Wednesday, November 26, 2008

Making Sense of Formant Tracking Charts part 1: The Tenor issue and a shoutout to Italian Tenor

Following the commentary from George and Baritonobasso I will work on changing the color scheme and deal with the phonetic/phonemic issue. With this post I wanted to kill three birds with one rock (stone. I used to say rock for years). 1) I wanted to start discussing the formant tracking issues now that we have the charts available 2) I wanted to address the C4 to C5 octave relative to some private talks I've had with Martin Berggren 3) I wanted to bring support to a concept that Italian Tenor (a frequent contributor to NFCS) has brought up often, "the deep" [u].

1) A. The first thing to remember about any vowel chart is that it is created with the purposes of the author in mind. Despite a rather broad international experience, my formative musical years took place in the United States. My chart therefore concentrates on the four operatic languages most sung in the U.S. In fact a different chart should be made relative to every national school. Italian and Spanish have 7 vowels-sounds, English 12 German 13, French 14. For Scandanavian Slavic languages and Asian languages, there are vowel qualities that are not represented here. I would welcome charts following our principles here that reflect a more accurate scheme for formant tracking (vowel modification)for those specific languages. Nevertheless the 16 vowels represented on my charts form a full enough spectrum to satisfy most linguistic situations.

B. A vowel chart is a beginning not an end. It should be understood that vowel modification is worthless if the phonation mode is not efficient. Achieving an inertial state in the vocal tract will only have an effect on a lean, muscularly balanced phonation mode. In the best case scenario, a resonant vowel adjustment will yield sensations that the singer will come to recognize and reproduce easily as long as the instrument is healthy. Once the singer has a clear proprioceptive experience of 1st and second formant resonance, then s/he will be able to track formants more spontaneously and instinctively.

2) A. The register shift for the tenor voice is probably the most important issue. Proper execution of this change reflects positively on several aspects of the instrument, including correct muscular balance of the voice, efficient phonation, appropriate laryngeal depth and of course appropriate tuning of the vocal tract (formant tracking). It is important to know that all of these conditions must exist simultaneously in order for a smooth and comfortable register change to occur.

B. The misconception about "pure" vowels. We only need to hear the average soprano sing in her upper range to know that vowels cannot be sung as they are spoken. This is true throughout most of the range. Depending on the circumstance, certain vowel qualities are not possible. In effect, every note has a different resonance necessity and therefore a different formant balance (vocal tract partition/shape). Furthermore there are acoustic regions where more than one vowel intersect. This means that one vowel adjustment will sound like two or even three different vowels depending on context (surrounding consonants, etc). In other words, intelligibility does not mean that someone is singing the spoken form of a vowel.

C. Providing all other conditions are met, the vowel chart can be followed to accomplish the acoustic shift from formant 1 to formant 2 between C4 and C5. The acoustic shift should occur around F4#. Contrary to popular belief, this change does not depend on voice type. Second formant tuning is more appropriate from F4# on, however, a lighter tenor will feel less stress singing a first-formant dominant F4# then a heavier voice. Regardless, the larynx will rise for an F4# sung in first formant dominance. The distinction is that the muscular shift happens earlier for a dramatic tenor than a leggiero, but the acoustic shift happens in the same place. This can get confusing. It suffices to say that the resonance adjustment needs to happen around F4# for any tenor, in fact sometimes on F natural depending on the vowel modification involved. The leggiero who keeps F1 dominance up until A4b like some do will experience a more difficult shift when they finally do go to F2 tuning.

3)Italian Tenor from NFCS has often spoken about the "deep [u]" relative to the tenor high range. According to the chart, Italian Tenor is correct relative to the two notes that define a tenor's existence. There are more Bbs and B naturals then high Cs written for tenors. Those are the notes that occupy most tenors' minds. It just so happens that the best second formant vowel for Bb and B is [U], and as previously stated, the larynx must be low to achieve F2 tuning. Italian Tenor's "deep [U]" is totally consistent with acoustic expectations.

This video of Carlo Bergonzi singing "O Paradiso" from Meyerbeer's L'Africaine is an interesting example of the deep [U] concept. The first Bb at 1:30 is perfectly tuned. Likewise, the final Bb is tuned to some form of [e] or [ø] very consistent with the tuning of B4b according to the chart. It is important to know that the Italian score (like the French original) has the Bb on an [a] vowel. Bergonzi changed the text setting in the cadenza (traditional practice) to accommodate a more resonant vowel. This means that he was very sensitive to F2 tuning.

By the same token, there is a tendency to make too much of a good thing. Lip rounding as with the deep [U] lowers the first formant and encourages F2 tuning. However, flexibility and subtlety is important. Bergonzi uses the same [U] on the first Gb at 0:49 which sounds comparatively slightly tense until he gently modifies to someting closer to [u]. The same Gb sung on an [u] vowel at 2:07 is perfect. The [u] vowel is far enough from the suggested [Λ], however some form of [u] lies precisely at the formant frequency that would coincide with the 5th harmonic (H5) 930Hz. Bergonzi has a tendency to round the lips consistently. This is not a bad strategy because in the [a] to [u] vowel spectrum, this strategy works well in the passaggio and above. However, more precise tuning is important on certain notes. My studying of Bergonzi's beautiful voice reveals some difficulties where the rounding lips actually takes him away from ideal resonance. However, on the whole the approach is logical and practical relative to F2 tuning.

