Thermal Metrics with the Same Units Can Describe Different Heat Paths

Identify the temperatures before reading the number

Junction temperature, TJ, concerns the semiconductor inside the package. Case temperature, TC, concerns a specified location on the package exterior. It does not mean the server chassis. The ambient reference, TA, belongs to the stated thermal environment; a convenient room or rack-inlet reading cannot automatically replace it.

For top and bottom case fields, retain the orientation relative to the printed circuit board. TI defines the top as facing away from the board and the bottom as facing toward it. A bottom-case temperature is therefore distinct from a temperature at a location on the board itself.

The units describe a temperature difference per watt: K/W and °C/W have the same numerical scale for that difference. They do not identify either endpoint or say which watts form the denominator. Device power dissipation, heat through one branch and the electrical input of a complete server are different quantities.

References: [O1][T1][A2]

RθJA includes the test board and environment

RθJA is the junction-to-ambient temperature difference divided by device power dissipation under the stated characterization conditions. Texas Instruments' SPRA953D, revised March 2024, explains that the board conducts and spreads heat from the device. Heat can pass through the device's connections into the board and then reach the surroundings over the board's area. Its construction changes the resulting metric.

Nexperia's IAN50019 identifies board size, material, copper planes, thermal vias and free or forced cooling among the relevant conditions. A package mounted on another board therefore has another surrounding heat-removal system, even when the package name stays the same.

For a catalog comparison, retain the board and environmental basis alongside the symbol. Matching units without that basis cannot distinguish a package difference from a test-condition difference. Even a properly matched characterization comparison addresses the reported conditions; transferring its result to a populated server board requires application-specific evidence.

References: [T1][N1]

RθJC qualifies a particular case heat path

Junction-to-case resistance concerns heat transfer between the junction and the specified case surface. The characterization described by TI directs almost all device heat through that surface. A top-case field and a bottom-case field consequently describe different paths.

That condition matters when total power appears in a junction-to-case expression. It approximates heat through the selected case path in that characterization. It does not establish that the same fraction follows that path in an assembled product.

If substantial heat instead leaves through the board or another route, multiplying catalog RθJC by all device power and adding an arbitrary case temperature loses the required relationship. onsemi's August 2022 AND8220/D emphasizes the case-location and heat-flow qualifications. A case-path metric alone also leaves the thermal environment beyond that surface unresolved.

References: [T1][O1][A1]

ΨJT keeps total device power in the denominator

ΨJT is a thermal characterization parameter: the junction-to-top temperature difference divided by total device power dissipation. In symbols, ΨJT = (TJ − TT) / Ptotal, with TT denoting the specified package-top temperature.

Analog Devices' June 19, 2025 explanation distinguishes a junction-to-case characterization with a dominant case path from ΨJT characterization with multiple heat paths. The latter quotient does not require all dissipated power to travel through the top.

A small ΨJT can therefore accompany a small top temperature difference while much of the heat leaves elsewhere. Its value does not reveal the top-path heat rate by itself. TI also identifies board construction and airflow dependence. Using ΨJT for an application estimate needs a justified relationship to the characterized heat distribution; changing the symbol from theta to psi does not remove that dependency.

References: [A1][T1][O1]

TM-282: one 4 K difference gives two different quotients

TM-282 is an original fictional steady-state example. Stipulate 2 W of total device dissipation, with 0.5 W leaving through a defined top branch and 1.5 W leaving through other paths. Stipulate a junction-to-top temperature difference of 4 K. These are explanatory inputs, with no named product, measurement or predicted operating temperature.

Dividing the temperature difference by total power gives 4 K / 2 W = 2 K/W. This is the ΨJT-style normalization: the denominator includes the 1.5 W leaving through other paths as well as the 0.5 W in the top branch.

Dividing by the stipulated top-branch heat instead gives 4 K / 0.5 W = 8 K/W. This is a quotient for that defined fictional branch. It is not catalog RθJC: the example supplies neither a product characterization nor the corresponding case-path test conditions.

The top branch carries one quarter of the stipulated total, so the total-power quotient is one quarter of the branch quotient. The difference between 2 and 8 describes the denominator, without any change in hardware.

Swapping them exposes the error. Multiplying 8 K/W by 2 W yields 16 K; multiplying 2 K/W by 0.5 W yields 1 K. Neither recovers the stipulated 4 K because each pairs a quotient with the other denominator. Those are demonstrations of mismatched arithmetic, not additional thermal predictions. Neither quotient ranks an actual component.

A valid test comparison still leaves the assembly question

Suppose two component offers each include a field labelled thermal performance. One reports ΨJT and the other reports RθJC(top). Selecting the smaller number would compare different quantities before considering any package difference. The first useful decision is to classify the fields correctly, rather than rank the offers.

If both offers instead report the same metric, the temperature locations and characterization conditions still need to match for the intended comparison. Analog Devices' July 30, 2007 package-characterization article explains the role of standardized JEDEC configurations in consistent package comparisons. Nexperia's mounting-base discussion illustrates why a favorable value at one package boundary need not produce the most favorable junction-to-ambient response across different package types. Heat continues beyond the reported boundary.

An assembled-system result answers a further question: how the identified device behaves with its actual board, other heat sources and cooling environment. Applicable manufacturer data or qualified system analysis must connect that configuration to the claim. A semiconductor field does not become a server-wide resistor merely because server power is also expressed in watts.

Keep time scope attached as well. onsemi distinguishes steady-state thermal data from transient behavior. A steady-state quotient alone cannot establish a junction-temperature excursion during a changing workload. The catalog comparison can remain useful while the operating-temperature claim remains unestablished.

References: [A2][N1][O1]

Bring the selected system into physical IT planning

The related chiplet and packaging guide helps identify whether an offer concerns dies, a packaged accelerator, a board or a complete server. Carry that identity into the thermal comparison. An internal component metric cannot supply the complete system's external installation requirements.

Once the equipment and applicable requirements are identified, Leviathan's Physical Infrastructure Planning & CAD service can discuss rack layouts, cable pathways, point-to-point schedules, labels and agreed installation documents. Authoring capability, reviewers and outputs are confirmed for each engagement. Semiconductor thermal characterization, cooling selection and thermal-performance approval belong to the appointed product and thermal specialists.

Use the linked planning service and project inquiry to share the equipment models and quantities, available rack and connection plans, and requested physical documentation. That supports a discussion of planning scope, requirements and availability after the relevant equipment decisions have been established.