Recording for simplicity vs Recording for a mix

I have been wondering recently what is the right way to record _______ (instrument X). I came to the conclusion that one needs to take a different approach to recording instruments depending on if that instrument is designed to stand by itself on the final product (for example, a solo piano piece) or if the instrument is fitting into a mix of different types of sounds (for example, piano on a pop track).

1) Levels
This thread:
http://www.gearslutz.com/board/so-much-gear-so-little-time/420334-reason-most-itb-mixes-don-t-sound-good-analog-mixes.html
Dismantles the theory that one should track as hot as possible ... you only need to track to around RMS = -18dBFS and all the parts will sum to a good level on the master buss without having to attenuate it.

This makes perfect sense to me - for a busy mix.

For a simple recording, say a stereo pair on a piano that is to stand by itself, you have the liberty to track a little louder, to the level you wish the recording to appear on the CD.

Normalisation is not good for your audio. From http://www.playgroundstudio.com/blog/?p=85:
2. Never ever “Normalize” your tracks, especially if it’s the entire mix. Not only are you introducing further processing into your mix, but you are inviting quantization errors and digital artifacts. Again, even if we receive mixes that only peak at -10db, 9 times out of 10, it will end up sounding better than mixes that reached 0db or hit the “red” once or twice during mixdown." - KEITH CLEVERSLEY

I would argue changing the gain on a single source after recording it digitally is effectively normalising. So record loud to start with!

2. Transient and frequency response in microphone choice

Rock engineers love dynamic microphones (and colored condensers) .. classical engineers love accurate condenser microphones. Why is this?

Dynamic microphones normally have a colored response which involves a slower response to transient sounds and a non-flat frequency curve. Engineers can use this to their advantage: they choose a mic especially tailored to the frequency area that instrument is to occupy in the mix. For example, the standard for guitar amps is an SM57, which rolls off below 200Hz and above 12kHz - highlighting the exact space you wish the guitar to sit in the mix (midrange) - leaving room for the bass below and the cymbals and vocal sibilance above. There is also a 'presense peak' around 7k that gives the guitars 'edge' and makes them 'stand out'. The slower transient response is known to smooth sound sources that are not designed to serve a percussive element in the mix. Another industry standard is the AKG 541 for drum overheads - which has a boost in the 8kHz - 20kHz, the region that drum cymbals live in.

For classical recording, the mics are often set as a stereo pair (whether it be MS, bluemline, or XY) - and there is very little post mixing involved. These engineers want the most transparent and accurate registration of what is in the concert hall - and they go for accurate, flat-as-a-board mics such as Earthworks, DPA, and Schoeps. The same goes for simple recordings - say, a acoustic guitar and vocal. You do not want the mics to themselves be colored, because you do not need to fit them into a mix - the source is free to occupy the whole frequency range rather than have to fit into a conflicting mix of frequencies.

So, to sum up using the piano example: if you were doing a solo piano piece, you would use accurate mics to capture the detail of the piano as precisely as possible, with all the highs and all the lows. For piano in a pop mix, you would typically EQ out a lot of the body below 250Hz to prevent clashes with the bass, scoop out some upper midrange where the clarity of the vocal sits, enhance some highs around 8-10k for highlight the attack of the hammers, and perhaps heavily compress it for aesthetics (see: Lady Madonna by the Beatles), and to let it sit in snugly behind the drums and vocal.

Recording and instrument to stand alone: Go for accuracy and transparancy
Recording for a mix: Record according to the contribution that instrument is to make to the final mix, so the final mix is as balanced as good-sounding as possible.