Access Official Technical PDF

Share Summary:LinkedInEmail

GaN power electronics offers designers a way to improve efficiency and power density in applications ranging from compact device chargers to data-center power supplies. The gains are not automatic: they depend on how the semiconductor is used, the converter topology and the control strategy. Comparing GaN with the strongest silicon alternatives therefore means evaluating the complete power-conversion system, not just the switching device.

Enhancement-mode gallium nitride high-electron-mobility transistors (eMode GaN HEMTs) have low gate charge and output charge compared with silicon devices, as well as virtually zero reverse-recovery charge. These properties can reduce switching-related losses and support hard commutation in circuits where silicon Superjunction MOSFETs may be limited. For designers, this can expand the choice of operating modes and make high-frequency, high-density designs more practical.

Why device characteristics affect converter design

Power semiconductor losses depend on how a device switches and on the energy stored in its output capacitance. In single-ended topologies, output-capacitance energy is an important consideration. In half-bridge circuits, output-capacitance charge and reverse-recovery charge are particularly relevant because they affect losses during switching transitions.

Silicon Superjunction devices have advanced through successive generations with lower on-state resistance and improved switching characteristics. However, their behaviour can constrain operation in some circuits. For example, hard commutation can create losses and voltage overshoots that make it unsuitable for certain Superjunction devices. Designers may instead select a control method that avoids hard switching, use a different topology, or distribute voltage stress across cascaded converter stages.

GaN HEMTs can give designers more flexibility. In a totem-pole power-factor-correction (PFC) stage, for example, GaN devices can support continuous conduction mode (CCM), triangular conduction mode (TCM), or a modulation approach that transitions between hard and soft switching according to operating conditions. Silicon implementations may need to remain in TCM to avoid hard commutation. The right choice still depends on the system’s efficiency, cost and size targets.

GaN power electronics in server supplies

Server power supplies are a demanding application because data centers seek both lower energy use and more power within a constrained equipment footprint. A typical high-efficiency architecture combines a bridgeless AC/DC stage, such as a totem-pole rectifier, with a resonant DC/DC stage such as an LLC converter. The output voltage and rectification arrangement influence the design of the downstream stage.

In the evaluated 12 V server-supply designs, optimization comparisons considered the PFC stage, including its EMI filter, as well as the LLC stage and the complete system. The analysis compared silicon designs operating in TCM with GaN designs operating in TCM or CCM. It found that GaN-based designs could improve the efficiency and power-density trade-off, with TCM offering an advantage in the highest power-density region of the PFC comparison.

The whole-system optimization indicated a potential path to approximately 3 kW in a form factor described in the study, with nearly twice the output power of off-the-shelf supplies delivering 1,600 W in that size. The analysis also reported an average efficiency increase of four percentage points without increasing dissipated heat in the power supply. These are design-study results, not guarantees for every implementation; actual performance depends on component selection, layout, thermal design and operating conditions.

Designing compact USB-PD chargers

Universal chargers must accommodate multiple output-voltage profiles while remaining small and efficient. For a 65 W USB Power Delivery (USB-PD) adapter, the evaluated topology options included several flyback configurations. The comparison considered efficiency and power density together, because a compact enclosure must also dissipate the heat produced during conversion.

The asymmetrical PWM flyback was identified as the strongest option among the evaluated candidates. Its operation uses the magnetizing current to achieve zero-voltage switching (ZVS) on the primary-side half-bridge and zero-current switching (ZCS) of the synchronous rectifier. These switching conditions help establish a basis for high conversion efficiency. The design varies switching frequency with input and output conditions.

A 65 W prototype using silicon MOSFETs supported USB-PD output profiles from 5 V at 3 A to 20 V at 3.25 A. Its reported maximum efficiency was 94.8%, with a lowest full-load efficiency of 93% at the stated low-line input condition. The analysis estimated that substituting GaN HEMTs could improve full-load efficiency by around 0.4 percentage points across the input-voltage range. The prototype and the estimated GaN improvement should be understood as distinct results.

Practical considerations for engineering teams

GaN can provide meaningful advantages, but selecting a device technology is only one part of converter development. Engineers should assess the intended operating range, switching method, magnetics, thermal limits and complete system volume. Multi-objective optimization is useful because it makes the trade-off between efficiency and power density visible across alternative topologies and switching frequencies.

  • Choose topology and control together: GaN’s ability to support different switching modes is valuable only when the control strategy uses it appropriately.
  • Evaluate the complete power stage: Include filters, auxiliary electronics, printed circuit boards and thermal constraints in design comparisons.
  • Separate measured and modeled performance: Prototype results, optimization outcomes and projected device substitutions represent different levels of evidence.
  • Match the design to the application: A compact charger and a high-power server supply have different priorities, even when both can benefit from reduced switching losses.

Conclusion

GaN power electronics can help designers advance efficiency and power density in both server supplies and compact USB-PD chargers. The greatest value comes when device characteristics, converter topology and control are considered as an integrated design problem. For industrial and power-conversion teams, the practical takeaway is to compare complete architectures against their real thermal, size and operating requirements rather than treating GaN as a simple component-for-component replacement for silicon.

Access Official Technical PDF

About The Author