Kawai K5000W

Ten years after the K5, Kawai return to the technology which gave them the most successful implementation of additive synthesis ever, and include something of everything else they've learnt in the meantime. Paul Wiffen wonders how it all adds up...

Give the SOS team their due: they do like to maintain editorial continuity. Once a writer has covered a particular product or technology, they like him to see it through all the developments the manufacturer may make from it. Hence, since I covered the Emu EIII and its successor, the EIV, they asked me to look at the E64 and then the E64K. So I suppose I shouldn't have been surprised when Assistant Editor Matt rang me up to tell me that there was a new synth whose predecessor I had reviewed for them, and ask if I'd like to cover it. I, of course, expected it to be something I had looked at in the last year or two. When Matt told me that Kawai had finally updated their unique implementation of additive synthesis (the first incarnation of which I reviewed almost 10 years ago, in the shape of the K5), I was both amazed and delighted. Amazed because if a manufacturer hasn't got around to developing a new model of a design within a couple of years, you can usually reckon that it has been consigned to the dustbin of history. Delighted because I had been a big fan of the K5 and was feeling the distinct need for a breath of fresh air in the smog of PCM‑based, effects‑laden synthesizers which assault our ears at every turn these days.

It seemed like only a few years ago that I was looking at the K5 additive synth in Argent's back room in Denmark Street, but Matt informed me that it was back in 1987. Although this made me feel horribly old, I was unable to deny the facts when he sent me a photocopy of my review from the December '87 SOS, to refresh my memory on my original comments on the K5. I approached this piece with some trepidation, sure that my earlier writing style would make the more mature Wiffen cringe, but apart from the terrible jokes (which haven't improved over the years!) many of the comments could have been made today, particularly the intro about music shops and their customers judging an instrument by its presets, rather than properly investigating the potential of its synthesis system for themselves. The only change is that now people judge a workstation by the demo song as well.

Preset Patrol

My initial reaction (and those of several of my colleagues) was that there weren't many presets in the K5000W which were going to grab people's attention. There was certainly nothing as immediately impressive as 'Big Time' on the K5, still the best and most playable brass sound I have ever heard from any synth, analogue or digital, sample‑based or purely synthetic. Many's the time I regretted not keeping my K5M for that program alone — the sampled version I have has all the power but lacks the touch responsiveness. If anyone out there has one they want to sell... Although I would never judge an instrument by them, the demo sequences supplied on floppy were somewhat insipid as well, except for the Techno one, which, if anything, erred on the side of brutal. This was the only hint, initially, of the potential of the machine, like a Mr Hyde suddenly emerging from a meek and mild Dr Jekyll.

However, I soon discovered the two mistakes I had made. To start with, only the 120 sounds in the first bank implement the Advanced Additive process at all, and as I had ended up very quickly in the 'B' bank, which is composed entirely of PCM‑based sounds, it was hardly surprising that the sounds weren't exciting my jaded ears (the third GM bank was little better, but then haven't we all OD'd on General MIDI sounds?). Secondly, the Advanced Additive programs in that first bank really only show their true colours with real‑time manipulations of the mod wheel and the data dial (linked via the softkeys below the display to the most critical sound parameters). Suddenly the whole machine comes to life. Incredibly broad timbre sweeps grab the ear, with swathes of harmonics phasing in and out, the machine alternatively growling and soaring, snarling and singing sweetly — all within single patches. The problem with most PCM‑based machines (and this is equally true of the K5000's B and GM banks) is that the only way you can make any convincing change to the sound is to change programs. But Bank AA on the K5000 contains sounds which are really Jekyll and Hyde combinations, depending on your playing style and the position of the wheels. This is how real instruments should be, responding directly to the input of the player, affected by his mood and giving voice to his emotions. Pianists or violinists don't need to switch between programs to go from soothing to vicious — they just change the way they play. Real synthesizers have the same characteristic, and stand out easily from the PCM playback machines by virtue of this ability to radically change their sound in real time and reflect the personality and mood of the performer.

