Wide Bandgap Technology to Maximize Efficiency and Power Density in High-Voltage LED Lighting

High-voltage LED lighting has proven to be a viable replacement for previous technologies such as high-intensity discharge (HID) lighting. With the adoption of high-voltage LED lighting, many manufacturers rushed to production and implementation in a variety of applications. While there was a significant increase in light quality and power density, efficiency has become an important aspect to address. Also, early applications saw failure rates that were much higher than expected. The main challenge of high-voltage LED lighting is to continue to increase power density and efficiency as well as making it reliable and more affordable for future applications. In this article, wide bandgap (GaN) technology will be covered and how it can address the efficiency and power density challenge for high-voltage LED lighting. This discussion will show how wide bandgap technology can be used to maximize the efficiency and power density, with a focus on the buck portion of the LED driver architecture shown in Figure 1.

Wide bandgap (GaN) semiconductors can operate at higher switching frequencies compared to conventional semiconductors like silicon. Wide bandgap materials require a higher amount of energy to excite an electron to have it jump from the top of the valence band to the bottom of the conduction band where it can be used in the circuit. Increasing the bandgap, therefore, has a large impact on a device (and allows a smaller die size to do the same job). Materials like Gallium Nitride (GaN) that have a larger bandgap can withstand stronger electric fields. Critical attributes that wide bandgap materials have are high free-electron velocities and higher electron field density. These key attributes make GaN switches up to 10 times faster and significantly smaller while at the same resistance and breakdown voltage as a similar silicon component. GaN is perfect for high-voltage LED applications, as these key attributes make it ideal for implementation into future lighting applications.

technology-fig1.jpg?la=en&ts=7884fd60-917c-4129-be5c-f721941801f4″ style=”height:176px; width:600px;” _languageinserted=”true” title=”System architecture of a non-isolated high-power LED driver” alt=”Wide Bandgap Technology to Maximize Efficiency and Power Density in High-Voltage LED Lighting”>Figure 1: System architecture of a non-isolated high-power LED driver. (Image source: STMicroelectronics)

Figure 1 shows a high-level architecture of an LED lighting application that will serve as a baseline example for applying GaN wide bandgap technology. Although wide bandgap materials can be implemented across the application, the high-voltage current generator buck, highlighted in green, will be the focus to leverage wide bandgap technology for maximizing efficiency and power density. Most lighting applications require high power factor and low harmonic distortion across a wide AC input voltage range. In this case, it is preferred to implement a PFC boost to provide a clean 400 VDC input for the LED driver and meet power quality requirements. There are multiple options for a front end PFC boost converter; transition mode (TM), continuous conduction mode (CCM) as well as others. Transition mode is characterized by variable frequency operation and zero current switching at turn on of the power MOSFET. Other advantages are simple design, small inductor size, and no reverse recovery of the boost diode. The main challenges are high peak and RMS input current, which also results in a larger EMI filter as the power increases. CCM, instead, provides fixed frequency operation. The boost inductor current always has an average component, besides near zero crossing points. The inductor is designed for 20-30% ripple, resulting in a smaller EMI filter compared to TM operation. This also means a larger boost inductor and a smaller EMI filter for the same output power when compared to TM operation. The main challenges are more complex control and the need for an ultrafast soft recovery diode or SiC diode. Consequently, the CCM PFC is generally more expensive than a TM PFC. Ideally, a zero reverse recovery switch can be used in place of the rectifying diode in CCM PFCs. This makes GaN transistors very good candidates for this application.

Isolation is optional and can be introduced between the input stage and the second stage of power conversion. In this example, isolation is not used, and the input PFC stage is followed by a non-isolated inverse buck stage with CC/CV control. In the cases where isolation is needed, a resonant power converter (LLC, LCC) or a flyback converter can be used depending on the output power requirements of the application.

The PFC boost converter generates a regulated DC bus voltage on its output (higher than the peak of the input AC voltage) and passes this higher DC bus voltage to the inverted buck converter stage. The stepdown operation is quite simple. When the switch in the buck is on, the inductor voltage is the difference between the input and output voltages (VIN – VOUT). When the switch is off, the catch diode rectifies the current and the inductor voltage is the same as the output voltage.

MasterGaN system in package (SiP) for LED drivers

Along with power density and efficiency, a key challenge for high-voltage lighting applications is the complexity of the design. With the use of wide bandgap semiconductors like GaN, the power density and efficiency of the circuit can be increased. ST’s MasterGaN family addresses that challenge by combining the high-voltage smart-power BCD-process gate drivers with high-voltage GaN transistors in a single package. MasterGaN allows for an easy implementation of the topology shown in Figure 1. It embeds two 650 V GaN HEMT transistors in Half-Bridge configuration as well as the gate drivers. In this example, the entire buck power stage is integrated into a single QFN 9×9 mm package requiring minimal external component count. Even the bootstrap diode, typically needed to supply the isolated high-voltage section of a dual, high-side/low-side, Half-Bridge gate driver, is embedded into the SiP. Consequently, the power density of an application that uses a MasterGAN device can be increased dramatically compared to a standard silicon solution while increasing the switching frequency or the power output. More specifically, in this LED driver application, a 30% decrease in PCB area was achieved and no heat sinks where used.

