Use models for portable-device audio amplification have evolved dramatically over time. For example, when a cell phone’s main function was simply to reproduce speech from a speaker held close to the ear, the ear speaker needed very little power. In addition, audio qualities such as total harmonic distortion (THD), noise, and signal-to-noise ratio (SNR) were of less concern.
Speech generally consists of high crest-factor, low duty-cycle signals. Thus, it requires low average power, which in turn minimizes efficiency concerns. Since RF and display sections dominated power consumption within cell phones, most efficiency issues involved non-audio electronic components.
More recently, however, cell phones and other portable electronic products integrate an ear speaker, a headphone speaker, and a near-field speaker (for hands-free operation). Furthermore, reproducing music (MP3 files) and movie soundtracks puts a heavy load on the audio channels. As a result, audio-channel power consumption is no longer a side issue. Rather, it has become a major power drain. Also, low-fidelity sound reproduction is a thing of the past, with audio transmission now demanding an SNR above 100 dB and THD less than 0.1%.
Acoustic audio power amplifiers are generally divided into two operational types: headphone amplifiers (HPAs) and speaker amplifiers (SPAs). HPAs must drive 32- or 16-Ω speakers up to 30 mW and maintain very high audio quality (typically 105-dB SNR, 0.01% THD, and 20-kHz bandwidth). However, 30 mW is a very high output power for headphone applications—it’s high enough to be painful. Typical listening levels are in the 100-µW to 1-mW range.
Generating 30 mW into a 32-Ω load requires a 1.4-V peak signal swing, with additional headroom margin needed for IR drops. Therefore, a supply voltage of ±1.8 V is generally used to reach 30-mW output power.
A typical headphone cable contains three wires, one each for LEFT and RIGHT drive signals and one wire for a common ground return. Other wire elements might be added for volume control, mute, or a microphone output. With such a configuration, a stereo HPA must have single-ended outputs.
But if the supply is a single-supply rail, this will lead to a large dc offset problem. To eliminate the need for large ac-coupling capacitors, most headphone amplifiers are powered from a split supply voltage, where an on-chip inverting charge pump often generates the negative rail.
Most headphone amplifiers employ a linear amplifier, such as a class A/B output stage variety, to achieve the high-quality audio performance required for an HPA. The venerable class A/B amplifiers combine class A and class B operation (Fig. 1). They are usually designed so class A operation dominates at low output power. Due to low crossover distortion, class A offers the best audio performance.
Class B operation, which comes into effect at high output levels, features higher efficiency than class A. However, it also experiences more crossover distortion. Overall, class A/B amplifiers can achieve very low THD, since the crossover distortion is largely attenuated by closed-loop feedback.
Class A/B amplifier efficiency is proportional to output swing for a constant supply value. To reclaim some of the efficiency losses at low output power, the “class G operation” technique is used to reduce the supply rail value for low-level signals.
A circuit is used to detect the input-signal level. If it crosses a predetermined threshold, the supply rails are raised to a higher value as needed. Most class G amplifiers have two supply rail values: a high rail value for large signal swings (VDD), and one for low-level signals set to some fraction of VDD (e.g., 1/2 of VDD) (Fig. 2). As a result, efficiency for signals at 1/4 full-scale output power approximately equals that of full-scale power signals (Fig. 3).
A variation on class G operation, appropriately named “class H operation,” is when the supply rail varies continuously with the peak signal requirements (Fig. 4 and Fig. 5). This maximizes efficiency at all signal levels. A minimum supply rail level is limited in class H operation, though, due to circuit design and process limitations.
Some manufacturers use the term “class H” for their HPA operation when, in fact, it’s really class G operation. True class H operation is rarely found in today’s IC HPA amplifiers.
Speaker amplifiers (for near-field applications such as hands-free and speaker-phone operation) in portable electronic products usually drive 8- or 4-Ω speakers. Typical listening levels fall in the 100- to 300-mW range, but IC amplifiers are commonly available at 1- to 2.7-W average output, with peak outputs at nearly twice that level.