Additionally this should show that the chart is a point of departure and not the final word on formant tracking. In the case of Bergonzi, he found a vowel sound between [U] and [u] that coincided with H5 of F4# (Gb in the score). The lesson here is that a vowel spectrum includes many formant frequency pairings for which there is no standard vowel name. Live singing goes beyond the limits of vowel definitions. However, the concept behind the chart can certainly explain something that may appear to be anomalous.

Happy Thanksgiving to all!

© 11/26/2008


chessoperaspirit said...

Very interesting stuff! A couple questions:
You make the distinction between the acoustical shift that happens at F# regardless of voice type and the muscular shift that changes. I've always thought of the mas very connected. How does that work for a lower voice type that has to make a muscular shift before its time for F2 dominance? Do they switch registers while still tuning to the first formant? Can you give an example of what that would sound like?
Also, is that acoustical shift different if you have to produce a different vowel? If you had to produce an /i/ on a note slightly lower than F#, would you switch to F2 dominance for the sake of intelligibility and clarity without losing resonance?

Thanks TS. I'm just trying to understand. Thanks so much for taking the time to share all this. I look forward to your response.

Martin Berggren said...

Dear TS,

Enjoying the posts, as usual!

I have some feedback for discussion about the very interesting Bergonzi example! To me, his Bb4 at 1:30 sounds rather light in comparison to the usual heroes of the acuti. I cannot hear the whistling-like character I usually hear above the passagio in my favorites. I would say that the larynx is slightly high, no? On the other hand, the vowel is unusually pure, more so than in the my recordings of the aria with other artists. I made a spectrum of that note, which confirmed what I hear, I think. The dominating partial is H2, but, unusually enough, also H1 (F0) is surprisingly strong (-5dB from H2). The singer's formant frequencies are about 10 dB below H2. (I checked a few other places in this aria, and it seems like Bergonzi does not have a particularly strong singer's formant). I believe that the second formant of his quite clean /U/ is used to amplify H2 here.

Then I compared with Björling (a recording from the 30's). For him, the singer's formant frequencies dominate strongly. Next comes H3 (almost 10 dB below the singer's formant). H2 is quite weak (20 dB below the singer's formant). The vowel, at least after a second or so, does not sound much like an /U/ however. It sounds like he modifies the vowel against some sort of /a/ to raise the second harmonic in order to better match H3.

del Monaco's strategy is very close to Jussi's here: a strong singer's formant followed by H3, with a similar vowel modification.

Corelli (in a Youtube clip from Tokyo 1971) is interesting, and much different from the above! His extremely powerful Bb4 cuts like a razor and sounds much more like an /U/ than Björling's or MDM's, although not as pure as Bergonzi's (but the sound is much more powerful). The high-pitched whistling sound I have learned to love is very apparent here. Now, how about the spectrum? It is quite telling! There is almost a perfectly even balance between H2 (H3 is weak!) and the singer's formant, which happens to be positioned around F7, which gives a perfect-5th relation that I believe is the source of the particular whistling-like timbre of his voice at this note.



Martin Berggren said...

A correction to my previous post. About Björling, I meant to say

It sounds like he modifies the vowel against some sort of /a/ to raise the second formant in order to better match H3.

and nothing else.



Toreadorssong's Vocal Technique Blog said...

Dear Martin,

Wonderful observations. I am reading an article currently that might help me figure out the wave length of the inertial zones between harmonics. Most tenors sing [a] on B4b when that is the written vowel. I wonder however if it is a real [a] or that it sounds like one because it is what we expect. It would be interesting to see your spetrograms. Do the first and second formants of [a] (cir. 800 and 1100 Hz bear out in the Bjoerling clip?

It is my belief that Bergonzi loses the singer's formant because the phonation mode is too thick. By that I mean that his voice always sounded to me like a lighter voice that is slightly thickened. Hence the limits of the extreme upper range. That thickness could have a bearing on laryngeal position. I believe that the constant rounding is to make up for a larynx that tends to ride a little higher than he would like. This is very subtle, but remarkable when we listen to people like Fleta, Lauri-Volpi and Gigli. Amazing efficiency of phonation. Bjoerling belongs in that category as well.

Martin Berggren said...

First, I have sent you the spectra and the sound clips in a private mail.

I am reading an article currently that might help me figure out the wave length of the inertial zones between harmonics.

What article is that? (I am not challenging you; I am just curious!)

My understanding is that the amount of supraglottal inertia for higher frequencies can vary substantially depending on the singer's technique. That is, two singers may phonate the same vowel, in the sense that the locations for first and second formants are very similar, but the supraglottal inertia of the singers can be quite different. Therefore, it might be difficult to say something definitive about inertial regions solely on the basis of the vowel.

Titze makes some quantitative comparisons in his JASA article from May this year. Judging from Titze's results, it seems to me that a long, narrow epilarynx tube together with a wide pharynx, the same mechanism that is behind the formation of the singer's formant, also serves the purpose of generally increasing supraglottal inertia for higher frequencies. That effect seems to me similar to the role of a mouthpiece in a brass instrument. The mouthpiece does not much change the locations of the resonances (the "formants") but it raises the inertia for higher frequencies, making the instrument easier to play and the sound brighter.


Do the first and second formants of [a] (cir. 800 and 1100 Hz bear out in the Bjoerling clip?

Please check from the sound clip I've send you what really to call that vowel! I find vowel naming difficult across languages... Bjorling does something here I have heard him do before. He starts out with one vowel quality and modifies it significantly during the duration of the note, to gain resonance, I believe. Towards the end, where I took the spectrum, the vowel is very clearly not a [U] anyway...