Why Peak Pulse Power Cannot Predict TVS Junction Temperature

A TVS diode's peak pulse power rating does not provide its junction temperature during a surge. The rating is tied to a specified waveform, starting temperature, and mounting condition. Junction-temperature estimation needs the actual power-versus-time waveform and a transient thermal impedance or validated thermal model for the specific device. Without those inputs, multiplying peak power by steady-state thermal resistance produces an unsupported number.

This mistake often appears after the voltage checks look complete. The selected TVS has a reverse working voltage above the normal rail and a stated clamp voltage below an absolute-maximum value. Someone then asks whether twenty or one hundred pulses will overheat the diode. A spreadsheet using peak pulse power and RθJA seems convenient, but it combines a transient electrical rating with a steady-state thermal quantity.

The familiar temperature-rise equation is being used at the wrong time scale

For steady dissipation under a defined board condition, temperature rise is often approximated from power and thermal resistance. A surge is not steady. During a short pulse, heat first remains in the active silicon and nearby material. It then spreads through the die attach, package, solder joints, copper, and board. Each stage has its own thermal capacity and time constant.

Transient thermal impedance, written as Zth(t), describes that time-dependent response. For a rectangular power pulse and a matching thermal reference, a first-order estimate may be written as:

ΔTj ≈ P × Zth(t)

The equation is not a license to use any Zth curve. The curve must correspond to the same device, thermal path, pulse duration, duty cycle, and mounting condition. A junction-to-case curve cannot be substituted for a junction-to-ambient calculation without a model for the remaining path. A curve measured on one copper footprint may not represent a compact production layout.

Peak power leaves out most of the waveform

TVS power changes throughout a surge. The instantaneous value is the voltage across the diode multiplied by the current through it at the same time. Measuring voltage in one experiment and using a current from a data-sheet table can create a combination that never occurred.

The peak is only one sample in that time record. Two pulses can have the same peak power and very different duration or decay. Their energy, thermal distribution, and peak junction temperature can differ. Two other pulses can have similar energy but deliver it at different rates, again producing different temperature histories.

The minimum electrical record for a thermal review includes:

• the source waveform, impedance, polarity, and tolerance;

• diode voltage and current measured on a common time base;

• probe bandwidth, delay alignment, and measurement location;

• pulse duration and the complete power waveform;

• the interval between pulses and total pulse count;

• ambient, board, or case temperature before the sequence.

Probe delay deserves attention. A voltage probe and a current probe may have different propagation delays. Multiplying unaligned waveforms can create a false power spike or suppress the real one. Deskew the channels with a suitable fixture or document the remaining timing uncertainty.

A thermal calculation needs a chain of evidence

For a nonrectangular power waveform, an engineer may divide the pulse into rectangular sections and apply superposition, or use a validated electrothermal model. The approximation has to preserve the relevant peak and energy behavior. A triangular pulse should not be replaced by a rectangle merely because the areas match if the peak-temperature error has not been bounded.

If the thermal model is unavailable, the honest answer is that junction temperature has not been calculated. That does not stop engineering work. It changes the validation method.

What to do when the TVS data sheet has no Zth curve

Many TVS data sheets provide peak pulse power curves and temperature derating but no transient thermal impedance suitable for a junction-temperature calculation. A practical evaluation can still establish whether a candidate is accumulating heat or drifting electrically on the target board.

Begin with the actual source network. Record open-circuit voltage, source impedance, waveform, polarity, and repetition interval. Measure diode current and terminal voltage on the production-representative PCB. Calculate the time-resolved electrical power and integrate it for pulse energy, while keeping measurement uncertainty visible.

Next, repeat the sequence at the relevant starting temperature. Surface temperature from an infrared camera is not junction temperature, especially for a short pulse. It can still be used for controlled A/B comparisons if emissivity, focus, frame rate, and viewing area are consistent. A small thermocouple provides another trend measurement but adds thermal mass and may not follow the fastest event.

After the required sequence, allow the device to return to the defined measurement temperature and recheck leakage current, breakdown voltage, appearance, and any application-specific clamp measurement. A diode that is not shorted may still have changed. A handheld continuity test cannot detect all relevant damage modes.

If calculated junction temperature is a formal qualification requirement, request one of the following from the TVS diode manufacturer:

1. device- and package-specific transient thermal impedance under stated mounting conditions;

2. a validated compact thermal or electrothermal model;

3. test data for the required pulse sequence and starting temperature;

4. the limits and uncertainty associated with any recommended calculation method.

ASIM's TVS portfolio contains different working-voltage, package, and peak-power groups. Those catalogue fields are useful for screening. They do not show that two nominally equal-power parts have the same temperature rise. The comparison has to return to waveform, thermal path, and post-stress behavior.

Repetition changes the starting point of every pulse

After one pulse, the junction cools through several thermal paths. If the next event arrives before that transient has decayed, it begins from a higher temperature. A statement such as “twenty pulses” is incomplete without the interval. Twenty pulses separated by a minute are not thermally equivalent to twenty pulses delivered one second apart.

For repeated events, record pulse width tp, period T, duty cycle, count, and any cooling break between groups. Then track at least three trends: surface or board temperature, leakage current, and clamp behavior at a controlled current. A rising trend can be more informative than the final binary result.

Do not continue increasing stress after an unexplained parameter shift. Stop, measure the device, and inspect the current path. A later catastrophic short can erase evidence of the earlier mechanism.

Derating curves answer a narrower question

A temperature derating curve usually reduces allowable peak power as an ambient, lead, or case temperature rises. It helps determine whether a data-sheet rating remains available at a hotter starting condition. It does not automatically report junction temperature after the target pulse.

A pulse-duration curve shows allowable peak power versus pulse width under its stated assumptions. Read the waveform definition and reference temperature before using it. Do not extrapolate beyond the plotted range. If the production board has less copper or sits beside a hot converter, treat the catalogue condition as a reference, not as proof of the final assembly.

Can pulse energy alone determine the TVS junction temperature?

No. Energy is useful for controlled comparisons, but the same energy delivered with a different peak or time distribution can produce a different temperature history. The thermal impedance and starting condition are still required.

Is infrared temperature the same as junction temperature?

No. Infrared equipment observes package-surface radiation and has finite spatial and time resolution. Use it for repeatable trend comparisons unless a validated correlation to junction temperature is available.

Does a larger TVS package always run cooler during a surge?

Not in every pulse regime. Die size, package construction, pulse duration, solder pattern, copper, and the part's electrical clamp behavior all matter. Compare complete device data under matched conditions.