In the final part of his short series on pushing back the boundaries of effects processing, Paul White explores many different applications of audio filters, as well as exploring the possibilities of granular synthesis. This is the last article in a two‑part series.
One effect that has migrated from the synthesizer world is filtering. The simplest synth filter has a low‑pass response with variable resonance at the cutoff point, however, there are a number of other filter types which can be useful, and details of all the different filters can be found in parts four and five our Synth Secrets series, in SOS August and September 1999 respectively. Just running audio through a synth‑type filter can be fun, particularly when you experiment with higher resonance settings, but things get more interesting when you manipulate a filter in real time. An envelope is often used to do this, and the trigger that this envelope requires can be derived from the audio signal itself.
The least complicated approach is to use the envelope generator to control the filter's cutoff frequency, and to cause the envelope to trigger whenever the incoming audio signal exceeds a certain threshold. This works well when sounds to be processed are separate and have clearly defined starting points, but is less effective for sounds that overlap, as triggers can become less reliable.
Remember that the envelope above (or a different one triggered in the same way) can also be routed to control a filter's resonance, for extra variety. What's more, other modulation sources can be used in place of an envelope generator. LFOs can be fun in this role, either used free‑running or with their rates sync'ed to tempo, and envelope followers can also be used. The latter could control the filter cutoff frequency according to the level of the input signal — in other words, the higher the input signal level, the higher the filter frequency (or vice versa if that's how you want to set it). The sound this creates can often be musically useful, though where the initial sound is modulated in level, the resulting undulating envelope can lead to the filter opening and closing in a unmusical and seemingly erratic way.
Where the incoming sounds are from a sampler or a MIDI synth, then MIDI triggering and control of any audio‑processing filter can be employed, and clearly this will work consistently whatever sound is being processed. However, it is important to note that, in a typical synthesizer, each voice will have its own filter while, in the case of an external filter module, all the voices will be processed via the same filter — this means some decision has to be made as to how the single processing filter will behave when a new MIDI trigger is received. Should the filter envelope start again as each new note is played or should it only reset after all keys are released? Some units make this choice for you while others provide switchable functions. Other points to note are that stereo filters are necessary for stereo signals and that filters should be inserted into the signal path, rather than being used in a send/return effects loop configuration.
Stand‑alone filters can be used in a number of different contexts to add interest and movement to a sound. An obvious application is to take an otherwise filterless synthesizer (such as a Kawai K1, Korg Wavestation or Alesis Quadrasynth) and treat its output with a filter. It's also very common to use filters to process sample loops and, in this application, the automation offered by many software plug‑ins and MIDI‑controllable hardware units is very attractive. Complex, tempo‑related effects can be created within your sequencer and then copied to as many bars as are required.
We've had resonant high‑pass, low‑pass and band‑pass filters ever since the first analogue synths, and though they're still widely used today, there's a lot more that can be done using more sophisticated filter types. We've already seen how you can use the complex filtering of a vocoder to create interesting new sounds, but that is only the start. There is a great deal of mileage to be had from processing different frequency bands separately (see 'Divide & Conquer' box on page 124)
One particularly complex multi‑band filtering effect is available within Emagic's Logic Audio — the Spectral Gate plug‑in. Although the documentation doesn't make it clear exactly how the filters are configured, a few minutes playing with the controls demonstrates the range of effects that can be created: everything from the more obvious filtering characteristics to sounds that appear almost resynthesized. It's very easy to create metallic, electronic‑sounding textures from quite conventional input material, though, as with all filtering, the more harmonically rich the input, the more interesting the output.
While on the subject, it's also worth trying out the 'Convolve' process in Bias Peak or any other software that offers the facility. This sounds to me like a filter‑based effect, and it allows the characteristics of two sounds to be merged in order to produce a new sound sharing characteristics of both. This process is often used in sound design to combine sounds in unusual ways.
