I have performed my analysis of Phrase #1 upon which doubts have been cast.
There are three purposes of this post:
1) To show that the
phrase 1 executed in the demo is perfectly valid for both aDANa and aThANa. There should be no confusion in any of the swaras used.
2) To caution forum members that
careful use of the spectral analysis programs is essential, and one should not simply believe any graph and numbers presented.
3)
Reality checks are essential, i.e. if something is obvious in the audio file but doesn't show up in the spectral analysis,
check the FFT analysis and its limitations first, instead of casting doubt on the singing.
My intention is not to waste time on FFT analysis which is useful for other purposes but not for the present thread. The features in the demo are obvious and valid. Any discussion on the validity of the swaras in the demo can be done offline -
[email protected].
Files and Tools used:
Phrase 1 from my demo:
http://rapidshare.de/files/21427502/Phrase1.wav.html
FFT analysis:
There are any number of FFT programs available to analyze audio files. I used the commercially available program Spectrogram 14 downloadable from:
http://www.visualizationsoftware.com/gram.html
FFT Parameters:
The results depend on these parameters. Especially for short files (such as this one), high resolution is critical. I chose 16384 FFT points. To avoid excessive noise, averaging of results over 20 millisec intervals was done. Results without averaging are the same, but have more noise and are harder to analyze reliably.
Calibration:
First I calibrated the swaras for a sruti of 1 kaTTa (C), using the FFT program with a tanpura recording set to various notes:
http://rapidshare.de/files/21538504/Cal ... n.mp3.html
The calibration results are (with an error of +/- 4 Hz):
Swara Freq. (Hz)
S 130
R1 138
R2/G1 146
G2 155
G3 164
M1 175
M2 183
P 197
D1 207 (shuddha dhaivata)
D2/N1 218 (chatussruti dhaivata)
N2 233 (kaishiki nishada)
N3 246 (kakali nishada)
S' 261 (tara shadja)
Analysis of Phrase 1:
The graph shows the Phrase 1 sung in HM aDANa (left) and CM aThANa (right). The FFT has nominal accuracy of +/- 4 Hz and there is additional error of +/-2 Hz or so when swara frequecies are determined by averaging around a small region. The real swaras are represented by the most intense (green) regions of the spectrum. Less intense regions represent FFT noise or almost inaudible swaras that are touched on during singing.
aDANa (left graph)
The initial part is the "p n S' R' ," phrase which ends at tara shadja (S',
Feature #1) at 260 Hz.
The final part is the panchama (p) at about 195 Hz (
Feature #4).
Feature #2: In aDANa, this is a
meend (glide) down from the S' to shuddha dhaivata (D1) touching kaishiki nishada(N2). S/ --> D1. The meend is nicely visible in the left graph, and ends accurately at D1 (207 Hz). There is no kakali nishada.
Blind averaging of the meend can give a false result that looks closer to the nishada. Thus, a reality check is most important - there is obviously no N3 in the audio file at this point.
Feature #3: This is the kaishiki nishada (N2) of the ending phrase "d1 n2 p", measured to be 226 Hz.
Thus all aspects of the aDANa phrase are perfectly clear.
aThANa (right graph)
Feature #1 and
Feature #4 are identical to the left graph and need no explanation.
Feature #2: In aThANa, I illustrated the S --> D oscillation two times, replacing D1 with D2 and singing it more quickly than in aDANa. There are hence two quicker
meends/gamakas , both beginning at S' and ending in chatussruti dhaivata D2 (measured at 220 Hz and 217 Hz respectively. There is no kaishiki nishada nor kakali nishada here.
However, blind averaging of these fast gamakas can give a false result that looks like a nishada. Again, reality checks are essential.
Feature #3: This is again the kaishiki nishada (N2) of the ending phrase "d2 n2 p", measured to be 229 Hz.
Thus all aspects of the aThANa phrase are also perfectly valid within limits of vocal capability.
Of course there are slight "imperfections" which go unnoticed during singing. e.g. the panchama (feature #4) on the left seems to start slightly lower nearing prati madhyama. This imperfection is not seen on the right graph. In superior voice-cultures like BMK and KJY these imperfections would be almost non-existent. But if I could sing like these guys, I wouldn't be here either.
It would be more worthwhile to focus on the discussion at hand, rather than cast doubts on the audio files based on superficial analysis using a mathematical tool that needs to be handled carefully.