by Chris Grigg of NEGATIVLAND.
Thinking of Digital Audio in Terms of Higher Fidelity is a Mistake.
We've all been conditioned by the guys in the white lab coats to think of digital audio as a great improvement in sonic purity, but this is a mistake. While there are obvious advantages to digital recording, processing, and distribution of sound, it turns out that digital technology is also a really great way to destroy sounds in new and surprisingly rich ways by using the technology not as it was intended to be used, but rather by finding the interesting new applications in the tools and then abusing them.
This article - not to mention a new CD-ROM sound library disc also called Digital Dysfunctions, which I hope will be out by the time you read this - is dedicated to the proposition that digital processing makes it possible to absolutely murder sound in lots of new, complex, grating, harsh, organic, electronic, alien, unpredictable ways. Here are a few tips on how to severely screw up audio sample files on any reasonable computer using any reasonable soundfile editor, plus maybe some software plug-ins and cheap shareware. I used Sound Designer II on an old Macintosh IIci in these examples because it was what I had available as I wrote this, but your personal favorite soundfile editor should work just fine.
FAVOR CRAPPY SAMPLERS, FAVOR ORGANIC SOUNDS
I don't use a fancy sampler or digital audio card to capture many of my own digitally dysfunctional sounds for this simple but unusual reason: In this kind of work, the cleanest sample isn't always the most interesting - or useful. In fact, lots of my best sounds started out in life being captured not only at a lowly 22kHz sample rate, with the okay-for-some-uses-but-basically-pretty-crappy MacRecorder 8-bit sampler, but actually with the even crappier cheap condenser mic built right into the plastic digitizer box, held up to my face as though it were an SM58 or something. More distortion? Massive aliasing? Horrific noise floor? You bet. So why do it?
Because all these supposed injuries to the signal are actually only injuries if your primary consideration is preserving the pristine quality of the original sound. In the case of many of my own projects, the original sounds are just random things like me blowing my nose or making mouth farts, and I just don't care about preserving the beauty of that. My aim isn't to capture the acoustic sounds, it's something very different: I'm trying to create spectrally dynamic, rich, complex source signal files not so much to listen to in their own right, but rather to use as excitation sources for the subsequent mutations, truncations, and other extreme transformations I know I'll be making. Viewed from this perspective, those characteristics of cheap samplers that are usually seen as flaws - all of which serve to introduce frequency-domain artifacts into the signal, many of which aren't even harmonically related to the signal - are actually benefits that help produce more interesting, dynamic, rich, and complex end results.
Likewise, if you choose sounds with dynamic timbres (i.e., overtone structures that change over the life of the sound) - organic sounds like recordings of vocal utterances, physical events, or certain kinds of complex synthesized tones (like some FM timbres, for example) - then your subsequent processing will produce more interesting effects. Apply any of the following processes to a sample of a steady-state sine tone (which has the simplest possible spectral structure) and the results will be a lot lamer than doing the same stuff on a sample of yourself making mouth farts, which is an easy-to-make example of a spectrally complex sound (see Figure 1 & Figure 2).
So much for generalities. On to the specifics.
DIGITAL DISTORTION IS YOUR FRIEND
You know how great/terrible it sounds when you turn an analog signal up way past the clipping point of an input? Well, guess what? You can do the same thing digitally, only (1) it's easier to do, (2) it sounds much harsher, (3) the signal may even intermodulate with the sampling frequency, and (4) if the spectrum and/or level of the sample you're turning up changes significantly over time, so will the characteristics of the distortion. Digital distortion has of course been turned into one of those forbidden lands in life where we've been conditioned never to go, but don't be a scaredy-cat; you can do some really wonderful stuff with it.
Note: If you're going to try out the techniques in this article as you read, you may want to turn on your editor's Allow Edit Undo switch at this point.
How to: In Sound Designer II, just select the whole file, then under the Edit menu select Change Gain. In the dialog, set the gain to something ridiculous, like 30dB or 50dB (come on, be daring - you can type in any number you like, the computer won't explode, and there's always Undo), then click the Change Gain button. But careful now! As you probably could have predicted, your sound just got a lot louder (as well as gaining a lot of new higher-frequency components due to the complex clipping), so either turn down your monitoring level quite a bit, or use Change Gain again to reduce the level of your sound quite a bit before you play it back to see how your brand new distortion sounds.
If it doesn't work the first time: If you didn't hear any dramatic change in the sound, either use a bigger number for the amount of gain increase, or normalize your sound before doing the extreme gain change. (Note, though, that since the entire dynamic range of the 16-bit wordsize is only 96dB, there's no point in typing in a number larger than that.) If on the other hand the change was too dramatic, just undo and then gain change again with a smaller amount of boost.
EXTREME DIGITAL EQ
Digital EQ has some advantages over analog EQ. You can get much more precise control, you can get much more extreme amounts of boost or cut without adding analog noise, and you can use it to generate frequency-responsive digital distortion. I like to use it for emphasizing (or creating) high-frequency content in sounds I've mangled in other ways. Very narrow bands, like 10 to 100Hz, and very large boosts, like +24dB, are good for this, especially in the 5 to 8kHz range, which can create a techy sort of fake clarity. Sometimes big bass boosts in an already bass-heavy signal can give you the most amazing speaker-cones-ripping-apart effects as the EQ makes the signal go into digital distortion. Note that since EQ is a gain change for only a part of the signal, carefully setting the EQ boost control can exceed the limits of 16-bit full-scale in sync with only certain frequency components (i.e., digitally distorting the whole mix on the bass note attacks only) - and I hope by now you find that prospect inspiring rather than frightening.
How to: In Sound Designer II, select the whole sound, then select Parametric EQ under the DSP menu. When the dialog appears, click on the Peak/Notch filter type (the middle one that looks like a bump), set the Center Freq to 8,000Hz, set the BandWidth to 10Hz, and set the Boost/Cut to 24dB. Now click on the Preview button to start playback, and tune the Center Freq and BandWidth until you like the sound; if there's too much distortion, turn down the Boost/Cut. When you have it how you like it, click the Process button and the EQ will be applied to the soundfile.
EXTREME PITCH SHIFTS
Movie sound folks have known for a long time that if you pitch a recording way up or down, it takes on a very different sonic identity. Add digital processing to the picture and you get two big bonuses: (1) Virtually unlimited transposition ranges become available because you can transpose the same sound over and over without picking up analog noise at each pass; and (2) Most contemporary pitch shifters actually sound pretty bad, adding all kinds of schmutz to the sound (remember, we're talking about digital dysfunctions here, so this is a plus). Transpose a sound several octaves down and most pitch processing software is almost guaranteed to turn it into something interesting, particularly if it's a not-too-low-pitched, organic sound to start with.
How to: In Sound Designer II, select the whole sound, then select Pitch Shift under the DSP menu. When the dialog appears, turn off the Time Correction checkbox, set the Semitones slider to something like 7, and click on the Process button. When the "Updating the soundfile" modal box disappears, click Preview to hear your slowed-down sound, then click Process again. Repeat until it's so low and weird that you like it.
If it doesn't work the first time: Transpose it farther, or use more stages of transposition to get the same amount of transposition (i.e., 1 semitone ten times in a row instead of 10 semitones once). If it doesn't sound messed up enough for you yet, transpose it down a long way (like down an octave twice) and then back up by the same amount (like up an octave twice); the net pitch will be the same, but more processing artifacts will be present now.
APPLY THE SAME PROCESS OVER AND OVER AND OVER AND OVER AND OVER AND . . .
As the previous technique goes to show, sometimes the controls on a processor just don't go high enough for what we'd really like to do. In the analog world, this would be the famous Spinal Tap "These knobs go to 11" problem. In the digital domain, we have the option of simply re-applying any given function to make its effect more extreme - particularly since preserving absolute fidelity isn't the point. Why limit your pitch shifting to ±12 semitones just because that's as far as your editor's pitch shift slider goes? If once is 12 semitones, then twice is 24 semitones, thrice is 36 semitones, etc. As long as you like the sound you're getting, no number of repetitions is too extreme. On some projects, I've applied a single filter to the same soundfile over 200 times in a row (but then I don't get out much). Of course, certain processes grow really unusual artifacts when reapplied in this way, but then, "unusual" is good in the digital dysfunctions book; and in fact, some delightful effects can only be achieved by letting these kinds of errors accumulate. If applying it once sounds kinda crummy, then applying it 20 times should be really interesting.
There's this guy, see, and his name's Tom Erbe, see, and he's written this great piece of shareware for the Mac, see, and it's called SoundHack. It opens files and lets you apply lots of different DSP processes to the sounds. It isn't real-time because it uses highly evolved floating-point arithmetic rather than the lowly integers most sound programs use, and that takes a lot longer - but boy, is it worth it. If you want clean digital processing, come see SoundHack.
Most of you reader-people out there won't already have SoundHack, but if you have a reasonable Internet connection you can get a copy via anonymous FTP from the directory ftp://shoko.calarts.edu/pub/SoundHack/. Also, an on-line manual is available on the Web at http://shoko.calarts.edu/~tre/SndHckDoc.
For purposes of digital dysfunction-style sound mangling, I especially like two of the "hacks" that SoundHack offers: Varispeed and the Phase Vocoder.
