A datasheet Thermal Information table hands you six numbers in degrees C per watt — RθJA, RθJC(top), RθJC(bottom), RθJB, ψJT, and ψJB — and each answers a different question. Understanding package thermal resistance means knowing which one applies to your board, because feeding the wrong metric into a junction-temperature estimate produces a number that can be off by 100 degrees C. This guide decodes all six and shows which to use for the temperature you can actually measure.
Key takeaways
- Theta JA (junction-to-ambient) is a comparison metric, not a prediction for your board. A thermally optimized PCB can cut the datasheet RθJA by 30 to 55 percent, per TI’s own PCB study.
- Psi JT and psi JB are thermal characterization parameters, not true thermal resistances. They carry the Greek psi precisely because heat leaves the die by several paths at once.
- To estimate junction temperature in your system, measure a temperature you can reach, top of package or board, and use psi JT or psi JB, not theta JC.
- Theta JC(top) is often 10 to 50 times larger than psi JT for the same part, so using it to back out junction temperature from case temperature gives a badly inflated answer.
- Every datasheet theta and psi value is measured on a specific JEDEC board in still air, defined by the JESD51 series. Change the board or add airflow and the numbers move.
What each thermal metric actually measures
Every metric is a temperature difference divided by device power dissipation, in degrees C per watt. What differs is the two points between which that difference is taken, and how the heat is forced to flow during the test.
The split between theta and psi is the part most engineers miss. A theta value is a true thermal resistance: it is defined only when you can name the single path the heat takes and force nearly all of it down that path. A psi value is a characterization parameter that accepts the real situation, where heat leaves the die through the leads, the board, and the top surface at the same time. That distinction, laid out in Electronic Design’s treatment of the two, is why psi JT and psi JB exist alongside the theta metrics rather than replacing them.
The formal definitions, following onsemi’s application note AND8220, are below. TT is the temperature at the center of the package top; TB is the board temperature next to the package; TC is the case temperature; P is total device power.
| Metric | Symbol | Between | Equation | Primary use |
| Junction-to-ambient | RθJA | Junction and ambient air | (TJ − TA)/P | Compare packages; worst-case bound |
| Junction-to-case, top | RθJC(top) | Junction and package top | (TJ − TC,top)/P | Sizing a top-mounted heatsink |
| Junction-to-case, bottom | RθJC(bot) | Junction and exposed pad | (TJ − TC,bot)/P | Exposed-pad parts on a heat path |
| Junction-to-board | RθJB | Junction and board | (TJ − TB)/P | Board-dominated heat flow |
| Junction-to-top parameter | ψJT | Junction and package top | (TJ − TT)/P | Estimate TJ from top temperature |
| Junction-to-board parameter | ψJB | Junction and board | (TJ − TB)/P | Estimate TJ from board temperature |
The datasheet Thermal Information table, decoded
Numbers make the relationships concrete. The table below pulls the actual Thermal Information blocks from three Texas Instruments documents: the LM2621 boost converter, a small SMD; one exposed-pad column of the LM3940 LDO; and the DC/DC example device in TI application journal article SLYT739. All values are in degrees C per watt.
| Metric | LM2621 (small SMD) | LM3940 (exposed-pad) | DC/DC conv. (SLYT739) |
| RθJA | 160 | 23.3 | 28.9 |
| RθJC(top) | 52.7 | 16.1 | 18.9 |
| RθJB | 80.1 | 4.8 | 4.1 |
| ψJT | 5.5 | 2.7 | 1.3 |
| ψJB | 78.8 | 4.8 | 4.1 |
| RθJC(bot) | N/A | 1.1 | 1.0 |
Three patterns fall out, and they hold across most parts. First, psi JT is always small — here 1.3 to 5.5 degrees C per watt — because only a sliver of the heat reaches the top of a surface-mount package. Second, psi JB tracks RθJB closely, matching it exactly at 4.8 and 4.1 in the two power parts, since most heat does leave through the board. Third, RθJC(top) dwarfs psi JT — 52.7 against 5.5 on the LM2621 — which is the single reason you cannot swap one for the other when estimating junction temperature. RθJC(bot) is meaningful only when the part has an exposed pad soldered to copper, which is why the LM2621 lists it as N/A.
