The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Synths for sale  Search Gearslutz Go Advanced
Analogue vs Digital - Diva and OB8 test Keyboard Synthesizers
View Poll Results: Which synth is First in the 8 bar cycle?
Diva
92 Votes - 51.11%
OB8
88 Votes - 48.89%
Voters: 180. You may not vote on this poll

Old 24th July 2014
  #601
Lives for gear
 
maisonvague's Avatar
 

Quote:
Originally Posted by SWAN808 View Post
thing is with Diva - with correct programming I think you can often get closer
That's just it, SWAN: vulnerability to accusations of "pilot error" will always exist making it impossible for any test to be conclusive. In situations where Diva is seen unfavorably, it would be because it was not programmed properly.

So further tests seem pointless. Just accept it. Diva is indistinguishable from genuine analog. Diva now sounds better than the instruments it was originally designed to emulate. Digital analog has become superior to genuine analog.

"Better - than - human. We - are - better - than - human. Made - by - human. Now - better - than - human."
2
Share
Old 24th July 2014
  #602
Urs
Lives for gear
Quote:
Originally Posted by pendejo View Post
To my knowledge Diva emulates the JP8 oscillators, filters and envelopes...so that's what I will use in straightforward comparison.

You want to try something else...knock yourself out.
If your JP-8 doesn't come with built-in MIDI, you may need to look elsewhere. Diva's parts are modelled after a Revision 2. Many of the popular "anti-Diva" comparisons are done with the earlier JP-8 model (or even an MKS-80) which has a completely different filter structure, more like the JP-6.
1
Share
Old 24th July 2014
  #603
Lives for gear
 
grumphh's Avatar
 

Quote:
Originally Posted by prontold View Post
it means you cannot reject the null hypothesis because you are too likely to have seen a result caused by random chance.
Sorry - but a 50/50 split is not "likely" to be a result of random chance, it is random chance.


The problem is that for every individual who thinks they got it right there is one who got it wrong even though he thought he got it right.

So statistically there is no reason to believe that the people who got it right did so by anything other than the same random choice than the people who got it wrong. heh

...of course, this is something you will never get those that happened to answer correctly to admit
1
Share
Old 24th July 2014
  #604
Urs
Lives for gear
Quote:
Originally Posted by jimmyklane View Post
Urs, this seems like an opportunity to learn from a master: I can see how a voltage controlled current source (OTA) can be modeled simply, but my above point above was relevant to direct DSP-based real-time COMPONENT LEVEL modeling. Johnson noise in carbon comps, tolerance dependent non-linearities of capacitors at audio rates, and the individual distortions and slew limitations of active devices....not just a thevenin equivalent or generalized equation as is currently possible, but literally an audio path determined with a true component level model. I'd love to see an "aging" knob where each component breaks down randomly per instance. I'm sure with an awful lot of work you could describe each component mathematically in the individual sense, but computing the audio and control paths in real-time can't be possible with current processing power available to anyone outside of university CPU farms! My specialty was analog filter design, so what's possible with modern DSP is not something I'm truly informed on.... So if I'm wrong, please educate me!
Well, one has to naturally choose a level of detail here. We only model aspects that contribute to sound, not noise or heat (and I haven't even heard of Johnson noise). You're absolutely right that one can't run a SPICE model with all known-to-mankind properties of electronic components in realtime. However, one can often model the effects of those underlying properties with a few simple tricks. As long as the bahviour is correct then, the question of whether or not the implementation was correct is purely academic.
1
Share
Old 24th July 2014
  #605
Urs
Lives for gear
Quote:
Originally Posted by Autumn Leaves View Post
I've just found another comparison:

Roland Jupiter 8 vs U-He Diva - Filter wide open: (1. JP-8 --> 2. Diva)

Yup, CoolColJ has a revision 1 Jupiter-8. That one has way more instabilities than the later models due to its poor 12-bit controller and it has a different filter architecture. The differences are much like the differences between certain revisons of the MKS-80.

Funnily, those examples have become the classics in proving that Diva doesn't sound right (even more so than the comparisons with early Minis). I'm still uncertain if these attempts are just unfair or if they are desperate.
Old 24th July 2014
  #606
Lives for gear
 
maisonvague's Avatar
 

@Urs

Just so you know, I mean you no disrespect with my joking. In fact, it's an honor to be able to participate in this discussion with you.

My honest feelings regarding Diva are this: it's an impressive bit of code that's a milestone in the history of analog emulation. But I find the notion that it sounds even better than what it was designed to emulate in the first place difficult to accept.

