Making The Most Of Your Kawai K1
How would you like to hear a filter sweep, or digital delay effects, coming from your Kawai K1? Greg Truckell looks at how to make a K1 do six impossible things before breakfast.
If I had been told 10 years ago that a synthesizer without even a filter to its name would be Synth Of The Year in a prestigious hi-tech magazine such as this, I would probably have chortled quietly and turned back to my creaking great modular system. Well, in the past few years we've had the Yamaha DX7, the Casio CZ/VZ range and, of course, the Kawai K1 - all with no filter. So, how can a synthesizer without a filter be so popular?
The answer, as always, is quite simple: the K1 sounds marvellous, and the price is right. Some have said that the K1's Variable Memory synthesis (not the most imaginative nor descriptive chunk of jargon to emanate from the land of the rising wavetable) is restrictive: in short, that the K1 is capable of only a limited range of sounds. That takes the argument, usually made by people who bought a different synth just before the K1 came out, into programmers' land. Let's assume, for the moment, that the factory presets represent the best possible combinations of the 256 internal waveforms. Where does that leave us with regard to the creation of new VM sounds?
A group of over 50 factory presets that I checked made use of just over 40% of the K1's waveforms. Two Wave Groups were used most: not surprisingly, one of those was the PCM Wave Group; the other was the Low Frequency Group. Just over 50% of the waveforms in each of these groups were used. The Basic Wave, Mid and Hi-Mid Frequency Groups came next with over 30% each, followed last by the Hi Frequency Range Group at a mere 25%. Over 80% of the patches used three or four discrete Waves, yet the overall picture is of just 1.8 Waves used per patch, due to the high degree of repetition (call it feature-sharing if you prefer) within the group of patches examined. Conclusion: the factory presets barely scratch the surface - assuming, of course, that the waveforms ignored by the factory voices are in fact of any real use. Well, I think they are.
On the whole, the K1's factory presets do a good job of showing off a limited range of VM's potential. That range can be broadly split into two categories: the instantly impressive PCM sounds, such as '6 string' and 'Ahh'; and the vaguely impressive but musically useless 'evolving' sounds, such as 'Nomads' and 'Terminator'.
Browsing through the 256 waveforms in the Wave List, you will notice that many waveforms have the same name, and that many of these similarly named waveforms appear in several Wave Groups. Now, you could simply ignore the waveform names completely, and let your ears be your only guide - if a Xylophone and a Rock Organ sound good in your Panpipes, then that's your business. Alternatively, you could use the waveform names as a firm guide to their intended purpose - to some extent, this seems to have been the path chosen by whoever programmed the factory voices, as many of the sounds employ only waveforms bearing names from the same group of instruments.
There is, of course, a middle path. There is no reason why you should not use the waveform names as a guide when searching for a particular contribution to a sound, provided you are prepared for that sound to bear little resemblance to the instruments bearing the names of the waveforms used. Indeed, browsing through the waveforms on a single Source set to a plain 'vanilla' patch, it will quickly become apparent that many of the waveforms do little to conjure up images of the instruments from which they take their names.
The K1 doesn't have a filter, and neither does it have the ability to change a waveform in real time, like the Casio CZ/VZ range and Yamaha FM synths. The only timbral variation possible within the K1 is created by altering the relative mix levels of the various Sources. Suppose you analyse a sound that you want to synthesize and find that at, or just after, the attack transient of a sound, waveform A is being generated, and that at the sustain section of the same sound, waveform S is being produced.
With any other form of synthesis, you might arrange your program so that the waveforms matched the original sound - waveforms A and S - as closely as possible at the attack and sustain sections of the patch. The decay section would smoothly evolve from A to S, covering all of the middle ground. The K1's VM synthesis method, however, doesn't offer any middle ground between Sources. Nothing evolves from A into S - A goes away, S comes in; one Source fades out, another Source fades in, and if you need something different from the intermediate balance, then you need to add an extra Source.
There is a good reason why so many VM patches exploit three or four waveforms, and it's the same reason why there seem to be so many duplicated waveforms. The similar waveforms appear in different frequency groups because they represent portions of the sound they purport to imitate, at such points in the sound which have these frequency components. If there is a given waveform in the Hi Frequency Group, then it is because the instrument in question (after which the waveform is named) has a point in its sound in which there are high frequency components. Your job is to envelope the Source so that the waveform only has a significant amplitude as and when it should. An example or two would probably be useful here...