So how are these startling timbre changes achieved, and how does the voice architecture of the K5000 compare to that of the K5?

Creative Addition VS Destructive Subtraction

Most forms of synthesis, from early analogue to PCM‑based, take as their starting point a waveform (whether a single‑cycle oscillator or a sampled sound), and this contains all the frequencies that final sound can ever have, as the only real tool they offer to change harmonic content is the filter, which just removes frequencies (hence the term subtractive synthesis), and in a very unsubtle manner. The results can be quite good with a big fat waveform to start with, such as you can find on the best analogue machines, but if you apply the same technique to samples, most of the time all this does is remove the life and character from the sound. The beauty of additive synthesis is that you add together the harmonics you want to create the required sound — so if you suddenly want to change the timbre, you simply add a different set of harmonics or radically change the level of some but not all. If you think of subtractive synthesis as sculpture (removing the unwanted bits to leave the required shape) and sample‑based synthesis as photography (capturing exactly what is in front of the camera), then additive synthesis is more like oil painting, building up the layers of paint until you achieve the required effect.

The scheme offered for this on the old K5 was already pretty comprehensive (especially since it was the first all‑realtime implementation), allowing the levels of 64 harmonics (or 128, if you wanted to annoy the local dogs) to be controlled individually or in groups (odd or even‑numbered, octaves or fifths, high or low groupings, etc) by envelopes, velocity, pressure or other modulation, with the amount of change being 'angled' to affect the high or low end more or less. Then digital filtering (a heretical concept for academic additive aficionados) allowed a more precise shaping than the blunt analogue equivalent. It was the most flexible implementation of additive I had ever seen (although the ill‑fated French‑Canadian Technox high‑end system of the late '80s was doing some clever things with sample analysis, before the company went bust).

Kawai have kept this same starting point for the K5000 (some of the harmonic editing screens look very familiar, obviously stirring some race memory), but they've added a host of refinements. You can still choose to work with 64 harmonics or combine two voices (known as Sources) to get 128 waveforms (switching the second to cover 65‑128). However, now you can have up to six Sources in a Program. This means that you can have six different sets of harmonics, all reacting independently under the various means of real‑time control we will cover in a minute. Alternatively, you can replace one or more of the Sources' harmonic series with a PCM sample — "a worse heresy than filtering," I hear all the additive purists cry!

Fortunately, these anoraks, the synthesis equivalent of trainspotters, are a dying breed. They used to lie in wait for unsuspecting journalists on the cheaper stands away from the main thoroughfares at trade shows, and having lured you into their lair, waffle on endlessly about how the pure additive system they had developed using mountains of public funding at some third‑rate university in the middle of nowhere, could theoretically reproduce any sound with the right programming. But when you finally got them to play you something, it always sounded like a rather cheap, thin drawbar organ (a primitive additive synthesiser in itself, but usually somewhat more cost‑effective than their monstrous prototype).

PCM: No Problem

As I had no problem with digital filtering on the K5, I have no objection to the addition of PCM sources on the K5000. There are some sounds which no amount of harmonics can build up, usually noise elements with unrelated frequencies, which occur most commonly at the beginning of natural percussive sounds. This is when the energy level is so high that it spills over into generating harmonics unrelated to the fundamental. By using PCM attacks, the K5000 borrows a technique from a contemporary of the K5, the Roland D50, and allows nature to lend a hand with the least musical bit of a sound.

The 266 PCM sources available for this use in the K5000 (not including the 225 percussion samples) encompass all kinds of blown and plucked instruments, as well as breath noises and other effects (rain, wind, thunder, cars and other natural phenomena). Whilst I would recommend they are used sparingly, there are a few programs where the added realism of the attack peps up the sound no end. I would have loved to hear the 'Big Time' brass from the K5 updated with a real sample attack from a trumpet player. I wonder if you can port K5 programs across to the K5000 via SysEx, or would some poor soul have to sit down and painstakingly enter every parameter manually?

