Lab measurements of the GTSM40N065D reveal the device’s conduction vs. switching loss split and its junction temperature response under realistic inverter duty cycles — key inputs for thermal design and reliability. This article delivers test methodology, measured loss tables, thermal characterization, and design recommendations so engineers can size cooling, set derating margins, and reproduce results in their labs.
Point: The GTSM40N065D targets medium-power applications where a 650V IGBT class balances blocking voltage and switching efficiency. Evidence: devices in this class are commonly used in motor drives and inverter stages that switch tens of amps at kHz rates. Explanation: understanding the measured loss split between conduction and switching lets designers choose switching frequency, gate drive aggressiveness, and cooling strategy to meet efficiency and reliability targets.
Point: Recommended use-cases include medium-power inverters, motor drives, and SMPS front-ends. Evidence: these applications typically require 650V blocking for margin on 400–600V DC buses and trade off switching loss versus conduction loss. Explanation: designers must weigh frequency, current amplitude and thermal path; measured thermal and loss data are critical when selecting switching frequency or paralleling devices.
Point: Datasheet parameters such as Vce(sat), gate charge, Ic max and Rth(j‑c) directly influence losses and thermal response. Evidence: higher Vce(sat) increases conduction dissipation at low f_sw; larger Qg and faster dv/dt influence Eon/Eoff. Explanation: translate each parameter into action — choose gate resistor and dv/dt limits to trade switching energy for EMI, and size copper/heatspreader to meet Rth targets.
Point: Reproducible loss measurement requires strict control of bus voltage, gate drive, temperature and measurement points. Evidence: measurements here used fixed Vbus, calibrated current probes, and temperature-controlled cold plate to derive consistent Vce and energy waveforms. Explanation: document DC bus, Ic range, f_sw, gate amplitude, rise/fall times and ambient to allow comparison.
Point: Key vectors include Vbus = 400–600V, Ic = 5–40A, f_sw = 20kHz and 100kHz, Vge = 15V, and controlled tr/ tf. Evidence: these vectors capture inverter and SMPS regimes. Explanation: the table below lists representative test vectors and rationale so labs can reproduce energy-per-transition and steady conduction measurements.
| Vector | Vbus (V) | Ic (A) | f_sw (kHz) | Vge (V) | tr/tf (ns) |
|---|---|---|---|---|---|
| Conduction | 400 | 10 / 20 / 40 | DC | 15 | — |
| Switching Low | 400 | 10 / 20 | 20 | 15 | 50/50 |
| Switching High | 600 | 20 / 40 | 100 | 15 | 20/20 |
Point: Use high-bandwidth oscilloscope, calibrated current probes and power analyzer; sample at ≥100 MS/s per transition. Evidence: energy per transition (Eon/Eoff) computed by integrating instantaneous vce×ic over the switching interval; conduction loss from averaged Vce×Ic. Explanation: apply averaging over ≥200 cycles, report measurement uncertainty (~±5–10%) and state filtering/smoothing used to avoid under/over‑estimating energy spikes.
Point: The device shows a conduction-dominant loss at low f_sw and increasing switching contribution at high f_sw. Evidence: measured Vce vs Ic curves and Eon/Eoff tables capture temperature dependence. Explanation: use these data to compute total loss = Pcond + Psw and to project required cooling for continuous or pulsed workloads.
Point: Conduction loss can be approximated by Pcond = Ic × Vce(avg) but integrate Vce(Ic) when non-linear. Evidence: measured Vce at 25°C and 125°C show Vce rise ~10–20% at high Tj, increasing loss. Explanation: sample values — at 20A and 25°C Vce≈1.2V → Pcond≈24W; at 125°C Vce≈1.4V → Pcond≈28W. Use table or curve fits for design automation.
| Ic (A) | Vce @25°C (V) | Pcond @25°C (W) |
|---|---|---|
| 10 | 0.9 | 9 |
| 20 | 1.2 | 24 |
| 40 | 1.8 | 72 |
Point: Eon/Eoff scale with Ic and Vbus and are sensitive to gate rise/fall times. Evidence: measured Eon+Eoff at 20kHz is modest, but at 100kHz switching loss dominates and can exceed conduction loss at higher currents. Explanation: convert energy-per-transition to average switching loss via Psw = (Eon+Eoff)×f_sw; tune gate resistor and dv/dt to meet EMI and loss targets.
