KVR Audio

chk071 · Post by **chk071** » Fri Sep 23, 2016 12:46 pm

Numanoid wrote:
v1o wrote:Get over it. Monark is a couple years old now. Technology has moved on and the modelling has gotten better.
And how old is the original MiniMoog ?

Does the Minimoog feature analog modelling technology?

Yorrrrrr · Post by **Yorrrrrr** » Fri Sep 23, 2016 12:53 pm

chk071 wrote:Oh, BTW, i don't know if anyone mentioned it yet, but, Legend will be 4-voice polyphonic. Small bummer for me, i expected it to be at least 32-voice polyphonic, but, from what i gathered from the Reasontalk thread, it was a design decision from Richard, and is related to the small CPU footprint of the plugin. Which actually could be a major argument FOR Legend, because other emus, like Monark or Diva, take quite a lot of CPU power (especially Monark).

Indeed:

"The key idea in The Legend is to process large parts of the voices, in particular the CPU-demanding VCF and VCA, entirely in parallel using SSE. Since SSE does 4 things in parallel, this means 4 voices poly or unison can be computed in parallel. Note that this is a different concept from developers using SSE to speed up this or that portion of the code. What we do is to run just one large circuit simulation, which then returns the outputs of 4 voices simultaneously. This all happens on a single CPU core, so you fully benefit from your other cores still (e.g. if you run many instances of The Legend).

We could double the voices using AVX to 8, but not everyone has a processor with AVX at this point, and it is not supported in RE just yet.

Richard"

chk071 · Post by **chk071** » Fri Sep 23, 2016 12:55 pm

Yep, that's the part i meant in my post.

chk071 · Post by **chk071** » Fri Sep 23, 2016 1:53 pm

So, the release of the VST/AU versions will be very soon now, right?

Cinebient · Post by **Cinebient** » Fri Sep 23, 2016 2:20 pm

http://somefun.net/fun/wp-content/uploa ... 8svjeo.jpg

Elektronisch · Post by **Elektronisch** » Fri Sep 23, 2016 2:29 pm

Interesting read. How does VST version CPU use compare to lets say Diva (with multicore setting on)?

Synthetic Wav · Post by **Synthetic Wav** » Fri Sep 23, 2016 2:43 pm

chk071 wrote:
v1o wrote:
Synthetic Wav wrote:it seems there is no oscillator phase retrigger ? that would have been great to not have to sample it to get the same attack on the patchs it matter.

monark is not modelled on the same mini so it s probably why it don t sound as close to the mini the legend is modelled on.
Get over it. Monark is a couple years old now. Technology has moved on and the modelling has gotten better.
Don't think that's ths reason for the discrepancies. Rather the Minimoog model which was used to A/B plugin and hardware. There were similar discussion about Monark and Diva, when Monark was released. Obviously, the model Monark was based on had much more distortion in the signal path, than the one Diva was modelled on. Apart from that, Monark is still very advanced technically. There's a vast amount of detail in there.

monark does the audiorate stuff damn good . the real mini sound slighty better, more alive but it s already close
https://soundcloud.com/nativeinstrument ... comparison

aMUSEd · Post by **aMUSEd** » Fri Sep 23, 2016 2:51 pm

Synthetic Wav wrote: monark does the audiorate stuff damn good

Does Legend work at audiorate? (particularly for the RE CV modulations)

Ingonator · Post by **Ingonator** » Fri Sep 23, 2016 3:55 pm

"The key idea in The Legend is to process large parts of the voices, in particular the CPU-demanding VCF and VCA, entirely in parallel using SSE. Since SSE does 4 things in parallel, this means 4 voices poly or unison can be computed in parallel. Note that this is a different concept from developers using SSE to speed up this or that portion of the code. What we do is to run just one large circuit simulation, which then returns the outputs of 4 voices simultaneously. This all happens on a single CPU core, so you fully benefit from your other cores still (e.g. if you run many instances of The Legend).

We could double the voices using AVX to 8, but not everyone has a processor with AVX at this point, and it is not supported in RE just yet.