Getting back to the additive side of things, each Source gives you 64 harmonics to play with, and the level of each can be set for the hardest and softest keystrokes, either individually or in groups. The graphic display allows you to see all 64 soft keystroke levels or all 64 hard keystroke levels simultaneously, and switch between them at the push of a button. The same groupings of harmonics for editing levels that were offered on the K5 are available here. As well as Each and All (changing just one or every harmonic respectively), you can select Odd, Even, Bright (the upper 32), Dark (the lower 32), Octaves (1, 2, 4, 8, 16, 32 and 64) or 5ths (3, 6, 12, 24 and 48). As on the K5, a dot appears beneath each affected harmonic in the chosen group. Once you have selected the group you wish to edit, moving the data dial raises or lowers the levels of the selected harmonics. Once you've adjusted the hard and soft harmonic levels, the velocity depth (ie. how much effect velocity has) can be set, and one of 12 crossfade curves chosen to govern the transition between these two. These start at 1 with quick changes between the lower velocities, and gradual changes occur in the higher velocities; 5 gives a straight linear response, and at 12 there's virtually no change in lower velocities, with all the dynamic response in the upper velocity range.

I can't honestly think of anything that Kawai have left out...

But velocity isn't the only factor which can change the level of the harmonics. Keyboard scaling can make higher or lower notes on the keyboard proportionately louder or softer. Levels can also be altered in real time using envelopes, but whereas the K5 only gave you four envelopes to alter the levels of each (meaning that each harmonic had to be assigned to a group first), on the K5000 each harmonic now has its own envelope all to itself, increasing the flexibility and subtlety sixteenfold. I guess that the computional power available for the price must have increased at least sixteenfold, because these envelopes are also more complex. They are four‑stage (attack, decay 1, decay 2 and release) level and time envelopes with the ability to loop between decay 1 & 2 stages. The K5000's Harmonic Envelope Multiview display lets you select the group of harmonics whose envelopes you want to edit, and lets you visualize the result. This saves you the hours of work it would take if you had to edit each harmonic's level envelope individually. Once you've made the broad sweep adjustments collectively, you can then use the Harmonic Level View display to select and edit each harmonic's envelope individually.

The ability to loop these envelopes means that the harmonic content of the overall sound can be continuously varied. This makes the timbre more interesting to the ear. The human ear is not very sensitive to the relative levels of harmonics when they remain static, but it can perceive the smallest of changes when they happen in real time. By cycling the envelope, these changes can be repeated continuously for as long as the note is held down. This is the real triumph of the K5000 — that it can manage such complicated changes in real time and thereby excite the ear. The closest analogy to this in analogue synthesis is Pulse Width Modulation, which creates this kind of harmonic change by varying the width of the positive section of a pulse wave using an LFO. As this changes the relative strengths of the different harmonics, the effect is very similar. However, PWM only allows one very narrow set of changes in the harmonic structure. In contrast, the cycling harmonic envelopes on the K5000 allow limitless variations on the harmonic content. Innumerable different timbral changes can be created very simply, each with a character all of its own. I found that once I had set up a envelope cycle of around a couple of seconds, I could quickly create lots of different timbres just by changing the groups that decay 1 and decay 2 levels affected. It was like having endless different variants on the theme of PWM.

Unfortunately, very few of the Advanced Additive programs which ship with the K5000 utilise this ability, and if they do it is extremely subtle. However, many of them use the modulation wheel to change individual harmonic levels, and with them the harmonic content of the sound. By slowly moving the mod wheel backwards and forwards while playing, you can easily hear the same effect. I would strongly suggest that you try this if you go to check out the K5000 in a store, otherwise you're likely to remain unaware of its sonic potential. And if you stop reading now, you would still not know the half of it. Kawai's boffins have come up with some other great short‑cuts to sonic interest.

