posted 23 April 2001 20:46            

So this posting is for those of you - if any you be - who want mostly to use Kyma as a small orchestra of about 8-24 rich, polyphonic voices, that can play a complete algorithmic composition in real time. My hope is there are a few of you who want to do this, and by joining in on this discussion, perhaps the good folk at Symbolic Sound will be persuaded to make this a reality for Kyma soon; alternatively if there are few or none who dream of an all Kyma orchestra in one box, then that will be a reality check for me.

Because while I DO think Kyma is the best soft synth out there - certainly there is no better sound design tool available at any price - it turns out that as of this writing NO soft synth is able be the flexible, modular, real time electronic orchestra I have chased after for 'lo these many years.

Kyma is close. Very close. All it really needs now is some decent filters (already covered, already promised), and a better way of making use of it's DSP power in a non linear way. Conceptually this is fairly easy to state: Make available a prototype Sound like WaitUntil that can turn a DSP branch OFF as well as ON, thereby allowing patches that would be far too large to play in real time, to be turned on and off in segments, one branch at a time. Sort of like the TimeLine, but parallel, instead of serial. I suspect this is an exceptionally difficult thing to do with the existing Capybara/Kyma, or Kurt would have done it already.

If you are happy with Kyma 5's Time Line, then you probably won't care much about any of this. The Time Line is a remarkable piece of work - ongoing work, I should say - and it elegantly and powerfully answers the need of doing linear time based composing. I have no quarrel with it, quite the opposite. But this is not the kind of composing I am personally involved in. What I need is to have a large number of parallel real time voices. Something like a sound equivalent of the color mapped video display. Remember, you could have 16 simultaneous colors out of some 4000 possible colors? Or today, 256 colors out of some 16 million? So for this synth orchestra, I want 16 to 32 real time voices - nice fat, rich, and evolving voices - instantly selectable out of a possible several dozen or even several hundred.

This is actually fairly easy to get these days with Emu or Roland sample playback synths, and in fact this is what I have used in the past. A rack of 4 Emu Morphius synths gave me a very solid 32 simultaneous voices out of several hundred online at a time. But these are sample playback voices, not the dynamically evolving, synthesized voices Kyma is capable of. Sure, dynamic filters like the Emu's help, but you still have playback voices, zombie voices.

There is a Renaissance of modular analog synths going on, which is very cool. Two companies are making very credible Moog like modulars just a few miles from where I live in Fort Worth. Once I thought this would do the job, if the modular synth was big enough. So I bought a very large Serge modular. 36 VCO's, 12 LFO's, 24 VCF's, etc. Very large. It was a great toy, perhaps my greatest toy ever. But I sold it a few years ago because I simply wasn't using it. The problem I ran in to was I could not keep a dozen complex patches going at once. I was constantly hitting a buzz note, then trying to figure out where the loose patch chord was, or what knob I had accidently changed. I had that thing for some 12 years, and I never made a complete real time composition on it. Like Kyma, it was great for making up really cool, evolving individual sounds, that could be recorded and mixed with other really cool evolving sounds. But it failed utterly as a real time orchestra. In fact, the modular analog synths being sold today are - as they always have been - sold as mostly single voice lead instruments. I'm glad to see them come back, but they do me no good at all.

Same goes for the DSP performance synths like Nord. Cool in their way, but essentially a single voice lead synth. So the only real hope I can see for a real time orchestra for algorithmic composition has to be with the soft synths.

Of the current crop of modular soft synths (I'm not even considering prepackaged synths like Reason or sample play backs like Gigasampler) there are the native synths such as Reaktor, VAZ, and Tassman; and the DSP based modular soft synths, Kyma and Pulsar/Scope. Except for Tassman, which I have only played the demo on, I own and have extensively used all of the above. Have even been a beta tester for three of them. Each has its strong points, and each fails in the end.

In a year or two or three, when we all have dual and quad 2GHz desktop PC's, the native synths are going to be the way to go. I especially think Reaktor is one to watch. Its ability to encapsulate modules almost without limit, while bringing the control surfaces to the front in easy-to-design panel displays, is the best I have seen. And with the new Moog inspired Ladder Filters - that sound great AND have the same kind of controllable (playable) resonance settings as the analogs - Reaktor moves to the head of the line for sound quality. Of course by this time I would expect Kyma to be available as a native soft synth as well, and most likely still the queen of sound morphing.

But for now, Reaktor, VAZ, and especially Tassman simply do not have the CPU cycles needed to do the kind of orchestra I hope for. Which leaves Pulsar and Kyma.

