SAN FRANCISCO " There is an old-guard sentiment among consumers suggesting that when it comes to home audio receivers, bigger is better and biggest is best-the most powerful systems ought to give the moving men hernias and should be financed by a second mortgage. But a new generation of audio receivers, DVD players and integrated home theater systems is replacing these monoliths, and examples are expected to show up under many a Christmas tree this year.
Using digital amplification-a technique borrowed from switching power supplies-the new-generation AV receivers are smaller and lighter than ever before. By promoting 94 percent power efficiency, digital amplifiers can eliminate the bulky metal heat sinks that ordinarily accompany power transistor output stages. Thus, they are able to package 50-watt stereo TVs in the plastic housing of a liquid-crystal display-and to enable shoebox-size 100-W/channel AV receivers and integrated DVD players with AM-FM receivers, Dolby Digital and DTS surround-sound decoders.
The multichannel audio systems enabled by these digital amplifiers-automotive entertainment systems, as well as home theaters-will result in sales of more than 1 billion audio amplifier channels between 2004 and 2006, according to market research house Forward Concepts (Tempe, Ariz.).
While entire six-channel home theater audio playback systems are available for little more than $100, some consumer audiophiles-and holdouts among consumer electronics OEMs-raise concerns about sound quality. The truth is that "Class D" digital amplifiers, as these devices are known, require carefully designed filters. The filter components (inductors and capacitors) can be as big and bulky as the metal heat sinks the digital amplifiers were intended to replace, but neglecting them can have disastrous effects on sound quality.
Thus, semiconductor suppliers of Class D and other digital amplifier types are positioning their products according to the competence they profess in filter design. In all cases, these manufacturers have an interest in minimizing the complexity (as well as the number and values) of external filter components for their potential customers.
One group of manufacturers-National Semiconductor and Maxim Integrated Products, for instance-is concentrating on low-power products, like headphone amplifiers for cell phones and PDAs, in which the power saving of Class D is absolutely demanded, the filter requirement is minimal and any compromise in sound quality will largely be unnoticed. Another, so far very small, group, typified by manufacturers like D2Audio Corp. (Austin, Texas), is trying to offload the user of filter design problems altogether, by supplying only pretuned drop-in boards offering 120 W/channel (see Nov. 3, page 4). Up to eight channels are available on a 4.3 x 6.8-inch board. But with OEM modules selling for about $100 each, they will be used only by high-end consumer OEMs like Harman Kardan.
But the most visible group-typified by Tripath Technology, Texas Instruments and Apogee Technology (with parts made by STMicroelectronics)-is targeting the consumer mass market with high-power amplifier parts. These manufacturers are implementing "zero-crossing," spread-spectrum and other noise-reduction techniques in an effort to minimize the external filter requirement. (Cirrus Logic announced its intention to attack this portion of the market, although its Class D line so far has centered on a low-power headphone amplifier.)
This camp of suppliers has argued, and occasionally demonstrated, that digital amplifiers can indeed approach audiophile quality. But they'll have their work cut out for them in getting digital amplifiers accepted by the "golden ears" crowd.
In operation, digital amplifiers use the width and frequency of a digital pulse train-pulse-width modulation (PWM)-to carry an audio signal. The pulses are generated by trip comparators, matching the voltage levels of an audio signal against those of a sawtooth reference to trigger the pulses. Thus, a high-amplitude, high-frequency signal is represented by a dense cluster of short-duration pulses; a lower-frequency, lower-amplitude signal will have longer pulses, farther apart.
The audio signal is produced by pumping the digital pulse train into a pair of power MOSFETs, typically wired in a push-pull configuration. An additional inductor-capacitor filter (as well as the speaker coils themselves) adds "persistence" to the signal, effectively smoothing out the pulses.
Transistors are not natural amplifiers; they are switching devices. To get a transistor to behave as an amplifier, as with conventional Class AB amplifiers, you have to apply a bias current, which forces the transistor to forever hang in a region between being completely on and completely off. That means the transistors of the Class AB amplifier will be sucking up current-and generating heat-even when there is no audio signal present.
On top of that, audiophile equipment manufacturers have learned that many types of power transistors perform most linearly (and are claimed to sound the best) when they are "hot biased"-that is, when they are given almost, but not quite, enough bias current to snap them into an "on" state. This will make the power transistors extremely hot to the touch, and it is no wonder (or exaggeration) that these high-end Class A amplifiers will come with 30 pounds of aluminum heat sink.
