Consumer adoption of the evolving digital lifestyle is posing serious challenges for traditional audio design conventions and driving a technological renaissance in audio system architecture. Those who adapt will thrive, and those who lag will fail.

Three trends are impacting the consumer audio market: 1) the emergence of digital media sources; 2) the convergence of audio and mobile communications; and 3) the economics of power and thermal management. Traditional analog architectures have struggled to accommodate these trends.

The Class A/B audio amplifier, a traditional audio system workhorse, is a power pig that requires expensive supply systems to provide the necessary current and thermal radiators to help dissipate the heat. Analog input Class D amplifier alternatives solve the power and thermal management challenges but add their own negative side effect of radiating electromagnetic interference (EMI) that causes problems with EMI compliance, radio co-existence and mobile phone compatibility.

By its very nature, the analog interconnection is susceptible to noise pick-up, requiring additional expense and design effort to mitigate noise ingress from nearby mobile phones that would otherwise result in the infamous “bumble bee buzz” and make media phone docks incredibly annoying. While analog system design is antiquated for today’s consumer audio market, adopting digital architectures with digital signal processors and data converters that enable enhanced fidelity, acoustic optimization and cool feature enhancements is cost-prohibitive for most mass-market consumer audio products, relegating these benefits to be enjoyed only in more expensive, high-end equipment.

Adapting and thriving through this technological renaissance requires a new system design paradigm, a new architectural approach for contemporary mass market consumer audio systems. Let’s consider the seven key elements for achieving affordable fidelity in this new market environment for consumer audio system designs: power efficiency, EMI compliance, radio co-existence, mobile phone compatibility, audio fidelity, acoustic optimization and system cost effectiveness. Each of these elements can be achieved with Class D amplifiers designed to manage and suppress EMI in consumer audio systems, as shown in Figure 1. 

Figure 1: Example of new audio system design paradigm based on Class D digital architecture.

1. Power efficiency
System designs with higher power efficiency, or alternatively lower power wastage, deliver several benefits. Besides the obvious “green” benefits, higher efficiency reduces heat dissipated by the amplifier and eliminates the cost and weight associated with metal heat sinks. Depending on thermal requirements, low-power systems using Class A/B amplifiers in the range of 5 W per channel require heat sinks that can cost 10 to 20 cents and add weight to the system design. Of course, the cost and weight increase proportionately with output power.

Because the amplifier dominates the power budget for audio systems, more efficient amplifiers enable use of smaller and less expensive power plants. For example, for a 5 W rms per channel stereo system using a Class A/B amplifier with a generous 50 percent efficiency rating, a 7-V power supply would need to provide continuous 0.29 A rms of current when playing content at full scale with a 10:1 crest factor. If dynamic range compression is employed to reduce the crest factor to 5:1, then the supplied current doubles to 0.56 A rms. With an 85 percent efficient Class D amplifier, the supply current requirement would be reduced by 44 percent, which translates into a substantial cost savings.

An obvious benefit of higher power efficiency and lower current consumption is longer battery life for portable systems. In an increasingly portable world, battery life is critical to usability, and in case of disposable batteries, critical to cost of ownership. Operating voltage requirements dictate the number of batteries. For example, many sub--W Class A/B amplifiers require at least 9-V supply voltage, if not 12 V. So in the case of alkaline batteries, stacks with six or eight batteries are required, negatively impacting weight, ownership cost and usability. Delivering adequate power with a reduced voltage level of 6 V requires a more reasonable stack of four batteries, resulting in a lighter and easier to use end product.

Operating time per battery refresh (whether a result of a recharge or a replacement) is directly related to system current consumption. One popular metric for portable audio devices is continuous operating time playing audio at a nominal listening level, such as one quarter of a Watt rms for all channels. Eight hours is considered good performance for standard alkaline batteries. At quarter-Watt rms output power, the impact of power delivered to the load diminishes relative to the impact of both operating current and quiescent current (power consumption with no power delivered to the load, or mute), thus providing a more meaningful metric for comparison. Thus, both operating efficiency at nominal output power and quiescent current are important parameters for assessing battery life performance.

2. EMI compliance
In too many system designs, EMI compliance is an afterthought, achieved through use of application band-aids such as ferrite beads, inductive chokes and metal shields -- all adding cost and time to market. Design for compliance begins by understanding the target requirements (e.g., FCC part 15 in the case of consumer products for the U.S. market, as shown in Figure 2), identifying those system components most likely to offend and implementing smart design techniques to mitigate radiation. 