My views are from the perspective of a professional musician. I'm not a programmer. I can't code algorithms. I can play synths, but not make them. Based on years of experience working with analog synths I trust my own ears and my own experience (and while I recognize the pitfalls of trusting ones ears, I still find them a most valuable tool as a musician. )

Diva is good. Very good. But it's not the final solution to all the problems involved in the digital emulation of analog sound. I'm sure, when pressed, you'd agree (though I recognize -- and appreciate -- the risks you take to participate in these discussions).

Obviously not everyone shares my views. Plus, it would seem more and more people are coming to actually prefer the analog idealism of digital instruments like Diva.

So, congratulations!

Clark
2
Share
Old 24th July 2014
  #607
Urs
Lives for gear
Well, I sure hope that Diva is distinguishable from analogue, in many aspects. Otherwise we wouldn't have done a proper job. It was a conscious decision not to emulate the noise of the Juno-60 chorus. It also was a conscious decision to change master/slave roles in oscillators. It furthermore was a conscious decision to increase the possible input gain of the MS-20-style filters.

I do think however that, in musical context she's indistinguishable from those classic analogue synths. If someone came into our studio and produced an electronic album twice, once with our analogue gear, once with Diva, we'd get a similar 50/50 outcome as in this thread.
2
Share
Old 24th July 2014
  #608
227861
Guest
Urs

Sorry to plug my video demo again but can Diva do every sound in my demo? It's the Juno 60 with a wide variety of sounds including warmish toward the end with a decent amount of bottom end. That's where I find soft synths to fail in addition to the filter sweeps. Please check it out and let me know.

I honestly think not but maybe you can prove me wrong. These aren't the most complicated patches but I do think they have the unique Roland character.

Again apologies for the self plug.

Old 24th July 2014
  #609
Lives for gear
 
maisonvague's Avatar
 

Quote:
Originally Posted by Urs View Post
I do think however that, in musical context she's indistinguishable from those classic analogue synths. If someone came into our studio and produced an electronic album twice, once with our analogue gear, once with Diva, we'd get a similar 50/50 outcome as in this thread.
That would be an interesting experiment!
1
Share
Old 24th July 2014
  #610
Lives for gear
 

Quote:
Originally Posted by Urs View Post
I haven't even heard of Johnson noise
Seriously?


But I've also got to ask why you choose to model the more modern revisions of synths, when people sonically prefer the older ones? And where can I listen to comparisons between those "golden" hardware analog synths vs. Diva?

Whenever someone uploads a comparison that shows an obvious difference, with the analog synth displaying the characteristics people like me really like in these synths, this always turns up to be just the version that has not been modelled or is different for some reason or another...
6
Share
Old 24th July 2014
  #611
Urs
Lives for gear
Quote:
Originally Posted by draven5 View Post
Urs

Sorry to plug my video demo again but can Diva do every sound in my demo? It's the Juno 60 with a wide variety of sounds including warmish toward the end with a decent amount of bottom end. That's where I find soft synths to fail in addition to the filter sweeps. Please check it out and let me know.

I honestly think not but maybe you can prove me wrong. These aren't the most complicated patches but I do think they have the unique Roland character.

Again apologies for the self plug.

Nope, not even our Juno 60 sounds like this. There's way too much saturation going on.
Old 24th July 2014
  #612
Urs
Lives for gear
Quote:
Originally Posted by living sounds View Post
Seriously?


But I've also got to ask why you choose to model the more modern revisions of synths, when people sonically prefer the older ones? And where can I listen to comparisons between those "golden" hardware analog synths vs. Diva?

Whenever someone uploads a comparison that shows an obvious difference, with the analog synth displaying the characteristics people like me really like in these synths, this always turns up to be just the version that has not been modelled or is different for some reason or another...
It's getting tiresome. Hans Zimmer has had 4 Minimoogs at the time. He sent us the one he prefers over all others. Which Minimoog do you think we should have used for reference?

The old Jupiter 8 sounds like a Jupiter 6 with analoge envelopes. Use the Jupiter 6 setting instead to get you pretty close. However, don't expect the same steppyness in as the JP-8 rev 1 for the Wheels and tuning of both VCOs and VCFs. Maybe that's why we didn't want to do it?
3
Share
Old 24th July 2014
  #613
Lives for gear
 
Eigenwert's Avatar
Quote:
Originally Posted by grumphh View Post
Sorry - but a 50/50 split is not "likely" to be a result of random chance, it is random chance.


The problem is that for every individual who thinks they got it right there is one who got it wrong even though he thought he got it right.

