allyn@milton.u.washington.edu (Allyn Weaks) (04/14/91)
gaspar@urz.unibas.ch writes: >There is in fact a relationship between vowels and overtones. If you sing >the vowel A and the vowel E at the same pitch it's the overtones that are >forming the different sound of the vowels. Nope. Vowels are recognizable because of formant frequencies caused by the shape of the mouth and vocal tract, and are largely independent of pitch. The formant spectrum is basically an envelope that encloses the pitch spectrum. Pitch is determined primarily by tension in the vocal chords. You can test it yourself by singing 'ah' (or any other vowel) at various pitches. The pitch overtones (and timbre) go with the pitch, but the 'ah' sound stays largely the same - the formant frequency (and _it's_ overtones, which aren't simple harmonics) stay the same for a given vowel. The 'ah' as in father has strong peaks at about 700, 1100, and 2600 Hz. 'oo' as in pool has its strongest peaks at about 300, 700, and 2500 Hz. All of this is somewhat oversimplified of course. I suggest looking up Arthur Benade's _Fundamentals of Musical Acoustics_ for more details. This is the reason that sampled voice sounds on a keyboard are so unsatisfactory - they only sound like a voice at the pitch they were sampled at and if you get more than a whole tone away it all goes to hell, since the sampler is shifting the formant envelope frequencies as well as the pitch. >That is also the way you distinguish different instruments, by the overtones. Only partly. Much more important is the attack - the wild ravings of the first several milliseconds of a note. If you hear only a sustained tone, it's often hard to tell a violin from a flute from a horn. Pianos are more distinctive because of non-linearities in the overtones. Allyn Weaks allyn@milton.u.washington.edu
sandell@aristotle.ils.nwu.edu (Greg Sandell) (04/14/91)
allyn@milton.u.washington.edu writes: > >That is also the way you distinguish different instruments, by the overtones. > > Only partly. Much more important is the attack - the wild ravings of the > first several milliseconds of a note. If you hear only a sustained tone, it's > often hard to tell a violin from a flute from a horn. Pianos are more > distinctive because of non-linearities in the overtones. The interesting thing is that in years of experiments spectrum always shows up to be the most prominent dimension, and onset comes in second place. If you look at the semantic terms listeners use to describe timbre, the most frequently used ones pertain more to spectrum (dark, bright, rich, brilliant, mellow, warm, nasal, dull). The ones pertaining to attack (biting, incisive, soft, hard) are less frequent in use. Outside of the laboratory, timbre is apprehended over time, over exposure to a number of notes. Certainly part of the apprehending process is hearing a number of different onsets over several different pitches. The attack's duration, noisiness, presence of inharmonicity changes a bit from one note to the next, but not that much. Not as dramatically as does the spectral envelope from one pitch to the next, which in fact show rather beautiful patterns of change across an instrument's playing range. And spectrum changes in interesting ways according to the force with which the note is played; the changes is onset caused by dynamic are salient, but not as rich in patterned information. Most of the seminal studies that revealed the priority of the attack portion were from around 1963, and were based on isolated notes. We could make more sophisticated experiments now. Here's my question: if you took a clarinet melody of substantial length and melodic span, and replaced all the attacks with trumpet attacks, would anybody think they were hearing a trumpet tune? Roger Kendall has a study (MUSIC PERCEPTION 4/2) where he played clarinet, violin and trumpet melodies in conditions with the attacks excised, just steady states. Listeners could tell what instruments they were hearing. They showed poorer performance in conditions where only the attack portions were presented. Oddly enough, when listeners had to identify instruments from a singly presented note, they did equally well with attack-only and steady-state only. So the priority given to the attack from those old 1963 studies needs to be re-evaluated. Greg Sandell -- Greg Sandell sandell@ils.nwu.edu
galetti@uservx.afwl.af.mil (04/15/91)
In article <1991Apr13.235927.5503@milton.u.washington.edu>, allyn@milton.u.washington.edu (Allyn Weaks) writes: > >>That is also the way you distinguish different instruments, by the overtones. > > Only partly. Much more important is the attack - the wild ravings of the > first several milliseconds of a note. If you hear only a sustained tone, it's > often hard to tell a violin from a flute from a horn. Pianos are more > distinctive because of non-linearities in the overtones. > I have to agree with you here. I see this a lot when I edit a sample. Without a descriptive attack the sample could be a simple analog synth. Vocal samples sound really stupid without some form of complex attack. For example, if you sample someone singing "ah" and then edit it so the attack is gone, it sounds pretty boring. If you sample "Tah" and preserve the attack, it sounds much better. The Roland D-50 uses this concept. It takes advantage of the primacy effect that attack sounds have on the human ear. When a distinctive attack is followed by a simple oscillation, the result is a realistic and pleasant sound. Well, I like it anyway! > Allyn Weaks > allyn@milton.u.washington.edu ___________________________________________________________________________ / Ralph Galetti Internet: galetti@uservx.afwl.af.mil \ | PL/LITT Interests: computers, music, computers | | Kirtland AFB, NM 87117-6008 and music, golf, sleep. | \__"No, they couldn't actually prove that it was HIS vomit" - Nigel Tufnel__/