FPGA Power Design to Drive GaN Adoption

Article By : Maurizio Di Paolo Emilio

Adoption rates for GaN are rising along with power density requirements. To ensure a high level of reliability and safety, thermal management strategies should be applied to any electronic device that generates heat...

Field-programmable gate arrays (FPGAs) are used in a variety of signal-processing applications, but engineers must ensure high accuracy to support the vast demands of the hardware. Form factor and thermal performance are two factors to consider when designing power supplies for modern FGPAs.

Gallium nitride (GaN) is a wide-bandgap (WBG) material that can be used to make semiconductor devices such as diodes and transistors but in a smaller, higher-efficiency form factor than is possible with silicon. GaN minimizes the losses inherent in traditional silicon and therefore does not require silicon’s rigorous thermal management strategies. Adoption rates for GaN are rising along with power density requirements.

To ensure a high level of reliability and safety, thermal management strategies should be applied to any electronic device that generates heat. Maintaining device temperatures within a safe operating area (SOA) is critical to ensuring reliable long-term operation; this becomes increasingly challenging as power densities rise.

Thermal management solutions aim to maximize the thermal efficiency and minimize the size, weight, and cost of the devices used. With the increasing electronic-device content in many applications, problems related to power dissipation in the form of heat have inevitably multiplied. Losses in a typical switching power fall into two categories: switching and conduction.

Thermal design should always be considered in the context of the whole system concept, the application, and whether you can use active or passive cooling methods. At first analysis, it is necessary to select power discrete components that have high efficiency and low power losses. DC/DC converters, widely used in power circuits for industrial applications, offer high conversion efficiency and reduced power losses.

DC/DC solutions

The TPS546D24A buck converter offered by Texas Instruments is capable of delivering up to 160-A output current at an ambient temperature of 85°C, enhancing thermal performance for industrial applications (Figure 1). The TPS546D24A maximizes power density in FPGA power supplies and enables engineers to reduce power loss by 1.5 W in high-performance data centers and enterprise computing, medical applications, wireless infrastructure, and wired network applications. The device offers a switching frequency of 1.5 MHz, with a 0.9-mΩ low-side MOSFET, achieving 3.5% higher efficiency than other solutions, according to TI.

Figure 1: Simplified schematic of the TPS546D24A (Image: Texas Instruments)

The device offers an output voltage error of less than 1% and pin-strapping configurability to monitor current more accurately for fault reporting. Moreover, its integration helps eliminate up to six external compensation components from the board. The buck converter’s PMBus interface provides a selectable internal compensation network, allowing engineers to reduce the overall power solution size by more than 10% for high-current FPGAs per application (ASIC). The TPS546D24A runs 13°C cooler than recent solutions, improving operating reliability in hot and harsh environments.

“PMBus is a serial interface and an open standard that is useful in many markets and applications,” said Rich Nowakowski, product marketing engineer at Texas Instruments. “The 40-A TPS546D24A allows configuration by pin strapping or by PMBus commands, and it reports current via the PMBus interface as a simple single address even when multiple devices are stacked up to 160 A.”

Figure 2: TPSM53604 evaluation board (Image: Texas Instruments)

TI’s TPSM53604, meanwhile, enables engineers to reduce the size of the power solution by 30% and reduce power loss by 50% over earlier devices (Figure 2). Because 42% of the quad flat no-lead (QFN) footprint of the TPSM53604 is in contact with the board, the package allows more efficient heat transfer than competing ball grid array (BGA) packages. The TPSM53604 is capable of operating at high ambient temperatures (up to 105°C) to support demanding applications in industrial automation, network infrastructure, test and measurement, industrial transportation, and aerospace and defense environments.

“The TPSM53604 is the smallest 4-A, 36-V input power module on the market,” said Nowakowski. “The device enables high power density, excellent thermal performance, and low EMI [electromagnetic interference] noise. We were able to achieve these benefits by integrating the inductor and other passive components into a three-dimensional module design and simultaneously solve several design challenges.”

The TPSM53604 power module addresses thermal design by integrating a very high-efficiency IC inside a routable lead-frame quad fan-type package.

GaN technology

In addition to improvements in conduction losses, the use of GaN enables a significant reduction in switching losses by increasing the speed at which the switch is turned on. Increasing the switching frequency also reduces the size of many large components (such as the transformer, inductors, and output capacitors). GaN has better thermal conductivity and can withstand higher temperatures than silicon. Both attributes contribute to reducing the need for thermal management components such as bulky heat sinks and cooling, leading to significant reductions in the overall size and weight of the power supply.

“The faster your frequency, the smaller your magnetics, the smaller your capacitors, and the higher the density,” said Steve Tom, product line manager for GaN technology at Texas Instruments. “The beauty of GaN is that we can switch faster without that thermal penalty, which is why power density and efficiency are so high.”

A major concern is the reliability of the devices. “We’ve done over 30 million device hours,” said Tom. “As of this year, we have processed 3 GW of power conversion to show that our devices have strong SOA. They are targeted very well for power supply and inverter applications.”

The family of GaN solutions integrates high-speed gate driver, EMI control, overtemperature, and overcurrent protection with a 100-ns response time. Integrated devices offer an optimized layout to minimize parasitic inductance, maximize common-mode transient immunity (CMTI, measured in dV/dt), and reduce board space.

“One of the unique benefits we offer at TI is our GaN supply chain,” Tom said. “TI owns the GaN process and operates the entire manufacturing flow, from [epitaxy] to packaging and testing. And then we couple the GaN FET with an optimized driver to enable the highest-speed, highest-performing, highest-reliability system.” TI has a complete portfolio of 150-mΩ, 70-mΩ, and 50-mΩ 600-V GaN FETs in mass production. The devices are enabling high- density designs in industrial, telecom, server, and personal electronics applications.

Figure 3: GaN applications (Image: Texas Instruments)

To extend GaN’s use in new applications such as automotive, grid-tied storage, and solar energy (Figure 3), TI is demonstrating its own convection-cooled, 900-V, 5-kW bidirectional AC/DC platform. The platform can provide a peak efficiency of 99.2% without a cooling fan, with a scalable, multi-level solution for 5 kW with natural convection. It includes the LMG3410R050 GaN and the C2000 digital controller, and it supports bus voltages up to 1.4 kV.

Leave a comment