[IMAGE 1: heat-flow diagram of a surface-mount package showing paths to top, leads, and board with the six metric labels | alt: “Package thermal resistance paths for theta JA, theta JC top and bottom, theta JB, psi JT and psi JB”]
Which metric to use: match it to what you can measure
The difference between theta JC and psi JT is what heat path each assumes. Theta JC(top) is measured with almost all the device power forced out through the top surface into a cold plate, so it only describes a part under an efficient heatsink. Psi JT is measured on a normal board with heat spreading naturally, so it, not theta JC, is the parameter for estimating junction temperature from a top-of-package thermocouple.
That test-condition split is defined in the JEDEC JESD51 series and spelled out in TI’s SPRA953. The practical rule: pick the metric for the temperature you can actually reach with a sensor.
| What you can measure | Metric to use | Equation |
| Only ambient air (no board data) | RθJA — comparison / worst case | TJ = TA + RθJA · P |
| Top-center of the package | ψJT | TJ = TT + ψJT · P |
| Board, right next to the package | ψJB (or RθJB) | TJ = TB + ψJB · P |
| Case under an efficient heatsink | RθJC(top) + heatsink model | TJ = TA + (RθJC + RθCS + RθSA) · P |
Theta JA: built for comparison, not for your board
RθJA is the number everyone quotes and the one most often misused. It is measured in still air on a JEDEC standard board, and TI notes it is the most commonly reported and most commonly misapplied thermal metric. Its value depends heavily on the board it was measured on.
How heavily? TI published a study varying PCB copper and construction and found a board designed for heat can lower the datasheet RθJA by 30 to 55 percent, letting the same die dissipate meaningfully more power. That swing is why RθJA belongs in package-to-package comparisons and worst-case bounds, not in a junction-temperature estimate for a real design.
It does have one honest calculation. The maximum power a package can dissipate before hitting its rated junction temperature is Pdmax = (TJmax − TA)/RθJA. Use it to reject a package early. On the LM2621’s 160 degrees C per watt, a 125 degrees C junction limit and 55 degrees C ambient allow only about 0.44 W before thermal shutdown, a hard ceiling worth knowing before layout.
Psi JT: junction temperature from a top-of-package thermocouple
When a part may be running hot, the fastest trustworthy check is a fine thermocouple on the top-center of the package. Psi JT converts that reading to a junction temperature: TJ = TT + ψJT · P. The equation uses total device power, so you never need to know how the heat splits between paths, a convenience that comes straight from the JESD51-2A definition.
A worked case: the SLYT739 device has psi JT of 1.3 degrees C per watt. Running 2 W with a measured top-center temperature of 85 degrees C gives TJ = 85 + 1.3 times 2 = 87.6 degrees C. Against a 125 degrees C limit, there is comfortable margin and no need for a detailed thermal model.
The catch is measurement technique. JEDEC specifies a 40-gauge thermocouple bonded to the top-center with thermally conductive epoxy and its wire routed along the package to avoid conducting heat away. Psi JT values are small, so a sloppy 5 degrees C error in TT is a 5 degrees C error in TJ. And psi JT is invalid the moment a heatsink is added, because the heat path it was characterized under no longer exists; SPRA953 is explicit that psi JT should not be used when a heatsink is intended.
[IMAGE 2: thermocouple bonded to package top-center with wire routed along the body per JESD51-2A | alt: “Measuring package top temperature for psi JT junction temperature estimation per JESD51-2A”]
Psi JB, theta JB, and theta JC: when you can measure the board or use a heatsink
If your sensor sits on the board rather than the part, use psi JB with the board temperature measured immediately next to the package; an average across the PCB will mislead you. Because most heat exits through the board, psi JB and RθJB are usually close, as the tables above show.
Metric choice can flip a pass into a fail. On TI’s TLV760, the datasheet lists RθJA of 275.2 degrees C per watt and RθJB of 56.8 degrees C per watt. For a 12 V drop at 60 mA (0.72 W), the board-based estimate gives a rise of 56.8 times 0.72, about 41 degrees C, so a 30 degrees C board puts the junction near 71 degrees C, in spec. Running the same power through the still-air RθJA implies a rise near 198 degrees C and an out-of-spec part. Both numbers are correct; they answer different questions, and only the psi JB or RθJB path reflects a real board whose temperature you measured.