Suppose that your instrument, whatever it is, has waveforms in the PCM Wave One Shot, Hi (or Hi-Mid) Frequency Range, Mid Frequency and Low Frequency Range Groups. The PCM Wave obviously constitutes the bulk of the instrument's attack transient, and some serious enveloping is required to fully exploit the remaining three Sources. Suppose for a moment that the instrument in question is a typical percussive plucked string type-guitar, harp, or bass guitar for example. One would then expect the higher harmonic content of the sound to diminish over time, and this can be achieved in VM synthesis by simply applying progressively shorter envelopes to brighter waveforms. Envelope keyboard scaling would be effected by applying some negative KS-Env Time modulation in progressively larger (more negative) amounts to higher frequency components. You can apply Amplitude key scaling to the various Sources to recreate the behaviour of the instrument's harmonic spectrum over its full range. If the instrument gets brighter as the frequency increases, then the high frequency components should have some positive KS-Env Level modulation. It is therefore possible to meet the requirements for a sound to be brighter up the keyboard, and also settle more quickly to a simple waveform at higher pitches.
The above is a description of how to achieve a certain kind of timbral change, but analogue filter sweep imitations are rather more difficult to achieve on the K1. The difficulty in imitating this effect with VM synthesis arises from the way in which the frequency components of the filter sweep evolve over time.
As a standard low-pass filter has its cutoff frequency swept, the low harmonics pass through first, then progressively higher harmonics as the filter opens up. After the filter envelope peaks and the filter starts to close, the first harmonics to be filtered out are the highest ones. Last in, first out.
Let's suppose that you have selected four waveforms to use as your Low, Mid, Hi-Mid and Hi Frequency components. The best way to do this is to compare the waveforms, through simple 'vanilla' envelopes, in pairs at first, and then together, to ensure that none disturbs the particular blend of harmonics you wish to use. A simple, symmetrical filter sweep might have the filter envelope's attack time equal to its decay time. This assumes that the filter envelope has a low sustain level (filter envelopes with a high sustain level are a doddle to imitate with VM synthesis - higher frequency components have longer attack times, zero decay, optionally lower sustain levels, and shorter release times).
If, however, you try repeating this symmetry on the K1's four Source envelopes, then things just don't work out - each frequency component has the same amplitude envelope, which means that there is no change in timbre over time.
"So what kind of additive synth is the K1? The presence of a basically unlimited frequency range, harmonic and enharmonic, from which to take one or two simple waveforms (they don't have to be sine waves) means that more than just a little additive enhancement is possible."
So, you try to get a bit clever - you set the attack and decay times of each component longer with respect to its high frequency content. That way the highest frequencies are last in. Unfortunately, they are also last out - an interesting effect, perhaps, but not a low-pass filter sweep (more like a band-pass). So, you try reducing the decay times more for progressively higher frequency components. That doesn't work either - it disturbs the symmetry we are trying to imitate. The highest frequency component has the longest attack time, but the fastest decay time, giving rise to an unnatural and undesirable sudden drop in high frequency content after the high frequencies peak.
The solution to the problem is to use envelopes symmetrical with respect to their attack and decay times, but to use the envelope delay parameter to delay progressively higher frequency components by longer times. Combine this with faster attack and decay times for progressively higher frequency components, and you should be getting close. (Make sure that the highest frequency component doesn't have too rapid an attack time, otherwise it will manifest itself as a delayed high frequency transient with a percussive edge.) With the ear's tendency to smooth things out, the perceived change in timbre over time will be roughly equivalent to an exponential low-pass filter envelope attack and decay.
Having discussed a programming technique which involves a correlation between the attack and decay phases, it is important to note the effects of modulation on the envelopes. Vel-Time modulation uses the Velocity curve for the Source in question, and only affects the attack time. Consequently, if your sound depends upon a correlation between attack and decay times, then you can't use Vel-Time modulation without disturbing that correlation. KS-Time uses the KS curve (there is only one KS curve per Single patch) and affects attack and decay times only. You therefore cannot keyboard scale release time, nor can you employ KS-Time modulation without disturbing any correlation between decay and release times.
