It turns out that Brian Halla, National Semiconductor's colorful president, is a country-and-western music fan. He asked me if there was a C&W title among the CDs I brought to demonstrations of the company's digital audio products at the Santa Clara Marriott last week.
My musical tastes tend to be eclectic: I had some Madonna albums, some Loreena McKennitt (in honor of St. Patrick's Day), a version of Mahler's Second Symphony, some Hildegard von Bingen (which is used synonymously with "New Age" these days)-even some Mark Isham film music, but nothing resembling C&W. I'm not even sure I know what that is, though I DO remember that when I lived in the Bay Area in the 1980s my pregnant wife was bemused by Gilroy Stations. I'd come home from work to find her humming things like, "She got the gold mine; I got the shaft" and "You're the reason our children are so ugly."
In any event, my Madonna CDs drew a pretty good crowd to demos of National's Class D amplifiers. A prototype 170-W subwoofer amplifier carried the baselines on the "Immaculate" album. Attendees also seemed to appreciate the drums and tabla on McKennitt's "Book of Secrets" album, even if they were unfamiliar with Celtic themes.
With the introduction of its LM4663 (at CeBit in February), a PC audio device that puts out 2 W per channel RMS into a 4-ohm load, National intends to do battle with Texas Instruments in the burgeoning market for power-efficient PC audio amplifiers, powered speakers and boom boxes. In fact, National's part goes head-to-head with TI's TPA2000D2, a 2-W Class D stereo amplifier that shows less than 0.07 percent distortion and noise (THD+N) at 1 W (0.5 percent at full power). The National part is rated at 0.2% THD+N (though it is unclear from the press release whether this is at full power). Such ambiguities didn't stop National's applications engineers from declaring, "Ours sounds better."
As good as that National system sounded, that IS a pretty tough claim to make. A prototype power amplifier built by Texas Instruments with the help of a company it just bought, Toccata Technology of Copenhagen, Denmark, and demonstrated in the back rooms at the Consumer Electronics Show in January, was specifically designed to combat audiophile prejudices about Class D amplification techniques. In a Hilton Hotel suite (next to the Las Vegas Convention Center), TI engineers and marketers set up an A-B comparison of their Class D power amp with a audiophile amplifier system (I believe it was an Adcom), using a pair of English-made B&W speakers.
TI allowed me to play and compare some rather long sections from a Mahler's Second recording I was carrying, a digital remastering of an analog tape on London. The French horns are so rich on this Zubin Mehta-led Vienna Philharmonic recording that they will make you cry, and twice we listened to a seven-minute segment of the music. In the end, I had to say the commercial system had the edge, but you had to be a real "golden ears" type to tell the differences. An audiophile might say that the Adcom put a little more air and space into the music, that the TI prototype had a more muscular tone (like a tube amp). But - with tears in your eyes from the shear joy of this musical listening experience - who wants to split hairs? TI's point, that Class D can come to audiophile quality, was made in spades.
This is interesting news for audio buffs who remember switching amplifiers as being shrill and difficult to listen to. Class D amplifiers use much the same technology as switching power supplies: a pulse-width modulator drives a set of MOSFET output devices, in a push-pull (or H-bridge) configuration. The width and speed of the pulses determines the frequency and amplitude of the analog signal pushed out of the MOSFETs.
The primary advantage of Class D amplifiers is that they are more power efficient than the Class A-B amplifiers typically used to drive speaker cones. Let me spell that out: Class D makes more efficient use of transistor switches than Class A or Class A-B. With Class A-B, you have to bias the transistor - feed it a current - that prevents it from slamming on and slamming off, but ideally keeps it hanging in a linear region between on and off states. The more bias current you give the transistor, the more linearly you can make it perform (Class A sounds best, audiophiles say), but this current, from 40 percent to 70 percent of the supply current, is dissipated as heat. Class D amplifiers, in contrast, require no bias current. The transistors just slam on and off as their makers intended them to do, but they use much less current and run a lot cooler.
This makes them really useful for modest-power audio amplifiers for appliances that need to observe space and battery life constraints such as powered USB speakers, patio speakers, portable radio-tape players (boom boxes) and laptop computers with multimedia playback capability. The high-power audiophile amplifiers that National and TI demonstrated to me are really prototypes of future devices. Both companies will develop 12-V devices to build 40-, 100- or 200-W amplifiers for use in car stereos. They will follow Tripath Technologies into this market: It is building high-power amplifiers using a variation on Class D, a frequency-spreading modulation scheme called Class-T, to build car amplifiers for Alpine. (Harris Semiconductor had conducted impressive demonstrations of its own car stereos using Class D, but they now appear dormant under the Intersil umbrella.)
The traditional problem with Class D amplifiers is they require an elaborate network of passive components - capacitors and inductors - to smooth out the pulses and strip off their digital artifacts. This is the sometimes audible stuff that makes audiophiles cringe. TI's claims that its recent introduction (the "third generation" TPA2000D2) has reduced the filter components requirement, but this is not quite correct. A glance at the applications circuit at the back of the preliminary data sheet for the part reveals (yes) far fewer inductors, but an R-C network as complex as any I've seen.
What's different about new-generation Class D parts is that they play tricks with the switching frequency to reduce the back-end filtering requirement. New power MOS, BiCMOS and linear CMOS technologies allow these devices to switch much more gracefully at elevated frequencies. Tripath does much of its work at 1 MHz, for example, which not only decreases the size of the output capacitors and inductors required, but also allows the systems designer to use a gentle roll-off filter, which is substantially cheaper and easier to implement than the steeper low-pass networks. TI claims it has phase-shifted the positive and negative pulses from the pulse-width modifier so that, instead of compounding their artifacts, they effectively cancel, leaving the user a much smaller filtering job. But audiophiles should know that even at 250 kHz most of the switching artifacts are well out of the audio range.
Still, at the National demo I had a hard time getting Brian Halla to feel comfortable with my musical tastes. I think he would have preferred party music. In fact, the ethereal female voices on my recording of Hildegard Von Bingen's church music decidedly revolted him. "You don't have many friends, do you?" he jested.