Ace programmer Greg Truckell begins a two-part power user's guide to this popular LA expander.

A funny thing happened to me several months ago: I bought a synthesizer I really didn't want. I knew that I didn't want it, you see, because I'm the kind of insatiable twiddler for the likes of whom programming power is everything in sound synthesis. The Roland D110 - the synthesizer in question - was, according to everything I had read and heard, a preset-basher's machine, sporting only a bunch of samples which you couldn't filter, and a couple of bog-standard analogue waveforms; a closed system. There is something about the D110 which says, 'Please don't try to experiment with me, you'd only get really frustrated'. I think it has something to do with the writing on the front panel. Calling a spade a spade has always made sense to me; my ESQ1 has 'Digital Wave Synthesizer' written on it, as has my CZ. My DX7 has 'Digital Programmable Algorithm Synthesizer' - hell, even my Korg MS20 has 'Synthesizer' written on it. Why then does the D110 have 'Sound Module' written on it? What does that tell you?

It doesn't tell you that the thing can be programmed, or that it is in fact a digital wave synthesizer of some considerable power; instead it hints that you should really consider it as a preset machine. Well, you can shove that right up your partial.

So why did I buy it? Simple: I'm a professional programmer, and at the time I was in the market for a new synth. The D110 was selling as fast as suppliers could get them, for all the currently fashionable reasons - it was new, it was multitimbral, and it offered built-in digital reverb. It was serious Linear Arithmetic synthesis to professional standards but without the arm and a leg price tag. It still is.

OK, on with the story. I installed the D110 in my rack, popped open a can, and sat down with the handbook. Now, Roland have always been market leaders in the highly competitive field of utterly incomprehensible handbooks; with the D-series Linear Arithmetic synthesizers, they have set new standards in user-unfriendliness. Maybe somebody did it for a bet? Briefly, they seem to say everything in three different places, in such a way that it never sinks in. I found myself discovering programming features through experimentation, which even after three complete cover to cover readings of the handbook, I was quite unaware of. Imagine my delight at discovering not just one, but four pitch envelopes! They are in the handbook - I checked. Somehow though, I don't think that's quite the way handbooks are supposed to work.

Here's where the plot takes the long-awaited twist. Cynical pro falls passionately in love with preset-basher's machine. They plot to murder the D110's husband, and run off into the hills to have lots of little partials. Let's talk seriously for a minute, though. Have a quick rummage through the features of a D110 sound, and think about it for a little while: four Waveform Generators per voice; up to 12 envelopes, each of them five-stage velocity responsive affairs; up to four filters; four low frequency oscillators; some 'really means business' type keyboard scaling features; more waveforms than you can shake a mouse at, and then some. There are over 240 parameters to a Tone. Think about it some more. That really is a heck of a lot!

In this, the first of two parts, we will concentrate on programming techniques to get the best sounds out of your D110 (or your D10, D20, and to a lesser extent your MT32 or D50). Next month we will cover some application techniques - for which you'll need a sequencer.

LFO SHORTCOMINGS

By this time, cynical pro and D110 have been together for a while, and they are expecting the patter of tiny partials at any time - and that's when the morning sickness starts, and Mr. Pro starts to wonder if LA is all he thought it was.

The D110's shortcomings are worth looking for. Once you know that they are there, you can bear them in mind in advance. Generally speaking, the problem is one of preset modulation routes. Many of those 240 something parameters apply to modulators, such as velocity or keyboard scaling, and you either use them in the configurations preset by Roland, or you don't use them at all. You have no choice about where to send an envelope or LFO for modulation effects; the connections are made, and if you decline to use them just as they are, then you quite simply can't use the modulation sources in question.

Let's turn our attention to the LFOs. Four of 'em per voice - that really is quite remarkable. Even more remarkable is just how very little it seems that you can do with them. Each LFO is tied to its Waveform Generator. Pitch modulation (vibrato) is fine. Amplitude modulation (tremolo) and filter modulation ('wah-wah' - don't you hate that expression?) are both out of the question. Even the fine old Roland tradition of LFO modulation of pulse width is gone, replaced by the slightly wacky velocity control of pulse width. There are, in fact, only 12 parameters associated with all four LFOs put together. Compare that with an Ensoniq ESQ1 or Yamaha DX7, which have at least eight or nine parameters associated with each LFO. Just three parameters cover each LFO on the D110; they only receive six lines in the handbook (page 61). The three parameters in question are rate, depth, and modulation sensitivity. And just as the destination is preset to one Waveform Generator, so the modulation source is preset, to the modulation wheel, and the modulation waveform is the simple triangle.

