Sidebar 1: Is Class-D Digital?
When I first read Mikey's introduction to this review, I was struck by his matter-of-fact statement that class-D isn't digital. Really? Are you sure?
Yes, I know the D in class-D doesn't stand for digital, but that doesn't mean the technology isn't digital. After all, in a class-D amplifier, the signal fluctuates rapidly between two levels, just as it does in DSD—what could be more digital than that?
Fact is, I've always thought of class-D as digital, and of class-D–isn't-digital claims as marketing propaganda propagated by people who want to sell class-D amplifiers to traditionalist audiophiles who think "digital" is a four-letter word.
But, as Mikey pointed out, digital-amplification guru Bruno Putzeys doesn't think class-D is digital; see the Sound & Vision interview Mikey refers to in the main article. Sure, as one of the main guys in Purifi, Putzeys has some class-D amplifiers to sell—but I can't think of many people I consider less inclined than Putzeys toward empty marketing-speak. So, I asked him to explain his position: Why do you say class-D isn't digital? The insights that follow are his; the mistakes, if any, are mine.
When a circuit works using two discrete states, it's perfectly natural to think of it as digital. The real question is whether thinking of it that way is helpful or not. In the case of a class-D amplifier, it turns out it's more helpful to put your analog glasses on.
Digital signals have a dual nature. They are analog signals—time-varying voltages—but they are also a sort of code. In their codelike guise, imperfections don't matter. It's like reading someone's handwriting: As long as you can make out all the characters correctly, you know exactly what the letter says—no errors. (Remember letters?) As long as your digital device—your cable modem, network switch, or DAC—can read the code correctly, then there's no error, even if the waveform itself is far from perfect. It's all or nothing, perfection or catastrophic failure. (I'm ignoring jitter, which is best thought of as an analog artifact riding on a digital signal. That's a different conversation.)
A loudspeaker, though, doesn't care about numbers. It cares about electrical signals—time-varying voltages that create magnetic fields that make speaker cones move. In that realm, errors matter. Errors change the sound.
"A class-D power stage," Putzeys wrote in an email, "just like a class-A power stage, produces distortion that depends on the power supply and on the output current that the speaker draws." Engineers deal with that distortion via what Putzeys called "highly analog countermeasures." If a problem has an analog solution, then it must be an analog problem. Besides, you can't make a loudspeaker's cone move with code.
"Choosing to call it digital or analog is really a crucial engineering decision innocently masquerading as semantics," Putzeys told me.—Jim Austin
But, as Mikey pointed out, digital-amplification guru Bruno Putzeys doesn't think class-D is digital; see the Sound & Vision interview Mikey refers to in the main article. Sure, as one of the main guys in Purifi, Putzeys has some class-D amplifiers to sell—but I can't think of many people I consider less inclined than Putzeys toward empty marketing-speak. So, I asked him to explain his position: Why do you say class-D isn't digital? The insights that follow are his; the mistakes, if any, are mine.
When a circuit works using two discrete states, it's perfectly natural to think of it as digital. The real question is whether thinking of it that way is helpful or not. In the case of a class-D amplifier, it turns out it's more helpful to put your analog glasses on.
Digital signals have a dual nature. They are analog signals—time-varying voltages—but they are also a sort of code. In their codelike guise, imperfections don't matter. It's like reading someone's handwriting: As long as you can make out all the characters correctly, you know exactly what the letter says—no errors. (Remember letters?) As long as your digital device—your cable modem, network switch, or DAC—can read the code correctly, then there's no error, even if the waveform itself is far from perfect. It's all or nothing, perfection or catastrophic failure. (I'm ignoring jitter, which is best thought of as an analog artifact riding on a digital signal. That's a different conversation.)
