Estimating DAB converter IGBT losses is an important step in designing a reliable, efficient bidirectional DC-DC power stage. Device selection affects more than the converter’s nominal voltage and current capability: engineers must also account for switching behavior, antiparallel diode losses, thermal limits, and the operating conditions that produce the highest electrical stress. A structured loss model can help compare candidate modules and connect datasheet information to the needs of the application.
A 20 kW, 540/125 V, 20 kHz dual active bridge (DAB) converter developed for an aerospace energy storage system illustrates this approach. Its high-voltage (HV) bridge connects to an aircraft DC link, while the low-voltage (LV) bridge interfaces with an ultracapacitor-based storage system. The design process examined the two bridges separately because their device current waveforms and operating stresses differ.
Why DAB converter IGBT losses depend on operating conditions
A DAB converter uses two active bridges connected through a high-frequency transformer. The phase shift between the bridges controls power transfer, and the converter can operate bidirectionally. In the evaluated design, the worst-case condition was associated with a 2:1 ultracapacitor voltage range. At that operating point, the analysis used 540 V on the HV side, 62.5 V on the LV side, 20 kW throughput, and a 20 kHz switching frequency. The calculated inductor current reached 427 A RMS and 640 A peak.
These values matter because losses depend on the actual current carried by each transistor and diode over its conduction interval—not simply on the converter’s rated output. The HV- and LV-side bridges have different current waveforms due to the phase shift, so the device-current models must reflect each bridge’s conduction intervals. Assuming piecewise-linear current waveforms allows those intervals to be used to estimate average device currents.
Estimating conduction and switching losses
Conduction losses in transistors and diodes
For an IGBT, conduction loss is estimated from its average on-state current, on-state voltage drop, and duty cycle. The forward voltage drop is not always a single fixed value: it can include a chip-level component and a contribution from the module terminals. At high power density, terminal losses may be significant enough that they should be considered separately when interpreting device data.
Antiparallel diode conduction loss is estimated using the diode’s forward voltage drop, average current, and conduction interval. The total bridge conduction loss then accounts for the four transistors and four antiparallel diodes on that side. In a bidirectional converter, the relative contribution of each device type changes with power-flow direction. In the reported LV-side analysis, diode conduction losses were prominent during forward, or charging, operation; when power flow reversed, the loss contributions shifted between the diodes and IGBTs.
Switching losses and soft switching
Where manufacturer data is available, switching loss can be calculated from turn-on and turn-off energy curves at the relevant switching current and frequency. If those curves are unavailable, a simplified estimate can use the device voltage, current, switching transition time, and frequency. Such approximations are useful for comparison, but they depend on assumptions about the switching waveforms.
Soft-switching conditions also influence which losses need to be included. In the analyzed DAB operation, zero-voltage switching (ZVS) and zero-current switching (ZCS) affect the transitions of the IGBTs and diodes. The modeled operation allows turn-on losses to be omitted for the IGBTs and reverse-recovery losses to be omitted for the diodes under the stated switching conditions. Turn-off losses remain important, particularly on the HV side, where the analysis found them to be a major contributor to device losses.
Device selection requires electrical and thermal checks
Loss estimates are only one part of choosing a power semiconductor. The selection process also needs to consider blocking voltage, current capability, output characteristics, switching performance, gate charge and driver requirements, leakage current, safe operating area, and thermal characteristics. Device ratings should not be exceeded during operation, including under foreseeable voltage transients.
For the converter examined, 1200 V IGBT modules were selected for the HV bridge, and 600 V modules for the LV bridge. Candidate devices from multiple manufacturers were compared using predicted losses at the design’s worst-case conditions. A Semikron SKM300GB125D module was selected for the HV side based on the loss comparison. For the LV side, several modules showed similar loss performance; the Semikron SKM600GB066D was selected based on its lower losses in the comparison and its stated maximum junction temperature of 175°C, compared with 150°C for the other modules considered.
LV-side device selection also illustrates why technology choices must reflect the application. The secondary side carries high current at relatively low voltage. Using multiple MOSFETs in parallel was considered, but doing so would add circuit complexity. IGBT power modules offered a practical alternative for the required high-current handling and module integration.
Snubbers can trade loss reduction for transient control
Experimental tests assessed the HV-side devices at 540 V and an 80 A peak current, and the LV-side devices at 125 V and a 300 A peak current. On the HV side, a 47 nF snubber capacitor across each IGBT reduced turn-off losses; the reported comparison showed a nearly 45% reduction in IGBT switching losses. The observed waveforms were consistent with soft-switching behavior under the tested conditions.
The LV-side tests showed a different result. A 100 nF snubber reduced the switching-transient rate of voltage change (dv/dt) and therefore reduced device stress, but it did not produce a significant reduction in switching losses. Snubbers also introduced high-frequency ringing, attributed as a possible resonance involving module, busbar, and connection parasitic inductances. This makes snubber selection a design tradeoff: lower transient stress does not necessarily mean lower switching loss, and the complete physical layout must be considered.
Use calculated losses to guide, then verify, the design
For engineers and system integrators, the practical lesson is to evaluate each bridge at its own worst-case operating point, model the actual device-current intervals, and include both semiconductor and thermal constraints in the comparison. Datasheet-based estimates can narrow the field of candidate IGBTs, while experimental measurements can check whether the model reflects the prototype’s behavior. In the reported work, calculated and measured losses showed close correlation, supporting the current models used for the analysis.
Snubber decisions should likewise be made against the intended objective. The HV-side capacitor reduced measured switching losses, while the LV-side capacitor primarily reduced dv/dt and device stress. Together, the loss calculations and tests show why power-device selection for a DAB converter is not a ratings-only exercise: it requires attention to current shape, switching conditions, thermal limits, and the effects of the implemented circuit.