Diva Vs. Real Analog

[zerocrossing]

As a u-he thread grows longer, the probability of Lotuzia mentioning a Xils product approaches 1.

[Echoes in the Attic]

As any thread grows longer, the probability of Lotuzia mentioning a Xils product approaches 1.

fixed.

[ho66it]

Check out HGSounds Arp, step sequencer, loads of different arp modes, scales and chord mode. I use it with everything.

Is there a similar app for Mac?

Well maybe not as good, but free.
http://www.artovaarala.com/
There is also 2 good one's, each about $100usd here.
Thesys
http://www.sugar-bytes.de/content/produ ... hp?lang=en
and XX5
http://uisoftware.com/XX/

I have all these, and I love Thesys, but haven't done enough with it.
The XX5 is amazing, if it wasn't so buggy.
Neither of those is an arp though, but can do arp type patterns, as sequences.
I would say try the XX5 demo out, and if you have problems, then just select in options, not to load AU's, then you can just export the midi and do it up in your daw instead. It has great pattern creation tools.
For just the arp thing, the Kirnu is a step above most.

Thanks for the info! I'll check these out. Regards.

[galaxiesmerge]

Analog modular synths can do true noise and chaos because they are analog --- the real problem is that this is not always (only sometimes) musically useful.

Analog gear (and I don't mean pre-packaged standard synths, whether Rhodes Chroma, Prophet-T8 or other vintage) that is modular *is* different in behavior to digital patch emulations.

Here is my personal and perhaps simplistic viewpoint:

1) The resolution of current digital in the mainstream is insufficient to model analog circuit chaos, especially, bifurcation and sensitivity to initial random conditions. This can be done on special hardware like the Symbolic Sound Kyma Pacarana systems and others but not with current 32 bit implementations. And you can hear the difference with 32 bit systems on *repeated* perfomances of the same piece while repeating a piece (i.e. aleatoric modular music) is impossible with analog: in other words, with a first or even second listing you cannot hear the difference between quality and behavior between digital synths and analog but on the seventh, eight and hundredth listening , you *will* hear the difference (I encourage the experimenters to try this). To get a good emulation you need to move into 64 bits, high sample rates and non-deterministic algorithms (which can be done in digital, its just uncommon in the world of synthesis, though, in modeling and simulation, it is done, and random sources are used that are analog, like this web-link: http://www.random.org/ ).

2) Sampling (Nyquist's theory) tells us all you need is 2x the frequency - so for 20KHz listening you sample at minimum at 40KHz. The reality is that you need to *spread* out the noise floor, and when you move to 96KHz the noise floor is widened so you start to get that analog "quality" back again. For myself, and speaking personally here, when I run Kyma to 192KHz, it sounds analog to me and indistinguishable. When I link in an analog sampled noise source into Kyma, then there is no more pseudorandom sound (which on repeated listening, our ears can hear). So high sampling rates and random sources are about the *quality* of noise which is what makes analog special, under certain circumstances.

3) The best demo I have heard of orchestra string synthesis is by Synful but that is due to the behavioral modeling and this is what gives a certain performance character that our ears expect and find musical --- doing the same thing with synthesizers is also possible but it does require better user-interfaces --- imagine how much more creative you could be in your *performances* if you connect some hardware controllers to DIVA? Diva would become musically alive because through the hardware knobs, you give it life, personality and character --- relying solely on the keyboard keeps you on a much smaller playing field. And controllers are getting really cheap!!! Try the Korg Nano series for example (I use a couple of those too).

4) I am not an expert, I am not a critic and I am not confident in my knowledge since I may learn something new that changes my perspectives radically, so I only offer a slice of how I see things now: that the gap between digital and analog will come very close, but, ultimately, the analog Steinway and the analog Stradivarius will be there still because of the randomness, because of the performances never repeating, because of a live human being playing. Remember, no two concerts by Jean Michel Jarre ever sound the same - and he has a lot of synth gear!!! It's all about performing: unfortunately, musically useful performance features, such as, for example, bidirectional portamento (i.e. portamento that tracks the hands separately) and inter-note transitional timbre control are just not attended to by developers to the extent that they should be in order to help the player achieve new realms of expression.

5) All my points, 1-4, are again about musicality --- but we cannot ignore technicaility either. And I will tell you this: sure we all want the good stuff, but, speaking for myself, my greatest creativity emerges within limitations because it the limitations that set the context for finding ways to jump "out of the box" of those limitations --- and I realize that this may sound like a contradiction, but it is not really.

Great forum, lots of intelligent and talented people on here - thank you all

Any reactions are most welcome.

[Ingonator]

Concerning improvement of the "Low end": I just tried using Drive and/or EQ FXs with Diva like e.g. Camel Audio CamelPhat 3.5 (i almost forgot i got that plugin...), Schwa Oligarc Drive and Stillwell Vibe-EQ. The first tests with those sound very promising and it looks i could get results comparable with the filter overload in my Moog Slim Phatty. I'll investigate this further.

Like i already mentioned the ultimate solution would be to have something like the filter overload of the Phatty and/or a proper Tube Drive inside Diva.


Ingo

[galaxiesmerge]

perhaps, but at the moment, more cores does not help run a single instance of a soft synth...

True of DIVA at the moment but not true in general. To the degree that a synth implementer can introduce multi-threading into the processing, multiple cores can indeed make a consequential improvement in how much sound can be delievered by a single instance, be it more consecutive notes sounding or higher levels of rendering quality.

Urs has stated on the u-he forum that his initial attempts at getting multi-threading into DIVA didn't meet with much success. Don't assume that some inspired breakthrough won't happen for a future release of DIVA, however.