So statistically there is no reason to believe that the people who got it right did so by anything other than the same random choice than the people who got it wrong. heh

...of course, this is something you will never get those that happened to answer correctly to admit
You are oversimplifying things. Take that example of the McGill students I mentioned twice:
At McGill they teach you to hear slight EQ curve appliance virtually NO ONE can hear before successfully going through their lessons.
A group of McGill can hit such an EQ curve above random chance. This is well-known, as it is a fact that the difficulty increases just as soon as the hit rate of a student gets above random chance (the whole program highly relies on science).
One McGill students among a group of GS posters would not contribute to the result enough to push the result above random chance.

This already has been explained to you several times before.
2
Share
Old 24th July 2014
  #614
Lives for gear
 

The bottom line is this....this was a fair exercise, we participated, some were right, some were wrong, but if you were wrong, don't be butt hurt about it, be an adult, and just accept you were wrong, and others were right. Some of you sound as though you want to take your ball and go home. Maybe, you should.
1
Share
Old 24th July 2014
  #615
Gear Guru
 
Yoozer's Avatar
Quote:
Originally Posted by maisonvague View Post
Diva now sounds better than the instruments it was originally designed to emulate. Digital analog has become superior to genuine analog.
I think the "superior" should be seen in context. As an example; some people have owned multiple 909s or 808s. Out of them, one may be preferred because it has been used/aged/calibrated in the way they like it. You know the tales - "the kick of nr 3 is better so that's why I kept it". These differences may be minute in terms of parts but people will pick the one they like best. Is that unit then superior? No, because the same kick with more bass may have a less snappy attack but it won't matter for that kind of music.

Likewise everyone's going to be influenced by their preferences; there is the "album Rhodes" sound or the "album Moog" sound which may be effected by EQ, compression or tape; if that album resonates with the listener it's going to be a kind of benchmark for the direction they push the sound towards, so if a choice needs to be made with regards to tweaking the model, that "imprint" of that sonic aesthetic may determine the direction. Hence "superior" is kind of an unfortunate word, but it still is the right term to use. After all, making music is all about producing something larger than life.

This is something I've wondered about myself; a synth or plugin dev is going to have certain preferences towards music. Put two coders together; one is a fan of Stevie Wonder, the other likes Einsturzende Neubauten, and you are going to end up with two - likely different - models of the same analog synth.
Old 24th July 2014
  #616
Lives for gear
 
grumphh's Avatar
 

Quote:
Originally Posted by Eigenwert View Post
You are oversimplifying things. Take that example of the McGill students I mentioned twice:
At McGill they teach you to hear slight EQ curve appliance virtually NO ONE can hear before successfully going through their lessons.
A group of McGill can hit such an EQ curve above random chance. This is well-known, as it is a fact that the difficulty increases just as soon as the hit rate of a student gets above random chance (the whole program highly relies on science).
One McGill students among a group of GS posters would not contribute to the result enough to push the result above random chance.

This already has been explained to you several times before.
But the entire point is that these people conduct experiments that do not result in 50/50 results heh heh heh

That is why they have credibility and GS'ers have none
1
Share
Old 24th July 2014
  #617
Gear Guru
 
Yoozer's Avatar
Quote:
Originally Posted by keybdwizrd View Post
I find myself thinking that you could post two recordings that were secretly identical, and a brawl would erupt here over which one sounded better.
Or the OP-X plugin vs Diva, and then see the "well I recognized it from a mile away" people disappear. This however is a really good way to get lots of people *really* pissed off at you (but it does reveal something about the psychological component of it all).

Edit: The initial 50/50 divide of the votes has no statistical significance since it is only one test with a 50% chance of guessing it correctly. Now, if you have a demo matrix with multiple variations and multiple sounds, you can point out the folks with the golden ears.

Edit 2: man it would probably really suck to get into that subtle EQ class with a cold. Everything sounds muffled!
1
Share
Old 24th July 2014
  #618
Urs
Lives for gear
Quote:
Originally Posted by Eigenwert View Post
At McGill they teach you to hear slight EQ curve appliance virtually NO ONE can hear before successfully going through their lessons.
I guess we're lucky then that the majority of people who use Diva make music for an audience that can still enjoy listening to music.

It must be really hard to have those super powers. The beauty of music must be completely lost to those people.
Old 24th July 2014
  #619
Lives for gear
but of course, an instrument is ALL about inspiration.
I have both, and know I will never be able to write or reach the sounds I do with a plugin, than I do with the OB8. Every time i sit down with the OB8 I come up with a new idea, and I reach far more depths of synthesis on the OB8 as well. It's a feeling as much as anything, it makes you want to create and explore, or sometimes just write.
I didn't start on HW and move over to SW, I have always used both... both naturally have their place.

