In this post, I’ll discuss some decisions which have arose as I begin constructing my piece. As I have mentioned, the main thrust of the piece is to showcase the different effects that can be achieved with uniform quantization in the MDCT domain.
Audio to process
I am immediately confronted with the choice of input audio with which to process. As I mentioned in class, I had initially considered a long recording of Romain cheerfully whistling and welding his Chanforgnophone in the garage behind the Knoll on a recent evening. However, I was unhappy with the results after processing. Since the signal was so complex (containing Romain’s singing/whistling, transient metallic clunks and drawers slamming, the low-frequency drone of the welding apparatus, and the intermittent high-frequency noise of the welding), the effect is somewhere along a continuum of slightly degraded (but annoyingly so) and extremely time-smeared and distorted (though not in a controllably harmonic or melodic way).
Instead, I found it much more fruitful to take very short audio samples (around a second) and batch-process them with (a) MDCT windows sizes from 256 samples (5.8 ms) to 218 samples (5.94 seconds), and (b) quantization bit depths from 4 to 14. Here are some examples which I find interesting.
First, some overtly glitchy, bit-starved examples:
The original sample is:
Second, here’s a good example of the pre-echo and reverberant effects of a huge window size:
The original sample is:
In the prototype piece I show in class today, I explore the combination of these extreme types of processed samples. However, I will continue to explore different types of short audio samples with which to process as well as the middle ground of window length and bit depth parameters.
One other decision I made regards spatialization, since I’d like to take advantage of the multichannel setup we’ll have at Bing. There are numerous ways to do this, such as ambisonics, vector based amplitude panning (VBAP), and the loudspeaker orchestra approach. (In fact, I think we could also make use of the WFS array, which seems like it will make its way to Bing for the concert.) I have chosen to take the loudspeaker orchestra route, in the tradition of the venerable Acousmonium and the BEAST of Birmingham. In fact, I believe Chris Chafe’s Tomato Music uses this spatialization technique (most recently at the Feb. 16th concert at Bing).
I believe that addressing each processed signal to one speaker is the ideal technique, since the elements of my composition are artificial and exist divorced from space. That is, I am not creating a virtual space with any cues regarding reverberation – the true reverberation from the input audio is highly distorted if existent at all, and I am not adding any digital reverberation since the long-window processed samples achieve a similar, if unnatural, effect. Additionally, I think of ambisonics and, to a certain extent, VBAP, as a way to sonically hide the speakers from the listener. For my purposes here, I’d like the audience to be aware of each speaker and the sound it produces. Moreover, I’d like each speaker to cultivate a personality or character through the piece. Perhaps one speaker tends to speak in the glitchy, narrow frequency language of the low bit rate processed samples, while other speakers tend to the higher bit rate, longer window samples (and thus speak at different time scales, with rich reverb and pre-echo dominating). Since this piece derives from a technological process, it seems entirely congruous and desirable that the speakers serve as focal points and performers.
I look forward to playing some of my ideas and getting thoughts or comments!