I bought a Pulsar card about 2 years ago, intending to quickly expand it to a 19-DSP Scope setup. But it had crashes, lots of them, so I waited until the software for it stabilized a bit, and while waiting I built modest single voice patches with the Pulsar Modular synths. Around the first of last year, after yet another software update became available, I was again ready to spring for the Scope expansion, and finally get down and dirty with an orchestra. Instead, I expanded my Capybara-66 to a 320LS, for about the same price. Two reasons, one that I figured out, and one was just pure dumb luck.

The reason I figured out was that, cool-looking as the Pulsar Modular synth modules are, they are not encapsulatable like Kyma Sounds. So the bigger your synth, the more voices you create, the more scrolling around on the screen you have to do because you can't get it to all to show on the screen at once. Worse than the scrolling, there is no overview.

Also, those curved, shaded 3D graphical patch cords, which look so cool in small patches, become a nightmare in big patches. In a real modular, with real patches, you can push them aside, and you can trace them by sliding your thumb & forefinger on them from output to input, and even that becomes a major headache. In the Pulsar virtual synth, you can only turn them all on or all off. Neither helps when trying to decipher a big modular patch you created a couple weeks ago.

Kyma, as well as Reaktor, have wonderfully encapsulatable connections, that give you an overview on a single screen, and let you zoom in as far as you like. And both have very elegant documentation tools, so it is not a burden to take notes as you go.

(Oddly enough, my own SoftStep modular sequencer software also suffers from lack of encapsulation. It's not an easy thing to do, as I am now discovering, as I work to put it into the next version.)

The other problem with Pulsar/Scope, the one I lucked into not stepping in, is that hardware problems still abound over a year later, In Dennis Miller's review of Kyma 5, he said he never experienced a crash with Kyma/Capybara. Most of you who read the review probably didn't think a lot about that, because it is true. Kyma's stability is a bit like the old HP printer ads: you have the luxury of not having to think much about it. Not so with Pulsar, especially if you add an expansion card. Take a look at the PlanetZ discussion board and you will see message upon message about how to deal with crashes. And unlike Kyma, which gets more stable as you add expansion cards, Pulsar tends to get less stable because of a bottleneck problem passing data between cards.

So, by process of elimination that leaves Kyma as the best candidate for a digital orchestra, just as I optimistically thought when I bought the Capy-33 five years and two upgrades ago. Sadly then, and frustratingly now, even Kyma can't do the job. But it is close, so close.

posted 24 April 2001 02:25            

so if anybody is interested in a reaktor 2.3.2, cubase 5.0, model e, lm4: - i sell it.

posted 24 April 2001 14:21            

Now if I had a *second* Kyma...

But how about the ability to split a Capybara into two virtual Kyma systems? Two would probably suffice since this is sort of like "double buffering" in that it solves a similar problem.

posted 24 April 2001 17:28            

Concerning my "novel," I guess I did get a little self indulgent in the length. But I wanted to establish that this is not a whim, it is something I have tried for years to achieve. As to what my problem is with, it is neither with Kyma or with the fact that Kyma is DSP based. It is....

David - Right, the parallelism I am looking for is rather like the patch bank on Virus/B, etc. - If there could be maybe two dozen of them all together in one Kyma patch. My Capybara has 8 cards, which gives me maybe a dozen rich simultaneous voices. But the way I work is more like painting than composing (I went to art, not music school, so it's OK for me to do this). I start up a riff on my step sequencer, then keep making changes to it in real time. Listening, changing, etc. My step sequencer program SoftStep was designed just for this kind of work, and once the new Kyma drivers that allow direct access to the Capy, I plan to include Kyma specific controls for even better control than you can get with MIDI.

Anyway, the problem I have is that while a dozen voices at a time are just fine, I need to have a lot more than that available in the musical pallet. Which is why Kyma, though close, will not really get the job done as it stands now. There needs to be a switch that can turn branches of a Sound off and on again, allowing Sounds to be built that - while they cannot play all voices at once - they could play a small dynamically changeable subset of them.

The Kyma sound matrix needs to load the whole sound at once. This means you can't really do it in real time - there has to be a complete stop in ALL voices. Then, upon continuing, either you have a completely new set of voices, or you have the same except for whatever you wanted changed - but try to do this and keep all the presets where they were. It's one of those things that work fine with just a few voices, but falls apart when you try to use it on several at once. At least, I have not been successful with it, even allowing for the complete stops required.

Dennis - I like your name for it, the "When" Sound. The "double buffering" idea is a bit of a kludge in the positive sense, as a practical if inelegant work around. Actually I have been experimenting with something like that, combining Kyma, Reaktor, and Pulser. It was my reward for giving my customers the multi port MIDI they had been clamoring for: I get to use it now, too. My main problem is that I have to keep switching synth paradigms, and my aging short term has a hard time tracking. Also, just for the aesthetics of it, I really prefer to work with one synth per composition. When I can.

John