The Return of the Modular Synthesizer
Digidesign continue to add to their innovative range of synthesis programs which use the flexibility of computers for sound creation and the hardware of MIDI samplers to replay the sounds. Paul Wiffen finds he loves the former but hates the latter.
With Turbosynth, Digidesign continue to add to their innovative range of synthesis programs which use the flexibility of computers for sound creation and the hardware of MIDI samplers to replay the sounds. Paul Wiffen finds he loves the former but hates the latter.
When I first heard of Digidesign's original computer synthesis program, Softsynth, I was impatient to get to grips with it. Talk of all the harmonics available with independent envelopes had me salivating. And when I finally did get to grips with it, it was even better than I thought. Previous experience of 'additive synthesis' had been restricted to horrendous implementations like that on the Series II Fairlight, which took ages to perform its calculations, or real-time versions like the Oscar and Prophet VS synthesizers, which were good as far as they went but didn't allow for independent enveloping of harmonics. The beauty of Softsynth is that the software implementation doesn't place the same restrictions on facilities that hardware-based additive synths do. This means that during the process of sound creation you can draw on much greater resources. However, there is a penalty to pay. To use the computer-created sounds on a keyboard, you have to use MIDI to transfer the sound across, either using the System Exclusive format of a particular machine or the MIDI Sample Dump Standard.
As a result the sounds, once converted into a stream of sample data, become like all samples, fixed and immutable. Of course, you can still filter and envelope them (always assuming your sampler has filters and envelopes), but you cannot induce changes in the levels of individual or grouped harmonics from velocity or aftertouch. You will also find that the duration of the sounds halves when transposed up an octave and doubles when you go down an octave. That is why the latest generation of hardware additive synths like the Kawai K5 (which I raved about in the December 1987 issue of Sound On Sound), which creates the envelopes of the harmonics in real time, are always going to be a better bet. Sounds don't change length as you transpose them up and down, and realtime interaction through touch-sensitivity and other performance control allows you to breathe life and expression into additive sounds.
Digidesign's new program, Turbosynth, can't suffer from these comparisons, because the sort of flexibility it gives you hasn't been available since the days of wall-to-wall modular synths. They were monophonic, unstable in both pitch and timbre, and you could never retrieve a patch again once it had drifted or been changed. The only synths which have come close have been monsters like the Matrix 12 or the Voyetra 8, which use digital techniques to provide for flexible routings and have expensive price tags to keep them out of the hands of impoverished musos like me and thee.
Now, at last, the sort of flexibility that was possible in the good old days is now available through the power of the Apple Macintosh computer (and shortly the Atari ST). The wonderful thing about software-based synthesis is that you don't need a piece of hardware for each stage in the operation (oscillators, filters, VCAs, etc) and you don't need to multiply these components by the number of voices you need for polyphony. When you open up a new file in Turbosynth, all you see is the output. This is because the number of available modules is not fixed, you simply grab the icon for each part of the sound (oscillators, filter envelopes, amplifier envelopes, etc) and open up as many as you need. In one particularly over-the-top Turbosynth patch I created, I managed to use 15 oscillators (mind you, I did run out of memory trying to create envelopes for each of them - in the end I compromised and made do with just five envelopes).
Turbosynth gives you 12 different modules that can be used to generate or modify sounds, and any combination of these (including several of each type) can be patched together in any configuration to create a unique synthesizer architecture for each patch. For one patch, you might want three oscillators mixed together through one filter envelope and one volume envelope, whereas on another patch you might want a more complicated arrangement with individual enveloping on six oscillators before they are mixed together and filtered. You can also add sampled sounds (MIDI sample dumped into the Mac) or use waveforms extracted from such samples on your oscillators.
The Oscillator module allows you to select from 15 preset waveform icons (including sine, triangle, sawtooth, and square as well as more exotic forms) and place the chosen waveform at any point on the oscillator's time scale. If this is the only waveform you select on that oscillator, then the oscillator will produce that waveform at the required pitch for the duration of the sound. However, if you select another waveform and place it elsewhere on the oscillator's time scale, then the program will automatically crossfade between these two waveshapes in the given time.
You can place as many of these waveforms as you like within the time set up for that oscillator to create a constantly moving effect. The only things I have come across in the past that were this much fun on a hardware synth were the PPG Wave's ability to move about within a wavetable via an envelope or LFO, and mixing four oscillators via the joystick or envelope on the Sequential Prophet VS. The Turbosynth system is much more flexible, as you are not limited to moving between just four waveforms (as on the Prophet VS) or the waveforms which happen to share the same wavetable (as on the PPG). I got completely carried away seeing just how many waveform transitions I could make in a second (if you set the waveform icons to overlap, you can fit innumerable changes into a very small time interval, giving you the same kind of effect as guitar distortion or sync sounds) and listening to the great sonic movement that is possible with more spaced out transitions.
