In the circuit description of the 5532, Doug says he doesn't understand the function of Q14. I am not the designer, but I think he was on the right track when he referred to clamping. It's plain to see that Node 3 rides at about 2Vbe above the neg supply (Q8 + Q9). If node 3 (the collector of Q9) starts to go below 1Vbe, then Q14 turns on and sucks current out of Node 2, limiting the drive to Q8 + Q9. In effect, this prevents the collector voltage of Q9 from ever going below about 1Vbe. In other words, it prevents Q9 from going into "hard saturation." Hard saturation causes slow recovery time - so the purpose of Q14 is to keep the circuit recovery time fast whenever the output stage has approached the negative rail. At least that's my guess.
What I don't understand is that the modern, real audio opamps: LME49990, OPA1611, and OPA211 were omitted! They all are superior in performance compared to those that have been presented.
With properly designed LME49990 based circuitry it is actually possible to do 24-bit quality analog work.
That's not quite accurate. You need to multiply the noise spectrum by the frequency response that it's exposed to, and then rms it up (square it and integrate over the bandwidth). If the response is a single-pole low-pass, that 1.57x factor (over sqrt(BW))pops up automatically from the integration. The methods are equivalent if the noise density is flat with frequency, but it often isn't.
To calculate the noise in the audio bandwidth you have to add all the noise components in an RMS fashion then multiply them by a bandwidth factor like 1.57 x sqrt (BW). The factor depends upon the slope of the filtering beyond the 3 dB points. 1.57 in this case is for single pole filtering. There are app notes for this on the ADI and Intersil web sites.
It isn't true that the LM4562 has no "single" version. The LM4562 is actually the dual of the LME49710 and is therefore exactly the same as the LME49720. The LM4562 was released first, and the single version was then developed. By the time the single was released, National had changed their numbering scheme; the original plan was to phase out the LM4562 number and use just LME49720, but the former had gained too much traction in the market place. Interestingly enough, the LM4562 is less expensive, in small quantities at least.
National have some other very impressive op amps in the LME series, most notably the LME49713, which is a current-feedback op amp with similarly ultra-low distortion, ultra-wide bandwidth, ultra-high slew rate, lower noise, and higher output current capability (at least ±93 mA)
Other op-amps worth mentioning are the new AD8597 (single) & AD8599 (dual) from TI. These have been released since Doug's book was published and are recommended by TI over the AD797; which is very nice of TI given that the AD797 is considerably more expensive!