For a long time now, I have put off breadboarding a Class-D amplifier.  A Class-D amplifier operates the output transistors as switches instead of as variable resistors, as in more traditional amplifiers.  In doing so, nearly all conduction losses in the output power devices are eliminated, leading to a very efficient amplifier.  Much more information is available at Wikipedia.

This circuit is based on a design by Johan Sorensen, who developed this circuit over a period of time in the early 2000s.  His work appears here and is a great read on the practicalities of Class D amplifier design. Notable differences between his circuit and mine center around my use of a MOSFET gate driver IC instead of his discrete component solution. His solution has the advantage of working at higher voltages and (likely) with cleaner, more precise switching waveforms; mine has the advantage of simplicity and the reduced need for many different supply voltages.

Let’s begin by having a look at the circuit, shall we?

The circuit begins its amplification by blocking any DC component that might be present on the input signal, then sending it to the non-inverting input of a common TL082 operational amplifier.  This amplifier accepts negative feedback from the power stage to better control the output waveform.  Next, the output signal is sent to a comparator.  If the signal is above 0V, it outputs a logic high; otherwise, it outputs a logic low.  This signal is then sent to the MOSFET driver which generates voltages appropriate for driving the MOSFETs.  The MOSFETs allow either +12v or -12V to flow to the speaker via the output filter.

The feedback signal is taken from before the output filter, which might not be what you expected.  The reason this is done is because the output filter would introduce new poles into the closed-loop transfer function and would destabilize the circuit.

Note that in the schematic, the VCC pin for the gate driver is connected to ground.  This is not a mistake.  The IC’s ground pin is connected to -12V, so it still thinks it’s connected to a normal 12V supply.  A comparator with an open-collector output was chosen.  During operation, the comparator output is either “on” and pulled to -12V via an internal transistor, or it is “off” and pulled to ground via the 1K resistor R6.  These two voltage levels, because of the gate driver’s power supply connections, are interpreted as low and high, respectively.  A second reason for this connection is that the gate driver must share a common connection with the “bottom” MOSFET.  The driver has a built-in charge pump to generate a high voltage to turn on the “top” MOSFET, so this solution works quite nicely.

Immediately after the circuit was deemed working, a short video was made to showcase its functionality:

Audio quality, as you can tell, is not crystal clear; however, there is a lot about this quick experiment that isn’t ideal. Given that it was made with spare parts I had on hand, I am glad it works as well as it does.

Update: There is an error in the schematic.  Pin 3 of IC1 (the inverting input of the LM311 comparator) should be connected to ground.  My apologies for the omission.

Update 2: For better performance, replace C2 with a 0.001uF capacitor and R4 with a 1M resistor (or just don’t populate one at all).  This has the effect of greatly increasing the available gain before distortion.