Thermal Profiler Setup: Getting a True Reflow Reading
A thermal profiler records the temperature history that a board actually experiences, and the value of that record depends entirely on how the sensors were attached. A reading taken beside a pad instead of on it, or routed so that heat drains along the wire, describes the instrument rather than the assembly.
Profiling is the measurement that proves a reflow profile is inside its window, and the window is narrow enough that a two or three degree error matters. Most of the error in practice comes from the fixture rather than from the electronics.
What a Profile Has to Establish
A usable profile reports four things: the ramp rate as the board enters the heating zones, the soak or soak equivalent that brings the assembly to a uniform temperature, the peak temperature reached at the coldest joint, and the time above liquidus.
Two numbers are usually quoted together with the curve. The first is the peak temperature at the coldest measured point, and the second is the difference between the hottest and coldest points on the assembly, which is the spread that decides whether the profile suits the whole board.
Thermocouple Type and Wire
Type K wire is the common choice for reflow work, with type N used where a wider range or better stability is wanted. Wire diameter matters more than the alloy choice for this application, because a thick wire conducts heat away from the junction and a thin wire is fragile.
Fine wire in the range of 0.1 to 0.25 mm is a practical compromise. The junction should be as small as possible and welded rather than twisted, since a twisted junction creates a second thermal path and adds contact resistance that varies with pressure.
Attachment Methods and Their Errors
High temperature solder is the most accurate method for a metal target, provided the alloy used melts well above the profile being measured. Silver loaded epoxy is a practical alternative, but its thermal conductivity is lower than solder and the layer thickness must be kept small and consistent.
Adhesive tape and cyanoacrylate are the least accurate and the most common. Both insulate the junction from the surface to some degree, and the measured temperature lags the true surface by an amount that grows with the heating rate. When tape is used, the junction must still touch the feature being measured.
Wire Routing and Heat Loss
Heat travels along thermocouple wire, so a wire that leaves a small pad and passes directly over copper will pull heat out of the junction. The standard remedy is to route each wire along a path that keeps it in contact with the board surface for a distance before it leaves the assembly.
Bundling all the wires together into a single loom improves the situation further because the bundle reaches a common temperature. The alternative failure is a bundle that passes over a large thermal mass and drags the whole set of readings toward one value, which hides the thermal management differences the test was meant to reveal.
Choosing the Sample and Sensor Locations
The sample should be a real assembly with real components, and the sensors should be placed where the risk is highest. A large thermal mass, a small package beside a heavy plane, a connector body and the coldest corner of the panel are all more informative than a sensor placed in the middle of an empty area.
A minimum of three sensors is needed to describe a spread, and five or more are normal on a mixed assembly. One of them should sit on a bare pad close to a large copper area, because the joints that fail are usually the ones that never reached the intended peak rather than the ones that ran hot.
Delta T Across the Assembly
The spread between the hottest and coldest location is set by the copper distribution, the component mix and the heat transfer inside the oven. A small, uniform product can hold a spread under 10 C, while a mixed board with a heavy connector can show 25 C or more.
The spread cannot be removed by raising the peak temperature, because the hottest location reaches its limit first. It is reduced by improving the balance of the design, by slowing the ramp so that the assembly equalises, or by accepting a longer soak, and each of those choices has a cost in flux consumption or throughput.
Reading the Profile Curve
The ramp from ambient to soak should be smooth, without a plateau that indicates a stalled conveyor or a sudden step that indicates a zone set point error. The soak does not need to be flat, but the rate of change through it should be low enough that the flux activates before the alloy melts.
The peak is read at the coldest sensor, not the hottest, and the time above liquidus is read from the same curve. For a lead free alloy with a liquidus near 217 C, a common window is a peak between 235 and 245 C with 45 to 90 seconds above liquidus, adjusted to the component limits on the assembly.
Repeatability, Calibration and Records
A profiler should be verified against a reference before it is trusted for a release. An ice point and a boiling point check, or a comparison against a calibrated reference instrument, establishes that the channels agree with each other and with reality.
Records should include the profiler identification, the calibration date, the attachment method, the sensor locations on a photograph, and the resulting curve. Without the attachment detail, a profile cannot be reproduced, and disagreement between two sites usually turns out to be disagreement about how the sensors were fixed.
Common Measurement Traps
The most frequent trap is measuring air temperature instead of board temperature, which happens whenever a junction is left exposed and unshielded. The second is measuring a small coupon instead of the real assembly, which understates the thermal mass and produces a profile that the product cannot follow.
The third is a sensor that was attached to a pad but has come loose during handling and now reads the air inside the oven. Continuity of the trace should be checked before the run, and the curve should be inspected for a step change that indicates a detached junction. Conveyor behaviour contributes to the same errors, which is why conveyor speed calibration belongs in the same maintenance plan as the profiler.
FAQ
How many thermocouples are needed for a reflow profile? At least three, and five or more on a mixed assembly. The purpose is to describe the spread across the board, so the sensors should be placed where the thermal behaviour differs most.
Does the thermocouple have to be soldered to the pad? Soldering gives the most accurate result on a metal target. Silver epoxy and tape are workable alternatives, but both insulate the junction to some degree and introduce a lag that must be kept consistent.
Why does the measured peak temperature differ between two runs? The usual causes are a different attachment method, a loose junction, a change in conveyor speed or a different sample. Profiler calibration and oven zone settings should be checked only after the fixture has been ruled out.
}
