Experimental Composition Using a Guitar-Controlled Synthesizer


Michael Horgan

Music 220D

6/9/2008


For my 220D project, I have created a composition which employs my 220C project as the principal interface for the computer-based synthesis of sound. My 220C project is a pd patch that generates an accompaniment based on its input. The core of the patch is a fundamental frequency estimator object called yin~, which I implemented as a pd extern. Two synthesizers in the patch receive the fundamental frequency estimate from the yin object and produce output accordingly. Since I am a guitarist, I use the guitar as the input interface to the patch; however, any analog or digital signal can also be used.


Listen to the composition


After going through the process of creating a composition with this software, I now have a solid understanding of how useful it can be as a tool for composing. Originally, I imagined the patch would work well in real-time, accompanying the guitar output in a melodic and harmonically pleasing way. However, as I mention on my 220c web page, the limiting factor of the yin~ object is the large buffer size it needs to accurately measure low frequencies.


This causes problems with real-time playback, so to get around this problem, I simply made recordings of the patch's output while I improvised on the guitar in a particular key, and then sampled these recordings. For my composition, I used the e_gypsy pitch rounding object, so the composition is in the E major “gypsy” key, which contains a flatted second and a flatted sixth. All mixing, editing, and additional synthesis was done using the Reason 3.0 production software.


The purpose of the piece is to showcase the guitar patch's usefulness as a composition tool. The results are less precise than I had hoped, particularly with regard to rhythm. The large buffer size (2048 samples) means that the rhythmic timing of the synthesizers isn't always perfect, which is audibly clear in the composition. The pitch tracking is also a little slow, so it doesn't always capture every note from the guitar, and because of the large amount of polyphonic interference from the guitar, it doesn't always capture the right pitch. This led to some interesting results, as the melodic material of the composition is not an exact replica of the input pitches. Thus, the melodies are the combined result of my input from the guitar and the guitar patch. This kind of computer-assisted, improvised composing style can then be useful as a way to generate unique melodies from spontaneous improvisation.

The main synth in the piece was recorded directly from the pd patch and then mixed in Reason. Since the rhythm is a little sloppy, a second synth, generated by Reason, repeats the melody half way through the piece. This second synth is much cleaner sounding, but it plays essentially the same melody as the first. I am pretty happy with this second melody, which I did not compose myself but rather used the pd patch to assist in creating it. I think this is the patch's most promising application as a compositional tool.


The other contribution of the pd patch to the composition is the accompaniment which plays beneath the main melody. This output is the response of the patch to a sequence of arpeggios played on the guitar. As a second application, the patch is pretty useful for generating what I would call “harmonic coloring.” As I play the arpeggios of E major, A minor, and D minor in succession, the patch outputs a colorful collage of the arpeggios—and sometimes neighboring pitches—in a somewhat unpredictable way.


In the future, I hope to improve upon the patch by reducing the latency and improving the rhythmic response of the synthesizers to the guitar input. This will, of course, be very difficult as yin~ object demands a large buffer size. However, there may be ways to get around large buffer sizes by instead measuring the harmonics of the fundamental and then rounding the pitch down the appropriate number of octaves.