Tako's vocal clip..... >At the beginning and end of the clip, I am singing in "chest". In the >middle of the clip, I am singing in "head". Notice that when I am singing >in "chest" in the 200 Hz range, the formant range (for me, 2800-4000Hz) is >heavily populated by the 13-20th partials.
Dear Tako, Sorry it's taken me so long to get around to downloading your recording and display.
I must say I'm both impressed and mystified by this spectrum presentation. I'm one of those tenors who has always maintained that I don't have a falsetto. As well as that, in my teen years my voice never 'broke' but only slightly lowered and filled out, so there's no way I can sing that final top F# if that's what it is. I'm impressed because it seems to me that you're going fairly smoothly from chest to head then falsetto and back. Do you agree with that? The mystery surrounds the richness of the harmonics. The only way I can explain such a dense display is by suggesting that there is a serious over-load somewhere in the signal path between the microphone and the spectrum analyser. I don't wish to throw cold water on your equipment (electronic I mean :) but I'm sure that after all our talk about the spectrum of a square wave, you will understand that if an overload occurs at any step in the signal path, the clipping of the signal, (the flattening of its peaks) will cause harmonics such as this to be generated and displayed accordingly. As Lloyd points out, most microphones are not built to handle such signals which is why these frequencies have to be judiciously excised. Does a trained listener hear the harmonic density the same live as in that recording? The C# to F# I mean. Make certain also that you don't have an A/D convertor that's exceeding its peak. 111111111111111111111 means a flat top.
More on the noise from the folds. If you've been able to follow the idea of harmonic analysis of square waves, it may have occurred to you that as the pulse width of the open time of the folds narrows to produce the highly sought after formant frequency, it will also produce still higher components of the frequency that its currently mimicing. This could even be lower than a formant frequency which is handy for lower voices. However derivation of these will depend on just how square the top of the physical excursion is; it may be round! As it's likely to be imperfect just as can occur in the closed time of the folds, the lack of co-ordination of these signals means that they fall into the definition of noise. If it's not needed, it's extraneous and degrades the " Signal to Noise Ratio." So the exclusion of noise from the signal by means of filtering, is a thoroughly acceptable practice. By the way, I've been quite intrigued how the 6th, 7th and 13th harmonics so neatly fit with the 6:1 ratio of that little gizmo. Just can't make it work ;) Maybe it's a pass filter for these frequencies and a noise stop? It would certainly appear to have the potential to frustrate the fundamental by demanding that all tone be delivered as groups of formant frequencies to be integrated in the vocal tract. Amazing. Surely not, why such an elaborate set of folds if we could have communicated with a series of honks? Doh!
No I'll stick with it being a formant frequency matching transformer for a while I think. That means it can have an efficiency role, a high frequency integration role and a noise stop as well as being comparatively transparent to lower frequencies.
Oh by the way, don't overlook the noise from the amplifier, it tends to be missed with speakers but it's always heard by a spectrum analyser and with headphones. That's why SNR readings usually include a roll-off filter, so we don't read signals we can't hear and thereby make the gear look worse than it really is; hmmm;) Besides, they open up the market for the commercial opportunists.
Regards Reg.
|