Audio Production Fun for Ethan

Here we go...

Bear in mind that all this is totally enwrapped in a particular aesthetic, i.e. mine. My hope is that you may be able to use some of these ideas in developing your own ability to Make Cool Sounds.

First of all, I like sound that yields new audio information to your ears, both at the macro- and the micro- level. What do I mean by this? I can perhaps explain best in the negative. An unchanging waveform on a single pitch is boring and annoying. Phrases of notes done with a static timbral palette -- same thing. Some people really like that good-old 'square waves forever' audio, but not me. Think about the sounds that us humans make: constantly changing. In fact, we can't make an unchanging sound. Human, all too human. Much of what I do in synthesis is an attempt to recreate the random or semi-random nature of human-produced audio.

One way to do this is to duplicate a single audio synthesis technique one, two, three or many times. Detune each one slightly -- this was what old analog synth pioneers discovered (hey, the minimoog has three oscillators!). Add the inherent pitch-drift of analog gear and you get... interesting sound!

Many of the basic DSP techniques can work to emulate that 'more than one' audio aspect. Digital delay, echo, for sure. Flanging and "chorusing", yeah. And reverb/room-simulation surely messes up the sound in an interesting, reduplicative way. Plus it sounds so gosh-darned mystical.

With this as my introductory background, I want to go through some pieces that exemplify some of what I listen for. These form the core of my 'production' lecture in my Intro to Digital Audio class, and as such they deal not so much with "synthesis", but more the overall sound of the music. It's a way of listening, though, and you can apply the same principles when thinking about synthesis on a more local level.


Reverb

The most dramatic example of reverb(!) that I use is this piece by the Irish new-agey/folk-rock band Clannad: Reverb! Note how it "fills out" the choir sound. And mystical?!? Oh yeah! It's in Gaelic, for goodness sake! That guarantees mysticness! And it's in an apparent cathedral-like setting that is so reverberant that the walls seem to be imparting energy to the signal (a very long reverb-time). And an organ chord comes on, only every once in awhile.

Added to the heavy Gaelic-imbued magic is a little trick they use on the voice and the harp at one point called 'backwards reverb'. This is done by taking the original track recording, reversing it, and then reverbing the reversed signal and recording the result. Take that result, flip it back around so that now the voice/harp is going the correct direction in time, but now the reverb appears before the signal. Cool!

I'm not going to go into the details of room-simulation, but it's essentially a way of adding a reverb-ambience to sound. Many of our room-sim algorithms also allow you to localize a sound in the stereo field. Giving a sound a precise location is another way to add interest to it.

Double-tracking

Another method to thicken/add interest to a signal is to do it twice, but be sure that there is some microrandomness between the first and second recording of the sound. This is easy if you're a human. Just record the same thing again (like I noted above, we can't do something perfectly). Give a listen to this track: This is another Irish folk musician, a guitarist named Pat Kilbride. See if you can hear when the double-tracking happens! :-)

Note that when he brings in the doubled guitar, he splits the two tracks hard left and hard right. In a stereo presentation, this tends to produce a really 'wide' field sound. I use this a lot, or I'll assign different doubled parts randomly through the stereo presentation.

Here's one more great use of double- (and more...) tracking:

The piece is Wallflower by Peter Gabriel. He wrote this while taking a prime role in helping mobilize anti-apartheid public opinion in the late 1970s. It's for Nelson Mandela. What he does is to sneak in pianos that double the piano(s) already playing. As the piece unfolds, you begin to hear more and more pianos all playing. The effect is of a huge upswelling of, well, stuff. Lots of pianos playing simultaneously is a big sound! Cool way to reinforce his message.

Delay/Flanging/Chorusing

You can kind-of duplicate the double-tracking effect by using a short, single delay (like < 100 msecs) and splitting the original and the delayed signals wide right and left. Very short delays will actually produce the effect of a sound seemingly coming from "outside" the stereo field because our ears use short delay differences to decode lateralization, so sounds with short inter-speaker (inter-aural) delay times can psychoacoustically appear in different locations.