SoundHack's Varispeed. The Varispeed hack lets you dynamically change the pitch of a sound over its life by drawing a time-versus-pitch curve in a graphic editor - sort of like a precision, scriptable pitch wheel on an ultra-hi-fi sampler. SoundHack maps the length of your drawn curve to the length of the soundfile, so even though all of the curves are the same length on the screen, you can apply any curve to any sound. Besides simple pitch glides, you can do stuff like record-stop sudden slowdown effects. More radical discontinuous curves can make a single soundfile sound like it's a tape edit made from many different sources. This is one kind of digital dysfunction where even fairly simple source files can come out sounding really interesting. A few useful Varispeed curve files are available for download too.
How to: In SoundHack, open the sound you want to process (start with a real short sound - SoundHack's great but it's really slow!), then select Varispeed under the Hack menu. In the Varispeed dialog, check the Varispeed box (this may seem redundant, but if you don't do this the dialog just acts as a straight sample rate converter/pitch shifter), then check the Vary by Pitch radio button. Now click on the 'Varispeed Function . . .' button to bring up the function editor dialog (see Figure 3).
If you have my curves from the Web site: In the function editor, click the Read button at bottom right, and use the Open dialog to find and open Harsh 3 short.function. Once it opens, click the Done button at bottom right and skip over the next paragraph.
If you don't have my curve files, then just draw a bunch of random shortish lines in the curve editor window so that the curve looks something like the one in Figure 3, then click Done.
Now we're back at the Varispeed dialog, where you can pick Best, Medium, or Fast quality - I like Fast because it sounds a little rougher - and, finally, click the Process button. SoundHack will prompt you for a file type (pick Sound Designer II 16-bit linear) and file name, and then go off to work its obscure magic for a while, leaving your original file alone and spitting out the new file when it's ready.
If it doesn't work the first time: The effect should be pretty damn noticeable with Varispeed functions like the ones shown above. If you created a Varispeed function of your own but don't notice much of a pitch-change effect after running the Varispeed hack, go back and look at your curve in the function editor and either make the numbers at the bottom right and top right of the graph (the minimum and maximum number of semitones) bigger and redraw the graph, or make the curve of the graph more randomly distributed. Also, if your original sound was too short and your Varispeed function curve had too many points that were at too high a number of semitones, the shortening effect of playing those bits of the soundfile that much faster can result in a really, really short soundfile. Either adjust the curve to something lower, or use a longer sound. One last thing to check: Make sure you're listening to the output file, not the input file.
SoundHack's Phase Vocoder. Phase vocoding is a technique whereby a soundfile is analyzed, and the results of that analysis are numerical characterizations of the envelopes and its many different frequency components - at which point it isn't really a soundfile any more. You can mess around with those envelopes however you like (stretch the time scale, maybe map the envelopes from some frequencies onto different frequencies, etc.), and then resynthesize it all back into a soundfile again - a soundfile exactly like the one you started with, except for your changes. You can do a lot with phase vocoding, and get a lot of strange results from it, depending on how you set the controls. I mainly use it for stretching really short sounds out into entire little soundscape-pieces.
How to: In SoundHack, open a spectrally dynamic but really short sound (less than a quarter second), then select Phase Vocoder from the Hack menu. For now, ignore all of the controls except the number to the right of the pull-down that says 'Scaling,' and set that number to, oh, say 10. Click the Process button at lower right, pick a file format (SDII 16-bit linear), name an output file, then wait for the program to analyze your old sound and figure out the new one.
If it doesn't work the first time: Try a bigger Scaling number, or use a sample with a wider spread of frequencies. I sometimes like to set the Scaling to 1,000 and let it run overnight on really short files.
ABUSE OF DINR
DINR stands for Digidesign Intelligent Noise Reduction. It's the name of some not-very-cheap software plug-ins that company makes. They're intended for removing various kinds of noise from audio files, and many people love 'em for that - but when the controls are set wrong, DINR tends to create its own interesting artifacts around certain kinds of well-defined features in the signal. Kinda sounds like burbling or robot voices if the sound has any sort of fast, steady modulation in it. A couple of files full of interestingly wrong DINR control settings are also available for download.
How to: Open your sound in Sound Designer II, then select Broadband Noise Reduction under the DSP menu. We have Contour and Settings files on our Web site available for loading, otherwise generate your own using the procedures outlined in the DINR manual. Click Preview. If you like the result, Process; otherwise, mess with the controls. Scoot the points in the red Contour line around (I like alternating very high and very low bands), turn up the NR Amount. Use the plus and minus keys to move the Contour line further away from the black noise signature display. The higher you put the Contour line toward the top of the Spectral Graph display the more robotic the sound will become. Click Process when you like it.
