357MagnumNinja
Shade7510
also under what conditions would one need to get a decent/very nice sound card?
Seriously? Audio production. That's about the only reason I can think of. Unless you have a game or music that's recorded in 24-bit, 96kHz... Which is still not too common these days. Or if you want special features like SPDIF-in. ...that's not exactly standard on onboard sound.
Onboard Azalia sound systems tend to have most of these features, so the only reason to get a discrete internal sound card is to get S/PDIF-In.
For audio production, an internal sound card is inadequate - there's too much electrical noise in computers. Instead, you want an
external card for that. Turtle Beach makes some nice stuff. USB is also inadequate for this, because USB latencies are too high.
But a USB device generally eliminates the noise issue.
357MagnumNinja
the limit of human hearing is around 24khz, and you need 48khz to accurately record a 24khz sample. So, typically, 16-bit, 48khz is good enough. the 24-bit MIGHT make a bit of difference... but I can't speak on that aspect, seeing as I've never had a 24bit sound card...
Human hearing stops at 20kHz. The human ear is incapable of processing sounds any higher than that.
44.1kHz was chosen because it accurately reproduces 22.05kHz sounds, and it's a nice round composite number (it's the square of the multiple of sequential primes). The Nyquist-Shannon theorem requires that the sampling frequency be greater than double the frequency you wish to sample. The problem with 44.1kHz is that it requires filtering out certain high frequencies due to aliasing. Higher sample rates reduce the amount of aliasing. That's why 96kHz is preferred for high-end audio systems.
16-bit audio is well more than adequate, due to the simple fact that the human ear is simply not sensitive to small differences in amplitude. You can't hear the difference between 16-bit and 24-bit audio. The
only reason to use 24-bit is for mixing. When you're mixing multiple audio samples, the math on 16-bit numbers results in a lot of data loss. Working with larger sample ranges provides more dynamic range that won't be lost in mixing. The ideal mixing mode is 32-bit or higher, and 24-bit samples are a good start.