Archives | Feedback

Class-D Amplifiers, Coming of Age

by Nicholas Holland, Texas Instruments

Over the last two to three years, designers have enhanced many end-equipments with new features, one of which is increased audio functionality. This trend has impacted many consumer products, such as flat panel displays, PDAs and mobile phones. As the requirements for performance have grown, so has the need for richer feature sets in audio amplifiers, including the basic need to drive better voice quality at higher power levels into lower impedance loads. Traditionally, the Class-AB amplifier has been able to cope well with the early performance and cost demands of these end-equipments, however, the characteristics of linear amplifiers no longer suit the needs of the consumer. As a result, Class-D amplifiers are challenging the Class-AB (a.k.a. linear amplifiers) for use in consumer products that provide enhanced audio capability. Today, many end-equipments, such as LCD TVs, plasma TVs and desktop replacement PCs are driving the necessity for more output power, at similar cost, while maintaining or even reducing the form-factor. This trend is fueling the need for more Class-D devices and opening up many markets previously served by the traditional linear amplifier.

The Class-D amplifier
The main reason why consumer product vendors are moving to Class-D amplifiers is there very high efficiency, which directly translates into half the heat compared to a linear amplifier. The graph in Figure 1 shows clearly that the Class-D amplifier reaches 85 percent efficiency quickly as power increases, compared to the linear amplifier whose efficiency increases more slowly as output power increases. Most people listen to TV in the 2-4W range. At that level, when comparing the Class-D to the linear amplifier, one can see four times increased efficiency in the Class-D amplifier. The efficiency improvement is inversely proportional to the amount of heat generated within the amplifier. Therefore, for the same output power, the linear amplifier needs a large heat-sink, which increases the form-factor compared to the Class-D amplifier.

Figure1

Figure 1: Efficiency vs. Output Power

To illustrate further the practical benefits of Class-D in terms of heat dissipation versus output power, Texas Instruments (TI) retrofitted the linear power amplifiers of an un-named brand of radio or head unit (HU) with Class-D amplifiers. The experiment called for a thermister placed on the heat sink with the ambient temperature measured at varying output powers. Figure 2 shows the performance difference between the linear amplifier and the Class-D amplifier, illustrating the point that Class-D technology can effectively increase output power while decreasing the size and cost of the consumer product.

Figure2

Figure 2: Temperature vs. Output Power

So how is this efficiency achieved? A Class-D amplifier works similarly to a switched-mode power supply, where the output MOSFETs are either fully on (saturated) or fully off (cutoff). This has the effect of reducing the power dissipated by the transistors, increasing the efficiency of the amplifier. Unfortunately, there will always be some loss during the switching (switching loss) and non-switching times (conduction loss).

During switching times, losses occur because the rise and fall times of the FETs are greater than zero. This occurs for several reasons. First, the output transistors cannot switch instantaneously. The channel from drain to source requires a specific period of time to form. Second, the transistor gate-source capacitance and parasitic resistance in traces form RC time constants that also increase rise and fall times.

During non-switching times, power is dissipated because of the RDS(ON) of each FET and the current flowing through the transistor.

But overall, the losses in Class-D amplifiers are minimal and it is due to this switching nature of the device that the amplifier achieves its much higher efficiency. The switching technique is Pulse-Width Modulation (PWM), which compares the input analog signal to a high-frequency triangle waveform (typically 250 kHz) to generate an output waveform. This waveform then drives the MOSFET H-bridge. The resulting differential waveform is a PWM square-wave signal with a duty cycle that is proportional to the amplitude of the audio signal. The signal from the H-bridge drives the speaker via an output filter or is connected directly to the speaker (see Texas Instruments TPA2000D and TPA3000D Filter-Free families). Figure 3 shows the typical configuration of a Class-D output stage for a Bridge-Tied Load (BTL) configuration.