Stacking up you start to hear the nuances less and less, but in more stripped back / ambient music, paired with a great pre amp, the OB does take on more 3 dimensionality, life, and general organic 'real instrument' sound.

Couldn't recommend either enough though, which is why blind tests are kinda silly IMO
Old 24th July 2014
  #620
Urs
Lives for gear
Quote:
Originally Posted by maisonvague View Post
Ouch.

I realize you're only human -- and humans like to play with fire -- but if you want me to continue cutting you some slack in light of your (understandable) lack of objectivity in this discussion, I highly recommend not going in this direction.
I don't get your point. Whoever brought up "only people with special skills (that no-one else here has) can really understand the problem" is the offender. It's yet another deus ex machina that's used as a straw. I just enjoyed making fun of it.
2
Share
Old 24th July 2014
  #621
Lives for gear
 
pointsource's Avatar
 

Quote:
Originally Posted by Urs View Post
I guess we're lucky then that the majority of people who use Diva make music for an audience that can still enjoy listening to music.

It must be really hard to have those super powers. The beauty of music must be completely lost to those people.

Awesome post, Urs.

If I were you I wouldn't even reply to those "super heros" hahaha


I am starting to sell all my modular and I'm going back to software again with DIVA and Zebra. No point in having all this gear if you can do the exact same thing with the software. The people who enjoy the music certainly isn't trying to figure if someone used a software or hardware to create a sound.
Old 24th July 2014
  #622
Lives for gear
 

Quote:
Originally Posted by maisonvague View Post
Ouch.

I realize you're only human -- and humans like to play with fire -- but if you want me to continue cutting you some slack in light of your (understandable) lack of objectivity in this discussion, I highly recommend not going in this direction.
**** just got real son!
1
Share
Old 24th July 2014
  #623
I tease my mother. She believes that at specific times, she has experienced a sixth sense. A clairvoyance if you will.

One time, when I was quite young, I was out late, and she felt a growing concern. She was certain something terrible had happened to me. Well, I had gotten piss drunk, and picked a fight with three men, all larger than I. Lets say, it didn't turn out well for me. She claims to this day, that she knew that it had happened. I have asked about other times that she had experienced anxiety about me being out late, and had come home unscathed. However, she only recalls the instance where her suspicions were confirmed, naturally. This is called CONFIRMATION BIAS.

I too made a choice. I felt that the bass on the second part sounded more precise....or something sounded a little looser on the first, whichever you like. Naturally, like my mother, when the reveal went down....had I been correct, I would say to myself, "I knew something about that bass seemed different." Confirming why I had made my correct choice.

However, with the results being as close as they were, I could no longer pretend that my selection was based on anything other than chance.

Its not about being right or wrong, its about interpreting data. Of course there are variables, and no study which involves human self-report is without some problems.
1
Share
Old 24th July 2014
  #624
Lives for gear
 
maisonvague's Avatar
 

Quote:
Originally Posted by Urs View Post
I don't get your point. Whoever brought up "only people with special skills (that no-one else here has) can really understand the problem" is the offender. It's yet another deus ex machina that's used as a straw. I just enjoyed making fun of it.
Never mind. It's not important. I deleted the post.

Old 24th July 2014
  #625
Lives for gear
 
Synthbuilder's Avatar
Quote:
Originally Posted by Urs View Post
Well, I sure hope that Diva is distinguishable from analogue, in many aspects. Otherwise we wouldn't have done a proper job.
Please then can we have the corrected Korg35 emulation put into DIVA? Call it number 3. I'd love to be able to have the bite that the MS-10 and early MS-20 have.

And... can we have another polyphonic assignment mode that re-triggers the same voice if the same note is pressed again? Like the Juno-6 for example. The current cyclic assignment in Poly 1 is brilliant but for DCO based synths and string machine emulations it is just a little too lively. Poly 2 doesn't do it because all new notes will re-trigger the same voice.

