software synths microtonal capabilities

In a discussion of microtuning capabilities of software synths:

>I've used Reaktor for microtonal work, but it requires a good degree of
>customization. I'm not aware of any microtuning capability in Absynth.
>I was likewise impressed with Midicode's built in tuning support. It is
>very well done.

I've recently been using Bitheadz Unity Session. It allows arbitrary
retunings via a (text editable) tuning table. Its not the most elegant
solution, but it does seem to work. Session's integration is well done
as well - after a few false starts I was able to run Performer
(a sequencer) into Max, and then into Session, and then route the sound
back into Performer - all at once.

The downside is that this seriously stresses a G3/500 mac and
only works for a few voices at once. The workaround is to record
a few voices, make some new ones, record them, etc...

--Bill Sethares

jacky_ligon wrote:

> Seriously though - this is the way I usually have to work with these
> things, and my CPU is well over twice this. There's just no way (yet)
> to use these softsynths in the manner that one would sequence
> hardware synths to have a whole arrangement going at once. I can
> sequence 32 tracks of midi on my 2 port card, but my machine would
> collapse trying to do it with softsynths.
>
> If there's anyone doing full arrangements without having to render
> the tracks as audio, I'd like to know what kind of monster machine
> they are using. I think eventually we'll get there though, but until
> then, I'm ecstatic (Hey Mike!!!) to be working with all this.

A base Kyma system can do around 40 voices of no-frills sampling. I can get 32 voices from a guitar
sample with guitar effects. One of the synths I've been working on tonight with an audio-speed envelope
and vibrato can only manage 5 voices at 48 kHz along with a filter and fairly good reverb covering all
voices. Extrapolating that up to 32 voices would mean around 8 expansion cards. Less if you keep the
reverb global, more if you want to do other clever stuff.

The mono legato problem's been half solved as well. You can always have the highest or lowest note
sounding, but not the most recent.

Graham

In-Reply-To: <acb6p5+t5ic@...>
Me:
> > A base Kyma system can do around 40 voices of no-frills sampling.
> I can get 32 voices from a guitar
> > sample with guitar effects.

Jacky:
> Sounds like a great tool. With that, one could seemingly do a pretty
> full arrangement. Now in this scenario, how complex would the timbres
> be?

Sampling is any timbre you like, but always the same one with linear
interpretation. The guitar sample with guitar effects is the melody line
on my magic demo, see <http://x31eq.com/music/> if you've
forgotten.

The numbers aren't that great. Note, though that Symbolic Sound claim 60
voices for a basic system

<http://www.symbolicsound.com/FAQ.html#anchor1689563>

whereas I can actually get 40 working. It's less than the claim made
about Reaktor with a standard PC.

Me:
> One of the synths I've been working on tonight with an audio-speed
> envelope
> > and vibrato can only manage 5 voices at 48 kHz along with a filter
> and fairly good reverb covering all
> > voices.

Jacky:
> What are the chances that you might be able to let us hear this? I'd
> love to check it out.

It'll escape sometime. The synthesis algorithm is the one used in my
IOCDrum piece, and in the decimal counterpoint examples. That's without
vibrato or reverb and does 20 voices.

Jacky:
> I'll bet 8 expansion cards would make one go bank-broke though -
> right?

It depends on what you have in your bank. If you're John Paul Jones, you
can carry two fully loaded systems on tour with you. The figures are:

$3300 for a basic system
$595 per expansion card
about $100 for delivery within the US

That makes, phew, $8160. For the difference it makes, you may as well
have the $270 firewire interface to make it $8330.

The prices haven't changed over the past four years (although firewire is
new) so it is behind commodity hardware of equivalent power. Good luck on
finding software to match.

> Oh - this is good, and useful. So it'll allow you to use portamento
> on your mono-voices to get woodwind-like gestures?

You can do portamento, yes.

> What do you have to do to tune it up? How complicated is it?

