A Little Theory of Class-D Amplifiers
A class-d amplifier uses Pulse Wide Modulation PWM (sometimes
referred to as pulse duration modulation). This means that the original signal
Vin at the input is modulated with another signal Vm which has a much higher
fixed frequency. The waveform which is used as carrier wave or modulation signal
is normally a sawtooth signal. The principle is actually quite simple to understand
if you look at the figure below:

The red signal is Vin and the blue signal is the modulation signal
Vm. The pulse width modulated signal Vd can be expressed as:
Vd = "1" if Vin > Vm
Now we have a discreet or digital signal which can only be either
"0" or "1". The signal has a fundamental frequency equal to that of
the modulation frequency fm (the frequency of the modulation signal Vm) but will also
contain the input signal and a band of frequency components around the modulation
frequency.
A Low Pass filter with bandwidth fc has to reject the switching
frequency fm and components around fm, passing the amplified input signal at
the output.

Schematics
Single push pull output stage:
Balanced output stage:

|