2
$\begingroup$

I'm experimenting with applying unusual 'spectral mangling' type special effects to audio.

1) What DSP techniques can be used to change the timbre of a complex tone, such as that produced by the voice, or a musical instrument?

  • Are there any ways to manipulate the level of specific frequencies in the spectral profile of a complex tone. E.g. Assuming a tone with a fundamental frequency of 440 hz, use some DSP method to augment, for example the 2nd harmonic overtone at 1320Hz Khz, or remove another peak at say, 3Khz).
  • What's the best way to remove specific frequencies, or as narrow bands of frequencies as possible using DSP?

I use mixing/mastering, offline, and real time DSP VSTs, standalone programs, outboard effects units, but am interested in any methods which could work.

2) What alternatives (if any), are there to using techniques which introduce significant ringing or other artifacts which degrade the sound?

3) Are there any ways to 'clean up' audio after processing, removing artefacts which have been introduced?

| improve this question | |
$\endgroup$
  • $\begingroup$ To change the harmonics of a tone, you first have to identify what the tone is. You're going to have to find the frequency of the tone before you can do anything to it. Depending on what you expect and what your environment is, this may be difficult or not. Also, this is a pretty open-ended ("overly broad") question. There are an infinite number of things you could do to a tone. Do you have something specific you want to do? $\endgroup$ – endolith Apr 27 '12 at 17:32
  • $\begingroup$ I mean assuming a given tone, and specifically, cutting or boosting the specific harmonic overtones which make up the timbre of the tone. I have edited the question to better reflect that. $\endgroup$ – Dale Newton Apr 27 '12 at 18:05
  • $\begingroup$ You'd also have to identify when an individual tone starts and stops, because you're applying different filters to different tones depending on their fundamental frequency. This isn't polyphonic, is it? :) $\endgroup$ – endolith Apr 27 '12 at 20:07
  • $\begingroup$ Yes exactly. And yes, to make things more complicated, much of the time its polyphonic. Eventually Id like to control the effect with a midi-controller/keyboard to make selecting the fundamental frequency easy. $\endgroup$ – Dale Newton May 1 '12 at 3:25
  • $\begingroup$ Modifying individual tones in a polyphonic mix is way hard, since the harmonics are interleaved or overlap with those of other notes. :/ Though if you know what the notes are because they're being played on an instrument that you can see, then it's a lot easier. $\endgroup$ – endolith May 1 '12 at 5:47
2
$\begingroup$

Phase vocoders with custom resynthesizers can be used for this, as in time pitch stretching code that attempts to preserve the spectral formant envelope. But phase vocoder analysis/resynthesis is not artifact-free.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Would you mind listing any of the artifacts caused by Phase vocoders? I imagine the heterodyning would incur some level of specktral leakage at least, as it uses a low pas filter which must have a cut off slope? $\endgroup$ – Dale Newton May 1 '12 at 2:49
2
$\begingroup$

Looks like a regular equalizer should do the trick. In your particular example you could simply boost 1320Hz and cut 3kHz with parametric EQ. The "Q" control of the parametric determines the frequency selectivity and will also affect any ringing. However, unless the Q is ridiculously high, there will be no significant ringing at all. There are many more DSP methods do achieve this but a parametric EQ is simple and I'm sure their VST plugins for that already. No need to re-invent the wheel.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks Hilmar. THe problem I have with the parametric EQs is the limitation of the cutoff slope. They generally cant cut cut and boost individual frequencies like this as surgcially as Id like. The freeform ones, which are prceise enough, suffer from ringing when the cutoff slopes are steep. What are the other DSP methods you mention which could be used to achieve this? $\endgroup$ – Dale Newton May 1 '12 at 3:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.