Monday, June 9, 2008

Inside the V-Synth, part 3

In this installment of Inside the V-Synth, we'll look at the V's onboard effects. The effects are divided into three blocks: multi-effects (abbreviated MFX on the panel and in the manual), chorus (which also includes some fixed delays), and reverb. Each of these blocks has a dedicated on-off button on the panel. All effects parameters are stored as part of the patch. One effect can be selected for each block at a given time.  The figure below illustrates how signals are routed to the three blocks:

All three of the effects blocks has a send that comes from the mix of all of the VCAs for each voice. (Any patch whose output is routed to the direct outs bypasses all effects, as in the case of the JD-800 and many other earlier Roland synths.) In addition, the MFX can send its output to the chorus and reverb inputs, and the chorus can send its output to the reverb input. It would have been nice if some alternate routings had been provided, but this routing works for most applications. Note that the signal path here (as in everywhere else in the synth) is stereo all the way through, but there are some MFX effects that will mix their inputs down to mono to run through the effect.  So the routing is straightforward, with only a few choices. But the MFX has lots of choices! To wit:

Multi-Effects Types

Filters and Equalizers: This category provides a 4-band parametric equalizer; with two sweepable and Q-variable mid-bands. (Note that there is also a 3-band parametric equalizer which effects all patches; its settings are stored as part of the system settings. To avoid confusion, I recommend keeping the system EQ set flat for most purposes.) There is a 2/3-octave graphic equalizer, which however for some reason only goes down to 180 Hz, which limits its usefulness. Also available are a resonant filter with its own LFO, an auto wah which can be driven by an LFO and/or an envelope follower applied to the input, and something called the "isolator and filter". This is an interesting chain of filters: it begins with a 3-band EQ that works like the tone controls on some guitar amps: the three bands cut severely when turned all the way to the left, and they overlap such that if you turn all three of them all the way down, it completely kills the output. Next are a low-band and mid-band "antiphase". When these are on, they put a bandpass tap on each of the stereo channels, invert the output, and then add the result back to the opposite channel. With center-panned mono signals, this just causes cancellation, but with mono signals that are panned to one side, or stereo signals (from a patch using a stereo PCM sample, or a stereo external input), the results range from weird phasing effects to stereo field expansion, depending on the nature of the input. This is followed by another filter, which can select 2-pole or 4-pole response, lowpass/highpass/bandpass/notch curve, and variable resonance. The chain ends with a bass cut/boost filter.

The final, and unusual, filter effect is the "humanizer". This is basically an envelope filter, following the amplitude of the input signal, but in place of a conventional filter it uses a formant filter tuned to various vowel sounds. As the signal level increases or decreases, the formant moves from one vowel to another, with the two endpoints settable. There is also an LFO to drive the formants, an overdrive effect preceding the format (which can make the effect more intense), and a 2-band EQ following.

Distortion effects and guitar amp simulators: There are two effects, an overdrive/distortion effect, and an amp simulator effect. Both are expanded versions of algorithms that are also available in the COSM filters. The former provides two distortion modes, four speaker cabinet simulations (ranging from a small practice amp to a double speaker cab), an option to mix the input down to mono prior to applying the distortion, and a 2-band EQ on the output.

The guitar amp simulator provides an extensive selection of options. The effect chain begins with a noise gate, with variable threshold. (Why, one might ask, does one need a noise gate inside a synth? Well, you might be using a real genuine guitar plugged into the external input. And it might be picking up real genuine AC hum...) There's an interesting selection of amp models with interesting names, like MS1959I... When you select the parameter, there's a little icon of the amp that gives the game away. Roland obviously didn't have permission to use other manufacturers' names, but the little icon for the above makes it clear that it's supposed to be a Marshall. There are others which I'm pretty sure are supposed to be a Vox, a Matchless, a Mesa Boogie, and a couple that I didn't recognize. Plus, there's the one amp whose name they could use, the Roland Jazz Chorus-120 (unfortunately, its wonderful built-in chorus effect isn't included in the model), and several Boss distortion stomp box models. There's a selection of tone controls typical of what you would find on a guitar amp, followed by a speaker sim with choice of 13 cabinet types, and (it's almost getting silly) a mic model with a parameter to vary how far the imaginary mic is from the imaginary cabinet! You can also mix the mic and direct sound.

Delays: The MFX provides an extensive selection of delays (don't forget that there are also a few available in the chorus). The most straightforward are the stereo delay and the multi-tap delay. The stereo delay is a straight-up digtal delay, with up to 650 ms of delay time, or 1300 if you put it in mono. (You can also sync it to the synth's clock source, by selecting a note value.) In stereo, it acts as two separate channels; there is also an "alternate" mode which cross-straps the feedback from one channel to the other channel. Low and high frequency damping is available in the feedback loops, plus a 2-band EQ on the output. The feedback can go very high in either the inverting or non-inverting direction; I have been almost been able to push it into runaway, but not quite. (You can hear it starting to slap, but then it fades out. I'm guessing there is an anti-runaway provision in the code that applies damping when internal overflows start to occur.) The multi-tap delay provides five mono taps (it mixes its input down to mono). Each tap can go to 1300 ms, or be synced to clock, and can be panned L-R in the output. Feedback is from the output of the combined taps back to the mono input.