Figure 2: Example of FCC EMI compliance limits and test results

High-frequency clocks with sharp edge transitions are the most prevalent EMI aggressors. Common techniques for managing this include minimizing clock trace lengths, softening clock edges with resistor/capacitor filtering and burying clock traces inside grounded conduits on multi-layer PCBs.

Class D audio amplifiers, when employed for their power efficiency benefits, are similarly a ready source of both radiated and conducted EMI. Unfortunately, effectively managing this EMI has proven to be more difficult and costly.

By nature, Class D amplifiers are unintentional radiators whose outputs consist of an audio modulated high-frequency carrier typically switching high voltages at frequencies ranging from 100 kHz to 1 MHz, with harmonics extending beyond 30 MHz into the EMI compliance frequency bands. To compound the radiation problem, the outputs typically drive long lengths of wires connecting speakers to the amplifiers and serving as an efficient, if undesired, radio antenna for both electric and magnetic fields.

Expensive external LC filters, or in some cases ferrite beads, are employed to attenuate these switching harmonic spurs on the outputs while demodulating the encoded audio signal. In some cases, shielded cables are also used to contain the EMI and prevent it from radiating. These after-the-fact measures, while effective, add cost to the system. 

Figure 3: Anatomy of Class D Radiation

Conducted emissions are a related but different requirement limiting the amount of noise allowed to be injected onto the line supplying power to the device for frequencies up to 30 MHz. Class D amplifiers by design switch large currents into and out of both the ground plane and the supply voltage rails, particularly during peaks of large audio signals, resulting in modulated square waves with duty cycle and amplitude proportional to the audio signal. (See Figure 3 for details.) In addition, the charging and discharging of large switch capacitance produce a narrow current spike on grounds and supplies during every pulse-width modulation (PWM) edge transition, independent of audio signal amplitude. The width of these spikes is correlated with edge transition time and amplitude correlated with switching capacitance inherent in the amplifier. The narrower the spike, the larger the resulting high frequency harmonics conducted onto the supply.

One technique for mitigating conducted EMI is to isolate the system supply from the amplifier supply with a series inductor, essentially attenuating the voltage variations emanating from the chip. The high frequency current loops are then minimized around the amplifier via its bypass capacitors. To minimize the voltage ripple caused by the switching current spikes, it is helpful to use low equivalent series resistance (ESR) capacitors.

3. Radio co-existence
If EMI compliance is a challenge in the quest to convert consumer audio system designs to Class D amplifiers, then radio co-existence might be considered mission impossible. To avoid desensitizing AM and FM radios when co-located very close to a Class D radiation source, the radiated emission levels must be 500 times lower. To quantify in an example, the world wide regulatory compliance limit of 37 dBuV per meter at 100 MHz in the FM band equates to a receive power level of 18.5 dBuV EMF with an FM antenna 1 meter in length located at a distance of 10 meters from the radiation source. At a distance of 0.1 meter as would be the case for radio co-existence, the receive power level would be 40 dB higher, almost 60 dBuV EMF and 500 times the sensitivity of a typical FM tuner.

For FM co-existence, audio developers have either defaulted to using Class A/B amplifiers or employing a combination of expensive LC output filtering, physical shielding and separation, and shielded cables. Some amplifiers employ lower frequency PWM switching rates to take advantage of lower amplitudes associated with higher order harmonics that result in the FM band between 88 MHz and 107 MHz, and slower slew rates on the PWM edges around 10 ns to leverage the resulting faster SINX/X roll-off of the harmonic spurs. These measures are effective, but must be inherent in the Class D amplifier design.

AM co-existence is even more difficult with Class D amplifiers given typical PWM switching rates and lower order harmonics blanketing the AM band between 500 kHz and 1700 kHz. For PWM fundamental switching frequency at 300 kHz, for example, four of the strongest harmonic spurs disrupt the AM band at 600 kHz, 900 kHz, 1200 kHz and 1500 kHz. At any of these frequencies, AM reception is rendered impossible without significant shielding and physical separation measures.

Some Class D amplifiers employ AM frequency avoidance schemes whereby the switching frequency can be adjusted to alternate frequencies where all its harmonic spurs avoid landing on the tuned AM frequency. This scheme mostly works, although not always uniformly across the AM band and not always for every AM frequency. In addition, because the PWM switching frequency moves, this technique can be problematic for digital input amplifiers with digital filter sample rates based on the PWM switching frequency, consequently requiring multiple coefficient banks for each different switching frequency.

