CMOS Power Amp

CMOS Invertor Power Amplifier
Not too long ago, I wondered how complementary MOSFETs in a CMOS inverter configuration would work as a power amplifier. Its common knowlege that CMOS invertors produce tubish rounded clipping when overdriven. The CD4049 CMOS Hex Inverter IC has been used in several overdrive circuits. However, existing CMOS inverters are low current logic gates, and certainly cannot be used to drive speakers.  I figured a IRF640 paired with a IRF9640 would do the job well.
This purpose if this design is mainly to test a concept that might be built upon later and is intentionally limited. It is intended as a small power amplifier rather than a complete guitar amplifier. There are no tone controls, and it needs preamplification. As a Class A amplifier, it is inefficient as hell and produces less power than one could get with a Class AB design. However, as shown, it does pretty well.
I used Multisim 10 Student Edition for designing and simulating the circuit. In simulation, the circuit drew about 750 mA. According to simulation, the overdrive characteristics are pretty good. Here is a screenshot of the Osciloscope window from Multisim showing the input signal on top and output signal on bottom.

 The first stage is simply a MOSFET source follower to provide a low output impedance signal source for the power stage. The output impedance of this stage will effect the negative feedback to input signal ratio of the power stage. Any low output impedance buffer should do, but you may want to avoid op amps or any non-Class A buffer if Class A sound is important to you. With a suitable preamp, the first stage might not be needed.
The power stage is a basic CMOS inverter using complementary power MOSFETs rather than a typical low current digital CMOS inverter. Unlike an ordinary CMOS inverter, this is an analog inverter rather than digital, but then again, so are the CMOS inverters in any CMOS overdrive circuits.
R6 and R10 limit the current to something managable. The MOSFETs are designed to handle 10A, but I imagine that requires large heatsinks and a power supply more powerful than the PC power supply I'm using. One would think that using a current source or simply limiting the supply voltage would do the job, but in simulation such setups seemed to adversly effect the output characteristics. The symetric placement of the resistors seem to matter too.
The zener diodes are there for gate protection. I figured 15V zeners would allow for a large enough input signal and still provide adequate protection. The 1 kΩ gate resistors are there to suppress oscilation, and the value was chosen because I have a bunch of  1 kΩ resistors. Different values might work better. The value of the input and output capacitors also reflect what I have available, and are large enough to prevent significant loss of bass frequencies. 1 mF might be overkill for the input capacitor.
Because R8 ties the output to the gate, it gives the output stage an approximate midpoint bias. This makes the difference between having a digital inverter and an analog inverter. Anyone who is familiar with using the CMD4049 in analog applications is familiar with this. It also provides negative feedback(which is why it midpoint biases the circuit). I chose a low value resistor for this, so the feedback would be significant.
R5 allows me to control the input signal to feedback ratio. This affects both gain and damping factor. A larger value pot would probably allow for more gain. With the pot at maximum value, the gain is highest, but the damping factor is lowest. Vice versa if the pot is set to 0 Ω. The ratio of R5 to R8 might matter more than the specific values chosen for these, but values that are too high might affect frequency response by limiting how fast the MOSFET's respond to signal changes.
The biggest difference between simulation and the actual build is the current drawn by the actual circuit built on my breadboard. It draws about 1.3 A and the heat sink(which I scavanged from an ATX power supply) gets pretty hot.  I guess FETs are like a box of chocolates, you never know what you're gonna get. It loads my power supply down to around 10.5 V. I really need to build a better power supply for testing.
To me (as best as I can remember), this sounds better than the 12V Mufu design. I used a booster circuit for the preamp. As would be expected, with R5 maxed out, I get more gain, and with it turned down a bit,  the bass response is more clear. I don't have a frequency generator, so I don't know if it clips quite the way the simulated version does, but I like the way it sounds. Its significantly louder than either of the two Ruby amps I've built. I did hook up my little handheld osciloscope and used my 7 string for a signal source, and it did seem to have rounded clipping. I had no problem with the frequency response, but I was using a car stereo woofer. It might possibly sound a little dark with guitar speakers due to the effect of the gate capacitance of the MOSFETs on high frequencies. With soft picking, I get clean sounds, and with harder picking, I get more distortion.
I hope you like this design. If anyone builds this, let me know what you think.