Published in February 2004

Class D Is In Session!
By Neal Weinstock

There's now a new choice in amp design.

     Elementary amp-think: Very high-end audio systems use Class A amplifiers, which offer the best fidelity but generate the most heat and consume the most power. Class B amps, in comparison, serve up lower fidelity but are more efficient. Class A/B is a compromise between the two. Its fidelity is better than B but not as good as A, and A/B amplifiers generate more heat than Class B amps. This is all pretty bad news (and old news) for installed audio. Here, power efficiency and low heat release can be far more critical than they are in home audio or even in a studio. And low-fi signals can degenerate quickly into audible mud up in the balcony, as this writer can attest from a couple of recent concerts where we were late ticket-buyers.

Class D
     In fact, there are a couple of applications where power efficiency and low heat release are even more important than in installed systems: laptop computers and MP3 players. Also important in these apps are amplifier size and expense. Originally for use in laptops, Class D (or PWM, for pulse-width modulation) amplifiers were developed in the 1980s. They met the laptop’s need for tiny size, power efficiency, low heat release and low cost; the “D” stands for “digital,” and the basic Class D engineering challenge has always been all about putting an amplifier on a chip. That has also been both the unrealized hype of Class D and the vast promise.
     PWM chips currently used in PC audio sell for a dollar or two. But then, there are analog amps that sound just as good, and put out as much power, that sell for even less. Analog systems in this area have, however, probably improved as much as they can (while costs may indeed keep coming down for the next few years). PWM amps—although they’ve been around for awhile—are just at the beginning of their improvement curve. This is almost always the way with old technologies replaced by radically new and different technology: Film is getting better even as it is clearly being replaced by digital recording; sailing ships got better as they were replaced by steam.
     PWM amps promise to allow high-quality amplification to be placed at hundreds or thousands of nodes in a digital audio network (and thus they promise a need for software that equalizes such complex setups), for radically more capable, dependable, less power-hungry and localized audio networks in stadiums, large public buildings and other huge venues. The promise of Class D is really at the heart of audio’s overall transition to digital technology.
     And that promise is beginning, finally, to be realized.
     Of course, it began 20 years ago. Class D amps were no sooner developed for laptop PCs than they were quickly put to use in installed audio by Peavey, among others, to mixed response. They were highly efficient. They just didn’t sound very good.
     Skip Taylor designed those first Peavey Class D amps back in 1984. He is now the CTO of D2Audio, a start-up specializing in Class D amplification. “The earlier Class D technology was not really totally digital,” he reminisced. “It was a hybrid. We had to go analog to do the processing. It was a stretch to do it; it was very complex and difficult to manufacture. Today’s technology is totally different. It’s totally digital, using very intelligent processing in the digital domain.”

More Than 100 Companies
     Just how intelligent is that processing? “More than 100 companies have some kind of Class D design,” stated Bob Adams, manager of Analog Devices’ (ADI) digital audio engineering group. “Anybody with an algorithm. But these amps are going to come up against real-world issues such as interference from FM signals, and a lot of them aren’t going to make it. There may be just three or four really strong designs out there.”
     Adams has long been a skeptic about Class D amplification in high-quality audio applications, but even he and ADI are now developing a Class D product, initially for use in Sharp’s LCD TVs. In general, it may be said that the biggest skeptics about Class D have been those companies that dominate the analog-to-digital (ADC) and, especially, digital-to-analog (DAC) conversion business. ADI is very prominent there. So is Asahi Kasei Microsystems (AKM). Both are now coming out with PWM chips; they may cannibalize their own businesses, so if ADI and AKM are in this class, it is for real.
     But the problems of Class D are real, too. “The biggest problem with PWM,” said Tony Rodrigues, manager of corporate development for AKM Semiconductors (AKM’s US subsidiary), “is that, as you decrease the amplitude of the pulse, you lower the voltage and come up against the rail on the switch. So linearity is a problem at low signal levels. Typical PWM designs come up against the limits of MOSFET technology….A lot of people are using current Class D designs
in subwoofers, where accuracy isn’t quite so important.”
     AKM has announced its PWM single-chip solution, which is to be available in a few months. Rodrigues said they get around the low-level switching problem by using six MOSFETs per channel instead of the usual four. That way, they get a low-voltage rail of 5V, which he claims is much lower than anybody else. The design has a 12V full voltage level (unusually, AKM gets 5V and 12V on the same die). It puts out 140W, or 70 per each of two channels.

Another Problem
     Another problem for Class D amps is radio frequency interference, as mentioned by Bob Adams. Pulse amplitude modulation switching frequencies are in the AM radio range, and harmonics get into the FM range. Many Class D amp-makers have responded to this problem simply by recommending that their products not be placed near a tuner. This is not so unreasonable for home audio as it may seem at first, because PWM amps are ideal for placing within powered loudspeakers, typically located at some distance from a radio. It is likely to be more problematic for many PA installations, however: Imagine someone carrying a transistor radio into range of a powered speaker, and interrupting important announcements with scraps of counterprogramming.
     AKM claims to have really good filtering, to avoid such a situation. (After all, Rodrigues offered, it is one of the world’s leading audio filter vendors.) D2Audio, also highly concerned about the potential for RF interference, took a different path. “That’s one reason why we’re not selling a single-chip solution,” said Skip Taylor. “We can shield a box properly. It’s a complete EMI-shielded environment.” D2Audio sells only as an OEM to name-brand audio systems vendors, like a chip supplier, but sells complete subsystems in a box, ready for deployment typically within loudspeakers or in amp enclosures.
     He added, “We also don’t sell silicon [outside of the complete boxed solution] because we’re selling a time-to-market advantage to our customers.”
     Taylor also touted D2Audio’s inclusion of an “overpowered DSP” in its solution, which is available to run software such as equalization or digital format decoding. Audio companies may find very good reasons to use that DSP’s spare processing power to decode streaming signals such as MP3 or Windows Media; to decode Dolby Digital, DTS or other surround schemes; or to process digital rights management or even run an operating system (Windows CE or Linux, most likely) within the amp module or the powered speaker. “This product is the culmination of the last 20 years of my experience,” said Taylor.

Speakers Running Windows?
     D2Audio is hardly alone in its DSP spare-horsepower capabilities. Many other companies have been carefully considering just which set of digital features they ought to puzzle together at the amp or powered speaker level, now that Class D amps provide the digital shelf space, and now that they allow amps to run cool enough so it is not ludicrous to think of putting a computer in one.
It may be hard to understand why anybody would want a loudspeaker to run Windows, but Microsoft probably can think of some plausible reasons. More likely market successes in digital amp features may include the following ramifications:
• Surround or streaming signals delivered to the amp or speaker would affect product licensing fees. Dolby, for example, charges fees per decoding device. So they’ve had problems selling PC vendors on surround sound; should HP pay a fee for every PC, even though relatively few are used as music listening devices? Should Apple’s fee per PC be higher, because Macs may be more likely to be used that way? It would be much easier for a Dolby to license loudspeakers and amps, instead of PCs.
• Variants of this strategy may also come to installed audio, where licenses per system, for the likes of Dolby’s and DTS’ theatrical products, now are built into products that simply aren’t used in many large PA systems. So it is hard to get surround sound into certain settings without buying into some components you may not otherwise need. Surround decoding built into powered speakers might allow for more better-sounding systems getting built, at more reasonable cost overall.
• Digital rights management and signal encryption to the speaker can make content owners happy by limiting access to digital programming. This is important especially in the case of live sound, where bootleg concert recordings are a serious issue.
• Very complex new adaptive EQ and latency control algorithms may be a radically important innovation—and soon a requirement—in an installation with hundreds of powered loudspeakers. One provider of such software that is betting on a big market in installed audio is the Italian firm, PowerSoft. For the home, Harman already has a receiver that includes a remote control featuring a small microphone, which feeds information to an equalizer to adapt sound for any listening position. Imagine system contractors being able to quickly install large numbers of speakers at only roughly estimated locations in a venue, then walk through the room(s) with a microphone that allows for instant, automatic EQ optimization.
• Combine the previous items with Class D’s relatively low heat dissipation and lighter weight, and it becomes radically less expensive to install loudspeakers just about anywhere that AC and network-connection wires can reach.

Many More Channels
     By far the most common development from Class D amp manufacturers in high-end home and small commercial installation markets, however, is to take advantage of the size, low-power consumption and low heat production of the technology to build many more channels of audio into a single amp enclosure. That way, multi-room surround systems can be powered from one or a few central location(s). Among the vendors that have recently introduced 12, 16- or 32-channel PWM amps are Crestron, Sonance, SpeakerCraft, Elan, Phoenix Gold, Audio Design Associates, NuVo and Oxmoor.
     Another radical new possibility is for the network interface for a digital audio network to be integrated with the PWM amplifier on a single chip. In other words, Peak Audio’s CobraNet, Gibson’s MaGIC, Peavey MediaMatrix, Netstream’s Digilinx, Digigram, Fostex’s NetCira, Yamaha’s mLAN, etc., might run from an Ethernet MAC (or 1394 PHY, in mLAN’s case) straight to a front section of the Class D amp chip that processes the communications protocol. This would save some money and some precious board space in products such as bookshelf speakers that have tight space constraints.
     But silicon manufacturers we spoke with think it is a level of integration too far for now and the near future. Such a system on a chip would involve highly complex mixed-signal design that will take a couple of years and high projected chip volumes to justify development expenses.
Those high chip volumes usually materialize only where consumer applications are involved. So we may see such integration happen first with wireless network links. This becomes especially interesting with the coming generation of Ultra Wide Band (UWB) technology, which is expected to offer tight jitter and latency characteristics compared with existing 802.11 radios. UWB may allow distribution of signals to multi-speaker settings with as accurate spatial imaging (in other words, tight timing control) as achievable with wired links.

Could Be - But They’re Not
     Any of these network connections could be built easily into in a box like D2Audio’s. But neither that company nor any of the other Class D amp providers is doing so…for now. Instead, AKM, D2Audio and others speak of providing I2S, AES/EBU and analog interfaces (ADCs) on their PWM modules, and leaving it up to their customers to decide how they want to feed signals. And, indeed, it is downright fascinating to look over the names of the Class D amp providers and those of the new digital networking providers, and see that the two lists hardly intersect.
     Texas Instruments (TI), the largest of all suppliers of chips to the AV business, may break this barrier. It is said to be developing a new Class D solution, but we couldn’t get anybody to go on the record about it yet. TI already supplies most 1394 chips; few are used for high-quality audio, although a recent TI design had raised some strong interest, according to insiders. But the hard line between digital networking and digital amplification proved too difficult for Cirrus Logic (CL) to cross. The owner of Peak Audio, and thus of CobraNet, briefly brought out a Class D solution in 2002, but discontinued it. Skip Taylor, who worked at CL at the time, said, “There was nothing wrong with the technology at all. There was just a decision to refocus on a lot of DVD technology.”
For brand-name amp and powered-speaker manufacturers, this implies a serious requirement to choose carefully just which innovations they want to introduce when. It is difficult to work closely with more than one new OEM technology at a time, so a clear pattern has emerged: Many pro audio manufacturers developed IT networking capabilities first and moved more deliberately—or are now moving—into Class D amplification. There are more than 40 CobraNet licensees, for example, and among them now several Harman companies, Peavey, Lake and increasing numbers of others are bringing out PWM designs for installed audio.
     The biggest advantage of all for PWM? Ask any jobsite employee at a sound contractor while he’s putting 20-pound PWM amps into position instead of 70- to 100-pound analog amps. New technology can sometimes be all about easing the schlepping issues.

Neal Weinstock, editor of the IT/AV Report, is the founder and president of Weinstock Media Analysis, a market research firm. He helped found BridgeCo, a Swiss maker of audio networking semiconductors, has authored two books on computing and design, edited magazines including TV World, and has written hundreds of articles.