Theta JC is the exception where case temperature works, but only under a heatsink that forces nearly all the heat through the case. Modern datasheets separate RθJC(top) for a top-mounted sink from RθJC(bottom) for an exposed pad on a thermal pad. The bottom value is measured with a cold plate under a 2s2p board per the transient dual-interface method of JESD51-14; the older cold-plate approach traces to SEMI G30-88 for ceramic packages. Some suppliers also give a junction-to-sink parameter, psi JS, which you use like psi JT but referenced to the heatsink surface.
Mistakes that produce nonsensical junction temperatures
The recurring errors are few and specific.
Using RθJC(top) to convert a measured case temperature to TJ is the classic one: because it is far larger than psi JT, it inflates the result, which is exactly the nonsensical outcome engineers report before switching to psi JT. Trusting the datasheet RθJA for a real board is the next; that value assumes a JEDEC coupon in still air, not your copper and airflow. Measuring board temperature far from the package, rather than right at its edge, breaks the psi JB assumption and understates TJ. And leaving an exposed-pad part’s RθJC(bottom) path unused, with no thermal vias to a plane, throws away the one low-resistance route the package was built around, so the small datasheet number never materializes.
FAQ
What is the difference between theta JC and psi JT?
Theta JC(top) is measured with nearly all device power forced out the package top into a cold plate, so it describes a part under an efficient heatsink. Psi JT is measured on a normal board with heat spreading through all paths. Use psi JT, not theta JC, to estimate junction temperature from a top-of-package thermocouple; theta JC is much larger and will overstate the result.
Is psi JT a real thermal resistance?
No. Psi JT is a thermal characterization parameter, marked with the Greek letter psi rather than theta because heat leaves the die through several paths simultaneously and no single path carries it all. It still has units of degrees C per watt and is used in TJ = TT + psi-JT times P, but it is not a physical resistance and cannot be back-calculated from theta JA.
How do you calculate junction temperature from case temperature?
Measure the top-center package temperature with a fine thermocouple, then apply TJ = TT + psi-JT times P, where P is total device power and psi JT comes from the datasheet. The equation needs only total power, not the heat split. For a part with psi JT of 1.3 degrees C per watt dissipating 2 W at 85 degrees C top temperature, TJ is about 87.6 degrees C.
What is theta JA used for?
Theta JA (junction-to-ambient) is best for comparing packages and setting a worst-case power ceiling with Pdmax = (TJmax minus TA) divided by theta JA. It is measured in still air on a JEDEC board and depends strongly on that board, so it should not be used to predict junction temperature on a real PCB, where a thermally optimized layout can cut it by 30 to 55 percent.
What is the difference between theta JB and psi JB?
Both relate junction temperature to board temperature. Theta JB is a true thermal resistance measured with heat forced into the board through a ring cold plate; psi JB is a characterization parameter measured on a normal board with natural heat spreading. Psi JB is slightly smaller than theta JB and is the one to use with a board temperature you measured next to the package.
Which JEDEC board are datasheet thermal numbers measured on?
Most vendors report values on the JEDEC High-K board, a four-layer 2s2p construction, per the JESD51-7 board definition and the JESD51 test methods. A Low-K single-layer board gives much higher theta JA. Because the reported numbers are board-specific, JESD51-12 exists to guide how they should be reported and applied.
What to do with these numbers
Start by deciding what you can measure. With nothing but a datasheet, use theta JA only to compare parts and to reject any package where Pdmax leaves no margin. With a board in hand, put a fine thermocouple on the top-center of the package and use psi JT, or on the board edge and use psi JB; both give a junction temperature you can defend. Reserve theta JC for parts actually running under a heatsink, matching RθJC(top) to a top sink and RθJC(bottom) to an exposed pad with vias to a plane. Never convert a bare-package case temperature through theta JC and expect a sane answer. If the psi JT or psi JB estimate lands within about 10 degrees C of TJmax, stop trusting hand calculations and bring in a full thermal simulation.
