Tuesday, May 27, 2008

Inside the V-Synth, part 1

So, now that I've had a few hours with the V-Synth, I will pass myself off as an expert. Actually, I did spend a fair amount of time studying the manual and online material before the synth arrived, so I was able to get going on it pretty quickly.

Roots of the V-Synth

Considering that the V is a bit out in left field compared to most of Roland's other offerings over the last decade, the question can fairly be asked: where did this idea come from? The roots of it can be seen in the JD-800/990 offering from the early '90s. That was Roland's first serious attempt to design a digital synth that would be attractive to synth programmers and sound designers. However, at the time it was introduced, it did not sell well. Part of this was the timeframe in which it was introduced (the synth market had turned heavily to romplers at the time). However, the attempt at building a knob-per-parameter digital synth turned out to be intimidating to some customers, as well as making the unit physically large. And, in the end they still had to put the effects, MIDI, and system parameters in menus. It helped that the 800 had two displays which were fairly generous for their time, and the 990 had an even better multi-line text and graphics display (derived somewhat from the S-series of samplers). Despite the flaws, it was state-of-the-art for a digital synth user interface at the time.

Now, the V-Synth came about as a compilation of everything that Roland had learned about synthesis and signal processing since the JD-800 was designed. And that was quite a lot -- the Variphrase technology which allows treating time and pitch as independent variables, the improved sample encoding and storage technology, and the COSM modeling filters, to name a few. Someone within Roland obviously desired to take all this new technology and make a worthy successor to the JD-800. A problem arose immediately: goven the complexity of this new synth, it was simply impossible to attempt the knob-per-parameter approach. However, the designers didn't want to send the sound designer down a twisty little maze of menus (having learned their lesson on that from the D-50). So they tackled the problem three ways:

1) They provided a set of dedicated controls for the most commonly tweaked parameters, including the oscillator tuning, the filter cutoff/center frequency, the ADSR values, and LFO rates. They also provided buttons to turn each component and effect block on and off -- not so much for performance as for convenience in patch editing, where it often helps to turn off some things in order to hear the impact of changes in other areas. For everything else, they added a large touch screen which graphically displays all of the parameters for a particular functional block.

2) They added an extensive set of performance controls, including some unconventional ones. Of all the polysynths I've ever seen, only a few very expensive models such as the McLeyvier and the Hartmann Neuron approach this same territory in performance controls.

3) They created an extensive modulation matrix to map these controls, and incoming MIDI messages, to a large slate of real-time-modifiable parameters.

We'll get back to the performance controls in a later installment, because some of them are difficult to understand the function of until you are familiar with the synthesis architecture. For now, we'll stick to what makes the noise. To start the demystification process, note this: for all of its complexity, the V-synth is still a subtractive synthesis machine. There's no exotic synthesis method at work here; although the oscillators and filters are far advanced from your old Juno, they are still basically arranged in the classic VCO-VCF-VCA signal chain.

The Voice Architecture

Each V-synth patch can make use of two digital oscillators, two COSM filters, a time-variant amplifier (digital equivalent of a VCA), plus three effects blocks and a "mixer" which can combine the two oscillator signals in various ways. For each patch, one of three "structures" can be chosen; the structures differ in which the filters are placed with respect to the mixer. Structure 1 mixes the two unfiltered oscillators, and then flows the mixed signal through the two filters in series. Structure 2 places one filter on oscillator 1's output prior to the mixer, with the second one remaining after the mixer. Structure 3 places a filter after each oscillator and before the mixer.  The photo below shows the three groups of buttons which graphically illustrate which voice elements are active in the selected structure:

Control groups for the voice elements: top left, structure selection and element on/off buttons; top right, effects on/off; lower L to R: oscillators 1 and 2, COSM filters 1 and 2, TVA envelope.  

The Oscillators: PCM Mode

The oscillators allow a choice of three different modes: PCM, virtual analog, or external input. The PCM mode plays samples, as you might guess. The usual choices of looped or one-shot playback are available. The factory load comes with about 700 samples, but you can also add your own. There are two ways to do this. For some reason, it isn't often stated in reviews of this synth that it contains a full-up sampling capability, similar to the S-series samplers, which choices of triggering mode for sampling start/stop, loop point, compressing, normalizing, and trimming options. Resampling is also supported; a note can be played using the full capabilities of a patch, and the result captured as a new sample. In addition, the user can also set markers called "events" inside the sample. Events divide the sample into segments, which can be used to divide the sample across a range of keys, or to play in a mode where playback advances to the next segment each time another note is played. Samples can also be loaded into the synth using the USB connection. In the V's "external disk" mode, its internal patch and sample storage will appear as a removable disk on the computer that it is connected to. (Both Windows XP/Vista and Mac OSX are supported.) Samples are stored in the V as ordinary WAV files, and new ones can be added by simply moving or copying them to the mounted volume.