For me, this is the best thing about Turbosynth. Many of the sounds that I made up were so good that I found I didn't need any further enveloping or filtering. On most synths nothing sounds more dull than the straight sound of one oscillator unfiltered, un-enveloped and not even chorused or detuned with another. But here I was making really exciting sounds on the Mac without even opening up any other modules.
As if this continuous waveform mutation isn't enough, you can also modify any of the 15 preset waveforms to create new ones of your own using the waveform window. Here you can squeeze, squash, stretch, slide or saturate the waveform at will, using various modification icons. If you prefer to get in there on a minute scale, you can pick up the pencil icon and simply redraw. Any waveforms that you create like this can be saved to disk so that you can call them up at any point in the future to use within other patches. These waveforms can also sit perfectly well in a series of crossfades, as can those created by the other method, that of extraction from a sample.
This is done within the Sample module, which allows you to play back samples loaded into Turbosynth via MIDI Sample Dump or any of the System Exclusive formats which the program supports (Emu, Ensoniq, Akai, Roland, etc). Once a sample is loaded, you can have it play back as is; re-edit its start, end and loop points (via a Loop window just like on sample editing programs); or extract a loop cycle from it to use as a new waveshape in the Oscillator modules.
Once you have decided on your combination of sound sources (Oscillators and Samples), then the real fun of routing the sounds begins. You can simply connect all the sounds via a Mixer module, set the relative levels with the sliders provided within this module, and then patch this straight to the output. This is fine, especially if you have really made use of the waveform crossfading we saw earlier. However, this means you have only used about 10% of the program's capability. The modification possibilities are huge thanks to further modules entitled Modulator, Waveshaper, Pitch Shifter, Spectral Inverter, Delay, Filter Envelope and Amplifier Envelope. These can be used in any order and quantity (within the overall memory limitations of your Mac). So, for example, you could envelope a waveform then use this to modulate another, mix the result with a sample and then filter everything. As far as the routing possibilities are concerned, I could take up the rest of the article just describing them, so let's pass straight on to looking at the individual functions of each module.
The Modulator allows for two inputs (modulator and modulated) and three types of modulation - FM, AM and PM, being Frequency, Amplitude and Pitch respectively. Most people are by now familiar with the kind of sounds FM is good at creating (seeing as Yamaha seem determined to keep releasing FM synths until everyone in the world owns at least one). The basic difference between Yamaha's implementation and that of Turbosynth is that the input waveforms on the DX series of synths is restricted to sine and cosine waveforms, whereas Turbosynth allows any of the waveforms or samples we saw earlier to be used as a carrier or modulator. Basically, this avoids the need for 4 or 6 operators like on Yamaha machines. Using more complex waveforms as the starting point means you do not have to perform so many FM operations to build up an interesting sound. In this respect, Turbosynth's FM implementation is more like that of New England Digital's Synclavier, where only two waveforms can be combined but you can make these as complex as you like. Of course, if you want to perform multiple frequency modulations on Turbosynth, then you can always bring additional Modulator modules into play.
As for the alternative types of modulation found on this module, Amplitude Modulation (AM) produces effects that range between tremolo and ring modulation, depending on the frequency and level of the modulator, whereas Pitch Modulation (PM) can be used to add vibrato and other slower frequency effects. Slider icons control the amount of modulated signal that is added to the original signal and the amount of actual modulation. One feature I found fascinating is that you can actually use a signal to modulate itself, by specifying that sound as both input and modulator. One for the experimenters, this module.
The Waveshaper module allows you to tinker with the phase and polarity of the input sound. This is done by specifying a curve which governs the way the signal is subsequently altered. There are eight preset curves provided, which you can alter, or you can draw your own curves. Again, the principle here is to experiment and come up with your own personal applications.
The operation of the Pitch Shifter module is a bit more obvious. It only works on one signal at a time (although you can use several for different signals) and is used to set up intervals or transpose sounds up or down.
This module takes advantage of the fact that all the signals within the program are represented as digital numbers, making it an easy task to completely alter their harmonic spectra. This is done by reflecting the frequencies around the middle of the bandwidth of the sample (set by the sample frequency). The actual operation is controlled by an envelope you draw which represents the mix between the original and inverted signals. Again, the effect is very pronounced and is a new field for the synthesist to get into.
Turbosynth's Delay module acts in much the same way as a stand-alone digital delay unit. It allows you to set a coarse and fine delay time as well as a wet/dry mix. An interesting addition is the ability to invert the phase of the delayed signal (particularly effective if the delay time is very small).
The last two modules will be familiar to most analogue synth aficionados, the Filter and Amplifier Envelopes. These can be utilised to pre-shape the volume or harmonic content of a sound source before you use it as a modulator (as you can insert these modules anywhere in the signal path). However, as most of the samplers into which you will dump your Turbosynth patches have their own filter and amplifier envelopes, you are probably better off leaving the final overall shaping of volume and harmonic content until after you have dumped the sound into your sampler.
The reason for this is that Turbosynth performs its filter and amplifier enveloping by permanently changing the sound data (as indeed it does with all other functions). This means that any time-related functions suffer from the shortening or lengthening process that takes place as you play the sounds back higher or lower respectively. The main things which suffer from this are envelope-related, so the annoyance of this effect can be minimised by using the real-time enveloping on your sampler (wherever possible). But don't let this blind you to the usefulness of envelopes which can be placed elsewhere than at the end of the signal path. This is a skill which has been all but forgotten since the demise of the modular synth, as there has been little choice about where the envelopes go in most synths. It is very effective to use multiple amplitude envelopes to mix different oscillators and samples together, and without the ability to envelope sounds before you use them as modulators, you can only achieve static effects. To introduce dynamic (moving) modulation, you need to vary the level or harmonic content of the modulator over time. This is where Turbosynth's Filter and Amplifier Envelope modules come into their own.
The envelope shapes of both are set up by using multi-stage envelopes, which seem to have as many stages available as you can be bothered to create (simply by dragging lines about). There are eight preset envelopes for you to use as starting points, plus some handy overall modifiers which shift all points of the envelope in either the X or Y axis.
When you have finished routing and modifying your numerous signals, you patch them to the fixed Output module. Here, the final calculations are made by the computer before your sound is ready to be transferred to your sampler. The most important function of this module is to set an overall loop, which your sampler will need if it is going to be able to sustain the sound indefinitely (or at least as long as you choose to hold down a note). This works in exactly the same way as looping within the Sample module, although you will probably need to put in a lot more work at this stage as your sound has probably multiplied in complexity and time-related effects, which make looping more difficult. This is another reason to leave overall Amplifier and Filter shaping to the envelopes on your sampler, as the output loop will prevent you creating a release by data-altering.
Transferring a finished Turbosynth sound to your sampler via MIDI (just like sending samples to the Apple Mac) is not a quick procedure (I thoroughly recommend the RS422 option with Emu samplers, as it is 17 times quicker than MIDI). The problem is that the data transfer rate of MIDI, though plenty fast enough to deal with anything you might play, takes rather a long time to deal with the enormous quantities of data that are involved in samples (which is what Turbosynth converts the sounds to for communication with your keyboard). However, it is well worth the wait, for once transferred the Turbosynth patches really do make your sampler sound like a new instrument.
Using the velocity control of volume and filter on your sampler means that you can add dynamics to your performance (something your modular synth could never do for you in the good old days) and its LFOs and pitch bend will add expression to your performance. But perhaps the biggest boon is that your sounds can be polyphonic. All modern samplers offer a polyphony far in excess of that which was available in the days of modular synthesis, so you will find yourself able to chorus and detune sounds, stack them against each other, possibly even use them in unison mode, which will further enhance the work you have done on Turbosynth itself.
Unfortunately, the fact that sounds are transferred as samples does mean that you may not get a tremendous keyboard range out of them on playback.
Remember that for every octave you go up or down in pitch, the length of the sound halves or doubles (unless you have got a good loop on it). Some samplers actually restrict the sound's playback range to an octave up or down (Emu in particular), but the way around this is to make several different files of the same sound by altering the pitches, and then transfer these across and map them over the keyboard as you would when multisampling.
The Turbosynth program I used (version 1.0) seemed pretty thoroughly debugged, compared to most version 1.0 software which is usually a minefield of potential 'bombs'. In a week's thorough investigation with a variety of different samplers, I only experienced two 'lockups'. Once, when working with a newly-transferred sample and using the Control key with the speaker icon to hear the single cycle loop, the Mac got stuck on the loop and I had to re-boot the program to stop it playing. The only other problem was that, in communication with a Yamaha TX16W, a MIDI receive error generated somewhere in the system (not necessarily on Turbosynth) meant that I had to re-boot the TX16W (a long and tedious process). However, I know how sensitive Yamaha gear can be to MIDI receive errors, and I was unable to make the problem re-occur, so I think it unlikely that Turbosynth was to blame. All in all, I found it very reliable for a computer program.
Once you have realised the limitations of the Sample Dump procedure which Turbosynth uses to transfer its sound creations to your sampler, and have discovered which functions are best left to the envelopes, filters, LFOs etc, of your keyboard, then Turbosynth really becomes a very powerful tool. Of course, it would be wonderful if these limitations didn't apply, but I think a hardware synthesizer which can duplicate all of Turbosynth's functions in real time is a long way off, both in terms of time and money (unless Digidesign's own Sound Accelerator card for the Macintosh II and SE gets around these problems by producing the sound internally).
Until that day, Turbosynth provides a real synthesist's dream, with (almost) endless oscillators and modules to process the sound, all provided for the typical cost of a piece of computer software. It has to be a bargain in anybody's book. I can't remember when I last had so much fun creating sounds.
Thanks to Argent's (01-379 6690) for the loan of Turbosynth and the Macintosh.
Macintosh version: £285 inc VAT. (Atari ST version due soon at £235.)
Evenlode Soundworks, (Contact Details).
Gear in this article:
Review by Paul Wiffen
Previous article in this issue:
Next article in this issue: