Andrew and I spent a bonus 3.5 hours sitting in the 350 lab today working on our audio amplifier and it seems that we did nail it in the end. Our power amplification is well above the 1 million times specified in the manual. In fact, we pretty much maxed it out. Instead of doing like EE 340 where I figured out everything before I did it we decided to go for it and start building amplifier stages with the hopes that we could just tweak them into working correctly rather than actually calculating everything properly.
We started out by asking for maximum amplification from out MOSFET and found that we could get a voltage amplification of about 30 from it. Because it was impossible to drive our 8 Ohm speaker directly from the output of the amp we needed to put it through a BJT amplifier with a bit more reasonable of an output impedance. Anyhow, maximum power called for approximately 860 Ohms resistance on the drain of the FET, and this drove the BJT into clipping rather severly. Our simple solution to the problem was to decrease the amount of signal being coupled into the BJT by decreasing the efficiency of the FET, what started at 860 quickly became 560, then 480, 360, 240 and finally 150 Ohms. This amounted to a signal just clear of the level of clipping onset. And left us with well more than power amplification of 60dB.
Another curious note which we didn’t understand immediately, but soon got a handle on as everyone else in the lab was running into the same problem was a massive drift in power gain. It turns out to be no surprise at all that when a 100uF cap is in series with approximately 500K your time constant is a whopping 50 seconds! So evenif you let the circuit “warm up” for a minute you’ve still got a fractional drift of ~1/e to deal with after 2 minutes (1/e)^2 etc… Based on PSpice simulation I can drop that capacitance by a factor of 10 (or even 100) and not lose any bandwidth so it shouldn’t be a problem in the future (Time constant is 5 seconds (or 0.5)), just a funny story.