Before a sample can be used by the V, it must go through a process called "encoding". The encoding process supports the Variphrase capability; it creates a sort of map of pitch changes and variations in the sample. Four encoding algorithms are available, each of which works best on a particular type of source. (If you choose the wrong one, you may get weird pitch anomalies when you try to use the sample in a patch. But intentionally mis-encoding a sample might also be a useful way of creating chaos patches.) One particular algorithm, which works best on solo voice and instrument samples, will also analyze the sample for "formant" information (in other words, information about the resonances within the body of the instrument or the vocal tract of the singer), which can then be varied when the sample can be used in a patch to change the tonal character of the sample.

Once a sample is loaded and encoded, you can use it in a patch.  The patch parameters will allow you to set the loop mode, retune the sample, set up the Variphrase to treat the sample the way you want, and so on.  When playing in PCM mode, the Variphrase provides a number of tricks that allow you to treat pitch and time separately -- you can make time go slower and faster while pitch remains the same, or pitch bend without effecting playback speed, and so forth.  You can make the sample play in rhythm to an internal clock or an external MIDI clock, and you can make the sample play forwards or backwards as desired.  

Virtual Analog Mode

In the virtual analog mode, the usual analog waveforms are available: sine, square, pulse, sawtooth, ramp, white noise. These are probably computed rather than sampled waveforms. 
(Note: the 2.0 OS added a number of waveforms and significantly enhanced the virtual analog capability.  If you get a printed manual with a used V-Synth, it will probably describe the version 1.0 OS capability.  Download the revised manual from www.rolandus.com -- it's free.)   There are the usual waveforms that one expects from an analog VCO, including sine, square, pulse, triangle, white noise, and sawtooth, with variations -- a waveform labeled "Juno" is a digital simulation of a sawtooth wave from the Juno-60.  "LA-Saw" and LA-Square" are, as you might guess, re-creations of the sawtooth and square wave algorithms from the D-50.  There is the "supersaw" waveform which is all the rage in techno these days; it's a simulation of several stacked oscillators all generating sawtooth waves with slight pitch offsets.  Two waveforms are labeled "high quality" saw and square; I'm not yet sure what the difference is between these and the standard saw and square.  There are subtle tonal differences, but I'll have to put the scope on them sometime to see what the differences actually are.

The really intriguing types are the cross-modulation and the feedback oscillator.  Selecting X-mod on oscillator 1 causes osc 2 to cross-modulate it.  Since it still works even when osc 2 is a PCM waveform, very complex waveforms are obtainable.  The feedback oscillator uses some kind of iterative algorithm to create sounds that can sound like feedback through a guitar amp or PA system.  They can also sound like whistling, various wind instruments, a glass harmonica, or a variety of sounds that might usually be expected to come from an additive synth.  

A sub-oscillator can add a square wave at the  same pitch, or 1 or 2 octaves lower.  The sub-oscillator can be detuned.  There is also a parameter labeled "impact" which can add a spike to the start of the waveform to simulate a percussive attack.  

External Input Mode

Finally, the sound source can be an external input, via rear-panel jacks. This allows the V-synth to be used as a real-time processor for signals from other instruments. As an interesting example, one of the algorithms in the COSM filters is an amp simulator, complete with distortion, overdrive, and several speaker cab choices. You could use the V-synth as a rather unlikely guitar preamp! There's a gain control on the panel for external inputs, but I haven't yet tried it to see if it can interface directly with a guitar. Likely the inputs are low impedence, so if you wanted to actually do this trick, you'd probably still need a direct box or some such to prevent the synth's input from loading down the guitar pickups.  Do note that you still have to play a note on the keyboard (or via MIDI) in order for sound to come out!  

Oscillator Parameters and Modulations

There's the usual pitch tuning and keyboard tracking settings. Unfortunately, the very useful "stretched" and "shrunk" keyboard tracking options from the JD-800 aren't present. However, it does have 50% tracking so that playing quarter-tone scales is possible, as well as zero and negative tracking. One feature that did carry over from the JD-800 is the pitch randomization parameter, which can add a desired amount of pitch uncertainty to each note. Each oscillator has its own dedicated LFO, with the usual choices of waveforms. Keyboard tracking and the LFO can be assigned to nearly any parameter. For PCM waveforms, time is also a parameter that can track the keyboard and be modulated by the LFO! Remember, Variphrase makes pitch and time independent. Additionally, if the SOLO sample encoding mode was used on a PCM sample, formant becomes yet another parameter that can track the keyboard and be modulated by the LFO. And, every one of these parameters has its own ADSR envelope. Finally, and incredibly usefully, each oscillator has its own TVA (in addition to the TVA at the end of the patch signal chain), with a dedicated ADSR envelope of course. There's enough envelopes inside this synth to start a post office.

In Part 2 of this series, we'll look at the COSM filters, plus the oscillator mod mixer and the TVA.

Did ya miss me?

For the past two months, I've been in a major crunch at work, which is why I haven't posted. But in the meantime, I got a new toy! Here's a couple of glamor shots. The last one gives it away: