Soak, Ramp and the Limits of Each

The soak brings the whole assembly to a uniform temperature before the alloy melts, which is exactly what a thick board needs. A soak zone of 90 to 120 seconds below the solidus gives the interior time to catch up, and it also allows the flux to act on the oxides before the alloy flows. A soak that is too short leaves a temperature gradient across the board at the moment of melting, which produces different times above liquidus in different areas and therefore different joint structures.

The ramp rate is limited by the paste and by the components. A ramp above about 3 °C per second is unusual outside a specific high-throughput process, and on a thick board the achievable ramp is limited by the oven anyway. Where the ramp is limited by the oven, extending the soak is the way to reach the required peak without exceeding the component’s ramp rating, and the profile should be adjusted in that order.

Peak Temperature and Time Above Liquidus

The peak has to be high enough for the alloy to wet and flow at every point on the board, including the coolest. On a thick assembly the coolest point may be 15 °C to 25 °C below the hottest, so the setpoint has to be chosen from the coolest thermocouple rather than the average. A peak of 235 °C measured on the hot side can mean a 215 °C peak in the middle of the board, which is barely above the melting point of a tin-silver-copper alloy.

Time above liquidus is the other half. Too little and the joint does not form a proper fillet; too much and the intermetallic layer grows and the components see more heat than their rating allows. A working range of 45 to 90 seconds above liquidus covers most assemblies, with the thicker ones at the upper end. The measurement should report the time for each thermocouple, because the difference between the hottest and coolest traces is the most useful single number on the profile record. The general method is set out in the reflow profile notes, and the specific figures should come from the paste maker’s data sheet.

Delta Across the Board and Between Sides

The temperature difference across a board during the ramp and the soak is the number that decides whether a single profile can serve the whole product. A delta of more than about 10 °C at the point of melting makes it difficult to satisfy both the hottest and the coolest joint, and the usual remedies are a longer soak, a lower peak or a slower belt. Where the delta cannot be reduced, the design should consider a thermal relief or a change in the copper distribution rather than accepting a marginal profile.

The two sides of a board behave differently. On a double-sided assembly, the second pass exposes the joints made in the first pass to a second thermal cycle, and the bottom side, being closer to the heating elements in many ovens, may run hotter than the top. Measuring both sides on a double-sided board is worth the extra thermocouples, because the limit is set by whichever side is hotter and by whichever side sees the second pass.

Fixtures, Pallets and Their Thermal Effect

A pallet or a carrier changes the profile. The pallet has mass, it shields the underside of the board from convection, and it may block the infrared view of the board. A typical effect is a cooler bottom side and a longer time to reach the soak temperature, which means a profile developed without the pallet will not produce the same result with it.

The pallet material matters as well. A machined aluminium pallet conducts and distributes heat, a composite pallet insulates, and the two produce different board temperatures for the same oven settings. The profile should always be taken with the pallet that will be used in production, and any change of pallet material should be treated as a profile change. Where the pallet has cut-outs for components, the shadowing they create should be visible in the thermocouple traces, and if it is not, the thermocouples are not where the components will be. A pallet that holds a thick board under load also changes its flatness, which is the same mechanism described in the warpage notes.

Profile Records, Revisions and Verification

A profile record is only useful if it identifies the product, the revision, the oven, the belt speed, the pallet and the paste. The trace alone does not describe the process, and a record that has to be matched to an unknown set of conditions cannot answer the question that will be asked after a defect appears. The record should also carry the acceptance limits, so that a later reviewer can see whether the profile met them at the time.

Verification on a fixed interval, and after any change to the oven, the belt, the pallet or the paste, keeps the record meaningful. A profile that was correct at the start of a production run is not evidence about the boards made six months later, particularly where a blower or a heating element has been replaced. The re-measurement is quick, and it is the only way the reflow profile on file can be said to describe the process rather than an intention. Records should be retained with the batch documentation, since the profile is part of the evidence that the assembly was made to specification.

FAQ

Can a thick board use a standard profile if the belt is slower? Often yes, but the soak and the peak both need to be checked with thermocouples on the board. Slowing the belt increases the total time but does not by itself flatten the gradient, and a long soak is usually needed to do that.

How many thermocouples are enough? Four to six covers most thick assemblies, placed to include the largest thermal mass, the smallest component, the coolest corner and the centre of the largest copper area. Fewer than four leaves the delta unmeasured, which is the number that matters most.

Does a thicker board need more flux? Not directly, but the longer time at temperature does consume the flux activity, and a soak that is too long can exhaust the flux before the alloy melts. Where the soak has been extended, the paste should be checked for its recommended maximum soak time, and the joint appearance re-verified.