The K1 is unusual in that it allows both velocity and keyboard scaling not only to reduce but also to increase the envelope times. This allows increased velocity to cause electric piano tines to be more percussive, say, or to cause string section attacks to be shorter. It also allows higher notes on the keyboard to reduce the decay times of plucked strings, but to increase the decay times of parameters that last longer at higher pitches - bowed strings perhaps, or some soaring lead line synth.
Having mentioned envelope modulation, it's worth having a look at what else the K1 offers in the modulation department: plenty of modulators, plenty of modulation destinations, for sure. The K1 is one of the few synthesizers around that allows real-time modulation of vibrato speed, although this is only possible from the modulation wheel. It might have been nice to be able to modulate vibrato speed with key pressure, but then again you will probably want to control vibrato depth with key pressure, and assigning both effects to key pressure isn't really practical. If you don't want to take a hand off the keyboard, but you want to use both pressure-controlled vibrato depth and variable speed vibrato, then you can always overdub the modulation wheel controller on your sequencer.
If vibrato is used at all, then it is applied in equal amounts to all Sources to which it is assigned - you can't apply different rates or depths to different Sources. Delayed vibrato is produced by using some positive value for Auto Bend time - even if there is no Auto Bend depth. KS-Time modulation of the Auto Bend time can therefore be used to vary the vibrato delay across the keyboard, perhaps allowing keyboard pressure to be reserved for some other effect.
Additive synthesis, as everyone is supposed to know, is awfully complicated stuff. The K1's 204 VM waveforms are taken from the Kawai K5 additive synthesizer and, for some reason, no less than 13 sine waves are included in the K1's Wave List. It might seem that some of these are quite simply not really required, even for the most devilish (four harmonic) additive synthesis you care to dream up. Sin 4th, for example, could be created instead by detuning Sin 1st or 2nd by one or two octaves (+12 or +24). Sin 4th isn't needed to create Sin 8th or Sin 16th - they are both there anyway. Extend this procedure and I'm sure that you could eliminate a few more Sins - good Biblical stuff this!
However, this argument has forgotten to look between the harmonics; only with a full complement of sine waves can VM hope to be able to add the occasional inharmonic, just where it's needed. If the Wave List is to be believed, then all 13 sine waves each take up an equal amount of memory, in which case each sine wave takes up as much memory as any other waveform. Just think, you could have had four more pianos, strings, brass, and still have had room for Sin 1st. Were the sine waves worth it? It's too late now, you ate the apple.
So what kind of additive synth is the K1? Well, I'm not about to entreat you to embark on pure additive synthesis with only four sine waves at your disposal (ie. one per Source). However, the presence of a basically unlimited frequency range, harmonic and enharmonic, from which to take one or two simple waveforms (they don't have to be sine waves) means that more than just a little additive enhancement is possible. For instance, fixed frequency sine waves can be used to add 'formants' (enharmonic frequency components which do not vary in frequency over the instrument's pitch range), to recreate the natural resonant frequency of such things as the body of an acoustic guitar or violin, etc.
Another additive trick might involve synthesizing some instrument whose harmonics don't behave themselves. It is not altogether uncommon for some of an instrument's harmonics to drift out of tune during certain portions of the sound. Some brass instruments will often have one or two fairly low numbered harmonics which are slightly flat during a note's attack and decay, but which come into tune during the sustain phase. This could be simulated by using Sin 3rd, for example, finely detuned flat (negative) and faded out with the Source's envelope by the time the other Sources have reached their sustain sections. This might seem to be a waste of a Source during the sustain, but the alternative would be to dedicate the Auto Bend, and therefore also the Vibrato, to just one Source - much more of a waste. Alternatively, some positive value for Prs-Freq in the Common parameters, combined with Prs-Freq On in that Source's Freq parameters, would allow key pressure to bring the flat harmonic(s) to unison at the sustain section - and allow decreasing key pressure to flatten such harmonics again. It may even be the case that the rogue harmonic only goes out of tune over a certain part of the instrument's range. This could be reproduced through judicious application of KS-Freq modulation, set so as to cancel the effect of detuning for that Source over whatever part of the keyboard range is appropriate.