By now, some of you may have noticed that there is yet another serious omission in the LFO department: no delayed vibrato. However, there's a way around this that is not documented in the handbook (not where I've looked anyway). If you are applying a pitch envelope to a Waveform Generator, then the LFO modulation on that Waveform Generator does not come into effect until the sustain portion of the pitch envelope. A sensible feature - so why have Roland not mentioned it?

Now that you know the limitations of the D110's four LFOs, what can you actually do with them? Apart from the obvious applications in over-the-top solo or brass sounds, they can be usefully employed to fatten up the sustain of a sound by setting the rates and depths to slightly different values - but go easy, otherwise you'll end up with a real mess. It would be very effective to programme the vibrato frequency to some integer multiple or division of the beats per minute (bpm) tempo of the piece of music in question. This is standard, though often unconscious technique, on acoustic instruments. Even if you don't notice the difference for what it is, you will notice an improvement. Sadly, the 0 to 100 range for LFO Rate has nothing to do with rates in Hz, so you're just going to have to use your lugholes until you find the correct setting.

I enjoy the challenge of finding out what a synthesizer can't do, then finding out how to do it. Delayed vibrato was my first conquest with the D110, and I'll remain pleased about that one until somebody telephones me with the relevant page number in the handbook. What about filter and amplitude modulation then?

CREATING FILTER & AMPLITUDE MODULATION

On the face of it, these two really do appear to be out of the question on the D110. Now, not everybody will be shedding buckets of tears over this; however - and I've said this before - it is easier to ignore a feature that you have, than it is to use one that you don't have. Let's suppose that you have these features on your 'other' synthesizer(s). So what do you use them for? Tremolo and timbral modulation, of course. It is worthwhile considering techniques to create these effects in the sustaining body of a LA sound. With most synthesizers, you spend much of your programming time working on the ever elusive 'interesting attack transient', to emulate an acoustic event. With the D110, you have over 100 sampled attack transients, and you can easily afford to use two partials out of a four partial sound to perfect your attack across the range of the keyboard. Much of the criticism levelled at LA synthesis concerns the lifeless-ness of sustained sounds. So let's fix that...

Tremolo (amplitude modulation) is reasonably simple, and to achieve it we need some means by which to reduce the amplitude of one or more Waveform Generators, in a cyclic manner. There are a couple of possible avenues to pursue on the D110. Most simply, some of the 'less successful' loops already exhibit tremolo. Slap one of these, such as any of the Spectrum waveforms (1/101-110), into a patch, and make sure that it is present at a reasonable level - more than half that of the loudest TVA - during the sustain. Also bear in mind that the pseudo-tremolo frequency will increase up the keyboard. Some of the Spectrum waveforms have more high frequency harmonics than others: 1, 3, 7, and 8 are somewhere between breathy and noisy, and could lend tremolo to a flute patch. Some are quite glassy - 2, 6, 9 and 10 - and could play the part of a high tine in an electric piano sound. The remainder, 4 and 5, are rather more unusual; they remind me of the formants from a male voice or from double reed instruments - but they work fine in brass and string patches, too. A little care will be needed with the Spectrum waveforms, as they don't take too kindly to being pitch modulated or transposed much beyond the average range, so watch out for weird harmonics and tuning drift. Unless you are looking for something well and truly 'odd', I suggest that you avoid slow pitch ENV or LFO modulation, too.

OK, so that is pretty basic; but remember that in making a decision to employ a wobbly loop for tremolo purposes, you are turning a wobbly into a feature. It's harder to ignore a feature than a wobbly, and that's why you probably avoid the wobbly loops. Of course, given the right wobbly, and the right combination of other Waveforms, the same principles can be applied to create timbral modulation effects.

ADVANCED EXPERIMENTS

Right, you've passed the basic course. Collect 200 partials, and throw the dice - it's advanced time! At the first level of complexity we have phase cancellation. At the second we have ring modulation. I don't want to get right into the bones of these techniques; that's beyond the scope of this article. Permit me to suggest some areas and values for experimentation, though.

You get phase cancellation on a multiple oscillator (Waveform Generator) synthesizer when there are more than one Waveform Generator tuned to near enough the same frequency. As the Waveform Generators trundle along, from time to time the waveforms line up in such a way that, instead of the usual fat, detuned, chorus-like sound you would expect from your patch, you get a thinner and quieter sound. Sadly, this effect is too subtle to count as useful for tremolo purposes. With the Waveform Generators detuned to a large interval - over 25 cents on the Fine control - the usual chorus sound prevails. With a finer detune, a discernible 'skying' slow phase is heard. Lovely as a pure synth patch but limited to slow timbral sweeps, and an unlikely candidate for acoustic-imitative synthesis.