Generally, when a radio is involved, designers avoid Class D amplifiers and use Class A/B exclusively, compromising the potential benefits provided by Class D technology.

4. Mobile phone compatibility
While an Apple iPod/iPhone dock interface is increasingly becoming a standard “must-have” feature for consumer audio products, achieving Apple compliance significantly increases the system design challenge. Compliance involves three primary performance elements when the phone is plugged into the dock, as shown in Figure 4: 
• Audio system interfering with phone’s transmission to a basestation; 
• Audio system interfering with phone’s reception from a basestation; 
• Audio system picking up audible noise from phone’s TDMA transmission.

The first compliance element is generally not problematic because the phone transmission power is far greater than anything generally radiated from the audio system. Unfortunately, compliance with the second and third elements is a far different story, presenting significant design challenges. 

Figure 4: Three challenges for smart phone OTA compliance

Protecting smart phone reception: As with AM/FM radio co-existence, the radiation requirements to prevent desensitizing a smart phone receiver is more demanding than achieving regulatory compliance. Care must be taken with every high-frequency clock and digital signal to prevent high-frequency harmonics from radiating and disrupting reception. And the challenge is compounded for designs leveraging the efficiency benefits of high-power Class D amplifiers with their inherent high-frequency switching characteristics.

Interference from digital signals can be easily mitigated using a simple first order resistance capacitance (RC) filter to slow down the edges and attenuate the higher frequency edges and minimizing the unfiltered PCB trace lengths. Additional measures include routing clocks inside grounded conduits, constructed by routing grounded signal traces on both sides of the clock trace, and even shielding the clock traces by burying them in the middle layers of multi-layer PCB between ground planes located above and below.

Mitigating the interference from Class D amplifiers is more difficult. In fact, many designs default to using power-hungry Class A/B designs, which inherently have no high-frequency switching noise to radiate. Those designs using Class D amplifiers employ expensive filtering and shielding to suppress the interference, whether radiated into the phone receiving antenna or conducted into the phones via the power supplies.

Protecting audio performance: The all too familiar and annoying “bumble bee” noise heard in analog-based audio systems from the characteristic TDMA transmissions from digital phones creates an insidious design challenge. At mobile phone transmission frequencies typically ranging from 850 MHz to 1.9 GHz, almost every analog PCB trace in the audio signal path doubles as an efficient receiving antenna, opening a welcoming entry for the “bumble bee.” Making an analog architecture robust against this interfering noise requires use of differential analog signals, in-line high frequency filtering using ferrite beads, shielding of particularly sensitive signals, and analog components with high power supply rejection at harmonic multiples of 217 Hz.

While successful noise mitigation has proven possible, the design challenge is formidable. Adopting a digital architecture and minimizing the number of susceptible analog traces in a design makes the challenge far more tractable. In practice, nearly all sources of audio in a system are natively digital signals, whether a DSP radio tuner, CD player, digital media controller, Bluetooth wireless receiver or streaming Internet processor. Once the dock’s interface for digital content residing in a media player (e.g., an iPod) transitions to digital, the only remaining natively analog interface will be the auxiliary input to external analog devices. Consequently, a digital architecture is a technically superior option for noise immunity and digital signal compatibility.

System cost remains as an impediment to this natural solution. Audio component suppliers have performed serious acrobatics to add analog interfaces to inherently digital devices and make them compatible with the traditional analog architecture paradigm. Digital architectures are possible today, but typically require the addition of an expensive audio DSP to orchestrate the digital interface, adding cost inconsistent for use in mass-market consumer audio applications. Audio devices with smart digital system integration features will enable cost-efficient, noise-immune solutions for a new digital architectural paradigm.

5. Audio fidelity
Audio fidelity is defined as the degree of accuracy or faithfulness with which sound is reproduced. For audio systems, fidelity is a subjective proxy for sound quality. Various performance specifications are typically used to indicate audio fidelity, and while all are important, none guarantee excellent sound. The key to affordable fidelity is to adequately deliver on the highest impact specifications without incurring undo costs.

Dynamic range is the range of usable signal levels from minimum to maximum, typically using criteria of signal to noise + distortion ratio (SNDR). Signals that are too small will get swamped by the system noise floor. Signals too large will begin to clip resulting in unacceptable harmonic distortion, typically considered to be unacceptable when distortion energy exceeds 10 percent of the fundamental energy. Dynamic range performance less than 80 dB is considered poor, while at least 90 dB is considered acceptable for most mass-market consumer applications to maintain a good sound stage and audio depth.

Total harmonic distortion is a measure of linearity or purity of an audio system. The output frequency content of highly linear systems is identical to that of the input signals. Nonlinearities create additional signal components at harmonic multiples of the input frequencies, which obviously distract from the purity of the output signal. Total harmonic distortion (THD) is the ratio of the additional signal energy at all harmonic frequencies to the energy at the fundamental frequencies of the input, typically measured at a nominal listening level such as 1 W of output power. While THD performance less than 0.05 to 0.1 percent is adequate for most non-audiophile audio applications, THD for many mass market audio systems can range as high as 1 percent.

Power supply rejection measures the attenuation at the amplifier output of an unwanted signal on the amplifier’s power supply, whether caused by an external source or self-inflicted from large load currents pulling on the supply. The bass response of amplifiers with weak power supplies and poor power supply rejection will sound muddy because the large low-frequency audio signals create sustained load currents that modulate the power supply voltage and ultimately self-distort the output signal.

For linear Class A/B amplifiers, the transfer function is linear, meaning that the induced output signal frequencies match those of the signal on the supply independent of the input signal. For Class D amplifiers, the transfer function is a nonlinear mixer that multiplies the desired audio input signal with the unwanted signal on the power supply, resulting in output signals modulated by the audio input with frequency components different from those of either the audio input signal or the signal on the supply. With no audio input signal, Class D power supply rejection will consequently be excellent, although with large audio input signals, it can be poor unless feedback or other compensation techniques are employed.

Power supply rejection performance for Class D amplifiers should measure the total integrated energy across the audio band injected from the power supply with an input signal applied, preferably at half power. Ideally, the total integrated energy injected onto the output from the largest expected supply noise signal should be half the energy from THD. Generally, 50 dB of rejection is sufficient in most applications.

Maximum output power is the amount of average output power typically measured with a single frequency test tone at an amplitude corresponding to the maximum dynamic range level. For linear amplifiers without dynamic range compression, this indicates an amplifier’s relative ability to generate loudness from a given speaker. While maximum output power is commonly used as a proxy for potential loudness, some care should be taken since actual loudness depends on the sensitivity of the speakers being driven (output sound pressure level measured at 1 meter distance with 2.83 V input voltage) and the crest factor of the audio content (ratio of the signal peak level to its average level).

Output noise level is an absolute measure of the noise floor level at the amplifier outputs with no signal input. For most speakers, a noise floor of 100-500 uV is inaudible from most normal listening distances, while a noise floor as high as 1 mV will prove to be quite annoying.

Crossover distortion is perhaps the most detrimental issue for audio fidelity for which a convenient specification has not been defined. Audio signals comprise a multitude of time varying tones with levels oscillating between positive and negative extremes and passing through zero on each excursion. When the signal does not faithfully pass through zero, the resulting error is glaring. One way to test cross over distortion is to measure the residual output signal level near zero crossings at the output of a notch filter tuned to the input test signal frequency.

6. Acoustic optimization
Advanced audio processing for acoustic enhancement has typically been reserved for high-end premium audio systems, requiring added cost and complexity of a separate digital signal processing device for functionality beyond basic signal multiplexing, volume and tone control.
Affordable signal processing can help enrich the sound quality of consumer audio systems while enabling value-add features for the end user. Three basic acoustic enhancement features that are highly beneficial include frequency response equalization, dynamic range compression and stereo spatialization.

Frequency response equalization benefits consumer audio systems in two ways. First, it enables the designer to optimize the system sound quality by compensating for acoustic non-idealities in the enclosure and speaker. The equalizer can be used to flatten peaks and nulls in the overall acoustic response, as well as to notch out mechanical resonance frequencies that otherwise cause unwanted vibration. The net benefit is either better sound quality from existing acoustics, the same sound quality from cheaper acoustics, or both.

Second, equalization can be used to customize the sound image for different effects by enabling user-selectable frequency response characteristics. Many premium audio systems provide options to select a “Rock” sound profile, a “Classical” sound profile or a “Rap” sound profile depending on user preferences.

