Bb
Rick Denney's discussion of tubas and their sounds is must reading as an introduction to the material below. There, he demonstrates some of the things that can be learned by using simple, inexpensive tools to analyze tubas.
Here, I use the same analysis tool (Cool Edit 2000), but a very much less expensive microphone (Audio Technica ATR-97). Early attempts showed that it is necessary to take great pains to avoid clipping. In the current experiment, recording was done in a roughly 20'x20'x8' room. The room is moderately live, but the tuba was located diagonally across the room with the bell pointing more or less directly into a ceiling of acoustical tile. The microphone was simply placed on a table, at the computer. It was located about 6" from one wall and about 18" away from a windowed wall covered by blinds. The controls on the Santa Cruz soundcard were set to their default values. Preliminary tests indicated that this setup avoided clipping when the tuba was played mp-mf. Inspection of the individual waveforms shows no evidence of clipping (if you want to see them, just ask).
The tuba and the tubist were also inexpensive. I played my YBB-621S, more or less straight out of the case, with no appreciable warmup and no attempt to "tune". Ambient temp was roughly 82degF.
I played a simple Bb Major scale, pausing between notes, and then put the tuba away. Part of my hypothesis is that this is all the recording that's really necessary. We might argue about whether it's better to also play at ff and pp, whether we want a complete chromatic scale, and over what range - but this is an good first approximation. Eight notes.
Sitting down to Cool Edit 2000, I followed a mechanical procedure to isolate each note and produce a spectrum. For each note, I first selected that note out of the scale (and noted which note it was). I then selected 0.25sec of waveform, from the middle of the note (trying to avoid attack and decay, where the faults of the player would be most apparent). This 0.25sec was submitted to Fourier analysis. A screenshot utility allowed me to clip the same piece of CoolEdit's display for each note - and the resulting images were saved as JPEG files: [BBb.jpg, C.jpg, D.jpg..., B.jpg]
Here are the spectra, from Bb down to BBb. Before writing more, I'll open the floor for discussion: what can you deduce from these spectra? Just as important, perhaps: what would you do differently (without markedly increasing the workload)?
You might also think about the question: what part of "the tuba sound" is not captured by 0.25 seconds of mf playing of isolated notes? How would you measure these aspects?
For starters, what's with the almost total lack of fundamental on D? (yes, it's real this time - I really did quadruple check it!)
Bb
A
G
F
Eb
D
C
BBb