吃瓜tv

Separate conductivity from resistance

Thermal conductivity describes a material under stated conditions. Thermal resistance describes a defined path and geometry. For a homogeneous slab with uniform one-dimensional heat flow, resistance is thickness divided by conductivity and area: R = t/(kA). Doubling thickness doubles that model resistance; doubling area halves it.

Real PCBs do not usually carry heat through a constant area from source to sink. Copper spreads heat, planes contain openings and contact interfaces introduce additional effects. Treat the slab equation as a local reference. It is particularly useful for detecting a unit mistake or comparing two dielectric thicknesses under the same assumptions.

Two-branch thermal network

Follow the heat before adding numbers

Elements in series carry the same heat in the ideal model. Their resistances add. A package-to-board path, board dielectric and interface to a cold plate may form a useful series approximation only if the chosen heat flow actually traverses all those elements.

Elements in parallel share the same start and end temperatures. Their conductances add. If a board route and package-top route terminate at different environmental temperatures, the simple two-branch equation on this page is no longer sufficient. Draw both endpoints rather than deciding that all visible cooling features are parallel.

Work through a parallel-path example

Consider illustrative branches of 4 and 12 K/W connecting the same nodes. Their equivalent resistance is 3 K/W. Add a common 2 K/W downstream path and the total becomes 5 K/W. With 5 W entering the network, the modeled source-to-boundary rise is 25 K.

The branch temperature difference is 15 K, so the ideal branch powers are 3.75 and 1.25 W. That division follows the specified resistances; it is not a fixed rule for heat leaving a semiconductor package. Changing either branch changes both the source temperature and the power split.

Recognize the shared bottleneck

Improving one branch can give disappointing results when a common interface dominates. If both routes must pass through the same poorly mounted heat sink, reducing a local board resistance cannot remove that shared limitation. Use the tool to change one number at a time and compare the total, not just the improved element.

This reasoning also explains why a higher-conductivity substrate may not deliver the expected system improvement. Review material thickness and conductivity and then the interface impedance. A good comparison retains the same source, footprint, contact pressure and downstream boundary.

Keep package metrics in their proper context

A semiconductor datasheet may list thermal metrics measured or characterized using a specified board and environment. They are not all interchangeable resistors. TI's explains why junction-to-ambient results depend on the measurement setup and why characterization parameters require careful use.

For a project calculation, identify the metric's endpoints and conditions before using it. Do not add a whole junction-to-ambient number to a board and heat-sink chain that already includes those paths. Where the actual heat split is unknown, retain the uncertainty and use a representative model or measurement.

Distinguish warm-up from the final condition

Thermal capacitance governs energy storage and therefore the time response. A thicker metal plate can slow warming without providing the same proportional reduction in the eventual temperature. A short demonstration can therefore favor a heavy construction even if both designs converge toward similar steady-state values.

State whether a result concerns a pulse, repeating duty cycle or continuous operation. The calculator here has no thermal capacitances and produces only a steady-state reference. For a time-dependent case, record the load waveform and initial temperature for an appropriate transient model and subsequent physical validation.

Check the arithmetic with units

Keep thickness in meters, area in square meters and conductivity in W/(m路K) when using the basic slab equation directly. A 100 碌m layer with 2 W/(m路K) conductivity over 100 mm虏 gives 0.5 K/W in that ideal model. Confusing millimeters with meters can change the answer by three orders of magnitude.

Temperature differences in kelvin and degrees Celsius have the same numerical size. Absolute temperature references must still be identified. The tool reports a rise, not a complete temperature, until a downstream boundary is explicitly added in an appropriate calculation.

Turn a principle into a controlled comparison

Create two candidate drawings with one deliberate change and record all conditions held fixed. Measure or model the same temperature location and heat input. A comparison of a bare coupon and a populated enclosure says more about their different boundaries than about the material alone.

Bring that comparison to the power PCB review when the remaining question concerns a manufacturable construction. Include the model assumptions and the test you intend to use. The objective is a defensible design decision, rather than the lowest isolated number on a material table.

From a thermal estimate to a reproducible review

Choose a control volume first: the component, local board region or complete enclosure. Identify energy entering and leaving that boundary. Keep the resulting heat sources at their actual locations instead of assigning all losses to whichever temperature sensor is easiest to reach.

Prepare a measurement sketch with source footprint, sensor locations, attachment methods, orientation and downstream boundary. State whether the test is transient or stabilized. Use consistent mounting and input conditions when comparing a prototype with a model. Measurement repeatability is part of the evidence, not an optional note after a result looks favorable.

If the model disagrees with the assembly, inspect loss assumptions, interfaces and boundary temperatures before choosing another laminate. Use the simplest model that answers the current decision, then increase its detail where the discrepancy or design sensitivity justifies it.

Before release, place the chosen requirement, the proposed construction and the evidence in one review record. Separate measured values from supplier nominal data and from your own assumptions. If a value is unknown, identify who will obtain it and what decision depends on it. An empty field is more honest and more useful than an invented typical figure.

During an A/B comparison, change one relevant variable where possible. Preserve the original inputs and the reason for the change. If several things change together, describe the result as a comparison between complete configurations. Do not attribute the entire difference to one material or geometry feature without supporting evidence.

For a repeat order, confirm that the relevant drawing, material and process revisions still match the accepted configuration. Record alternatives explicitly. This small discipline helps purchasing avoid an apparently equivalent substitution that changes the engineering basis of the original decision.