Decades of development and high-volume production generated mature amplifier technologies that satisfy the requirements of many market segments. Class A designs address more demanding high-end applications, while Class B and AB amplifiers serve the consumer market with cost, power, and performance tradeoffs.
Class D technology, however, has changed the balance. With Class D, the power efficiency of audio amplifiers-once constrained to a range roughly between 30% and 50%-has just about doubled, reaching 85% or even 90%. For a given output power, this allows the supplies, heatsinks, and housing to be much smaller. Alternatively, a given budget will buy a louder amp.
This vast improvement stems from the fact that power transistors in Class D output stages are switched, rather than controlled, by analog signals-as is the case with traditional amplifiers. Because each transistor either has zero current flowing through it (in the "off" state) or zero voltage across it (in the "on" state), the power dissipated (given by the product of current and voltage, P = IV) is close to zero at all times-in theory. In the real world, the transistors' "on" resistance causes a small voltage drop in the "on" state. Other causes of power dissipation in Class D amplifiers include the power transistors' relatively large gate capacitance and system components preceding the output stage.
In practice, it takes additional circuitry to produce analog waveforms using a switching device. A modulator transforms digital or analog source signals into a pulse-width-modulated (PWM) signal prior to the output stage. A low-pass reconstruction filter converts the amplified PWM signal to an analog waveform for the loudspeaker.
In small amplifiers, the modulator and output stage are sometimes combined in a single piece of silicon. However, the output power produced by these chips remains limited due to inherent incompatibilities between the manufacturing processes for DSPs and high-power switching devices.
A number of circuit topologies have emerged, including fully integrated devices and multichip solutions where the modulator is separate from the output stage. The latter has the advantage of easy scalability. The same design can be reused for amplifiers with different output powers, because only the output stage and power supply need to change (Fig. 1).
By segmenting the system in this way, the output stage can be built in a robust, high-voltage technology suited to power switching devices while employing a generic CMOS process for the modulator. Digital-signal-processing functions such as equalization, dynamic compression, or digital volume control then can be more easily integrated with the modulator.
Power-Supply Issues And Remedies
Class D technology introduces some new issues of its own. It's often forgotten that Class D amplifiers are more sensitive to the quality of the power supply. Unlike their linear counterparts, whose bandwidths are limited to the audio range, Class D output stages switch at frequencies of several hundred kilohertz.
To accurately produce these square pulses, the supply must be able to increase or decrease its current output very rapidly without ringing or a drop in the output voltage. This requires reservoir capacitors that can hold enough charge to prevent a current surge from causing the supply voltage to drop. Because any parasitic resistance or inductance will impede a rapid delivery of the stored charge, these capacitors should feature a low effective series resistance (ESR).
Adding a small, low-ESR capacitor in parallel with a larger, conventional electrolytic capacitor isn't sufficient. Because all of the output power is delivered in short bursts, all capacitance must be low-ESR. Parasitic resistances and inductances in pc-board copper tracks are equally detrimental and should be minimized by placing the reservoir capacitors as close to the output stage as possible.
Arranging for the transistors in the different output stages to switch on sequentially rather than simultaneously can alleviate demands on the supply's transient behavior. To this end, advanced PWM modulators feature a built-in "PWM phase-shift" function that introduces a fixed delay between the PWM signals for each output channel.
At a fraction of a PWM cycle, this delay is far too short to make an audible difference to the output, but it spreads switching transients around the PWM cycle (Fig. 2). This technique diminishes instantaneous load-current changes by a factor that increases with the number of channels.
Many Class D amplifiers also provide little or no power-supply rejection (PSR), effectively using the supply as a voltage reference. To prevent mains or audio-band ripple from modulating the signal, the supply must be regulated. A switching power is the best answer in this case.
With their fast built-in load regulation often operating up to fairly high frequencies, these supplies eliminate the need for linear voltage regulators and help shrink reservoir capacitors. Essentially, the capacitor only needs to store enough charge to keep the supply voltage stable under load until the regulator kicks in. Moreover, switching supplies are more power-efficient than linear supplies, further reducing cooling requirements.
Class D And EMI
Another common problem with Class D amplifiers is electromagnetic interference (EMI) caused by the rapid switching of currents in the output stage. The simplest way to reduce EMI is to keep pc-board tracks or cables leading to the output stage as short as possible. Supplies and output stages should be located on the same pc board if at all feasible.
Speaker cables pose more of a challenge. In systems with built-in speakers, such as boom boxes, short speaker cables are an effective measure against EMI-and it won't add to the materials cost. Wherever external speakers are used, though, the cables' length is outside the designer's control, and EMI filtering becomes indispensable. Still, achieving sufficient EMI suppression without affecting sound quality can be difficult because the EMI spectrum is relatively close to the audio range. It consists primarily of the PWM switching frequency and its harmonics.
Consequently, designers face a difficult choice. Should they choose a filter with a low cutoff frequency that will suppress EMI but also attenuate the treble? Or, should they use a filter with a high cutoff frequency that will preserve a flat audio response at the cost of increased EMI (Fig. 3)? Higher-order filters can better separate the different frequency ranges, but often the cost of the high-quality inductors is prohibitive.
Adding a digital speaker equalizer to the PWM modulator chip offers a way out of this dilemma. When programmed as a treble boost, such an equalizer counteracts the effect of a basic low-order, low-cutoff EMI filter on the audio signal, keeping the frequency response flat across the audio range. Different equalizer settings also can be used for matching an existing filter design to speakers with different impedances, maximizing design reuse.
Beyond Class D
A number of additional techniques can help deliver a superior listening experience on a given budget. With a separate bass speaker or subwoofer, the remaining speakers needn't reproduce the entire frequency range, so they can be smaller and less expensive.
Because the lower end of the audio spectrum carries little or no directional information, this won't impair the stereo image. In addition, a dedicated subwoofer will make it easier to achieve the extended bass response preferred by many consumers. These considerations have led to a migration from plain stereo to a "2.1-channel" (i.e., stereo plus subwoofer) configuration in the consumer market.
Dynamic peak compression makes audio signals sound louder without resorting to a more powerful output stage. It relies on the fact that the instantaneous amplitude of audio signals at most times is far less than their peak amplitude. The signal is amplified in the digital domain by multiplying audio data with a gain constant.
During signal peaks, when the product of this multiplication comes close to the maximum or minimum values that can be represented in the signal processor, distortion is prevented by temporarily reducing the gain value. After the peak subsides, the gain is ramped back up to its original value. To avoid introducing step changes into the signal, all gain changes must take place during zero crossings. Moreover, they must be gradual, with well-defined attack (gain ramp-down) and decay (ramp-up) times, rather than instantaneous (Fig. 4).
Unfortunately, this creates a dilemma. A short decay time is desirable because it squeezes more power out of the amplifier. But the decay time also must remain large compared to the signal period. Designing for the worst case-the lowest-frequency bass signal-results in an excessively long decay time that limits the usefulness of the peak compression technique, while setting a shorter decay time causes distortion in the bass.
Recently introduced PWM modulators are overcoming this problem by making the peak compressor's decay time frequency-dependent. As a result, the gain can step up rapidly for high frequencies and more slowly in the case of bass signals.
As component vendors implement increasingly sophisticated schemes to address common design issues in Class D design, system designers can now take full advantage of this technology's benefits while mitigating its inherent drawbacks. By combining the power efficiency of Class D with digital signal processing to enhance perceived audio quality, designers today can improve the price/performance ratio of audio amplifiers substantially beyond what was possible with traditional linear technologies.