Why only 2 Oscs?

[Timo]

I'm a bit confused (I hark from the Trinity V3 era).

On my Trinity a "Program" was essentially a patch, based on one or two waveforms and these went through the ADSR filter and amp settings, and finally through the effects to produce the final patch. It was all self-contained.

So if I get my head around this, Program has been renamed to EXi (and can take the form of any instrument type), and two EXis (two old Programs) now make up a new Program (effectively a mini Combi [duet])?

Right, so, if each EXi is self contained (they have their own ADSR filters, amps and modulation stages etc., like an old Program would have), what architecture does the new Program (aka mini Combi) have? Does that have it's own ADSR filter, amp, modulation, stages etc. too?

I wish Korg would make their mind up what a Program is, lol. On a Trinity a Program was a simple self-contained patch and a Combi was several Programs; meanwhile on the Radias a Program is four timbres (which would class as a combi or ensemble to me); on the Kronos a Program is two EXis (a duet, or mini combi!).

So, does polyphony relate to the number of EXis you use, or the number of Programs you use?

Ok, going back to the first issue, I share the original poster's puzzlement about the restricted sub or the lack of a third oscillator in an AL1 patch (just as I do on the Radias). How come the sub is so restricted? Only triangle or square, and fixed to Osc 1 pitch -12 semitones. The Sub on the MOSS back in 1995 had saw, square, tri and sine, and the pitch (octave, semitone, and detune) was completely independent of osc's 1 and 2 as it had its own pitching page, along with being able to be modulated separately. It was like a third oscillator, and extremely useful too.



^ Dedicated sub-oscillator pitch page



^ Sub-osc pitch modulation sources

And the mixer page allowed independent amp modulations for the sub. You could attach any of the 4 LFOs or 5 EGs to it: http://i53.tinypic.com/52ylmu.jpg

On the Radias (and seemingly Kronos) when I need a third oscillator or a proper independent sub I have to duplicate the patch to a new patch and use it to home the sub and/or osc 3. But then when I tweak the overall sound I have to keep repeating my programming on both patches (like filter settings) to keep them together (osc 1/2/3/sub), as if they were one. Unfortunately this also halves the polyphony as you're using two patches at a time.

At least from the Oasys AL1 manual it appears Osc 2 retains it's own independent modulation of the waveform/pulse-width (like Moss did). On the Radias you can't modulate pulse width or anything on oscillator two, so oscillator two on Radias is like the Moss's old sub-oscillator.

[EvilDragon]

So, does polyphony relate to the number of EXis you use, or the number of Programs you use?

It's both. You might want to read the OASYS manual to get more insight.

[peter_schwartz]

FWIW, each main oscillators in AL-1 feature a sawtooth wave that actually outputs two sawtooth waves that can be detuned. In that case, two oscillators = 4 oscillators. And then there's the sub. So when it comes to sawtooth waves, that's 5.

Not too shabby.

[Randelph]

This is a bit off topic, but I never see the issue of 2 osc. discussed.

Having been in the Motif XS world, one of the things that seemed like an obvious advantage over the M3 was that the XS had 8 elements (osc) for each Voice (Program). This can have an advantage as far as having more articulations possible for each Voice/Program, like key off type noises.

But overall, my experience is the M3 feels more "expressive" than the XS, and my guess is that the main reason for this is how the two oscillator architecture is ideally suited for stacking up 4 way and even 8 way velocity splits. (not that the XS is incapable of this, but I didn't notice that happening that much, and because its laid out in a linear fashion, it's not as intuitive to create 4/8 way splits).

And what about the mono function where you can stack and detune a voice? Would that satisfy the OP?

And if AL-1 has an additional sub osc within a Program, does that eat one more voice of polyphony?

By the way Sina- you're enthusiasm about the voice architecture was inspiring to read- I hope Korg sets up a 3d party marketplace so that talented sound designers give us a lot to work with! As much as I love sound design, I want to spend most of my time building combis and composing, not getting super involved on the Program level. Gotta pick my areas of main interest, don't have time for all of them!