The reverse delay tries to create a reverse-echo effect. It works well on short percussive sounds, and on continuous sounds. Sounds that vary a lot, like vocals, get it confused and cause pops in the output. The band pass delay is like the multi-tap delay, except that it has a phaser effect preceding the delay line, and each tap has its own bandpass filter on the input.. The tape echo simulator does a pretty good impression of a Roland Space Echo, down to the tape saturation, and wow and flutter as might be caused by a lopsided capstan. The "vocal echo" is supposed to simulate the slapback reverb from a karaoke machine; it's basically a mono delay with a lowpass filter on its input.

Choruses and Flangers: This rather large category starts with a basic digital stereo chorus/flanger, with pre and post EQ and a pre-delay. The "hexa chorus" is a six-stage chorus whose outputs can be spread across the stereo field to produce pseudo-stereo effects (it mixes its input to mono). The "space chorus" is a re-creation of the famous Dimension D spatial effect. Since I've not had access to a real Dimension D, I can't say for sure how accurate it is, but it does seem to re-create the descriptions I've heard of it. There's also a re-creation of the Boss CE-1 Chorus Ensemble stomp box, which is actually the closest of all of the chorus algorithms to what we think of as the classic Roland chorus effect. Flanger algorithms include a simulation of the SBF-325 bucket-brigade based analog flanger, an algorithm that simulates several of the Boss stomp box flangers, and an odd "step flanger" whicn advances the effect in a stair-step pattern and can be synced to the V's tempo clock.

Phasers, Rotaries, and Trems: Two phaser algorithms are provided, one that simulates an analog phase shifting circuit, and one that is purely a digital implementation. There's a rotary speaker emulation (which behaves rather bizarrely if you attack it with the Time Trip pad), and tremolo/auto pan effect.

Lo-Fi: This rather strange category includes a bit crusher, which can both reduce bit depth of samples and down-sample to lower sample rates. It also includes algorithms that simulate the distortions and frequency losses characteristic of AM radio transmission and phonograph. The latter goes to some considerable trouble to provide all conceivable means to mess up your sound, including simulations of the three most common types of phonograph disc, surface noise, and wow/flutter (pitch irregularities that occur due to a record being warped and/or out of round, or due to worn or poorly made parts in the turntable drive).

Other Single Effects: The stereo pitch shifter can shift to +/- one octave, with a pre-delay and post EQ. It provides five choices of "fineness"; higher values produce less distortion but at the cost of additional inherent delay. (Unlike the COSM filter effect, this is a true pitch shifter, not a frequency shifter.) The "pseudo stereo" effect creates a simulated stereo field from a mono source.

Effects Chains: There are a number of selections that combine several types of effects in series. The dynamic processor effect combines a compressor/limiter, a spectrum enhancer (which adds second-harmonic content, somewhat like an Aphex box), an EQ, and a noise gate. This one is stereo all the way through; the "Vocal Multi" is similar except that it mixes its input to mono, and it replaces the noise gate with a delay. "Guitar Multi", which also mixes its input to mono, provides compression, distortion, chorusing, and delay in sequence. "Bass Multi" (mixes input to mono) provides compressor, distortion, EQ, and chorusing. "Electric Piano Multi", which appears to be designed mainly to be used with Rhodes pianos or similar sounds, provides spectrum enhancement, phaser, chorus, and tremolo/pan. This one is peculiar in that the spectrum enhancer and trem/pan are stereo, but the phaser and chorus mix their inputs to mono (bypass remains stereo) and provide stereo output. "Keyboard Multi" provides ring modulation (against an oscillator which is internal to the algorithm), EQ, pitch shifter, phaser, and delay. The latter three effects are mono with outputs pannable across the stereo field; the bypass remains stereo. Finally, there are chain effects that combine various combinations of distortion, chorus, delay, phaser, and spectrum enhancement.

Chorus and Reverb Types

Whew... the chorus and reverb blocks are simpler. There are five chorus algorithms, all but "Feedback Chorus" of which are fairly standard. "Feedback Chorus" is kind of weird; I'm not sure how to describe it. It's sort of like a phaser with, say, 16 or 24 stages. Or, you can choose a flanger or two short delays. The reverb block provides three hall reverbs (Hall 1 includes a chorus effect), three rooms, a plate reverb simulation, "garage", and "non-linear". The halls and Room 1 all sound kind of metallic to me; maybe I haven't spent enough time playing with the EQ parameters. The plate reverb is very decent, if you like the plate sound in general (I'm not a big fan of it). "Garage" pretty much sounds like what you think it sounds like. "Non-linear" is interesting; as best I can figure, it applies an arbitrary four-segment envelope (totally user configurable) to each sample of the reverberated sound. Some of the things it can do almost come out sounding like granular synthesis. If you want to render a perfectly good sound utterly unrecognizable, this is the algorithm to use.


The modeling capabilities of the MFX block go considerably beyond the standard built-in effects found on most synths.  The chorus and reverb blocks are useable, although for studio use an outboard reverb with more algorithms and more flexibility might be preferred.  The V's synth architecture provides a set of dedicated DSPs for effects processing (unlike all of the voice processing, which uses a shared pool of DSPs), so switching effects in or out has no impact on available polyphony.  One thing that's a nuisance for multi-timbral sequencing (via an external sequencer) is that all parts are routed through the same effects; the effects settings are determined by the patch selected in part 1.  The way around that is to direct the other parts to the direct outs (which bypasses the effects), but that's a setting that has to be made in the patch.  

In Part 4, we'll look at the V-Synth's innovative selection of performance controls.

No comments: