Only the peak levels (loudest points) of your examples should necessarily be above 0.5 (see the image below).  Ideally these peaks would be as close to 0.0 as possible without clipping.  There will, however, be a lot of variation from low to high (in an ideal recording environment spaces between words or syllables would be absolutely silent).

The most important thing, in the end, is the sound qualilty (good old subjectivity!) but you want to be sure to be getting a nice "hot" signal in order to reduce the signal-to noise ratio.

If you find you are speaking abnormally loud (or shouting) or having to speak artificially close to the microphone, you need to adjust the TRIM (pre-amp) level on the MOTU 828.  Boost it until you hear the background noise in headphones and then back it off slightly, just until the point you hear that it has mostly disappeared.  Too much background is worse than the noise created by subtle quantization errors.  You might also try recording in a part of the studio which is away from the fan and computer noise (some students have discovered quieter areas at the rear of the studios near and the fabric-covered/noise-cancelling walls).

One thing we have working against us is the use of a dynamic microphone, designed to work onstage with very loud signals like a singer or a trumpet player.  This is not the ideal microphone for spoken vocal signals but it has other advantages for us: 1) not being as sensative as many mics it is, therefore, not as "touchy" to learn on, 2) having a small field pattern, it is well suited to recording a signal in a room with other background noise such as the fans or computers in our studios.

== Digital Music Faculty