Conventional delay and echo effects are a mainstay of music production, though last month I suggested a few ways to make these more interesting. A further way of adding interest to effects is afforded to us by the recording process, which enables us to make use of negative time. By this, I mean effect sounds which are audible before the sound that they process even starts playing — something that the physical laws of real life don't allow, unless in close proximity to a concentrated source of tachyons! Once something is recorded, temporal rules can be broken. Sounds can be reversed, they can be treated with reverb or delay that starts before the sound itself, or reverse reverb can be added to a 'right‑way‑round' sound. This latter trick used to be popular during the '70s for music production and is still used extensively in film work to create demonic voices.
When analogue tape was the standard recording medium, reverse reverb was accomplished by playing a tape backwards, feeding the desired track through a reverb unit, then recording the reverb to a spare track. Once the tape was replaced on the machine the correct way around, the reverb track would start playing before the track it was derived from, with the reverb sound's envelope building up slowly in a suitably eerie manner. The same trick can be achieved in a tapeless environment (such as within a MIDI + Audio sequencer) by reversing a section of audio, adding reverb or delay, bouncing the processed result to a new track, then reversing both tracks again. I covered this procedure in some depth in SOS December 1998, for those who'd like to try it. For a more ambitious project, try one of the chopping techniques described last month, using reversed sections of sound or effect for some of the segments.
Of course, the other thing that's very easy to do in a tapeless environment is set up a conventional reverb, record this to a spare track, then slide it a beat or two ahead of the track being processed. This produces reverb that's the right way around, but which still comes before the sound that supposedly created it — you could almost think of it as negative pre‑delay!
One of the current buzz words in sound creation is 'granular' synthesis, which simply means taking very small slices of sound, often from very different sources (or from different times within the same source), and then joining them together to create a new sound. The difference between granular synthesis and wavetable synthesis seems to be mainly the length of the individual segments used. Granular segments may only be a few tens of milliseconds in length. Most of the results I've heard from granular synthesis remind me of a day out at the dentist, but for those on the cutting edge of techno, it might be exactly what is needed. I don't know of any granular effects boxes, but it should be possible to fake your own using a sequencer if you have the patience and a suitably sadistic mind. There are also software packages that can help, such as CDP's GrainMill.
At its simplest, you could set up two audio tracks with the level automation programmed to gate the sound on and off very rapidly. This is easiest to achieve using your graphical editor to draw nice neat square waves linked to MIDI continuous controller seven. By arranging it so that one channel is off while the other is on, and vice versa, you should end up with something very crunchy indeed. You should probably be aiming for something like 64 transitions per bar or more, so make use of the copy and paste facilities of your sequencer when creating the granular 'chopping' templates. When you get a good one, save it in your default song so that you can use it with other sounds. If you have a plug‑in filter that offers a sample and hold facility, you can get some pretty granular‑sounding effects out of it if you can persuade the sample‑and‑hold rate to go up high enough. If you can get it fast enough so that the individual steps blur into a continuous noise like somebody cleaning their oven with an angle grinder, you've probably got it about right!
Another way to approximate a granular type of effect is to use the manual digital scrubbing facility on a piece of recording hardware or software, and then record the results. Often you'll find there are two scrubbing modes, one that works rather like tape varispeed and another that constantly plays back a tiny loop of audio as you move through the audio file. It's the second one of these you need for real granular‑style emulation — just move through a sound slowly and linger on any sections that sound particularly interesting. Again, sample anything that sounds useful.
There are some spectacular granular treatments in Native Instruments' Reactor and Dynamo that work on short samples of sound, so you can load in your own audio clips and mangle them mercilessly before recording the results as a separate audio file. It's more of an instrument than an effects processor, but if you get the results you want, does it really matter? What's more, because VST 2 virtual synths can be automated, you can set up your granular effect within a song and know that it's always going to come back sounding the same each time.
While there may not be as many new effects as we'd like, it's often possible to create some extreme‑sounding treatments by either combining existing effects or by making existing effects dynamic in some way. Those reclusive people who design the sounds for sample CDs often set up huge loops of sound processors and delays that feed back on one another, then they record the output and select the best bits. If you have a few pieces of outboard, give this a try using lots of delay feedback so that everything is just below the point of breaking into self oscillation. Sometimes you don't even need an input signal to set the whole thing off!