Morfing Made Easy

If you don't want to spend any time at all adjusting envelopes, even in groups, there is an even faster way to come up with interesting timbres. Kawai refer to it as the morf display (their spelling, not mine). Here you can select four source harmonic spectra from other programs and have the K5000 interpolate between them to create new timbres. You simply specify which Program and Source you want to take the 64 waveform snapshots from (it uses both the soft and loud harmonic setups), together with the timing intervals between each phase. It is also possible to have the morf cycle between phases 2 and 3 to keep the sound moving. Once you have specified the sources and the timing (with or without looping), you press Exec, and the K5000 does all the hard work for you, creating all 64 envelopes based on the level of each harmonic in the four phases. The resulting morf is then instantly ready to be used as a Source in a new timbre.

There are two ways you can use this. The scientifically‑minded amongst you may want to try and plan the resultant sound by carefully matching similar harmonic spectra. Those who place more faith in serendipity may want to try entering Programs and Sources at random and seeing what they get. I tried both approaches. Planning was never as predictable as you might think, although I started to get better at mixing and matching harmonic spectra, so that the results weren't too wild. The random approach produced horrible noises most of the time (although some of them were splendidly horrible, if you take my meaning), but every so often it produced a real gem of a sound that you couldn't have planned or imagined in advance.

The best thing about this feature was that the results it produces are quick and often give completely unique sounds, which recall very few other synths (OK, maybe the odd hint of PPG here, a touch of Prophet VS or Korg Wavestation there, but nothing reminiscent of the current crop of tired PCM timbres). It's a breath of fresh air to ears stifling in stagnated sample layers. Even the heaviest sounds are without the leaden stodginess you get when you throw a load of PCM sounds together and smother them in effects.

Mix & Match

Once you've made yourself a few harmonic envelope‑based waveforms and created some morfs (or pinched them from existing patches using the numerous Copy functions, if you're feeling lazy), you're ready to build a Single patch. This is a bit of a misnomer, as up to six different sources can be combined in a Single patch. Of course, many of the individual Sources are complex enough on their own without needing to be combined with others, but following the Wiffen motto, 'Too Much Is Never Enough', I experimented with throwing tons of stuff at the wall and seeing how much stuck. Having picked a few enveloped harmonic Sources and a morf or two, I chucked in the odd PCM sample for good measure. Now that's what I call a sound. The effect reminded me of the Ensoniq VFX's more complex voicings, with lots of separate elements coming and going in the sound. One nice feature is that at any point in the editing process you can mute and un‑mute any of the Sources using the F2‑F7 keys below the display, so if you're working on one Source you don't have to be confused by the sound the others are making.

Just when I thought I couldn't cram anything else into the Single Mode, I discovered the AM parameter in the Single Common menu. The manual told me this stood for Amplitude Modulation. This lets you use any Source to modulate the Source to either side in the display. Like most forms of high‑speed modulation, this is great for adding enharmonic elements to a sound, ie. making incredibly complex sounds even more complex (and therefore unlistenable nine times out of 10). However, used judiciously, I found it had enormous possibilities, particularly if just used for a fraction of a second at the beginning of a sound, where enharmonic elements sound quite natural. There's certainly plenty of scope for future experimentation here.

Formant Filtering Made Better

In my original K5 review (all those years ago), I heaped praise on the Digital Formant Filter and how its 11‑band display made tailoring the pitches at which it was operating simple, by referring to them in octaves, instead of centre frequencies (unless you're a sound engineer, the only musical pitch you will know the value of will be the A below middle C — ie. A440). Well, Kawai have really taken advantage of the drop in the cost of DSP horsepower by expanding this to a 128‑band formant filter — one band per semitone instead of one per octave — and providing one per Source (ie. up to 6), when previously it was one per voice. The manual even boasts a table telling you the centre pitch of each band (band number 70 is A440) and the corresponding frequency.

For those of you not familiar with the concept of formant filtering (also known as comb filtering, because of the appearance of the resulting frequency response), it uses lots of very narrow bands to cut or boost very tailored frequency ranges, ideal for making precise and subtle changes to the sound. If you think of the normal low‑pass filter as a machete, think of the formant filter as lots of surgeon's scalpels. Each of the 11 bands on the K5 was tuned to an octave of C, but on the K5000 each band is assigned to its own semitone division.

