US Department of Defense Design Criteria Standard: Human Engineering, better known as
MIL-STD 1472, has a pretty complete section (5.3) on audio signals. A quick read yields the following standards that seem to bear directly on your question (edited here):
188.8.131.52 Signal meaning. Each audio signal shall have only one meaning.
184.108.40.206 Apparent urgency. The attention gaining characteristics of the signals in a set (e.g., rapidity of pulse pattern, frequency,
intensity) should match the relative priority of the signal.
220.127.116.11.1 Range. The frequency range shall be between 200 and 5,000 Hz and, if possible, between 500 and 3,000 Hz.
18.104.22.168.1 Attention and avoidance of startle reaction. To minimize startle reactions, the increase in sound level during any 0.5 sec
period should be not greater than 30 dB. In addition, the first 0.2
sec of a signal should not be presented at maximum intensity, use
square topped waveforms, or present abruptly rising waveforms
22.214.171.124.1 Use of different characteristics. When several different audio signals are to be used to alert an operator to different types
of conditions, discriminable differences in intensity, pitch, beats
and harmonics, or temporal patterns shall be provided. If absolute
discrimination is required, the number of signals to be identified
shall not exceed four. Signal intensity shall not be used alone as a
means of discriminating between signals.
It discusses in detail characteristics of warning signals but keep in mind the military defines “warning” as, “you need to respond now or you’re going to die.” Not everything will apply to your case, I don't think.
It also provides guidance on whether you should have audio signals at all and what controls to provide for users.
The latest Windows UX Interaction Guidelines also provides guidelines on the use of sound (p686 in the Win 7 PDF version). They emphasize that all sounds provide only supplemental information -that users should never need to rely on sound alone for feedback. Thus, all sounds, including those for errors and warnings, should be brief, subtle, and calm, if they're present at all.
Use sound only when there is a clear advantage. When in doubt, don’t
use sound... The ideal sound effect is one that users barely notice,
but they would miss if it were absent.
As for the quality of each sound, that's more art than science. The Windows guidelines recommend:
Light, pure tones, and glassy and airy sounds, with a soft fade-in and
fade-out (soft “edges”) to prevent abrupt, jarring, percussive
effects. They are designed to feel subtle, gentle, and consonant.
Windows sounds use echo, reverb, and equalization to attain a natural,
Simple beeps and buzzes of yesterday's computers are right out. You should exploit all the advanced sound reproduction capabilities found in modern systems. This means you need to create your sounds like you're creating a one or two note piece of music. If fact, maybe you should hire musician skilled with digital synthesizers, describe the information and effect you want to convey, and let him/her create some candidate sounds for you.