That's a stretch, part 2

Christmas time was upon us, and so out came the Christmas decorations, including Pengie, our inflatable Christmas Penguin. Part of Pengie's garb is a $2 Christmas tie, which plays a series of chopped up Christmas songs in full 4 bit glory. Unfortunately, the tie is showing its age, and the battery is going, and it sounds even squeakier than when new. However, closeup recording seemed to disguise the wavery warbling quality the weak battery produced. Here's what the tie sounded like this year (who knows what it will sound like next year, with further battery decay?):

Christmas Tie 4-bit sound maker recorded closeup

And here's a photo of me on Christmas morning wearing the Christmas tie, with Pengie and our tiny live Christmas tree (which normally lives in the back yard, without tinsel), and a host of holiday jolliness spread around the tree.

 

Warren Burt and Pengie on Christmas morning 2009, photo by Catherine Schieve

One of the first things I do with found sounds these days is to time-stretch them, hearing what they sound like when expanded. Back in the "old days" of the late 60, early 70s, there was only one way to time stretch sound without changing its pitch - a device called the "Springer Unit" which had rotating tape heads which allowed you to, to some degree, fragment and reassemble the fragments of sound, stretching the sound somewhat. I remember when I was a graduate student at UC San Diego, the department had one, but it was always broken, or breaking down, and students weren't allowed access to it anyway, so the idea of time-stretching sound then was pretty much just a dream, expressed in theoretical essays such as Steve Reich's "Slow Motion Sound," or in elaborate (mostly unsuccessful) experiments involving slowing down a tape while using an analog frequency shifter to transpose the sound up the required amount. Now the situation is vastly different. There are many software tools which allow you to time-stretch sound in many different ways. Most of them rely on some form of granulation (a way of chopping up a sound into tiny pieces and reassembling them on the fly), but they all sound different. Just off the top of my head, I can think of 7 programs on my computer that do time-stretching in one way or another: AudioMulch's Bubble Blower; Plogue Bidule's Looper; Adobe Audition's Stretch(process); Reaper's Elastique time-stretching; the many different time-stretching routines of the Composers' Desktop Project (whose wavecycle processes offer a unique method of time stretching - and hats off to Trevor Wishart for developing those!); and in the realm of software synthesizers, Linplug's CronoX, and Camel Audio's Alchemy both offer different ways of stretching samples in time and pitch. So the question these days is not "can we stretch a sound?" it's "which time-stretcher will give me results that interest me?" All time-stretching algorithms produce artifacts - in fact, the very notion of "natural" time-stretching is a contradiction in terms. What fantasy of yours corresponds to what you would like to call a "natural" sounding result to a totally artificial process? Nonetheless, some of the results of time-stretching sound more "mechanical" or "electronic" than others, and depending on the purpose your sound is intended for, you might prefer the results of one method or another.

Naturally, having recorded the Christmas tie, I tried stretching the recording in various programs. The program that gave me results I liked best, for this sound, was CronoX, a very nifty software synthesizer that not only allows you to stretch sound (to extreme lengths, too), it allows you to set the amount of duration stretching that will occur at different pitches, AND will allow you to then tune the whole thing into some microtonal scale. So here's the Christmas tie recording, stretched and doubled pretty extensively, into a 4 minute sequence.

Warren Burt: A Stretched Christmas Tie, December 25, 2009 - duration 4 minutes

I think it's pretty neat how a cheezee 4-bit chip can make a lovely sequence of bell-like, festive holiday sounds. And if you listen closely, and slowly, you can still hear the Christmas melodies coming through. For those of you tech-minded, here's the faceplate of CronoX with the settings used for this recording:

CronoX faceplate with settings for stretching the Christmas tie recording

For each key pressed, there are 2 copies of the recording playing - one stretched to 4.03 times original length, the other to 4.07 times length. The "time-track" (how much a sound is time stretched at different transpositions) is set to .63 for one layer, and .61 for the other - this means that a one octave pitch transposition down will only result in an approximately 1.3 times time stretch, as opposed to the normal 2x we're used to with tape recorders (for those old enough to remember tape, that is!). The keyboard is tuned into a 23 note just-intonation scale (not shown in the picture), so if the "key" of middle C is pressed, the sound will be at it's original pitch, but a "key" of C two octaves lower than that will produce a pitch just a little bit more than one octave below the original. The "chord" option in CronoX is used here, with notes (descending from middle C) of C, F Ab and C pressed. This means there are 8 different versions of the Christmas tie recording being played (2 per key with 4 keys pressed). This sound is not looping, so when the recordings reach their ending, there is silence. The almost 4 minute duration of the piece was determined by how long the slowest version of the recorded sound took to be played.