Of course, just as with adjusting individual harmonics, setting the amount of boost for 128 different bands is incredibly time‑consuming, so Kawai have used the same grouping concept as for the harmonics. The following groupings are available for simultaneous editing: Graphic EQ, 20‑band, 15‑band,10‑band, 5‑band and All. The Graphic groups the entire range into a classic 8‑band graphic format, producing the typical elliptical EQ pattern within each group (ie. the centre frequency of each band boosted the most, with progressively less boost as the frequencies move away from the centre). The others move the number of bands stated with equal boost to each within the group. This gives you the ability to set the formant filter shape as coarsely or as finely as you like. Obviously, you would use the wider groupings first to get roughly the shape you want; the narrow groupings, or even individual band editing, will then let you tailor it more precisely.

There's a bias control which allows you to shift the bands up or down one octave (+/‑12 semitones). This means that if you have set up an effect to accentuate particular pitches in a particular key (say root notes and fifths) and then you decide to change the key, you can re‑tune the comb filter to the new key. However, much more useful than the ability to set a fixed bias is the ability to shift the comb filter with an envelope. This means that some pretty outrageous sweeps can be created, and as the envelope is the same as those for controlling the volumes of the harmonics, it can be looped in the same way between the two Decay stages, to create a constantly moving effect. Do this on a Source which already has looping harmonic envelopes (preferably at a different speed) and the effect gets really wild. Shame they never made any presets which use this facility.

Common As Muck

So far everything we have looked at is available at Source level — so you could (if your taste is for excess) have six of these voice channels all doing wildly different things for each note triggered (with the attendant reduction in polyphony, of course, as each Source takes a note of the 32 available at any one time). However, I found that two or three Sources, used to maximum effect, created such a complex sound that anything else would have been gilding the lily.

From now on, things get much more ordinary, as all parameters described henceforth (referred to quite significantly on the K5000W as Common) are pretty standard for any PCM‑based synth.

These Common parameters are also available for editing both the PCM‑only programs in Bank B and the General MIDI presets in the third bank. However, all the sounds you create yourself must be saved to the 48 user editable locations in Bank B if PCM‑based, or to the 60 user locations in the first Bank if they contain additive stuff as well. The General MIDI sounds cannot be overwritten (which is just as well — there's no point in having General MIDI if some of the sounds may not be available because they've been overwritten).

As far as the quality of the PCM samples is concerned, there's nothing actually wrong with them — indeed, it was their 'sensible' workaday nature which caused me to lose interest. After exploring the innovative Advanced Additive archictecture of bank A, it was back to business as usual in Banks B and GM. You know the sort of thing — take a sample or two, bung it through a filter and envelope, and stick some effects on it. The K5000W brochure claims that Kawai have "developed a totally new library of PCM samples" for the machine, and that may well be the case, but I didn't hear anything that made me sit up and take notice. Having said that, the couple of General MIDI files I played back on the machine (one pop and the other orchestral) sounded absolutely fine.

You must excuse me if the description of the remainder of the sound‑generating architecture is a bit perfunctory, but we still have the effects and the sequencer to cover. Each Source used (up to six, either Additive or PCM) has its own coarse (set in intervals) and fine (set in cents) pitch controls, as well as keyboard scaling for pitch (this is most useful to prevent a noise element in the patch from being transposed across the keyboard). Next comes a pitch envelope to add an auto‑bend or warp (if required) to the pitch of each source on trigggering, useful to imitate the pitch variations in stringed or wind instruments, caused by the actual pluck or the player's embouchure. This is a two‑stage envelope covering attack and decay only. Once the pitch has moved through this envelope, it returns to the actual pitch of the triggered key. Positive values for the envelope take the pitch sharp and negative ones flatten it, so by making the attack level one and the decay level the other, it is possible to move the pitch above and below the point at which it starts and ends.