If you have read any of my previous 'Making The Most Of...' articles, then by now you should know that I have a fondness for exploiting those features of a synthesizer which aren't mentioned in the handbook. I must admit that the K1 had me stumped for a while - until I discovered the 16 digital delay lines with MIDI performance control...
But more on these later. First of all I'd like to deal with the Auto Bend feature. Set to a fast time and a depth in double figures, this feature can be used to create multiple percussive attack transients by exploiting the aliasing characteristics of high frequency components.
"I have a fondness for exploiting those features of a synthesizer which aren't mentioned in the handbook. I must admit that the K1 had me stumped for a while - until I discovered the 16 digital delay lines with MIDI performance control..."
As the Auto Bend sweeps the high frequencies past the frequency around which aliasing occurs, a glitch or several glitches can be heard. Experiment with slow Auto Bend times in order to get just the glitch you want - you will have to increase the decay time of the Hi Frequency Source in order to be able to hear the effects. When the Auto Bend time is decreased, then the glitches fall in rapid succession. Combined with a short percussive envelope, a fragmented attack transient can be created.
Applications for such transients include strummed or fuzzed guitars, double-tongued wind instruments, and anything which you want to have a percussive attack with more than one edge. Of course, this could be done by simply applying some delay to the envelope of a Source used in the attack transient, but the glitch technique can create a fragmented attack from only one Source, leaving three Sources free for the decay and body of the sound.
It is important to note, however, that there is an aspect of Auto Bend which can interfere with almost any application of it. Briefly, the Auto Bend will not function when any notes are being held. There is no explanation of this in the handbook. While this 'feature' may interfere with some of the sounds described above, it can be put to useful effect where a slower Auto Bend is used. In a lead line sound, for instance, this facility allows legato style playing to prevent the Auto Bend from having any effect. As usual, all it takes to turn what seems like a shortcoming into a useful feature is knowing about it in advance.
So what about those 16 digital delays? If you employ a PCM Wave as an AM modulating Source, then it will loop - even if it is a PCM One Shot. The effect I am about to describe works best with a percussive or transient PCM One Shot Wave as the modulating Source, but will only work with a looping waveform (all except waves 205-234) as the base Source.
Set up a pair of Sources in an AM configuration, with the modulator muted. Both Sources should have envelopes with fast attack times, no sustain, and long decay and release times - 75 or over. To create the simplest effect, set the base Source to Sin 1st - a quick jab at the keys will give you a repeating echo that sounds almost exactly like the PCM One Shot, with the repeats decaying in amplitude as the envelope closes. The frequency of the echoes is a function of the pitch played (performance MIDI or what?). Play middle C, the G above, the C above, and the G above that, and you should get an interesting little polyrhythm going, since the ratios of the frequencies are nice round values: 1, 3/2, 2/1, and 3/1. Remember to set Poly Mode to 'PL2' if you don't want repeated pitches to cut off previous notes.
Turning Freq Key Track off for the modulating Source will create an effect where the echo frequency is constant across the keyboard, and again the frequency is controlled by the pitch of the modulating Source, which can be edited with the Freq Coarse and Fine controls. However, the effects of ring modulation become rather dominant here, as each note has a different pitch interval between the base and modulating Sources.
A further possibility would be to copy the first pair of Sources to the second pair, then to tune the second modulating Source to some simple multiple of the pitch of the first modulating Source (again with KS-Freq off). If the first modulating Source is tuned to a C, then tuning the second to the G above that C will create an effect whereby the second pair of Sources echo three repeats to every two repeats of the first pair of Sources - a one-note polyrhythm, and a lot more tuneful than the PCM Omnibus Loops in my opinion. Experimentation is the key here - remember that anything which varies the pitch of the modulating Source will alter the echo frequency. Auto Bend and Vibrato are possible candidates; KS-Freq is another, more subtle, possibility.
That seems like a good place to end this exploration of the Kawai K1. Experimentation and a clean pair of lugholes will get you a long way with the K1. So do I really miss the filter?
No. I'm glad that the K1 has no filter, I'm glad that the DX7 and the CZ synths don't have filters, and I'm glad whenever an instrument makes me think a slightly different way about creating new sounds. And I'm especially glad when I can afford them.
Gear in this article:
Feature by Greg Truckell
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!