To generate 1.7-W into an 8-Ω load, the SPA must deliver 5.2 V peak, or approximately 3.7-V rms into the speaker load. Adding some headroom for IR drops, a typical supply rail is 5.5 V for a 1.7-W SPA. If lower IR drops can be achieved with larger switches, a little over 1.8 W is possible. These output power numbers have 1% THD. At 10% THD, larger output power can be produced.
Generally, in portable audio products, near-field speakers don’t reproduce high-quality audio. Therefore, an SPA usually needn’t achieve an HPA’s audio capabilities. Typical audio performance is 1% THD at full power, 10-kHz bandwidth, and 94-dB SNR.
Efficiency becomes a more important factor with an SPA versus an HPA, since an SPA’s power levels are so much higher. HPA efficiency is generally less than 50%—not great, but a small power loss compared to a battery with 4.7 Wh of energy capacity (about 0.01% of battery capacity for normal listening levels). However, the same 50% power loss in an SPA operating at 1 W equates to 0.5 W, or about 10% of the battery’s energy capacity.
Class D SPA
The importance of HPA versus SPA operation efficiency is a function of how much time is spent in one or the other listening modes. A cell phone in, say, speaker mode consumes more power, so efficiency becomes crucial. A linear amplifier (e.g., class A/B) can be used to drive a speaker (which was often the case in the past), but today the preferred speaker driver is a class D amplifier (Fig. 6). A class D SPA maintains high efficiency over a broad range of output power levels and only starts dropping off at power levels below 1% to 2% of full power.
A class D amplifier is not linear, but rather a type of switching amplifier. In a switching amplifier, a high-frequency carrier (relative to the audio band) modulates the audio input signal, typically from 100 kHz to 1 MHz. As a result, the output stage can be switched “digitally” (rail to rail) to put the output-power devices in either an On or Off state, which are the points of highest efficiency.
Switching amplifiers are usually configured in bridge mode to differentially drive the speaker load. This eliminates the need for output ac-coupling capacitors. Because a bridge-mode amplifier uses four power switches per channel, it’s twice the size of a single-ended output-stage amplifier. However, a bridge-mode output stage develops four times the output power for a given supply rail compared to a single-ended amplifier.
Class D amplifiers achieve quite high efficiency, typically over 90%. Using this type of amplifier does have its drawbacks, though. Since the audio content is now a modulated signal, it must be demodulated by some type of low-pass filter (LPF) to drive a speaker load. A high-power LPF that doesn’t cause efficiency losses or distortion problems is large and expensive, and thus not used in portable equipment.
However, a speaker in and of itself is an LPF and offers high impedance to typical carrier frequencies. In portables like cell phones, it’s common practice to let the speaker act as the LPF and demodulate the switching-amplifier output signal. Sometimes, a few ferrite beads are used in series with class D outputs to reduce electromagnetic interference (EMI) generated by the high-power switching output. Due to a speaker’s high impedance, the speaker’s modulation signal dissipates very little energy, maintaining good efficiency.
EMI can become a serious issue with switching amplifiers, though, when long wires are used between the SPA output and the speaker load, and there’s no separate LPF. For this reason, an HPA won’t use class D amplifiers if the headphones are at the end of a long cable. Consequently, a class D amplifier should be located close to the speaker load to avoid generating excessive EMI radiation.
Other types of speaker amplifiers have been used, but most are variants to the linear and switch-mode amplifier designs described here. In modern portable electronic products, demand for more battery power continues to heighten. High-resolution large color displays for video content, high-resolution camera and flash, and high-power audio outputs all take their toll on battery life. Increasing the efficiency of audio speaker amplifiers then becomes an important design consideration when attempting to improve battery runtime.
Tim Dhuyvetter is a senior member technical staff engineer at Fairchild Semiconductor focusing on audio and power-management product development and architectures. He has worked for more than 30 years in the semiconductor industry focusing on technologies such as power management, illumination, and signal processing system developments. Prior to Fairchild, he worked at companies such as Philips, Sipex, Telcom, Leadis, and Akros, where he worked on diverse products for commercial, consumer, and high-rel IC systems. He graduated from Cal Poly SLO with a BSEE and from Santa Clara University with an MSEE.