It's all done with Smalltalk expressions. I give a fairly complex example
at <http://x31eq.com/miracle/keyboard.html#kyma>. For a static
tuning table, something like

!Pitch is: (
`MIDIKeyNumber -60 //12 * 15
+ (
(`MIDIKeyNumber - 60 mod: 12)
of: { #(-1 0 1 3 4 5 6 7 9 10 12 13)}
)
"corrected to give pure 24-note octaves"
+ (`MIDIKeyNumber -60 //24)
* 12/31) nn + 4c nn + (!PitchBend*2) nn.

which is also more complex than it need be because there was a bug
stopping 24 element arrays working when I originally wrote it.

I can't remember how portamento works, but you can do it. Probably add
smoothing to the pitch parameter. Basic legato is an option on the
envelope. The trick for getting the highest note to re-sound when you
lift up a lower key is complex. It was given in a recent example.
Hopefully, if enough of us ask for it, it can be built into the
MidiMapper.

Graham

In-Reply-To: <acbisp+b9te@...>
jacky_ligon wrote:

> Ah - so it is able to do a kind of 'resynthesis'? So this would mean
> that one cannot assign samples freely to keygroups in the manner of a
> hardware sampler?

You can do that. I don't usually because I don't like abrupt changes in
timbre when you change keygroups. So I don't know how it affects the
polyphony. There are a whole load of other sampling algorithms that take
up a lot more of the processor. You can record a high note and a low note
and morph between them. Or you can hold the formant envelope constant
while changing the pitch -- this works well with live vocals. It also
does granular resynthesis with a lot of variations, which really needs a
better system than mine.

> How does the resampling sound as you play up in the higher ranges?

The notes get shorter if you've only got one sample. I've never seen this
as a problem.

> Not a possibility. Try not to buy gear that cost more than the car I
> drive!

I couldn't have afforded the base system *and* a car, but I obviously have
different priorities to you.

> J:L:
> Interesting. So when you do this, are you interfacing your computer
> to the Kyma, or other?

There's software running on the computer that lets you edit the presets.
Then it compiles to DSP machine code and loads it into the Capybara. If
you don't want to change presets, you can shut down the PC software so it
gets out of the way of other things you have running. So long as you
aren't streaming from the hard disk or anything like that. The firewire
box is supposed to remember the preset so you don't need a laptop for live
work, although I'm not sure of the details and may have misunderstood.

Any parameters you want to change in real time can be wired up to MIDI
faders.

> J:L:
> This is a fascinating issue. Robert Walker and I had some discussion
> a while back about the trouble with doing this with GM synths. If one
> can't hold down one key, and trill above or below the held down note
> without retriggering the attack of the sample, then it is not doing
> real mono-legato. In an acoustic instrument analogy, imagine how bad
> it would sound to do trills on real a trumpet or flute, and have the
> attack sound at the beginning of every note. My old analogs do it
> naturally, as do the TX series synths, but most GM modules are simply
> incapable of mono-legato, making them for all practical purposes
> useless for either emulating the sound of early analog synths, or
> monophonic wind instruments. Not only is this important for these
> things, but overall, it is a very important musical effect or
> gesture. Of course, there is some utility in the 'note-on-note-
> off/retrigger the attack on every note' type synths. One may need to
> mix them with other and more capable modules to keep things
> interesting though.

Kyma does have this "legato" option in the standard envelope to stop
retriggering. But the trigger doesn't have to be on a key down anyway.
You can trigger on a sustain pedal and only use the key presses for pitch
information if you like. Or use a pedal to turn legato mode on and off.

Remembering the other keys held down is harder, because the onboard
Smalltalk compiler doesn't have any kind of queue structure. I'm hoping
when my ZTar arrives (whenever that'll be) I'll find it resends note-ons
and I don't have to worry about it. I'm guessing the problem would
similarly disappear with a wind controller.

Graham