The digital amplifier, in contrast, uses power transistors much more efficiently. The pulse-width modulator switches them on or off and tries not to spend too much time (or power) in the in-between state. Consequently, the chain among PWM, driver stage and power transistor output in a Class D amplifier is significantly shorter than it is for multistage AB amplifiers.
The problem is that the output of the audio amplifier must be carefully filtered to block the residue and harmonics of the digital pulse train. That puts the semiconductor supplier in the filter design business, and there is no guarantee that the customer-the consumer electronics equipment maker-will adhere to the reference design or follow the semiconductor maker's recommendations. Thus, if the end consumer is unhappy with the sound quality of a home theater system, he or she might well blame the semiconductor supplier, when in fact it might be the fault of an underdesigned filter.
Tripath Technology Inc. (San Jose, Calif.) was one of the earliest companies to attack the filter problem head on. Using a variable-frequency sawtooth oscillator to spread the pulse about a 1-MHz region-a form of spread-spectrum technology the company called Class T-Tripath effectively reduced the complexity of the filter network required. For example, it found that the electrolytic capacitors used with switchers operating at 250 kHz could be replaced with smaller, lower-value capacitors when the switcher operates from 600 kHz to 1.5 MHz.
Tripath offers separate amplifier chips with low power ratings (up to about 10 W), integrated amplifier-drivers with power ratings up to 70 W and assembled modules with power ratings up to 150 W per channel. The company's TA2024 two-channel (15-W/channel) amplifier, for example, powers Sanyo's Vizon 42-inch-wide plasma screen television as well as Sanyo's 30-inch LCD screen and JVC's 26-inch HDTV. Tripath's 90-W/channel TA2022 is used in Sanyo's 5.1-channel DVD receiver package.
Tripath, in fact, boasts that it has already shipped 25 million channels. Its current customer list includes Aiwa, Apple Computer, Denon, Hitachi, Motorola, Onkyo, Samsung, Sharp, Sony and Toshiba.
For its part, Texas Instruments Inc. (Dallas) counts design wins that include Samsung's DVD receiver combo. Sanyo is said to be using a chip set consisting of the company's TAS5026A, a six-channel PWM, and TAS5110, a 50-W digital power amplifier. A separate stereo H-bridge driver, the TAS5182, extends TI's digital amplifier power range over 100 W.
But filter complexity can be the Achilles' heel of digital amplifier de-sign. By shifting the timing of negative-going pulses at the zero-crossing point, TI believed it could eliminate many of the audible harmonics associated with Class D technology and substantially reduce the filters required. Good or bad, the Class D amplifiers promoted as "filterless"-such as TI's three-year-old TPA2000D1 or Maxim's just-introduced MAX9700B-are low-power products intended as headphone amps for cell phones or PDAs, in which MP3 playback is an "extra" rather than the main selling point. Even then, such components are not entirely capacitor-free.
Where TI earns its stripes is in inserting a digital signal processor into the digital amplifier chain. Initially, in TI's conception, the DSP could be used to convert the pulse-code modulated signal coming off a DVD or compact audio disk into the PWM signals used by digital amplifiers. This would effectively bypass an A/D conversion stage. But it soon became clear that the use of DSP could effectively shape that pulse stream in front of the amplifier in ways that made filtering much easier on the back end.
Competitor Apogee Technology Inc. (Norwood, Mass.), in fact, uses this pulse-shaping technique to create "damped ternary" pulse-width-modulated timing signals. This algorithm is used to control the switching of power transistors in a full-bridge configuration. A simple passive low-pass filter is then utilized to remove the high-frequency component of the PWM pulse train before the signal goes through the audio speakers.
STMicroelectronics (Lexington, Mass.) manufactures Apogee's pulse-shaping DSP controllers as the STA30x series (where "x" is the number of channels). The STA5xx series (in which "xx" denotes power ratings) makes up the power output stages, which range from 50 to 130 W. A single-chip device (the STA314) with a two-channel output, 15 W per channel, gain controls and five-band equalization will appear in early 2004, the company said. The digital amplifiers are geared toward AV receivers, DVD receivers and televisions.
Even Tripath is using a form of DSP-sigma-delta A/D conversion-on the front end of its audio amplifier modules. The sigma-delta converter reconstitutes a 44.1-kHz audio pulse stream as if it were being sampled at 12.5 MHz (or 256 times 44.1). This "oversampling" effectively flattens audio amplitudes into a 1-bit-plus-sign digital stream-easily manipulated by a digital amplifier. The oversampling also positions the quantization noise and harmonics way up in the frequency spectrum, where they can easily be lopped off by a capacitor filter.