Re: Simulating Vibrato With Sine Wave
Re: Simulating Vibrato With Sine Wave
- Subject: Re: Simulating Vibrato With Sine Wave
- From: Brian Willoughby <email@hidden>
- Date: Tue, 2 Feb 2010 04:11:49 -0800
On Feb 1, 2010, at 08:20, Richard Dobson wrote:
On 01/02/2010 15:55, Hank Heijink (Mailinglists) wrote:
On Feb 1, 2010, at 10:24 AM, Richard Dobson wrote:
For vibrato (pitch variation) a plain sine-modulated (and
interpolating) variable delay line will do it very nicely. For
tremolo (amplitude variation) you simply have to set your median
level and mod level, and multiply signal with your slow sine.
Median level +- sine amplitude must be all positive, and< 1.0.
E.g. set sine to range between 0.25 and 0.75, to get a 6dB
variation : vibsig = 0.25 + (0.5 * sin(w)); out = in * vibsig. To
sound like real human vibrato, the frequency of the sine and/or
amplitude should be a little randomised; which is where things get
"interesting".
Randomisation won't do it. Things get interesting indeed - vibrato is
the topic of shelves of research theses, and depending on the
instrument (voice included) can be very tightly controlled and
modulated. I have yet to hear a convincing synthesized vibrato, which
is why I suggested tremolo. Not the same, but a lot easier to
synthesize convincingly.
Well, it will get you quite a way. Sundberg IIRC considers both
slow drift and faster "jitter" as measured aspects of vocal
vibrato, both of which can be reasonably modelleld by summed random
variations of the primary modulation source (probably need 1/f
noise rather than white). I think it is fair to regard these as
aspects of most instrumental vibrato too. All sorts of further
complications arise in modelling note transitions. Overall the
principle is that if you immediately notice it, it is too much. But
if it is not there, on the other hand, you pretty well immediately
know it is synthesised or in some sense artificial. All much easier
to manage of course when it is integrated into the synthesis engine
itself. Adding it externally inevitably brings compromises, most
obviously in respect of timbre, formants, etc. Vocal vibrato (as
violin etc) is almost entirely pitch variation, so for that tremolo
will not get you there at all.
Another important factor to consider is psychoacoustics. Generally,
the human ear+brain system cannot distinguish small variations in
pitch from small variations in amplitude. Thus, if a noticeable
amount of vibrato is too much, then a subtle amount of tremolo should
not be significantly different from a subtle amount of vibrato. This
can work to your advantage, since you have sample code available for
tremolo in CoreAudio.
Brian Willoughby
Sound Consulting
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden