I wanted to finish up our little series on how Class D amps worked today, tomorrow we’ll enjoy Christmas with our families and then onto new and fertile grounds for audio the day after.
We’ve learned that the output of a PWD amp is very simple: a series of on or off pulses that vary in length. If the signal is getting louder the pulses stay on longer and if the signal is getting quieter they stay on for less time. If all this happens fast enough the connected loudspeaker won’t be able to reproduce exactly what the PWM amp is doing (turning on and off rapidly) and the end result is a smoothing out of the on and off pulses into music.
So timing and speed are key elements in the amp technology. With the on and off pulses happening at hundreds of times faster than we can hear, the speaker ignores the stop and go of the output. There’s also a smoothing filter at the output of the class D amp that ties everything together and removes the gaps between the pulses. It is this filter that can have a major impact on the way a Class D amplifier sounds and is, in particular, the single most difficult and problematic design problem of Class D amps. Designs live and die by the way this filter is implemented.
How does the music get converted to these long and short pulses in the first place? That happens at the input of the Class D stage in perhaps the simplest of elements: the comparator.
The input comparator is a wonderfully simple analog timer that runs at a fixed speed and has one very simple task: look and see if the music is louder or softer than a moving reference signal. So imagine a scale to weigh something: the type with the balance beam and a fixed weight on one side. The fixed weight is the music we want to encode and, for just a moment, imagine we assign a weight to loudness. Using a weight scale of 1 to 100, let’s choose 20. Now on the other side of the scale, we start adding weights until the scale tips and then no matter how many more weights we add the scale cannot move any further. This is similar to what happens on the output of our Class D amp.
One the timer side of the comparator is a rising voltage and on the other side the music. When the timing voltage is lower than the music side, the output of the Class D amp stays on. As soon as the timer side voltage is higher than the music voltage, the output of the Class D amp goes back to zero and waits for the next timer cycle. The timer side voltage is a simple triangle wave that goes up and down the same amount and speed every time.
The beauty of this scheme is that instead of breaking the music into small quanta (bits) like building blocks of finite size, the Class D encoder is 100% analog and the fineness of the comparison is every bit as infinite as analog itself. No longer are we required to break apart the analog into predefined bits and hope for the best. Like analog, the only restriction is frequency and maximum loudness – where PCM and digital audio have a third restriction, fineness of the numbers used to recreate the finite building blocks of the music.
Class D is 100% linear when it encodes the music into PWM and with a proper analog front end and a well designed output filter, has the possibility of outperforming any analog power amplifier. But those are a lot of ifs and most analog power amplifiers sound better than most Class D amplifiers: so difficult is the challenge of designing one.