Then comes a more standard filter than the formant one described, just in case you need a machete instead of a surgeon's scapel. It is, of course, also digital but apes its analogue ancestors rather well, with the choice of high‑ or low‑pass, with resonance and velocity, keyboard scaling and envelope control. Again, the four‑stage envelope features two decay stages, but sadly with no looping this time. The results of this were distinctly ordinary after the formant filtering, but then I guess sometimes you have to eat bread and cheese. (Chamapagne and caviar every day would get boring.) The final stage in voice editing is the DCA (or digitally controlled amplifier). This can also be modified by key scaling, velocity, or its own dedicated envelope. Again, the envelope has two decay stages as well as Attack and Release.

That covers all the synthesis parameters available within single voice mode. It would be true to say that there are significantly more than on any other synthesizer I have ever used (64 harmonic envelopes per Source, just to start) and there's no doubt that while the display gives you a pretty good clue as to what's going on, and the grouping of harmonics and envelopes provides many shortcuts to sound programming, being able to see and edit all these parameters on a computer screen would be really helpful. Doubtless profiles for the most common editor programs will emerge in due course to allow you to do this.

Combi‑Ning Singles

As if the Single, with its potential of six Sources, isn't enough sonic grist to your mill, you can always combine four of them into a Combi. This allows up to 24 sources to be triggered from a single note (although don't expect much polyphony if you decide to push things to this level, as one voice will use 24 of the available 64 voices).

The real reason for using Combis would be for the keyboard‑mapping potential (different Singles in different areas of the keyboard or in different velocity ranges). You can set zones up and down the keyboard or in the velocity range to bring out all sorts of switching or splitting layouts. This is ideal for the live performer who has to cover bass lines and polyphonic parts, or for the more advanced imitation of entire orchestra sections, where you could bring in extra instruments on high velocities. For sequencing, however, you're restricted to using Single patches, because of multitimbral access considerations (see 'Sequencing Potential' box).

Conclusion

The only real problem with the K5000W is that it suffers from a split personality (if references to schizophrenia are not politically incorrect). On one side it features a technology which should really appeal to sound designers looking for inspiration and new ways of approaching things. On the other hand, the styles and auto‑accompaniment features (see 'Auto Phrase Generation' box) are more for the domestic built‑in speaker market. It seems unlikely to me that there will be too many cross‑overs between these two markets. As far as the workstation side of things is concerned, the sequencer is very worthy example of the genre if your sequencing needs are not already taken care of. Whether you need the Auto Phrase Generator and Chord Advice depends on the type of music you're doing and how confident you are in your own musical abilities. My guess is that the K5000S (the synthesizer‑only version) is going to be of more interest to SOS readers.

As I sit playing with the K5000W in the back room at Sutekina (just next door to the old Argent's shop in Denmark Street — funny how life goes in cycles, isn't it?), how do all these improvements to the original K5 synthesis architecture add up? Have they brought the concept up to date for the '90s, or made it look even more anachronistic in an age of General MIDI sound‑sets and FX‑swamped presets? All I can say is that once I started delving into the K5000, I was constantly amazed at how much has been added to the K5 concept in Advanced Additive. Where there were once four envelopes, there are now 64. Where there was once a single sound source, there are now six. If all this isn't enough to allow you to generate complex sounds, add in the ability to blend in PCM sounds, cross‑modulate them with AM, and then put them through the four effects. In terms of sound quality, it's every bit as good as the K5, but with so much more flexibility. The real key to the exciting sounds available is the real‑time modulation of harmonic content. This means you can use envelopes, LFOs, or even the mod wheel to really bring sounds to life. If you get the chance to try the K5000W, make sure you give the mod wheel a good workout when playing the sounds in the AA bank.

For me, reviewing the K5000 was like revisiting an old girlfriend and finding that she had grown even more beautiful than I remembered. I can't honestly think of anything that Kawai have left out of this stunning synth.