Point: Thermal resistance and impedance define steady-state and transient Tj under dissipation. Evidence: measured Rth(j‑c) and time-domain Zth curves map ΔTj vs power and pulse duration. Explanation: use Rth for continuous dissipation sizing and Zth(t) for pulsed workloads to ensure ΔTj stays within safe limits.
Point: Measured Rth(j‑c) on the package and Rth(j‑a) with recommended mounting allow ΔTj calculation. Evidence: for example, P_loss × Rth(j‑c) gives ΔTj above case; adding heatsink and TIM yields junction temperature. Explanation: designer should compute Tj = Tambient + P_loss×Rth(total) and verify Tj
Point: Zth(j‑c)(t) curves from μs to seconds show how short pulses create smaller ΔTj than steady power. Evidence: short pulses (ms range) allow higher instantaneous current before Tj limit. Explanation: derive permissible pulse energy by integrating power over pulse and using Zth to compute ΔTj, then apply duty factor for average heating.
Point: PCB mounting, sufficient copper and proper TIM reduce Rth and extend continuous current capability. Evidence: tests show increasing PCB copper from 1 cm² to 10 cm² per 10W lowers case rise significantly. Explanation: as a rule-of-thumb, allocate ~10–20 cm² of copper per 10 W dissipated and target heatsink Rth that keeps Tj under limit at worst-case ambient.
Point: Use flat, clean mounting surfaces, specified torque, many thermal vias and thin TIM layers. Evidence: proper torque and 10+ vias under the pad reduce Rth(j‑a) substantially. Explanation: recommended: 8–12 M3 torque, ≥12 thermal vias, and TIM thickness
Point: Convert measured losses and Rth into continuous current limits at target ambient. Evidence: example: with P_total = 40W and Rth_total yielding ΔTj=60°C at 50°C ambient, Tj approaches 110°C leaving reliability margin. Explanation: apply a safety margin (e.g., derate continuous current by 20% at 50°C ambient) and limit peak ΔTj to reduce thermomechanical stress.
Point: Consistent measurements require a pre-test SOP and standardized benchmark dataset. Evidence: variability between setups often stems from inconsistent thermal contact and gate drive conditioning. Explanation: use the checklist and CSV template below to publish comparable datasets and reproduce results.
• Verify flatness and torque of mounting; • confirm TIM thickness and via population; • calibrate probes and scope; • set gate drive amplitude and measure tr/tf; • pre-condition device with 10–50 warm-up cycles; • log ambient, case and measured Tj sensors; • average ≥200 cycles.
Point: Publish a minimal dataset: test vector table, Vce vs Ic at Tj, Eon/Eoff vs Ic and Zth curves. Evidence: consistent CSV headers enable cross-comparison. Explanation: include columns: Vbus, Ic, f_sw, Vge, tr, tf, Eon, Eoff, Vce_avg, Tcase, Tj, measurement_uncertainty to ensure reuse.
Measured conduction and switching losses combined with junction thermal impedance determine cooling and derating decisions for the GTSM40N065D; engineers should use the provided loss calculations, Rth curves and Zth pulses to size heatsinks and set conservative continuous-current derates. Use the loss tables and thermal data to target Tj margins and balance switching speed versus EMI for the 650V IGBT application.
Measured conduction loss at 20A is typically ~24W at 25°C when Vce≈1.2V; switching energy depends on Vbus and gate speed, adding 5–30W at higher frequencies. Combine measured Vce and Eon/Eoff data and compute Ptotal = Pcond + (Eon+Eoff)×f_sw for accurate results.
Use Zth(j‑c)(t) to convert pulse energy to ΔTj: ΔTj(t) = Ppulse × Zth(t). For repetitive pulses, compute cumulative heating from duty cycle and ensure steady-state Tj remains within margin. Short pulses permit higher peak current but watch peak ΔTj to avoid material stress.
Practical derating: reduce continuous current by ~20% at 50°C ambient compared with 25°C baseline and target Tj