By creating repetitive changes that happen too fast for the human ear to perceive, it's easy to emulate granular synthesis or complex modulation where the result is often dissonant and mechanical sounding. What counts is that the result is musically useful, and even treatments that result in atonal mayhem can sound good in context, especially when used as part of a rhythm track. Don't rule out cheap effects pedals either, as some of these produce surprisingly musical sounds, even if they don't have the best technical spec. My best advice is to break a few rules. If a box says 'guitar processor', try it out on vocals or drums to see what it does. You have nothing to lose but your sanity!
Although there are a few genuinely innovative effects available, most effect treatments are based on combinations of distortion, filtering, delay, reverb and pitch‑shifting. One way to get new mileage out of these old effects is to make them dynamic — use MIDI controllers to change the character of the effect as the sound progresses. For example, what starts out as a flange might end up as a simple repeating delay — all you need to do is reduce the modulation depth, increase the delay time and reduce the amount of feedback over the space of a few seconds.
Most sequencers with graphic editors make this kind of thing fairly easy, but there are also dedicated hardware processors capable of creating morphing effects. Some have morphing built in as a specific function, while others allow you to assign a number of parameters to a pedal, then change them all simultaneously during performance — as the pedal position is moved, the parameters change and the effect morphs. Sometimes you'll find that what happens in the middle of the transition is more interesting than what happens at either end, so allow yourself time to experiment if this kind of thing appeals to you.
A further avenue worthy of exploration is multi‑band processing. If a crossover type of filter is used to separate the audio being processed into three frequency bands, roughly corresponding to bass, middle and treble, then each band can be processed separately. Applying distortion to the individual bands is quite instructive as the sound holds together much better than when distorting the entire signal all in one go. At lower levels of distortion, the signal sounds compressed and more energetic, making this a good choice for treating drum loops, and even when the sound starts to get noticeably crunchy, it still retains enough integrity to be useful. Try this process on bass sounds as well as drums, and also try varying the amount of distortion in each frequency band. The Steinberg Quadrafuzz VST plug‑in offers four‑band distortion and demonstrates this principle very nicely. Another multi‑band processor to check out is the Electrix MoFX, which provides a distortion block, and a three‑way crossover to which flange, tremolo and delay effects can be applied in various ways.
If you're using a system such as TC's Spark that allows you to combine VST plug‑ins rather more flexibly than most systems, you could try applying different processes to the various frequency bands. For example, compress the high end to get more high‑frequency energy, distort the bass end for more deep‑down power and, while you're at it, add delay or modulation to the mid‑range. It is also worth placing a high‑cut filter after the bass‑band distortion to prevent too many harmonics in the mid‑range. The block diagram below illustrates this configuration.
Intonation correction such as that provided by Antares Auto‑Tune was never designed as an effect but, by adjusting the parameter settings, you can still coax some very interesting sounds out of it. The famous vocal sound on Cher's 'Believe', for example, may be evoked using Auto‑Tune, just by setting it to correct at the fastest possible speed. This effectively quantises the vocal pitch to the nearest note in the selected scale, and is becoming a more and more fashionable sound with each passing Top 40. If you're wanting to try this yourself, be aware that anything more than mild vocal vibrato can cause unwanted trilling. Not to say that this can't be turned to your advantage and used creatively: the trills or slurs (depending on the correction speed chosen) produced by heavy vibrato and pitch‑bending can be great for adding a hint of eastern promise. What's more, this trick isn't only useful for vocals — it can also be great for lead guitars and synth solos.
Pitch correctors can also get really fun when they allow you to feed MIDI note information into them from a keyboard or sequencer, so as to force sung notes to new pitches. With a fast correction speed, this creates the familiar quantizing effect, but because the degree of pitch‑shift depends on the difference between the original note and the target MIDI note, you may also hear more obvious pitch‑shifting artifacts, which can be used creatively. Using a slower correction speed causes the vocal or instrumental line to portamento to the new pitch — an effect which can be good within pseudo‑world music.