If a very short-delayed signal is added directly back into the original (and also 'echoed' so that more than 1 repeat occurs), it yields an effect called 'comb filtering'. The delayed signal destructively interferes with the original because of the phase inversion at particular frequencies (inversely related to the delay time; I can explain this better in person or you can google it). If you dynamically sweep the delay time up and down, it creates an interesting 'swirly' effect called flanging. This also sounds somewhat like doubling, and using several flangers in parallel produces an effect called "chorusing".

Here's a track by guitarist Steve Tibbetts with several of these effects applied to the lead guitar solo:

Listen especially to the lead beginning at time 2:17.

Comb-filtering isn't used much in production of music, but it is a powerful effect. You can really crank up the regeneration on a comb filter and produce a very pronounced pitch, even when it is applied to relatively pitchless source sound. Paul Lansky uses this a lot in many of his pieces (no examples here, sorry!). It's also the effect that made Darth Vader sound Darth Vader-like.

Echo can also be a very useful effect, especially used with synthesis. I have an example here:

This is by the group Porcupine Tree, and the echo on the voice endows it with an otherness that's pretty cool.

One more example, from Jeff Beck on his latest album "Loud Hailer":

Note the split-delayed guitar sounds. He's also using some ring modulation on the guitar -- a very interesting effect -- but I won't talk about it here.

Filtering/EQ

Metal bands often use double- (and often triple- or quadruple-) tracking of guitars split wide in the stereo field to get a 'big', wall-of-sound effect. They also employ filtering in production to shore up that 'big' sound.

Filtering (or EQ -- equalization -- another word for the use of audio filters across the sound spectrum) is used a lot in production. It isn't emphasized so much in synthesis, except as a part of a particular synthesis technique (subtractive synthesis, for example). Much of the time spent in remixing and post-production is spent with EQ/Filter knobs to get the sound set.

Here's an example of a metal band (Megadeth) trying for a 'big' sound by multiply-tracking Dave Mustaine's guitars:

Notice that the sound is a little "thin" though, especially when compared to the this multiply-tracked guitar piece by Soundgarden: Megadeth was trying to give the impression of 'bigness' by boosting the low-end of the sound (below 150 Hz or so). It takes more mass to move low-frequency sounds, so our ears decode it as 'big'. They also boosted the high-end (probably 5000 Hz and higher) to give the 'sizzle' to make the music sound bright, and close-by. But they way overdid it and wound up 'dishing out' the middle [sorry for all the scare-quotes, but lots of colloquialisms here]. Soundgarden left a fair amount of the low-midrange intact while still boosting the lows and the highs, yielding a more substantial sound. Or at least that's the way it sounds to me!

One of the snazziest use of filtering is the creation of an 'artificial' acoustic environment in Björk's There's More to Life Than This:

There's a lot I could say about how they built a 'fake' party (including the music seemingly coming from a party/live sound system), but the amazing filter fun comes at time 1:36. Hey! Drop in the sound of a door closing, dump all the high end from the program material, close-mic Björk and all of a sudden you're in the bathroom with her!

Another artificial-impression use of filtering is on the vocal parts at the beginning of Värttinä's piece Pojaton:

By applying a band-pass filter (and purposefully singing slightly out-of-tune), the production sounds as if the recording was made "in the field" with an old tape recorder on an early 'collect the folk music' expedition. By dropping the filter (and tightening up the harmony) at time 0:47, they move the song into the present era. I'll tell you more about this song sometime. Knowing the words, it seems a paen to the death of their culture. Daniel tells me that the tale in the song is from the Kalevala about the coming of Christianity to Finland towards the end of the epic poem.

Vocoding/Pitch Correction/Vocal Modeling

Vocoding is a way of constructing a complex, time-varying filter and then applying it to an arbitrary input sound. Often the analyzed signal (for the filter) is done with a voice ("vo"-coding), and then that filter can be used to impart voice-like characteristics to other sounds ("talking" traffic noise, etc.). Also often, the program that the voice-filter is applied to is a synthetic texture; a melodic instrument or a chord. Here's a whole bunch of examples, starting with Wendy Carlos' use of it for Kubrick's A Clockwork Orange soundtrack -- one of the first popular-consciousness use of vocoding: You don't have to use voice/synths to do vocoding, though, and I believe an untapped potential exists for the cross-filtering of many different sounds. Usually you want the 'program' sound (the one you are filtering) to have a fairly rich spectral content, though, in order for the filters to have energy in the spectrum to filter. But you don't have to.

