1. Register NOW and become part of this fantastic knowledge base forum! This message will go away once you have registered.

Intermodulation with two coloratur sopranos

Discussion in 'Vocals' started by bogdank, Apr 26, 2005.

  1. bogdank

    bogdank Guest

    I'm new to the recording art. I've been recording a coloratur soprano that doesn't use much vibrato. I recorded two voices for a duet with the same soprano. At times she would get her voice to be more resonant (a characteristic of coloratur sopranos I guess). When that happens within a voice that is part of a duet I would get weird intermodulation with the other voice. It would sound almost like the third (deeper) voice would appear like artifact in the background.

    I analyzed the recording and found the following:

    a) if I pan two voices far left/right, the effect goes away

    b) if I mix them mono or pan them less than 100% L/R the problem occurs

    c) the problem can be eliminated through DSP in a following way: you take the second harmonic of the lower voice and reduce it by about 18dB which reduces the intermodulation of that harmonic and the higher voice's base frequency component

    d) the solution in c) is done with a notch or highQ parametric filter, but you have to do it manually at all places where intermodulation might happen, so it is kind of time consuming

    Is there a recording technique that could treat this problem acoustically? Is there a way to record coloratur soprano so that their second harmonic ends up lower than usual?

    When this characteristic resonance happens the base frequency component and the second harmonic are of the same energy (equal in power). Normally, the second harmonic and higher harmonics would decay in energy. The problem here is that the resonance happens in their head (within sinus cavity or somewhere) so you can't fix it in obvious ways I guess. As long as they sing solo, it is fine, great, very beautiful. But, when you record a duet that can kick resonance like this it sounds weird and almost like a distortion.

    Any suggestions?

  2. David French

    David French Well-Known Member

    Sounds like the psychoacoustic effect of the difference tone to me. This is where if you play two pure tones at X hz and Y hz, you will sometimes perceive a tone at X-Y Hz. You could verify that this is what's going on by finding the frequencies that are sung and the one that is perceived. Not sure how to get rid of this; I've never encountered in in recording vocal harmonies, but I haven't recorded coloratura sopranos either. Maybe someone else has some more insight.
  3. bogdank

    bogdank Guest

    You're right. The difference between two base frequencies is about 124Hz so it appears like a low pitched tone. E5 (658.8Hz) and G5 (782.9Hz). Interestingly, when the second harmonic of E5 is reduced by 18dB or so, the effect goes away!

    Since this really seems to be a psychoacoustic effect, it probably cannot be avoided during recording. In the piece I was mixing the effect is later masked by the orchestra, so I didn't process it. The processing would remove the effect, but would change the quality of the voice somewhat. Like a catch-22. If I run into this within some other piece that has just vocals and no background, I would probably have to process since it sounds weird...

    Thanks for the reply.

  4. moonbaby

    moonbaby Mmmmmm Well-Known Member

    This may be a dumb question, but have you tried reversing the "phase" of one of the tracks? I ask this because your comment was that when the 2 tracks are panned away from each other, the "intermodulation" appears to disappear...
  5. David French

    David French Well-Known Member

    Thought so. When you say you reduced the 2dn harmonic, I suppose you mean that you performed a narrow band cut at 1,318 Hz? This of course won't always be the second harmonic, and as you said, thai would be unacceptable in terms of sound quality. I'd like to venture a guess that phase won't help because perception of steady tones is immune to phase unless you change the phase while listening.
  6. bogdank

    bogdank Guest

    Correct. I cut it by 18dB to completely remove the effect. It could be cut less (9-12dB) for a considerable improvement.

    Correct. Although, when two voices sing, it might be OK to process one of them somewhat without major degradation of overall quality. Again, psychoacoustics (masking) may help there too. :)

    Correct again. Changing phase on any of the channels did not help. Well, flanger or chorus might help, but I don't think classical music would sound appropriate with such effects. :)

Share This Page