Randelph

[Biopharmer]

All of Korg's VAs, including the AL-1, only have 2 oscillators. Why? Would it kill them to add a third?

MOD-7 has 7 oscs (including PCM oscillator).

I know that. But MOD-7 isn't a VA.

[Sina172]

...

[ozy]

this thread is totally nonsensical.

a) an oscillator with a double sawtoooth is not two oscillators

b) a program which layers two couples of oscillators (with separate filters) doesn't model a 4 oscillator synth

c) a stack of 12 mono-oscillators synths is not a 12-oscillator synth.

d) a suboscillator is not an oscillator. it extracts a same-phase wave from the same OSC. It's a function of the same OSC, not another OSC.

e) synths (modular VCO, studio electronics SE1) which allow to extract and SUM more than one waveform per OSC don't have as many oscs as they have waves.

f) jump back many years in time: a single-oscillator synth with a chorus unit is not a multi-oscillator synth.

g) a polyphonic 8 voices synth with two OSCs, when put in "unison2 mode is NOT a 16 OSCs synth.

h) a layer of two 2-OSC synths in unison is NOT a 4 OSC synth.

etc etc etc

"Voices", "OSCs" and "waveforms" are NOT the same thing.

If you ever tried syncing oscillators, FM-ing them, etc,

you'd know that a two oscillator synth is a two oscillator synth, a three oscillator aynth is a three oscillator synth. Stop.

If you check the phase of the different waves generated by the same OSC, you'll see that stacks of the same oscillator are NOT different oscillators.

The answer to the thread's question is:

Korg VAs model two-osc synths because that's Korg's choice.

Their model was originally the Korg analogues (especially the two-osc MS20), and that architecture they emulated. Stop.

That architecture has its distinct sound which is different from a minimoog's (3 non-sinc osc) or from a prophet's (2 sincable osc).

Finally: "more OSCs" doesn't always equal "more synth sonic power", at all.

ESPECIALLY on VAs, where all oscillators are indeed generated by ONE oscillating device (the system's digital clock).

One single well built analogue OSC can sound fatter than 6 layers of VA "detuned supersaws".

Sorry for spoiling the long-distance spitting contest, but "number of oscillators", based on paper spcifications, has nothing to do with a synth's fatness.

Synth fatness is a function of harmonics (hence the relevant impact of filters on fatness). It is deeply related to phase cancelation.

So, the number of theoretical oscillators can mean nothing, or can even mean that a synth with "more oscillators" can sound thinner.

Ask yourself why it's so difficult to emulate a old, cheap, string machine with a VA synth, no matter how many saws you detune and how much chorus you add.

Try emulatins a ARP solina with a Roland jp8080 (the "supersaw" VA by definition), use your easr, and then do the math.

Rest my case.

[danatkorg]

I'm a bit confused (I hark from the Trinity V3 era).

On my Trinity a "Program" was essentially a patch, based on one or two waveforms and these went through the ADSR filter and amp settings, and finally through the effects to produce the final patch. It was all self-contained.

Yup. Same as a Program on the Triton and M3, and an HD-1 Program on the OASYS and KRONOS: two "Oscillators" in a Program. Each "Oscillator" can play samples, has its own LFOs, EGs, filters, amp, etc.; the "Oscillator" is a complete voice chain, and you have one of these in a Single Program, and two of them in a Double Program. Just like a KRONOS Program.

Personally, I wish that the term "Oscillator" wasn't used here ("Patch" might be better, or "element," or "layer," or...?), but it has a long history and we're stuck with it for now.

So if I get my head around this, Program has been renamed to EXi (and can take the form of any instrument type), and two EXis (two old Programs) now make up a new Program (effectively a mini Combi [duet])?

EXi are the EXpansion Instruments: AL-1, CX-3, STR-1, PolysixEX, MS-20EX, MOD-7, EP-1, SGX-1. Everything except for the HD-1.

EXi are played by EXi Programs, which are very similar to HD-1 Programs. Instead of two sample-playback "oscillators," an EXi Program has two slots for EXi. Like HD-1 "Oscillators," each EXi is a complete voice chain. This should all be familiar to a Trinity user.

The definition of what a "complete voice chain" is will vary from EXi to EXi, just as it does with different hardware and software synthesizers. In the PolysixEX, for instance, there's a single oscillator with a sub-osc, a 4-pole self-resonating lowpass filter, an amp, and a chorus effect. In the MOD-7, there's a sample playback oscillator, a noise generator, 6 VPM/waveshaping oscillators/processors, an audio input, two independent multimode filters which can be linked or cascaded, a bunch of modulatable amp stages and summing, 6 panners, and so on. Each SGX-1 voice plays two stereo samples at once, plus some other stuff. A single note on each is a single "voice," but MOD-7 voices cost about 4x more than PolysixEX voices, and the SGX-1 is roughly in-between the two. More on this below and in the next post.

Right, so, if each EXi is self contained (they have their own ADSR filters, amps and modulation stages etc., like an old Program would have), what architecture does the new Program (aka mini Combi) have? Does that have it's own ADSR filter, amp, modulation, stages etc. too?

Programs don't make or process any sound - no filters, amp, etc. They do have some modulation sources which can be used by all the voices in the Program's Oscillators or EXi, such as a "Common" LFO and the Vector EG. EXi Programs also have a Common Step Sequencer.

So, does polyphony relate to the number of EXis you use, or the number of Programs you use?

It relates primarily to how many voices you play (there are other factors, but in general they matter much less). One note may play many voices - especially in Combis and HD-1 Programs, or when using Unison in EXi Programs. The different synth engines use different amounts of processing power per voice. For instance, the SGX-1 uses approximately 1% of the available processing power per voice, while the MS-20EX uses approximately 2.5%.

See the separate post below for more info.

Hope this helps!

- Dan

[danatkorg]

I posted this elsewhere a little while ago, and since there were some questions about polyphony I thought I'd share it here.

* * * * * * * * *


The KRONOS has a Performance Meters page, which shows:

* Percentage of voice CPUs in use, and the amount of this used by dynamic voices and fixed resources, respectively. ("Fixed" means that part of the specific synth engine starts using processing power as soon as a sound is loaded, before playing any notes. These include effects built into the engine, such as the amps and effects in the CX‐3 and EP‐1, as well as features shared by all voices, such as the MS‐20EX’s external signal processor.)
* Voice stealing: meter shows realtime voice stealing activity
* Number of sounding voices: shows, well, the actual number of sounding voices
* Percentage of effects CPU in use
* Whether or not Smooth Sound Transition is in progress

Some manual excerpts:

* * *

The KRONOS uses a complex system for voice stealing, which can stop voices from one synth (such as the HD‐1) to make room for another (such as the PolysixEX). It also makes sure that you never have to worry about drop‐outs from CPU "overs."

* * *

Voice CPUs

This meter shows the amount of CPU power being used for playing voices. It uses two different colors to show the relative costs of voices and EXi fixed resources. The relative percentage of each is shown numerically to the right of the meter graphic.

The figures shown are a real‐time view of the actual CPU usage. Because of this, they may vary slightly even if you aren’t changing anything. Also, due to the complex way in which the KRONOS uses the CPU, the figures may sometimes not change by much (or at all) when the number of voices changes by a small amount, or when you add or remove EXi with fixed resources.

* * *

Number of Sounding Voices

This shows the total number of currently sounding voices. Voices will have different CPU costs depending on which of the synth engines they use. With double Programs, various HD‐1 options (see below), and layers and crossfades in Combination and Sequencer modes, a single key on the keyboard may play a large number of voices.

In the HD‐1, stereo samples use two voices each, Wave Sequences double the number of voices (two for mono, four for stereo), layered Oscillators (normal Double Programs, for instance) use separate voices for each Oscillator, and velocity cross‐fades between MS use the total number of voices for both of the crossfaded sounds.

* * *

Effects CPU

The KRONOS uses multiple CPUs (or cores, depending how you look at it) for playing synthesizer voices, and a separate CPU (or core) for effects processing including IFX, MFX, and TFX. Note that effects integrated into EXi use the voice CPUs instead, and show up on the meters as EXi fixed resources, as described above.