For high-power LED lighting applications, CCM is the best operating mode to use. When implementing CCM with GaN devices, there will be the high-level benefits previously discussed as well as a reduced cost. There would be no need for very low RDSON to serve high power applications due to the reduced switching loss contribution to overall power losses. GaN also mitigates a major drawback of using CCM by eliminating recovery losses and reduced EMI, as GaN experiences no reverse recovery. CCM operation with Fixed Off Time control also makes the compensation of output current ripple dependency on VOUT very easy. It is clear that GaN switch implementation using CCM is a great combination for high-voltage LED lighting applications, as well as many others.

The basic scheme of an Inverse Buck topology is shown in Figure 2 along with an implementation that uses the MASTERGAN4.

technology-fig2.jpg?la=en&ts=6ee5b8f7-2cce-492b-a590-1f07dd743d8c” style=”height:616px; width:600px;” _languageinserted=”true” title=”Inverse buck topology implemented with STMicroelectronics MASTERGAN4 (click to enlarge)” alt=”Wide Bandgap Technology to Maximize Efficiency and Power Density in High-Voltage LED Lighting”>Figure 2: Inverse Buck Topology implemented with MASTERGAN4. (Image source: STMicroelectronics)

MASTERGAN4 embeds two 225 mΩ (typical at 25°C) 650 V GaN transistors in Half-Bridge configuration, a dedicated Half-Bridge gate driver and the bootstrap diode. This high level of integration simplifies the design and minimizes PCB area in a small 9×9 mm QFN package. The evaluation board that is shown in Figure 3, was designed with the MASTERGAN4 in an inverse buck topology has the following specifications: it accepts up to 450 V input, the output voltage of the LED string can be set between 100 V and 370 V; it operates in Fixed Off Time (FOT) CCM with a switching frequency of 70 kHz; the max output current is 1 A.

technology-fig3.jpg?la=en&ts=2cd58920-b291-47e0-91ca-eb7b856e2a6f” style=”height:477px; width:381px;” _languageinserted=”true” title=”Inverse buck demo with STMicroelectronics MASTERGaN4″ alt=”Wide Bandgap Technology to Maximize Efficiency and Power Density in High-Voltage LED Lighting”>Figure 3: Example of Inverse Buck Demo with MASTERGaN4. (Image source: STMicroelectronics)

The controller in this solution, the HVLED002, is used to generate a single PWM control signal. An external circuit based on simple Schmitt Triggers is then used to generate two complementary signals to drive the low side and high side GaN transistors with a suitable dead time. Two linear regulators are also included to generate the supply voltages needed by the MASTERGAN4. The inverse buck topology implemented with MASTERGAN4 creates a solution for increased power density and efficiency, but let the results discussed below speak for themselves.

Experimental Results:

The efficiency plots in Figure 4 show the advantages of the proposed solution vs. a traditional silicon solution as a function of the LED string voltage for output currents of 0.5 A and 1 A.

technology-fig4.jpg?la=en&ts=16bab7f1-92af-4dbb-ba5d-054b7904e1f7″ style=”height:373px; width:600px;” _languageinserted=”true” title=”Efficiency vs. LED voltage for MasterGaN and silicon MOSFET” alt=”Wide Bandgap Technology to Maximize Efficiency and Power Density in High-Voltage LED Lighting”>Figure 4: Efficiency vs. LED voltage for MasterGaN and Silicon MOSFET. (Image source: STMicroelectronics)

The efficiency of MASTERGAN4 stays at or above 96.8% across the entire LED string voltage range. It is possible to observe that across all power levels the gain in efficiency is maximized thanks to the low conduction losses as well as the minimal driving and switching losses of the GaN solution.

MOS + SiC Diode MASTERGAN4
Power devices area 0.66 cm²
Diode DPAK or TO220
0.81 cm²
Copper area for thermal management 33 cm²
Copper area to have 19°C/W
19.7 cm²
Copper area to have 24°C/W
Power inductor footprint 11.2 cm² 11.2 cm²
Overall Area 45.5 cm² 31.71 cm²

Table 1: Size comparison for GaN and Silicon MOSFET

Table 1 compares the silicon solution with the MASTERGAN4 based solution. As can be seen, more than 30% overall PCB area reduction is shown with the GaN design implementation. The results show one path that can be taken with GaN in this inverse buck topology. Increasing the switching frequency above 70 kHz can decrease the output inductor and capacitor size at the expense of higher driving and switching losses. At a higher frequency and reduced filter size, electrolytic capacitors can be replaced with more reliable and larger ceramic capacitors. The tradeoff between filter capacitor and buck inductor size can be optimized based on the switching frequency required by the target application.

Conclusions

This article discussed the implementation of an inverse buck topology for LED lighting applications based on MASTERGAN4. The system in package configuration has 650 V, 225 mΩ GaN transistors in half-bridge configuration and dedicated gate drivers. The GaN solution vs. silicon shows higher efficiency and reduced PCB area. MasterGaN is the ideal solution for a compact, high efficiency and high-power inverse buck implementation for lighting applications.