That leaves ring modulation. Most programmers dismiss ring modulation as a source of metallic, clangorous dins - like they used to get out of their DX7 before they either gave up or got smart and worked at it. Sure enough, slap in a wide interval - we're talking mucho semitones here - and bung on the ring mod, and you get a scrapyard racket. All very impressive and quite fun, but what else can ring mod do?

Try selecting Structure 10, which consists of two synth partials through ring modulation. What we actually need now, to start off with, is some plain vanilla flavour settings for one complete structure. It is a sensible idea, whenever you are about to try some programming experiment, to start off with a vanilla patch - and it makes sense to keep a vanilla partial handy somewhere, either in your D110, or on a disk if you use a computer.

Turn off partials 3 and 4. Set partial 1 WG Pitch Key Follow to '1', Waveform to 'Sawtooth', and Pulse Width and WG Pulse Width Velocity Sensitivity both to '0'. TVF Frequency should be opened to at least '80' with TVF Frequency Key Follow ('filter scaling' to anyone who remembers... ) at '1', which is one octave per octave. A TVF Freq Key Follow setting of '0' does not correspond to no modulation in the natural sense, but corresponds to each note on the keyboard generating pitches which have the same pool of frequencies from which to generate their harmonics. The result is a sound which has less harmonics further up the keyboard.

The vanilla partial should have no added filter resonance, so set TVF Resonance to '0', and the TVF Bias level to '0'. TVA Bias levels should also be set to '0'. The TVF and TVA envelopes should both be simple 'gate' type affairs, with all time values set to '0', and all levels to '100' - but set both T5 values to around '30', to avoid a nasty 'clunk' when the key is released. Incidentally, the envelope times are really rates, but expressed inversely; you can tell that they are rates because you cannot programme a time during which an envelope will stay at any level other than the sustain level, nor can you accurately work with adjacent levels of similar values. The TVA and TVF should have no Velocity Sensitivity or Key Follow modulation, and there should be no Pitch ENV or LFO modulation for the moment. You now have a vanilla partial - put it somewhere safe before it melts.

Copy the vanilla partial to partial 2, and then start playing with the coarse tuning to create different ring modulated timbres. Play a few notes and listen to the effect each new detune interval creates. Incidentally, if you grow tired of holding down the fiddly buttons on the front panel while looking for an interval, or if your mouse has to go to the little mouse's room, try this trick. With Structures 10 to 13, mute the second partial and play a low C. You should hear silence. Still holding down the C, play another note on the keyboard. What you hear is what you would hear if you turned the second partial on and tuned it to the second note you played. This way you can try out various widely separated intervals without denting your fingertips - and it's not even mentioned in the handbook! Using Structures 2, 4, 5 and 7, with the second partial muted, the same technique may be applied; the first note you play sounds, but is not ring modulated. Ring modulation may be brought into effect by any number of successive notes - ring modulation of chords, as opposed to chords of ring modulated notes!

So what do you get from all these huge detunings? Well, more rackets than a tennis club, of course - but now and again things smooth out nicely. And surprise, surprise, the nice bits turn out to follow the harmonic series. Well, most of the time anyway. I know I said I wasn't going to delve into the theory, but I lied. Very briefly, ring modulation concerns itself with the arithmetic generation of sums and differences of harmonics. New overtones (sometimes called 'partials', but don't let that confuse you) are generated which may not be musically related to the perceived pitch. Other harmonics may even disappear. In truth, the theory is the sort of thing that would slow you down if resorted to constantly. Far better to make a guess at the interval, and then home in on it. For your reference, a table of intervals is included in Figure 1. I'll be talking about harmonics again later.

First, time for some pictures. The 2D samples below show what happens to a simple plain vanilla pair of sawtooth partials when they make the sortie through Structure 10's Ring Modulator. Figure 2 shows a few pages of a sample exhibiting considerable modulation in timbre, brought about by ring modulation and merely detuning the partials by 20 cents.

Figure 3 shows a glimpse at the all-but-static waveform generated when two sawtooth waveforms are routed through ring modulation, and detuned by 17 semitones - an inharmonic interval. As you can see, there is no real resemblance to a sawtooth waveform. What you can't see is how much more interesting than a couple of bog-standard sawtooths they sound.

CALL YOURSELF A PROGRAMMER?

It recently came to my attention that there are still some people out there who genuinely believe that they are synthesizer programmers, but who freely confess to not knowing about the harmonic series. This is like calling yourself a pastry chef but admitting that the use of flour is completely beyond your understanding. You could get away with simply throwing together PCM partials that fit the type of sound you want; add Slap Bass (Loop) (1/082) to Slap Bass (1/060), that sort of thing. However, that is no better than eating MTVDs (Microwaved TV Dinners). You wouldn't want to do just that for every meal, would you? Well, my little digital pastry chefs, read on...

