Class AB Amp slight distortion at any level

Question

I wanted to build an amplifier for a spare set of 4 ohm 20W speakers I have laying around. After three days of searching circuits on the internet and trying them out on a breadboard not having much knowledge in amplifiers, I've finally managed to piece together this circuit that actually amplifies the signal instead of doing nothing. The problem that I'm having is, there's a slight amount of distortion at any input amplitude. RV1 has been tuned to drive the opamp just enough to 0V before the output starts clipping, directly hooking the inverting input up to 0V distorts the audio like crazy. Not having and oscilloscope and only a cheap multipeter how would I go about improving this circuit? Changing R1 and R2 to something lower makes the problem worse, to something higher and nothing happens, above 10k and the amplification fades away signifficantly. And the LM358 is the only op-amp I have

The reason why I used MOSFETS is because I've had them laying around and nothing else. And yes the power source is sketchy atm but that'll be changed out for a proper power source once the circuit is finalised and works without distorting.

Answer 1

Q2 is upside down (S and D are reversed): -

You should also use a 2.2 kΩ pull-down resistor on pin 2 of U1A to 0 volts and use non-polarized 10 μF decoupling capacitors. The 100 uF electrolytic will act like a diode when Q2 is conducting so change it to 10 μF non-polarized.

And, given that you say that Q2 is connected up correctly and that only the schematic is wrong, you will need to increase the bias across the LEDs more than likely to something like 3 volts for each. This is based on the transfer characteristic graph in the IRF540N data sheet: -

This is the gate-source voltage that just about causes the MOSFET to begin linear conduction. Around 2 volts the MOSFET will be relying on the negative feedback of the op-amp to jump over the hump of cross-over distortion and, given that the op-amp you have used is fairly slow, you will get distortion if the biasing is much below 2.5 volts. Start with 2.5 volts is my recommendation.

You might also try a 330 Ω resistor from the op-amp output to the speaker output to allow the op-amp to handle the low power signals feeding the speaker. This will linearize things too. It's quite a common-practice.

If you are using red, green or yellow LEDs, they will only produce a forward drop of about 2 volts and, that is a little low. I'd say aim for 2.5 volts.

Answer 2

(After reading your comments, I realize you got it MADE properly, unlike what the schematic shows, so ignore this first paragraph.)
First of all, your P-channel MOSFET drain and source pins need to be flipped around; the source pins of both MOSFETs are tied together to the output. If your transistors conduct and get hot when you swap the drain and source of the Q2, the Q2 might be damaged. You should check it for shorts across any two of its pins, and it should act as a diode between drain and source (drain positive, source negative). Just make sure you don't turn it on with the instrument and then get a false short reading.

Second of all, you will have crossover distortion because the gate-source threshold voltage could be anywhere between 2 and 4 volts, and each of the two LEDs for the MOSFETs need to match their particular MOSFET to which they are connected (upper LED for the upper MOSFET, lower LED for the lower MOSFET).
The idea of the LED here is to keep the MOSFETs around their threshold voltage at slight conduction to prevent crossover distortion.
So, measure each MOSFET's threshold voltage (at which it starts turning on), and then get an appropriate LED for it: a RED LED is typically around 1.7V, GREEN is about 2.0-2.3V, BLUE and WHITE can be from 2.6V to 3.4V. You can also combine two LEDs to get appropriate value for each, the D1 and D2 diode.