I'm using some DC-blocking capacitors to get the MP3 player output (which is grounded at -5V) to match the audio amplifier (grounded at -12V.) Unfortunately, I keep getting 1KHz noise from the switching supply. I tried going to the linear supply but that didn't work either. I even tried a big filter capacitor right at the MP3 player but no joy.
I figured out that the MP3 player was generating clean output but the speaker amplifier had issues with its ground being noisy. I tried a number of filtering techniques and such but came up dry. I decided to try with the LA4445 amplifier again: I figure I can bend the 2mm pins to fit the 0.10" spacing on the circuit boards I have. It dawned on me that I could probably make some kind of common-collector amplifier in a class A-B configuration.
I figure I can connect a resistor from emitter-to-base on each pole and one from base to a common input. The ratio of resistors defines the quiescent current and the values define the amplification. I may avan be able to do a push-pull with the 5-volt source for lower power and higher efficiency. I could set up the output to center around the -5V which is signal ground for the MP3 player. I thought about it for a while and decided to go with a common-emitter design with some biasing resistors and an op-amp — using the op-amp to compensate for the transistor switching.
First I tried the LA4445 because I thought it would be simpler. I spent an hour or so working, got everything wired up, but I took some shortcuts and the thing just won't work. Crap. Back to square one — again.
I decided to tinker with the transistors a little. I got to the point that I could generate output, but they won't stay on all the time in a true A-B configuration so it gets all noisy and such. I considered using some resistors and diodes to keep the transistors biased all the time.
I fianlly got a viable solution. I used the same crappy design I used for the amplifier I built years ago which just drives a couple opposed Darlington transistors with an op-amp as feedback, although I did a better job of fixing the high-frequency oscillation problem. It ends up drawing about 400 mA (5 watts) with the MP3 player in-circuit and the amplifier at volume with both speakers driven in parallel. I suspect that the output is around 2 watts total or so, based on around 2 volts RMS (as measured by the oscilloscope) into 2 ohms. If it hits 2.5 volts, it's at 3.1 watts which I think is about peak.
Based on the circuit I used, the input is already self-centering around the midpoint of the power supply. The transistors need 1.2 volts peak-to-peak to maintain quiescent current, so the theoretical maximum output (going rail-to-rail) is +/- 5.4 volts; RMS voltage is 3.8 volts or 3.6 watts into 4 ohms. I tried to get away with a single op-amp to do voltage gain and to power the main amplifier. It seems to work but is even more twitchy than before … although I guess the noisy volume level is so quiet that you'd have to be right next to the speaker to discern it. Here's the circuit:
I had an op-amp I was using to generate a solid signal ground, but I didn't need it so I took it out. Anyway, at the 12-volt source, the audio section (with the MP3 player draws about 50mA and when it's at full volume, it peaks around 200mA, so I assume the output power is around 1.8 watts total consuming some 2.4 watts at peak volume. It kind of sounds like crap with a bit of fuzzy distortion from the transistor switching noise — if you know circuits, it should be pretty evident why:
I got the amplifier wired and it works okay from the linear power supply but it's not working off the battery and I'm not sure why. It dawned on me that I can just use one of the op-amp's as a differential amplifier off the MP3 player and feed that into the prefabricated amplifier? I ended up with this simplified circuit:
The first experiment worked great so I rewired the circuit to take out all that extraneous crap to deal with the transistors and got it working again. It looks like I can get something like 3 watts of audio out of the thing, and the whole powered system typically takes around 2 watts at moderate volume.
It's just too bad I had to ditch the power transistors and heat sinks because they really looked cool.