Yesterday we mentioned that DSD and Class D operating systems were similar in that each can be played directly into the loudspeaker without conversion; unlike digital audio which cannot. Both are bit streams - DSD a series of fixed width bits that get more or less dense to make music and Class D a string of single bits with varying lengths - yet require no conversion to make music. In effect, they are both efficient analog schemes where digital audio (PCM) is an efficient digital system requiring precise translation from its numeric values back to analog.
Unlike DSD, however, Class D (PWM) is better suited to a real time use (as in a power amp) than a store and playback scheme - something DSD and SACD are quite good at. Because DSD uses a fixed width bit - and we simply need to record the number of those bits in any given amount of time to capture audio - it is perfectly suited to the recording and storage process for later retrieval on a DVD because the retrieved quality/precision of an individual bit doesn't matter. PWM, on the other hand, depends on the precise length (or width) to put more or less energy into the speaker to make music - and this lends itself far better to a direct playback situation rather than storing and later playback because in this case we do care about the quality/precision of the bit.
This limitation of PWM is not a big deal for us because we're not looking to use it for recording and playback, but it will become a big deal when we take a look at some of the reasons why Class D amps have, in the past, gotten a (well deserved) bad rap for their sonic performance. But more on that later.
Right now let's finish up understanding how the Class D amp works. We covered the input comparator, a simple on/off switch that compares the music to a reference up and down voltage, resulting in longer or shorter bits of energy we can use to make music. The bits of longer and shorter energy are called pulses and we are varying the length of time each pulse is on or off - time is referred to as "width" and thus we get Pulse Width Modulation as our scheme.
The output of this comparator is controlling a couple of power transistors that connect the big and high wattage power supply to the loudspeaker for longer and shorter periods of time. When the transistor is on, the entire power supply is directly connected to the speaker to put energy into it (which is why you need a near perfect power supply for good sound). The longer it is on the more energy goes into the speaker and we get movement. If we do this quickly enough, say a hundred thousand times a second, this choppy process seems perfectly smooth to our ears - just like a choppy series of still pictures, moving at 24 times a second, can fool our eyes into believing a moving picture is "real" and smooth.
This high speed choppy on and off process does have a drawback we need to deal with before we're finished - noise. When you switch high power on and off quickly you create noise and, if the power is high enough, this noise radiates out into the world just like a radio transmitter. What we'd like to do is smooth out the choppiness before we connect up a speaker cable and loudspeaker. This smoothing out is the job of the output filter, one of the biggest problems Class D amps have.
Remember a few days ago I wrote about how the last thing I would ever do as an amp designer is stick an output transformer on the output of a power amp like just about every tube power has? The same can be said about this output filter, it too is a necessary evil that has to be dealt with if we're going to get good sound from our amp.
The output filter of a Class D amp has been, in the past, one of the most troubling aspects of this amp class - until several really smart designers got involved and solved the problem.