Timbre is usually defined in terms of tone colour or 'brightness' of a sound. What gives a sound its particular timbre is the presence of harmonics, or 'overtones'. A sine wave has a pure timbre - it has no overtones, consisting only of the fundamental frequency. Jean-Baptiste Fourier famously deduced that any complex periodic waveform could be broken down into combinations of sine waves at whole number multiples of the frequency of the fundamental. When working with synthesized waveforms, we are usually working with harmonics, but some sounds also contain inharmonic overtones. These are often 'formants' - overtones which are not a whole number multiple frequency of the fundamental, and which appear at the same fixed frequency at all pitches from the instrument. These are largely responsible for the phenomenon of munchkinisation, where a sample played out of its natural range appears unnaturally coloured - and you can bet that there are plenty like that in the D110's PCM Banks.

The presence of many harmonics in a sound will make that sound 'rich'; brass or string sections are typical examples. The presence of high frequency, widely spaced harmonics at high enough levels will make a sound appear 'bright'; bells, for instance. The whole story is a lot more complicated than this. Most sounds change in timbre over time (especially during their attack phase); most sounds also settle to a more-or-less stable waveform. That's why the D110 has samples for both attack and sustain (loop) sections.

Formants can easily be created on the D110, too, by setting Waveform Generator Key Follow (Pitch) to '0', and by employing resonant fixed frequency filters. Bear in mind that formants are created by resonant bodies (eg. the body of an acoustic guitar), chambers, and cavities (eg. the f-holes in a cello's body), and you should get the idea fairly quickly. The human voice is a typical example, where the chest and head cavities, and even the nasal cavity, affect the timbre. Transpose a vocal sample up an octave and you half the spatial volume of the resonant cavities.

An understanding of the applications to which large harmonic intervals can be put is essential if you are ever going to get those smoochy percussive FM electric pianos out of your D110. Dispensing with technical terms for a moment (like 'wobbly' and 'munchkinisation'), the climb through harmonic spectra goes roughly like this: at unison or so we have 'rich'; as we approach loud 3rd harmonics we have 'woody'; next is 'synthy'; then we get 'metallic' up around the 8th or 9th harmonic, beyond which are 'glassy' and even 'tiney'. Broadly speaking, a 'tine' harmonic needs to be in or near the 13th-19th harmonic region relative to your fundamental. If you feel you can spare the partials, a percussive 'ping' from the Claves (1/027) or such like, a few harmonics above the main tines sound can help add presence to an electric piano patch, while the main tines partial itself must be synthesized or a looped PCM. Transposed so high, the PCM attack samples don't last long enough to ring out at all. Don't forget that Spectrum 2, 6, 9 or 10 can make interesting contributions to your mock DX7.

EXTENDED ATTACKS

Lingering for a moment at that part of a sound just after the PCM attack transients, there is a simple way to extend the 'interesting attack transient' further into the body of the sound. This is where those four pitch envelopes come into their own.

Let's assume that we have at least two partials for the body of the sound. With ensemble-type sounds such as strings, brass, or voices, an effective trick is to employ the pitch envelope that you would normally use if you only had one pitch envelope to play with for the whole sound. Use this pitch envelope on one sustaining partial (the most important one, if there is such), then copy it to a second sustaining partial, reverse the polarity of the modulation, and reduce the modulation amount a little. You can, of course, make things more complicated. I would suggest that the reverse polarity pitch envelope should be slightly faster to settle than the other, particularly in the upper midrange - though throwing a slightly odd pitch envelope in the uppermost registers can create an impression of tension and straining to reach those high notes. It all depends how laid back you want your ensemble to be.

Another interesting possibility exists for extending the 'interesting attack transient' even further into the body of the sound. It involves those good old wobbly loops again. PCM sounds 031 to 094 in Bank 2 are looped attack transients, and are certainly only featured because they take up no extra PCM memory - they just duplicate Bank 1 sounds. However, if you restrict the number of repeats to two or three, by using a short TVA envelope (congratulations to Roland for getting the 'V' back into synthesizer jargon, by the way), then a tremolo attack can be created. You could even have a slow attack, perhaps modulated by key velocity, so that your few repeat attacks might appear slightly later in the sound. Bear in mind that, unless you are looking for a sound effect, it would be best to stick to using attack transients from non-percussive instruments - flutes and the like.

That should keep you going for now. See you next month.