Stay in touch…

Blog

Read the latest Bitstream

RSS Feed

LinkedIn

Look for us at LinkedIn

Twitter

Follow us on Twitter

Mix Magazine

This installment of The Bitstream column appeared in the June 2005 issue of Mix Magazine.

The Bitstream

This column discusses Class D amplifiers…

DJ Classy D

This month, I’m starting on a rant about commoditization and one aspect of audio production and reproduction is the increasing integration and digitization of electronic functions formerly given over to Ye Olde Analog methods. One longtime holdout has been amplification which, as with most other applications has finally begun what some view as the downward spiral toward, blech, digital.

Let’s take a look at at these so called “digital” devices, to understand why people think they’re digital, and maybe give some respect to a worthy technology that sure could use some…

Digital amplifier; the phrase conjures up the vision of either an esoteric piece of hi–fi gear or a cool, compact, rack mount amp for reinforcement. What you may not know is that “digital” amplifiers are now cheap, ubiquitous, and available in virtually all power ratings.

First off, let’s get one thing straight; they ain’t digital, not in the generally accepted sense. Now, they ain’t Class A either but they ain’t digital, at least not from the pulse code modulation world view anyway. They are digital in the binary sense, however, but PWM power amps are more in the delta sigma vein of DSD than the brute force approach taken by PCM.

OK, Pulse Width Modulation; a modulation technique that converts analog amplitude variations into duty cycle variations of a very high frequency square wave. If the amplitude is zero, the square wave is off, not present. If the amplitude is low, the square wave spends most of its time off with an eensy weensy proportion of its time on. As the amplitude rises, the duty cycle or amount of “on time” increases and, at full power, the output devices are at 50% duty cycle — a full power square wave — dumping all the current the power supply can muster into the load.

At the final output sits a high current low pass filter. It filters the spurious harmonic energy, leaving the amplified, reconstructed audio program to pass unencumbered…Sounds familiar? It should, as the LPF serves the same purpose as an anti–image filter in a D/A.

This PWM thang is basically the same approach taken with cheap, SCR or triac–based lighting dimmers, the kind that are banned from any self respecting audio facility. What you may not know is why they’re banned, and that’s all about RF–induced noise in your studio, which we’ll look at later…

Most of you do know that, in the realm of amplification, Class A is the holy grail and only accessible to those who genuflect to the efficient, indirect radiator Horned (loaded) God. If you are a Class A acolyte, you should grab the prayer beads now and run screaming for the next article because, PWM amps are philosophically anti–Class A…Where Class A seeks to confine the output device within a small but linear operating range, a PWM amp’s output devices are purposefully designed to go non–linear and pretty much stay that way until they’re shut off altogether. One is extremely wasteful while the other is efficient to the max.

Efficiency is one thing but, as we all know, there are many other metrics by which one should judge an amplifier and, Class D has its share of problems. Since we as an industry have yet to see fit to concoct a standardized testing regime that actually correlates to our subjective impressions of an audio system under test, we have to blunder our way through a quagmire of operational parameters in the hope of finding something meaningful. For Class D, slew rate, power supply current limiting and band limiting output filters all have a major stake in the overall operation.

I talked earlier about RF noise, and I’d like to come back to that for a moment. For those of you that have listened to a 1 kHz square wave, you know it doesn’t sound much like a 1 k sine wave. The reason is that damnable gotcha, harmonics, and the distortion they create. You see, square waves are composed of an infinite summation of odd harmonic sine waves of decreasing amplitude (6 dB/octave) and that odd harmonic energy is what lends that air of, shall we say, annoying timbre. Notice I said “infinite” and, it’s that harmonic energy marching out to the microwave band that makes SCR dimmers an audio pariah. Whenever you switch an electronic circuit off, then full on as fast as you can, the circuit will generate radio frequency noise and that radiated noise can and will, if allowed to, worm its way into poorly shielded and grounded analog inputs everywhere nearby. So, Class D amplifiers, like their distant relative, the switching power supply, require inclusive Faraday cages to contain the RF beast.

That same RF–rich output took me by surprise the first time I hooked up a Class D module for testing. I thought the damn thing was oscillating the way an uncompensated analog amp will, until Ken, my wiser and more experienced colleague, pointed out that that’s precisely what Class D amplifiers do for a living, they oscillate! The output was an amplified version of the input all right, but the waveform had a massive amount of “fur” on it, the noise created from the switching operation. We all know that good grooming dictates that extraneous hair should be removed prior to presentation (you Unibomber look–alikes out there take heed) and, as I mentioned earlier, a low pass filter is a necessity at the output of a Class D product. That filter should be a simple as possible, a RL network suffices and keeps the phase response within reason, but the passband must be controlled or the intermodulation products, not to mention the ultrasonic energy, will ultimately take their toll on either subjective sound quality or your tweeter’s voice coil.

Looking at the power aspect of a power amplifier, Class D amps behave differently from Class A or A/B when it comes to maxing out. When the supply rails reach their limits, Class D amps tend to compress, rather than clip since they’re already clipped, but the power supply’s current capability is what shapes the overload behavior of the amplifier to which it’s attached.

Something related to this is the input signal and, if the amplifier is of the digital–input variety, either DSD or LPCM, then the input can clip when presented with continuous digital overs. If the amplifier is a hybrid sort of creature, with an analog input, then that can, of course, also clip in the old school way.

Some manufacturers refer to Class D amps with digital inputs as “pure digital” or some similar moniker but, don’t be bamboozled. Since the output of a Class D amp is analog, admittedly a very strange analog but analog nonetheless, you cannot really call them “digital” amplifiers. As I said at the head, they’re analog amps, just a wee bit nonlinear in their operation.

The final gross arbiter of performance with Class D products is slew rate or the rate of change capability of the output devices. The device’s finite rise and fall times, how fast the transistors can switch on or off, determines how faithfully they can track the pulse width commands given to them. If the transistor is still slewing up or down when a new command is given to head in the other direction, it isn’t able to complete the last command, so to speak. This results in slewing–induced distortion very different from that same distortion in an “analog” amplifier. There are other factors that also govern performance, such as stability of the modulator’s clock which, along with power supply voltage regulation, dictates linearity by the same mechanism seen in an analog to digital converter.

When all of the above considerations are carefully balanced, the result is a physically small, efficient and good sounding amplifier whose deceptive form factor belies the power trapped within. Nowadays, everything from iPods and cell phones to self–powered studio monitors sport versions of Class D amplification. Some of the many manufactures providing products, from single chip versions to complete modular OEM assemblies, include Apogee Technology, Bang & Olufsen, Carver, Cirrus Logic, D2Audio, DEQX, Maxim Integrated, Microsemi, Mitsubishi, Mueta, Power Physics, Pulsus Technologies, Sanken, Texas Instrument, Tripath, Wolfson Microelectronics and Zetex. Next time you’re at a hi–fi store or trade show, take a moment to recognize the changing face of amplification. Class D, It’s not just for reinforcement anymore!

Many thanks to Bang & Olufsen for their illustrations, as well as Ken Kantor for his patient guiding in the ways of The (Electromotive) Force. Until next month, keep on rockin’!

Bio

OMas, as Director of Brands at Sonic Studio, has had more than enough trade shows on his mind, but still found time to chill with the tranquil alt rock of The Album Leaf’s One Day I’ll Be On Time on Tiger Style Records.