If it doesn't work the first time: Either click Process again to apply the same thing another time (or another two times, or another ten times), or twiddle with the knobs until you get an effect you like. Do keep the NR Amount large. I like to use the Highpass filter as a boost. Note that the DINR artifacts are most noticeable in signals that have a lot of spectral dynamics; if your sound is too smooth you might not hear a big effect - try something more scrambly. Or just keep re-applying DINR to the same sound until you start to get something you like.
CRUMMY TIME EXPANSIONS
Some sound editors I could name (like Sound Designer II v2.6 and earlier, for example) did not use the best possible time expansion algorithms, and tend to add a tremendous amount of mid-to-high frequency artifacts to the soundfiles that they time-expand, as well as a sort of fast stuttering effect. Fortunately, it sounds really horrible - the more so, the more generations of it you apply to a sound - and this particular dysfunction has become a favorite of mine. (SDII v2.8 still adds artifacts, but seems 'better').
How to: In Sound Designer II, select your whole sound, then select Time Comp/Exp under the DSP menu. In the dialog, set the Time Ratio to 2.00000 (in SDII this is the maximum expansion factor), set the bottom slider to Accurate Timing, and click on the Compress/Expand button.
If it doesn't work the first time: Monkey with the Min. Pitch slider and make sure the bottom slider is set for Accurate Timing - there's a strange interaction between the Time Ratio, Min. Pitch, and High Quality/Accurate Timing controls that can result in a file being expanded to something much shorter than your Time Ratio number asks for. And, of course, you can just re-apply the time expansion if you like. Don't worry, it'll work eventually!
USE A DYNAMICS EXPANDER TO SIMULATE INTERMITTENCY
Almost any interesting sound's volume constantly varies, whether slightly (as in a held sung note) or dramatically (individual percussion hits falling away to the silences in between). If you use a software dynamics expander process and set the sensing controls just right, you can gate a more or less constant signal so that only the loudest bits come through - and because the volume peaks in many more-or-less constant signals are often essentially randomly spaced in time, the rhythm of the resulting gated sound often sounds similarly random, like a signal on a crummy cable that's shorting out. (Just to be crystal clear in our new era of endlessly ambiguous digital terminology: We're talking about a volume expansion in the audio signal here, not about size-expanding data-compressed files.)
How to: In Sound Designer II, select your sound and then pick Dynamics under the DSP menu. Select the Expander function, set the Ratio slider as high as it'll go - this gives a quick cutoff when the signal is below the threshold - then set the Detect slider closer to Peak than Avg, and click on the Preview button. As you listen to the expander work on your signal, adjust the Detect, Thresh, Attack, and Release sliders to get the sonic effect you want. Longer Releases tend to give more nearly acoustic-sounding effects; shorter Attack and Release sound more like electronics cutting off and on.
If it doesn't work the first time: The Thresh setting can be critical, so play with that. For even sharper rejection of the silence areas, just apply the expansion to the sound again.
USE LOTS OF DIFFERENT TECHNIQUES TOGETHER
If you've been listening to the example soundfiles as you read, then by now you should be getting the idea that there's a very large number of different, specific ways to wreck sounds interestingly with your computer. Well, it gets better: I think the real fun is in combining these techniques. Take a mouth fart sound, digitally distort it, Varispeed it, pitch it down four octaves, make it warbly with DINR, give it some excessive narrow-band EQ boost, and you're bound to come up with something interesting. If you're starting from a file you like and want to keep as is, then in some programs (like Sound Designer) this means making a copy of the source file to work on at every step and applying the new process to the copy. Yes, this is a major irritation, and it can produce a whole lot of soundfiles before long (although you'll quickly learn not to mind simply throwing away the bad and so-so ones). Think of the chains of files you produce this way as paths of exploration. The potentially infinite process of change-listen-change-again tends to be quite rewarding because almost every time you apply a new tweak you'll notice some new, unanticipated quality in the result, and after you've been doing this for a while, each result, whether you find it a usable product in its own right or not, will suggest a good next step.
TAKE ADVANTAGE OF THE INTERESTING FLAWS IN YOUR TOOLS
Much of this hyped-up "digital dysfunctions" talk is ultimately about discovering the unexpected, unintended sonic opportunities that your tools can provide, and about being brave about finding ways to use the new, weird, unpredictable sounds you'll find in them. All of the above techniques are intended to get you started down this path, but exploring your own gear and finding your own nooks and crannies may mean lots of experimentation. Don't be afraid to set controls to extreme positions, to chain processes together in unusual (or unusually long) sequences, or to do things you aren't "supposed" to.
But why would anyone want to cook up such weird, horrible noises as these? Well, it seems pretty natural to me to be making and using terribly harsh electronic sounds right now - after all, we're living through a terribly harsh electronic time.
P.S.: Oh yeah, one last thing: I expect these "flaws" to be fewer and farther between as the tools mature, so remember to save all your early-1990s sound editing software!