Eldartford brings up an interesting subject in that many of the amplifiers with more humble aspirations in the 1970s were capacitor-coupled . . . Johnnyb53 mentioned that his Altec integrated stereo had issues with reduced power output at 20Hz, which is a strong indicaion that it had an output capacitor. Most self-respecting "high-end" amplifiers, on the other hand, were DC coupled, possibly taking this to an extreme in the Kenwood L-07 that actually passes DC, and has DC gain.
Crown was proud of this in the late-1960s; I think that the DC-300 got its moniker because it was "DC Coupled". And nowadays, this power-supply configuration is standard, and an output capacitor is an anachronism . . . ah, progress. But wait a minute . . . a quick re-examination of the current path in a typical "DC coupled" power amplifier (including even the L-07M), and the speaker is still coupled via capacitors - it's just that now they've moved from the positive side of the speaker terminal to the negative, and we call them "power supply filter capacitors". If their only purpose was to smooth ripple, there could be just one tied between the + and - rails . . . but then there'd be no way to return the loudspeaker current to the supply. The power-transformer's center tap does this in the case of faults and on startup (and for extreme subsonic noise/content on the L-07M), but most power amps can in fact purr along nicely with this disconnected, and no DC ground for the power supply at all.
It's true that the "DC coupled" topology doesn't (or at least shouldn't) suffer from nonlinearities associated with the coupling/supply capacitors, the reason is that they are effectively inside the feedback loop, whereas in the "Capacitor coupled" topology they typically aren't. In the "DC coupled" scenario, there's still usually an electrolytic grounding the feedback loop (keeping DC gain to unity), and this can be a measureable source of distortion.
So that leaves me wondering a bit why it was cheaper to build a capacitor-coupled amp in the 1970s? The extra cost of front-end transistors is tiny. You still need two big electrolytics, or three for a stereo amp, so it's actually more expensive in that regard. There is a bit of savings in the lack of an output relay and protection circuit, but that doesn't seem like enough to offset the extra big capacitor.
This leads me to the conclusion that the real big chunk of cost was the talent to design an input stage with good offset characteristics, and perhaps the production technician to adjust the offset on every single amplifier, making up for the inconsistency in the semiconductors of the era. With an output capacitor, the whole thing is much more tolerant of both substandard parts and design mistakes. Now that transistor quality has improved dramatically, and there are plenty of existing designs to plagarize . . . those costs are much less significant for inexpensive systems, and capacitor-coupling has fallen by the wayside.
Some of those old designs were pretty good and while the art has moved forward on a lot of amp features (protection, thermal and short monitoring, metering) sound was quite good in many of them.
Indeed it was . . . and the idea of swapping in fancier parts to "improve" them to today's standards has some appeal. There's of course usually plenty of room for sonic improvement by simply correcting the ravages of 30-40 years of age, and technology and manufacturing of all electronic parts has moved forward innumerably in the past few decades. But the parts themselves are certainly NOT the real story, and the real improvement to be have comes from applying a modern design perspective to the older circuits, and making major or minor tweaks as needed. Some of them end up being pretty amazing just as they are, and others . . . well there simply aren't enough Black Gates and teflon caps in the world to help.
