X

Chapter 6 Filter eects 12 9
Freeze the input signal
m Click the Freeze button to hold, or sustain, the sound spectrum of the analysis input signal.
By freezing the input signal you can capture a particular characteristic of the signal, which is then
imposed as a complex sustained lter shape on the Synthesis section. Here are some examples
of when this could be useful:
•
If you are using a spoken word pattern as a source, the Freeze button could capture the attack
or tail phase of an individual word within the pattern—the vowel a, for example.
•
People cannot sustain sung notes indenitely. To compensate for this human limitation,
use the Freeze button. If the synthesis signal needs to be sustained but the analysis source
signal—a vocal part—is not sustained, use the Freeze button to lock the current formant
levels of a sung note, even during gaps in the vocal part, between words in a vocal phrase.
Tip: The Freeze parameter can be automated, which may be useful in this situation.
Set the number of lter bank bands
m Vertically drag the Bands eld to set the number of frequency bands the
EVOC 20 TrackOscillator’s lter bank uses.
The greater the number of frequency bands, the more precisely the sound can be reshaped. As
the number of bands is reduced, the source signal’s frequency range is divided up into fewer
bands, and the resulting sound will be formed with less precision by the synthesis engine. You
may nd that a good compromise between sonic precision—allowing incoming signals such as
speech and vocals to remain intelligible—and resource usage is around 10 to 15 bands.
Tip: To ensure the best possible pitch tracking, it is essential to use a mono signal with no
overlapping pitches. Ideally, the signal should be unprocessed and free of background noises.
Using a signal processed with even a slight amount of reverb, for example, can produce unusual
results. Processing a signal with no audible pitch, such as drum loop, also delivers unusual results,
but the resulting artifacts might be perfect for your project.
EVOC 20 TrackOscillator U/V detection parameters
Human speech consists of a series of voiced sounds—tonal sounds or formants—and unvoiced
sounds. The main distinction between voiced and unvoiced sounds is that voiced sounds
are produced by an oscillation of the vocal cords, whereas unvoiced sounds are produced by
blocking and restricting the air ow with lips, tongue, palate, throat, and larynx.
If speech containing voiced and unvoiced sounds is used as a vocoder’s analysis signal but the
synthesis engine doesn’t dierentiate between voiced and unvoiced sounds, the result sounds
rather weak. To avoid this problem, the synthesis section of the vocoder must produce dierent
sounds for the voiced and unvoiced parts of the signal.
EVOC 20 TrackOscillator includes an Unvoiced/Voiced detector. This unit detects the unvoiced
portions of the sound in the analysis signal and then substitutes the corresponding portions
in the synthesis signal with noise, with a mixture of noise and synthesizer signal, or with the
original signal. If the U/V Detector detects voiced parts, it passes this information to the synthesis
section, which uses the normal synthesis signal for these portions.