Nasal sounds are produced by momentarily closing off the mouth, thus redirecting sound transmission up through the nose. This alternate transmission path introduces new spectral peaks and, most characteristically, a pronounced anti-formant; that is, a spectral trough. For more background on nasals, see James Kirby's PowerPoint, Nasals, approximants, rhotics, and laterals. See also Robert Mannell's page Nasal Stops, which includes spectrograms showing how nasal consonants transition into vowels.
Anti-formants can be synthesized by notch filtering with the FilterBandReject2 unit. Issues surrounding the use of digital filters to produce nasal sounds are explored in Taming the Notch Filter and Emulating Nasal Resonances and Anti-Resonances. Insights gained during these explorations have been incorporated into Instrument #122: Nose1, which is detailed in Figure 4.
Instrument #122 works in cascade with the previously described Instrument #101: Buzz1 and Instrument #121: Mouth1. Figure 5 abstracts out the components of Instrument #122 to illustrate the mechanism used to interpose nasal resonances into ongoing output from Instrument #121. An EnvSust unit (#6) generates a sustained envelope which ramps the nasal sound in when the note begins and ramps the nasal sound back out when the note concludes. By subtracting Unit #6's output from unity, an inverse envelope is obtained which ramps Instrument #122's unprocessed input out as the processed output ramps in, then ramps the unprocessed input back in at the conclusion.
Notice that Unit #1 of Instrument #122 introduces the spectral peak designated F0 in Table 4 below. This fixed peak aside, Instrument #121 and Instrument #122 taken together require eight values supplied to identify a particular nasal sound:
Wikipedia identifies sixteen nasal consonants used in languages worldwide, but English employs only the three consonants listed in Table 4. Other nasals are not included simply because I was unable to obtain synthesis data for them.
|IPA||Description||Place of Articulation||Examples||F0||F1||F2||F3||NF||NB||Gain|
||alveolar||between the tip of the tongue and the alveolar
ridge (the gum line behind the front teeth)
|no knit or sudden||270||450||1340||2470||1200||75||33|
||velar||between the back of the tongue and the velum
Listing 6 presents the fourth-iteration synthesis of “Daisy Bell”. New indications for nasal sounds are color-coded in brown. Note ID's have been resequenced since Listing 4 owing to the need to activate Instrument #122: Nose1, the nasal extender. These new notes are #18 (me), #27 (an-swer), #37 (I'm), #51 (and), #84 (won't), #103 (mar-riage), #112 (can't), and #148 (u-pon).
Each note invoking of Instrument #122: Nose1 is accompanied by new
ramp statements controlling
contours #1, #3, #4, and #5.
The heart of the sound is a steady-state segment during which nasal resonances and anti-resonances speak fully.
When the nasal sound initiates a word (for example “me” at
note #18), the attack duration
for Nose1 (note parameter #7) is zeroed out and the formant ramps dispense with an onglide.
Likewise when the nasal sound concludes a word (for example “I'm” at
note #37), the release
duration (note parameter #8) is zeroed out and the offglide is skipped.
There are no instances in Listing 6 where a
nasal occurs in the midst of a word; however, we will get to that with words like “and” (
when plosives come into play during Iteration #6. For now you should just take
it on faith that for internal nasals both the attack and release durations for Instrument #122: Nose1 will
be around 100 msec., that the formant contours will contain both onglides and offglides, and that the onglide and offglide
durations will be consistent with these attack and release durations.
Next topic: Fricatives
|© Charles Ames||Page created: 2014-02-20||Last updated: 2017-06-12|