Operational Amplifier – How to Make a Bad Audio Amp

edit: What are important parameters in audio op-amps?

First there's noise. All components have some level of noise and there are several types of noise. While noise levels can be very low our ears are very sensitive to it. Noise is expressed in \$V/\sqrt{Hz}\$. That's a strange unit, but can easily be explained. Noise has a continuous spectrum and is defined as power over a specific bandwidth \$W/Hz\$. To get the voltage (in a specific load) you take the square root of that.
Next there's distortion. Probably the most published parameter is harmonic distortion, and it's the one manufacturers draw the most attention on. The reason is simple: it's relatively easy to obtain spectacular-looking figures like 0.01%. But these figures are rather meaningless, because the weakest link, the speaker, often adds several procent harmonic distortion extra, and our ears aren't that sensitive to it.
Then transient intermodulation distortion (TIM) is far worse. It occurs when a higher frequency component modulates a lower frequency, and because their product creates non-harmonic frequencies this is much more audible. TIM was discovered rather recently because measurements were originally done with single sine waves, and then this kind of distortion can't occur. High slew rate op-amps have low TIM levels. Despite being much more annoying than harmonic distortion TIM levels are hardly published, because it's harder to get the same fancy looking figures as for harmonic distortion.
Bandwidth is also important. Op-amps have a gain-bandwidth product (GBW) which indicates that the bandwidth depends on the amplification; a higher gain (amplification) results in a lower bandwidth. GBW is closely related to slew-rate, and you want to have a much wider bandwidth than the 20Hz-20kHz of audio to get high slew-rate values.

I've found a few interesting parts at Analog Devices:

[OP275](http://www.analog.com/static/imported-files/data_sheets/OP275.pdf): very low noise (\$5nV/\sqrt{Hz}\$), high slew rate and low distortion  
[AD823](http://www.analog.com/static/imported-files/data_sheets/AD823.pdf): comparable specs, a bit higher noise  

edit
Looks like I have to answer my own question... :-)
I read somewhere that the LME series is National Semiconductor's selection of high performance, high fidelity devices. You find the lot of them on National's site; there's too many to list here.

There are several power op amps, and also audio amplifier ICs, which would meet the stated requirements. However, this approach may be overkill for the kind of fidelity likely from the signal generation source.

As mentioned in a comment to the question, adding a transistor buffer stage is perhaps the best way to achieve the desired results.

Also, do note that a dual-rail (+/- x Volts) supply at a relatively high voltage is needed for these op-amps / chip-amps. While the dual-rail requirement can be circumvented with a suitable virtual ground implementation, voltages involved are still not your run-of-the-mill logic voltage range.

In any case, to answer the question...

Some suitable Op Amps:

• PA85: High voltage power OpAmp - Obsolete, I think, but still sold. This used to be a classic.
• OPA541: High Power Monolithic OpAmp
• OPA549: High-Voltage, High-Current OpAmp
• OPA2544: High-Voltage, High-Current dual OpAmp
• NE5532: Low noise dual OpAmp

Some suitable Audio Amps:

• LM3886: 68 W continuous average output power into 4Ω at VCC = ±28V
• LM3875: 56 W continuous average into 8Ω
• LM4780: Stereo 60W, Mono 120W (parallel-able) Audio Power Amplifier