Figure3

Figure 3: Typical Configuration for a Class-D Amplifier

The modulation scheme plays an important role in determining the type of filter required. For example, the first generation of TI Class-D amplifiers required an LC filter. Figure 4 shows the first type of modulation scheme used. In this scheme, when no input signal is present the differential PWM output signal has a duty cycle of 50 percent. The 50% duty cycle does not produce an audible sound, because the average waveform is zero. However, it sinks and sources a large current through the speaker, which leads to unwanted power dissipation. Now, as the input voltage increases, the duty cycle of the positive terminal OUT+ increases and the duty cycle of the negative terminal OUT- decreases.

Figure4

Figure 4: Output Voltage and Current Waveform for Traditional Class-D Modulation

For this type of modulation scheme, a second order Butterworth low-pass filter should be implemented. Shown in Figure 5, this filter uses two inductors and three capacitors for a typical bridge-tied load output. Primarily, this filter acts as an inductor to keep the output current constant while the voltage is switching, which reduces the power dissipation at low or no input signals.

Figure5

Figure 5: Typical Second Order Butterworth Filter Design

The main disadvantage of the filter is additional size and extra component cost. This type of modulation scheme can be used without a filter without affecting the fidelity. However, gains in efficiency would be lost due to the nature of the speaker being both resistive and reactive and the fact that the Class-D switching waveform applies a large voltage across the speaker. This leads to a higher supply current and cancels the advantages that Class-D brings in efficiency. Higher inductance at the output yields lower quiescent current (supply current with no input), because it limits the amount of output ripple current. Inductors L1 and L2, and capacitor C1 form a differential filter that attenuates the signal with a slope of 40dB per decade. The majority of the switching current flows through C1, C2 and C3, leaving very little current dissipated through the speaker. The filter also greatly reduces electromagnetic interference (EMI). EMI is caused either by an instantaneous change in current resulting in a magnetic (H) field or by a differential voltage resulting is an Electric (E) field. Since the filter in Figure 5 includes a common-mode and differential filter, this attenuates both the H and E fields. In the new generation of Texas Instruments Class-D amplifiers, the TPA2000D and TPA3000D family, the modulation scheme has been modified so that only a very short differential power pulse occurs to prevent ıshoot-throughı when there is no input signal. This results in a supply current increase of less than 3mA, for the TPA2005D1, with a load that appears inductive or resistive at the switching frequency. Figure 6 shows the output waveforms of TIıs Filter-Free Class-D modulation scheme.

Figure6

Figure 6: Output Voltage and Current Waveform for Next Generation Class-D Modulation

This innovative modulation scheme removes the need for the second order Butterworth low-pass filter, thus reducing system cost and solution size. EMI may still be an issue, but actual laboratory tests show that a ferrite bead and capacitor to ground placed in series to the amplifier output actually acts as a common-mode filter and therefore attenuates the E field or, in other words, reduces the amplitude or the switching harmonics in the MHz range (see Figure 7). Typical capacitor and ferrite bead values used are 1 nF and 100 ohms at 100 MHz respectively. This is good for circuits that have to pass FCC and CE because FCC and CE test radiated emissions greater than 30 MHz.

Figure7

Figure 7: Next Generation Class-D Filter Design

By using this modulation approach, where the positive and negative output signals are in phase, the differential voltage across the load is 0 volts throughout most of the switching period This greatly reduces the switching current, which eliminates power losses in the load.

Conclusion
Choices of advanced Class-D amplifiers are increasing in breadth and enabling designers across consumer end-equipments such as flat panel displays, PDAs, smart phones, mobile phones, car radios and many others to increase power performance, while maintaining or even decreasing form-factor and cost. The age of the Class-D amplifier is enabling designers to implement high-performance audio in a myriad of products, making the consumer experience even ıcoolerı than before.

About the author
Nicholas Holland is in Products Marketing with Texas Instruments, responsible for Power Audio Products. He holds a First Class Masters of Engineering degree from University of Bristol

Analog Main | Product of the Week | Columns | Editorial | Tech Notes

		Click here to get your listing up.