How it is possible to make bad audio amp?

Question

After discovering how great op amps are, and that some extremely good ones--especially at low power levels--are available at reasonable prices, I wonder why all audio amplifiers, big or small, don't achieve excellent performance by simply combining a good small-signal op amp with a simple output stage.

I mean, with opamp there is no need to worry about all these bias voltages & temperature stability, just stick opamp and any unmatched darlington transistors, and you are good to go.

Any pitfalls?

Answer 1

The use of opamps in amplifiers can drastically simplify their design, but opamps aren't perfect. If they would have infinite amplification over their full bandwidth they would tend to oscillate, so they are internally compensated, which limits their bandwidth. A limited bandwidth makes the amplifier prone to Transient Intermodulation Distortion (TIM), a type of distortion much more annoying than harmonic distortion (HD).

The reason only HD is published, and TIM never is, is that it's much easier to get good-looking HD figures. Who wouldn't be impressed by a figure like 0.01% harmonic distortion? Most customers don't realize that this figure is totally irrelevant because the total system's distortion is for the most part determined by the speakers, which easily add a few percent distortion.

The power stage isn't without its problems either. Class A amplifiers are hardly used because of their low efficiency. Class B or AB amplifiers have a crossover distortion where one transistor takes over from the other. This is a non-linear distortion which can't be compensated by feedback. May not be true. If someone can enlighten here I'd love to hear it..

A final quote on opamps:

"There is no such thing as an unconditionally stable op amp unless it lies on the table with power disconnected" [1]

---

Further reading
[1] Intersil appnote AN9415: Feedback, Op Amps and Compensation

Answer 2

Interesting question - the answer (well my answer) is you can make a great audio amp this way. You would still have to pay some attention to the output stage, and overall design, but the use of opamps is no problem (and very common nowadays for basic, cheap amplifiers with good performance)
Although opamps are convenient tools and there are some excellent modern ones available, there are certainly still plenty of ways you can use them to achieve a poor result if you don't pay attention to the details.

This doesn't mean people will buy it though, and designers know this, so you still get high end valve based "Hi-Fi" amps costing >£2000 with 2% THD. You could maybe say the intention was to make a "bad" amplifier here as (ironically) it will make more money - unfortunately "great" means a lot of different things to a lot of different people.
You have some in the subjectivist camp that have basically decided the human ear is more accurate than any measurement tool, and can hear things none of them can see. So they can always say "Yes, your THD+n is indeed <0.001% from 20Hz-20kHz but you are not allowing for unmeasurable effect x with your design, and this is why it does not sound good to the ear"

If the desire for technical perfection were all that mattered then things like oxygen free cables costing hundreds would never get to market :-)

I think you may want to read Douglas Self's "Small Signal Audio Design" and "Audio Power Amplifier Design Handbook"
I have found him to be quite an authority on such matters. His books discuss the use of both opamps and discrete transistors. He weighs up the strengths/weaknesses which includes plenty of real life test data, and gives examples where you can obtain better performance with discrete transistors.

Answer 3

There are the drive requirements of the output stage to keep in mind as well. An amplifier rated for 100 watts average output power into an 8 ohm load, using a standard emitter-follower output stage, will require a swing of about +/- 40 volts peak to peak from the driver stage. Opamps that can output these 'high' voltages are significantly more expensive than ordinary audio op amps. In addition, there is still the issue of biasing the output properly and ensuring the biasing is temperature-stable; using an opamp as the driver does not magically solve this problem.

There are ways to use discrete transistors in the driver and output stages, along with associated biasing circuitry, and use an opamp as the driver, such as the applications note here. These circuits seem to be primarily for high-speed applications, however, and what advantage they might have for hi-fi audio (where stated goal is usually to have as few gain stages as possible, and make each of them as linear as possible prior to feedback being applied) is unclear.