Richard"

I guess it is quite interesting how Legend scales the CPU use with additional voices. When activating the Unison (= 4 voices at once) or using the 4-voice Poly mode the CPU use is not 4 times that of a mono patch but less than 2 times the CPU use. While i got a quite old Core 2 Quad Q6600 CPU even at 4 voices and using a single Core this is not even remotely enough to overload the CPU here.
Actually the CPU load shown in Live 9 does not reach 30% here, even with effects and long release time. Also with 4 instanes of the same patch it stays the same. Only beyond 4 instances the CPU use inctreases but that is normal with 4 CPU cores.
Even 8 instances, each playing 4 voices (= 32 voices at once), are stil playable here. With more modern CPUs than this it should work even better. Soon i could maybe test it on an i5 Quad core processor.

Ingonator · Post by **Ingonator** » Fri Sep 23, 2016 4:14 pm

aMUSEd wrote:
Synthetic Wav wrote: monark does the audiorate stuff damn good
Does Legend work at audiorate? (particularly for the RE CV modulations)

I guess he means using Osc 3 and/or the Noise as a mod source and this of course works in Teh Legend too.
Would be quite strange if Richard work on small deatisl and did not get such essential things working properly.

As mentioend earlier like other Synapse synths (e.g. DUNE 2.5, Antidote RE and the RE filter module) this should use a zero-delay feedback filter where audio rate modulations usually work nicely (and filter self-oscillation too).

pdxindy · Post by **pdxindy** » Fri Sep 23, 2016 4:23 pm

Ingonator wrote:
As mentioend earlier like other Synapse synths (e.g. DUNE 2.5, Antidote RE and the RE filter module) this should use a zero-delay feedback filter where audio rate modulations usually work nicely (and filter self-oscillation too).

Plenty of audio rate modulations are not related to the filter so ZDF has no bearing on that... but filter stuff, yeah

v1o · Post by **v1o** » Fri Sep 23, 2016 4:23 pm

Numanoid wrote:
v1o wrote:Get over it. Monark is a couple years old now. Technology has moved on and the modelling has gotten better.
And how old is the original MiniMoog ?

The original MiniMoog is an analogue emulation?

Cinebient · Post by **Cinebient** » Fri Sep 23, 2016 4:31 pm

Could someone model a Telharmonium next?

Ingonator · Post by **Ingonator** » Fri Sep 23, 2016 4:55 pm

pdxindy wrote:
Ingonator wrote:
As mentioend earlier like other Synapse synths (e.g. DUNE 2.5, Antidote RE and the RE filter module) this should use a zero-delay feedback filter where audio rate modulations usually work nicely (and filter self-oscillation too).
Plenty of audio rate modulations are not related to the filter so ZDF has no bearing on that... but filter stuff, yeah

Cross modulation of Osc 3 with the other OScs seem to soudn quite similar in Monark and The Legend.
Anyway there seemed to be small corrections of some waveform compared to Monark so there could be differnces when using those too (which does not mean it is done "wrong" in The Legend compared to Monark and the real thing). As BigTone mentioned too certain waveforms could be have sligtl ydiferent in earlier and later Minimoog revisions and you should be able to use that with the Early/Late switch in the advanced settings.

Using Noise for audio rate modulation could sound quite different while i prefer the result in The Legend.
As BigTone mentioned a few pages ago emulating the Minimoog Noise properly, especially when used for audio rate modulation, was a quite difficult task and when done wrong does not sound anywhere near to a real Minimoog.
It seems to be obvious that in The Legend they managed to get this closer to the real thing than it was doen in Monark.
I am not even sure how far in Reaktor Core that is used to progeam teh modules for Monark it is even possible to properly emulate the Noise (opposing to write teh code from scratch using C++).

Some people might think you only need to get the filter right to model a Minimoog properly but obviously there is a LOT more involved to do the job properly.

chk071 · Post by **chk071** » Fri Sep 23, 2016 5:30 pm

A video i found on Youtube showing some more audio, and a general overview of the RE (with very silent commentary, as usual):

Really wished he would have done some clean sound examples though, instead of cranking up feedback and drive for every sound he makes...

Synapse Audio Minimoog emulation "The Legend" for VST/AU and RE released!