The problem with Multicore support is that I always optimised the memory usage of my synths for sequential crunching of voices, and the modules within the voices. This works well as long as instruction memory is small and audio memory is considerably large (Zebra's voices for instance render into 14 buffers, due to the modular approach).

DIVA however is strictly mono (small audio buffers) and has an excessive level of instruction memory. This actually favours per-module rendering, voice after voice. Like when one type of module is finished for all voices, render the next module for each voice.

Thus dividing cores by voices as we did is, uhm, suboptimal for DIVA (while probably great for Zebra). Hence we're investigating ways to restructure the voice scheduler into a different approach, which we think will scale very well across 2 or 3 cores for DIVA. As in, one core does VCOs, one does HPFs and another one does LPFs/VCAs. Something like that.

#---

While we're on the topic of Moores law, we're also looking into deploying the AVX instruction set of Sandy Bridge processors. This stuff can give our filters maybe a 30% performance boost.

We'll see...

Urs

Speaking naively on the subject, I would assume that you could profile your execution flow and perhaps take a monadic view of the problem: this means that you look at a thread of computation in terms of a flow graph between data and operators (as in data flow) and that the partitioning is done functionally by small groups of virtual "engines" (that act as your unit transformations) in runtime cooperative schedulers (and you need to look at making "engines" a first class entity upon which threads are built. In any case, the benefit is that unit operations (engines) stay in the CPU cache and they are restructured within the CPU cache to serve differing functional needs. This may be a bridge too far to cross now but something perhaps for the future to think about.

Another strategy is to refactor your mathematical basis --- you are likely using an approach in which you are doing a lot of number crunching: some of this might be eleminated and solved symbolically (and discretely) much faster. In mathematics this is referred to as a coordinate-free approach to simulation. For a good example, go and see www.geomerics.com on the Wayback Machine online at http://wayback.archive.org/ rather than their current site (they use the coordinate free Geometric Algebras to achiev 100x and 1000x speed increases in vector graphics processing for games).

And finally, to improve DIVA, well beyond hand unrolling of loops and other *tricks* in assembler, a smarter approach to giving people the tools to design modulators that operate on how notes transition - it is in modulation and transition in performance that musicality is most pronounced in my view. Of course, you will need to have a statistical model to help, but the lag of a note or two should not hamper things too much and it may open up a lot more creative freedoms.

Just ideas

[Nokenoku]

btw i am one of the ones who has been into arguments about analog vs softsynths for years on this forum..and always in favour of analogs.
softsynths ,till Diva came along, were not good enough for certain sounds (for 80s punchy sounds -which are my passion- i was FORCED to buy some analog).

What "80s punchy sounds" are you talking about exactly? Do you have some examples?

Edit:
To me this sounds pretty much 80s:
http://soundcloud.com/daataa

[Urs]

Could Diva be programed to run on CUDA cores haveing the graphics card help out would be a huge boost to computing but I'm not an expert. Just wondering about the posiblity of it. My i7 3.5 ghz with multithreading turned off already handles Diva pads in Divine mode at 48khz without a problem.

We'll try it. That stuff might require bigger latencies though, maybe 512 samples and up.

[Urs]

Just ideas

Good stuff!

The flow graph idea is something that we have in mind.

About the question of randomness - I think it's possible to get it there. I am however afraid of going too far, because people have gotten used to get the same result twice when hitting the play button as often. A software that does truely unique performances will need to communicate this to the user.

I'd love to have an in depth conversation about these issues once we've ironed out the last bugs and issues

Cheers,

Urs

[hakey]

As any thread grows longer, the probability of Lotuzia mentioning a Xils product approaches 1.

fixed.

Indeed.

That said, there's likely plenty of threads in which he hasn't posted anything.

However, outside of the u-he forum, I doubt there's a u-he thread in which he hasn't made an appearance in the last 6 months (usually with a reference to Synthix, or PolyKB, or more generally to drip poison - see the Diva thread at Gearslutz).

Anyhow, I'd like to offer the following slight amendment to my law:

As a u-he thread grows longer, the probability of Lotuzia spamming Xils-ware approaches 1.

[galaxiesmerge]

Just ideas

Good stuff!

The flow graph idea is something that we have in mind.

About the question of randomness - I think it's possible to get it there. I am however afraid of going too far, because people have gotten used to get the same result twice when hitting the play button as often. A software that does truely unique performances will need to communicate this to the user.

I'd love to have an in depth conversation about these issues once we've ironed out the last bugs and issues

Cheers,

Urs

Anytime
Just write me or my email (since I own Diva and Zebra) direct on deep technical issues.

[dsynth27]

Try this...http://www.moogmusic.com/products/tauru ... #demos-tab

Real Analog Vs. Diva(Vst's)

[Urs]

Try this...http://www.moogmusic.com/products/tauru ... #demos-tab

Real Analog Vs. Diva(Vst's)

Amazing, before listening to the product video I couldn't hear 20kHz - now I can!!!

[JoeCat]

...and inter-note transitional timbre control are just not attended to by developers to the extent that they should be in order to help the player achieve new realms of expression.

If I understand correctly, I think ACE's (u-He) mapping generator can be used for that - not random of course, but it's a step in the right direction.

Better midi-mapping software also would certainly help bridge the gap for hands-on control. I can't believe synth software development for plugs is being advanced so rapidly by small, independent developers (u-HE, xils, Madronna, etc.), while we struggle to get our midi controllers to fit into our workflow properly. It's a lot like smartphones - they can analyze a song on the radio, tell me who it is and offer it for purchase, tell me what street to turn on to get to Coyote Ugly, take HD video - but still can't make a phone call half the time without dropping it

[ZenPunkHippy]

Try this...http://www.moogmusic.com/products/tauru ... #demos-tab

Real Analog Vs. Diva(Vst's)

Amazing, before listening to the product video I couldn't hear 20kHz - now I can!!!