* * *

Smooth Sound Transition

This shows whether or not the system is currently in the middle of a Smooth Sound Transition (SST). While an SST is occurring, the Effects CPU meter will generally show higher levels of activity than normal.

* * *

Btw - note that the KRONOS uses a multi-core CPU, and there are corresponding differences from the OASYS both under the hood and visible/audible to the user. The biggest change is that on the KRONOS a CPU core is dedicated to effects, and so effects usage does not generally affect polyphony.

Hope this info is helpful!

[danatkorg]

Ok, going back to the first issue, I share the original poster's puzzlement about the restricted sub or the lack of a third oscillator in an AL1 patch (just as I do on the Radias). How come the sub is so restricted? Only triangle or square, and fixed to Osc 1 pitch -12 semitones.

As Ozy noted, that's how actual sub-oscillators work: they derive their frequency from the main oscillator, and so they're limited to octave multiples. This is exactly the way they work on analog synths (e.g., Roland Juno 6/60/106). If something has a "sub-oscillator" with independent pitch control, it's not really a sub-oscillator but rather a third (perhaps limited) oscillator.

So, why two oscillators + sub + noise generator as the AL-1 signal sources? Simple: because this is more than enough for creating most classic synth sounds. Two oscillators gives you FM, ring mod, and sync (and the AL-1's sync is excellent - most other VAs I've tried have real problems with aliasing when using sync). The Prophet synths had two oscillators; the third oscillator on the minimoog was *usually* purposed as an LFO (since that was the only LFO available).

We could have added third or fourth oscillators, but they would end up increasing the cost of normal voices. Instead, our thought was that if you wanted more oscillators, you were more likely to be creating a layer of different sounds, for which you'd want separate envelopes, LFOs, filters, etc. In that case, using a second AL-1 (or a different EXi, for that matter) makes more sense. It's absolutely true that if you want three independent oscillators going through the same filter, this structure means some duplicated effort and CPU inefficiency - but the assumption is that it's an unusual case.

Best regards,

Dan

[Megakazbek]

EXi are the EXpansion Instruments: AL-1, CX-3, STR-1, PolysixEX, MS-20EX, MOD-7, EP-1, SGX-1. Everything except for the HD-1.

What I can't understand is why HD-1 is made separate from all other engines. It doesn't make any sense at all, but only causes problems - we can't layer HD-1 with other engines in the same program, we need separate banks dedicated for it, etc - basically it completely doesn't follow the "rules" of the rest of the Kronos, for no reason at all.

[EvilDragon]

Layering is done better in Combi mode.


However, if we really need to have banks dedicated to one engine type, that's a serious drawback...

[SpIdErWeB]

Did Korg ever considered a modular system such the Legacy Cell but deeply (a-la Arturia Origin), to be able to use a HD-1 oscillator with a MS20 filter and the AL-1 step sequencer, etc...?

I know there's already some modular system with MOD-7, but I was just trying to think modularity in a large scale.

Phil

[danatkorg]

EXi are the EXpansion Instruments: AL-1, CX-3, STR-1, PolysixEX, MS-20EX, MOD-7, EP-1, SGX-1. Everything except for the HD-1.

What I can't understand is why HD-1 is made separate from all other engines. It doesn't make any sense at all, but only causes problems - we can't layer HD-1 with other engines in the same program, we need separate banks dedicated for it, etc - basically it completely doesn't follow the "rules" of the rest of the Kronos, for no reason at all.

It's an anachronism, I agree.

[danatkorg]

Did Korg ever considered a modular system such the Legacy Cell but deeply (a-la Arturia Origin), to be able to use a HD-1 oscillator with a MS20 filter and the AL-1 step sequencer, etc...?

I know there's already some modular system with MOD-7, but I was just trying to think modularity in a large scale.

Phil

Yes, we've definitely discussed this at various times.

(Legacy Cell doesn't let you mix and match in a modular way, actually; it just lets you layer together two sounds with effects, similar to to a KRONOS Program.)