Dynamic range compression adds gain to the average signal level and not to the signal peaks. For most audio content with typical crest factors ranging from 4:1 to 10:1, this technique allows the amplifier to overdrive the speakers without distortion, effectively increasing the average loudness while maintaining good fidelity. Compared to a linear analog amplifier, a digital amplifier with dynamic range compression delivers similar loudness without the added cost of higher power electronics, as illustrated in Figure 5. 

Figure 5: Increasing loudness with dynamic range compression

For compact audio systems such as boom boxes and media player docks, the stereo speakers are physically located close together. Consequently, each ear hears both speakers almost equally well, resulting in what sounds more like mono than stereo. Stereo spatialization leverages audio beam forming techniques to reduce the stereo crosstalk and enhance the stereophonic effect.

The affordable fidelity trend demands that these types of audio process enhancements be made accessible to mass-market consumer audio products, and not just the purview of premium audio systems. Achieving the necessary cost level will require integration of the DSP hardware with other functions and adoption of a digital architecture to eliminate the cost and performance degradation of digital-to-analog domain conversions.

7. System cost effectiveness
Achieving cost-effective system design requires holistic considerations up front in the development process. Force fitting new technology into traditional platforms, while expedient, is rarely optimum. A multitude of design issues need to be addressed beyond the basic core audio path. The challenge is to efficiently incorporate these elements into a system concept such that primary ICs mesh together cleanly and seamlessly with few external discrete components.

Power supply and regulation subsystems must be designed and sized to properly support all system requirements. In a typical audio system, there may be two to four different voltage nets to be generated depending on detailed system specifications. For example, the audio amplifier output power requirement might drive a high supply voltage greater than 5 V, while digital and mixed-signal devices may drive a low supply voltage in the 3 V range; a USB interface may require a 5V supply; and a large system-on-chip (SoC) processor may require 1.8 V or less. At a minimum there is usually at least a power amplifier supply voltage and a 3.3 V supply voltage for all other devices.

System clock requirements and distribution strategies are essential for both cost and EMI management. A single clock system with a centralized master clock is best to avoid synchronization issues across clock domains. When both a Class D amplifier and a switching regulator are employed in the same system, it is particularly important that their respective switching frequencies be synchronized to avoid down-mixing audible beat tones into the audio path. For EMI management, clocks routed between chips should be the lowest possible frequency with edges slowed by low-pass RC filters.

In considering the seven elements of affordable fidelity, the following four conclusive requirements emerge about the new audio system design paradigm:
* EMI mitigating Class D amplifiers must displace Class A/B amplifiers
* Digital architectures must displace analog
* Digital audio processing must be integrated in a cost–effective manner
* All seven elements of affordable fidelity must be advanced.

Unfortunately, evolution in consumer audio technology cannot possibly bridge the gap between what is and what’s required. Achieving the affordable fidelity paradigm requires a revolutionary leap from traditional consumer audio system design to a radically different solution that effectively incorporates these four tenets. It is not a matter of if but when this new affordable fidelity solution emerges and who will win first mover advantage during this technological renaissance in consumer audio.

The Si270x Class D audio amplifier from Silicon Labs represents the first Class D solution to effectively address the need for affordable fidelity by providing a digital architecture designed to suppress EMI in consumer audio systems. To learn more about the Si270x Class D amplifier and how it provides a new paradigm for the audio market, visit

* * *

Rick Beale
Director, Audio Amplifiers
Silicon Laboratories, Inc.

Rick Beale is the director of audio amplifier products at Silicon Laboratories, with responsibility for strategy, marketing, systems and applications. Mr. Beale joined Silicon Labs as a director in 2007 to provide leadership for new product initiatives in the company’s Broadcast Audio business, ultimately resulting in the creation of innovative technology and the new product line vector for audio amplifiers. Prior to Silicon Labs, he served as marketing and sales vice president for numerous startups, including Jam Technologies developing Class D amplifiers and RF Magic (now merged with Entropic) developing RF systems-on-a-chip. Mr. Beale began his career with Bell Laboratories designing mixed-signal integrated circuits for communication systems before transitioning to marketing and business management responsible for DSP ICs targeting cellular applications with AT&T Microelectronics (Lucent/Agere) and subsequently Motorola (Freescale). Mr. Beale holds degrees of MBA from Wharton, MSEE from Stanford University and BSEE from the University of Virginia.