GaN-based Class D amplifier design

High-fidelity sound reproduction enthusiasts are the latest beneficiaries of Gallium Nitride’s (GaN) fundamental qualities, as it provides these enthusiasts with a respite in a challenging environment. GaN solves their conundrum about what constitutes an optimal home audio setup.

The basic categories of audio amplifiers are Class A, Class AB, and Class B, which utilize the linear region of their transistors while attempting to recreate a perfect input audio signal with minimal distortion. It has been shown that such designs can achieve theoretical efficiencies of up to 80%, but in practice they are around 65% or less. In today's world of battery-powered smartphones, Digital Enhanced Cordless Technology (DECT) phones, and Bluetooth speakers, this linear approach has become a thing of the past because of the dramatic impact it has on battery life. As with most other areas of the electronics industry, audiophiles have found that using a switching approach offers better promise than linear.

For users who stick to the classic amplifier topology categories, their requirements will focus on accurate audio reproduction with little regard for the overall electrical efficiency of the solution. While this is perfectly reasonable in a home audio environment, many applications require high amplifier efficiency. This may be to save energy and extend battery life, or to reduce heat dissipation, making the end product denser and more compact.

Introduced in the 1950s, the Class D amplifier has always used a pair of switching devices in a push/pull configuration (Figure 1). The duty cycle of the pulse-width modulated (PWM) signal, controlled by the incoming audio signal, ensures that the switching device is either on or off, keeping operation in its linear region to a minimum. This provides the potential for 100% theoretical efficiency and zero distortion.

Figure 1: Basic block diagram of Class D amplifier design

Then, it turned out that the only available germanium transistors were ill-suited to the demands of this switching topology, and as a result, early amplifier designs proved unsuccessful. However, the advent of MOSFET technology gave Class D design auspicious traction. Today, Class D amplifiers find a home in a wide variety of applications for their electrical efficiency. It is also popular where compactness is a design requirement, such as in today’s flat-panel televisions and car head units, because bulky heat sinks are usually not required.

GaN-based high electron mobility transistors (HEMTs) offer a new technology for use as switches in Class D designs with higher efficiency and improved audio quality.

Meets the requirements of Class D amplifiers

In theory, high performance of Class D switching devices requires providing low on-resistance to minimize I2R losses. GaN offers much lower on-resistance than Si MOSFETs and can be implemented in a smaller die area. This, in turn, translates into small packages, which designers can use to bring more compact amplifiers to market.

Switching losses are another factor that needs to be fully considered. At medium and high power output levels, Class D amplifiers perform exceptionally well. However, due to the losses in the power devices, the lowest efficiency is at the lowest power outputs.

To overcome this challenge, some Class D amplifier approaches use two operating modes. This multi-level technique limits the output voltage that the power device can switch to when playing low-volume audio. Once the output volume reaches a predefined threshold, the output voltage rail of the switch is increased, providing a full voltage swing. To further reduce the effects of switching losses, zero voltage switching (ZVS) techniques can be used at low output volumes and hard switching at high power levels.

When implemented with Si MOSFETs, the hard switching mode results in charge buildup in the body diode due to the non-zero voltage at the output when the power device is turned off and on. The reverse recovery charge (Qrr) that then needs to build up needs to be discharged and its time needs to be factored into the PWM control implementation. In designs utilizing GaN, this is not an issue as these transistors have no inherent body diode and therefore no Qrr. The result is overall higher efficiency, improved distortion, and cleaner switching waveforms.

When the amplifier is operated in ZVS mode, the switching losses and the resulting switching power losses are effectively eliminated because the transition of the output is achieved by commutating the inductor current. However, as with all half-bridge designs, there is a need to consider the shoot-through issue, which is the moment when both the high-side and low-side switches are turned on simultaneously. A short delay, called the blanking time, is usually inserted to ensure that one of the switching devices is completely turned off before the other is turned on. It should be noted that this delay can affect the PWM signal, causing distortion in the audio output, so the goal is to make it as short as possible to maintain audio fidelity. The length of time for this delay depends on the output capacitance Coss of the power device. Although GaN transistors have not completely eliminated Coss, it is much lower than the Coss of Si MOSFET devices. As a result, the shorter blanking time allows the amplifier to have less distortion when using GaN.

Despite the improvement, the energy stored in this capacitance still needs to be processed and dissipated during the next on-cycle. However, because the impact of these losses is particularly pronounced at higher switching frequencies, GaN-based designs show higher efficiency than Si-based amplifiers.

Learn How to Realize the Benefits of GaN

GaN HEMT transistors have the same terminal designations as Si MOSFETs, with gate, drain, and source. Their extremely low resistance is achieved by the two-dimensional electron gas (2DEG) between the gate and source, effectively creating a short circuit due to the pool of electrons provided. When no gate bias is applied (VGS = 0 V), the p-GaN gate stops conducting. Unlike its silicon counterpart, GaN HEMTs are bidirectional devices. As a result, reverse current can flow if the drain voltage is allowed to drop below the source voltage. It is important to note that their clean switching is due to the lack of a body diode common to Si MOSFETs (Figure 2). This is responsible for much of the switching noise associated with the PN junction.

Figure 2: Structure of a GaN HEMT transistor

Figure 2a: Superior switching characteristics of a Class D amplifier over Si MOSFETs

Class D amplifier designs have been implemented that can deliver 160 W into 8 Ω without a heat sink. One such prototype uses the IGT40R070D1 E8220 GaN HEMT with the 200 V class D driver IRS20957S (Figure 3). This particular switch has an RDS(on)(max) of only 70 mΩ. If used with a heat sink, the amplifier can output up to 250 W and achieve a very respectable THD+N of 0.008% at 100 W. Switching from ZVS to hard switching causes a hump in the THD+N measurement. Operating at 500 kHz, the design shows no noticeable change in distortion (which occurs at a few watts), and the hard-switching region remains quiet and very clean.

Figure 3: 250 WD Class Amplifier Design

Figure 3a: THD+N measurement

Summary

Si MOSFETs have served Class D amplifier designers well over the years, thanks to advances in optimizing performance. However, further advances in their characteristics are challenging to achieve. Additionally, further reductions in RDS(on) would result in larger die sizes, making it more difficult to build compact audio amplifier designs. GaN HEMTs, however, break through this limitation while also eliminating Qrr. This, combined with reduced Coss and the ability to operate at higher switching frequencies, means that small, compact designs can be created, often without the need to resort to heat sinks. Final THD+N measurements also demonstrate the excellent audio performance that can be achieved with this new technology.