A related technique to vocoding is convolution. The way it works is different, but the results are similar -- the application of the overall "characteristics" of one sound to another. Convolution works by essentially multiplying two FFT analyses together. Many of the more popular 'high end' digital reverbs use convolution. I can tell you more about this if you're interested.

Pitch-correction is a more recent DSP technique, used by many rap artists now (I Am T-Pain). It's also used for most singers these days, because THEM DANG KIDS CAIN'T SING LIKE THEY DID WHEN I WUZ YOUNG! One of the first uses of it as a true technique-in-itself was by Cher, but I like this early use by Madonna:

Cute.

Some of the most recent vocal-synthesis/processing algorithms are fairly sophisticated. A few years ago, Yamaha released a plugin that used vocal physical-modeling technologies to produce a 'virtual chorus' around a single sung melody. The only example I know of is this cut from Porcupine Tree's release The Incident:



A Few Random Effects

Compression is used a lot in production, but not heavily in synthesis (except for hugely-compressed dub-step/techno/edm stuff). It essentially tries to keep the level of the program material constant, lowering it when it gets loud and raising it when it gets soft. It tends to add an aura of "excitement" or immediacy to the sound. A lot of radio stations go berserk with this on everything they broadcast. This cut by Peter Gabriel is one of the more interesting uses I've heard: The cymbals, especially.

Here's a cute use of tremolo, rapid amplitude fluctuation, but at LFO rates so it doesn't turn into AM:

I love the wide split on the tremolo-ed guitars in the chorus, also synced with the 9/8 tempo. This is also by Peter Gabriel, do you know where it's from? :-)

Trent Reznor uses a compression-like effect in this one:

This is from The Downward Spiral, and his increase of the noise underlying the vocals along with a subsequent decrease in the vocal track amplitude makes it sound like the noise is taking over.

Sounds

Sometimes I really go for a piece because of one specific sound in it. Check out the "solo" instrument sound by this cut from Los Lobos' CD This Time: The main passage featuring it starts at time 1:58 (and again at 2:30). Mitchell Froom produced this, and the guy is a genius at constructing weird sounds. I'm not sure exactly how this one was made.

Another solo instrument is from the soundtrack to the American version of the movie Traffic by the production duo Kruder and Dorfmeister:

The sound I mean is towards the end, starting at 4:03. I think it's a brass physical model synthesis algorithm, but not 100% sure.

I also love the sound that travels underneath this cut by How to Destroy Angels (Trent Reznor, again, with his wife and a few others):

It swells towards the end startting at time 4:55. Again I think it's a physical model of some kind, but it could be FM or something run through a distortion unit. I like the plucky-thing instrument too.

Overall

Obviously all of these example pieces have a lot of snazzy production effects going throughout each, and there are many examples I'm not including here that I use in my "production" lecture. As I said above, I generally like to hear 'interesting' things unfold, and often this means presenting new, unique sonic information during the unfolding of a sound, a phrase, or a piece. The Wendy Carlos Clockwork Orange example above is a good example of what I mean. All of Wendy's work has this in operation -- her Switched on Bach pieces are masterpieces of subtle (and sometimes not-so-subtle) changes to each sound/note during the construction of a piece. Every phrase of the Bach is split into several different patches, unfolding as the phrase evolves. A lot of this is necessary to duplicate the subtle audio alterations that humans make with "real" instruments -- a violin sound is always evolving as it is formed.

Along those lines, check out the following piece by Imogen Heap:

What Heap did that was really cool, though, is to include the instrumental parts only on an extended CD you could download: She won a Grammy for her engineering/production on this CD. Pretty amazing to hear all the subtleties she is putting into the backing tracks.



So there you have it -- my "production" lecture in highly compressed(!) form. I hope it helps with some happy listening. There are many pieces I haven't included here, and I haven't said everything I normally say about the ones I have included. More listening ahead!