Tony
Old 24th July 2014
  #626
Quote:
Originally Posted by Urs View Post
Well, one has to naturally choose a level of detail here. We only model aspects that contribute to sound, not noise or heat (and I haven't even heard of Johnson noise). You're absolutely right that one can't run a SPICE model with all known-to-mankind properties of electronic components in realtime. However, one can often model the effects of those underlying properties with a few simple tricks. As long as the bahviour is correct then, the question of whether or not the implementation was correct is purely academic.
So, what I'm getting from you here is a bit of "diminishing returns doesn't justify the effort", and "It's is possible(and done by uHE)to model the aspects of all these devices that affect the sound generating circuitry." Everything else is geek-speak and not truly necessary for the task at hand. My own takeaway is that I need to learn some DSP technique from this decade!
Oh...and also, Johnson Noise is the term engineers use for the random movement of electrons(self-noise) in a resistor(or any passive element with resistance, so anything :-) )...
Old 24th July 2014
  #627
Urs
Lives for gear
Quote:
Originally Posted by Synthbuilder View Post
Please then can we have the corrected Korg35 emulation put into DIVA? Call it number 3. I'd love to be able to have the bite that the MS-10 and early MS-20 have.

And... can we have another polyphonic assignment mode that re-triggers the same voice if the same note is pressed again? Like the Juno-6 for example. The current cyclic assignment in Poly 1 is brilliant but for DCO based synths and string machine emulations it is just a little too lively. Poly 2 doesn't do it because all new notes will re-trigger the same voice.

Tony
Yep, the Korg35 was messed up, but mostly the high pass. The lowpass should still be pretty good. Wasn't the MS-10 rather similar to the older Korg 700? Not sure... in any case, the oscillator gain knob needs to be much lower than on a real MS-20 to get the right sound.

The voice control code is currently being rewritten from scratch. This might make it easier to implement new methods, and funny enough, it currently works the way you describe (that's a bug, but I'm onto it).
1
Share
Old 24th July 2014
  #628
Lives for gear
Quote:
Originally Posted by grumphh View Post
Sorry - but a 50/50 split is not "likely" to be a result of random chance, it is random chance.


The problem is that for every individual who thinks they got it right there is one who got it wrong even though he thought he got it right.

So statistically there is no reason to believe that the people who got it right did so by anything other than the same random choice than the people who got it wrong. heh

...of course, this is something you will never get those that happened to answer correctly to admit
grumphh, the correct statistical interpretation is that you can't reject the hypothesis that all guessing was due to random chance, not that you can say with certainty that all guessing was due to random chance. So your first sentence is analytically incorrect. Your third is closer to the correct interpretation.

Statistics is the theory of making estimates based on incomplete knowledge. The whole point is estimating how certain you can be based on the data available. It would never make any sense to conclude from data that shows no strong association between independent and dependent variables (which are kind of hard to identify here) that you are 100% certain there is no pattern of association in the real world. There could be a pattern that your data structure fails to capture that a different choice of variables would reveal (e.g. what happens if you add in a poll choice which reflects a less complicated differentiation between the two sounds (eg brighter or darker)... does this allow patterns of consistent identification by listeners to emerge? The point being that differences may be audible, but everybody may hold a different conception of what an OB8 or a Diva is supposed to sound like). The correct interpretation is to say that you can't reject the possibility that there is no pattern in the real world.

Anyway, musically I think this is mostly irrelevant. I wouldn't care if I was listening to a track made with Diva vs OB8 (though I would be more excited personally about playing the Oberheim). I just get a bit ticked off when people make arguments that sound educated/logical without properly representing the bodies of thought they are referencing.
Old 24th July 2014
  #629
Lives for gear
 

Quote:
Originally Posted by pointsource View Post
Awesome post, Urs.

If I were you I wouldn't even reply to those "super heros" hahaha


I am starting to sell all my modular and I'm going back to software again with DIVA and Zebra. No point in having all this gear if you can do the exact same thing with the software. The people who enjoy the music certainly isn't trying to figure if someone used a software or hardware to create a sound.
Anyone inspired to make music will want to know what the people who inspired them are using. I'm quite used to audience members asking me what equipment I'm using after shows and even what software I'm using.
All I get from the rational that the audience doesn't care is musicians are making uninspiring music for a less than passionate audience.
1
Share
Old 24th July 2014
  #630
Urs
Lives for gear
Quote:
Originally Posted by jimmyklane View Post
So, what I'm getting from you here is a bit of "diminishing returns doesn't justify the effort", and "It's is possible(and done by uHE)to model the aspects of all these devices that affect the sound generating circuitry." Everything else is geek-speak and not truly necessary for the task at hand. My own takeaway is that I need to learn some DSP technique from this decade!
Oh...and also, Johnson Noise is the term engineers use for the random movement of electrons(self-noise) in a resistor(or any passive element with resistance, so anything :-) )...
Hehehe, well, we haven't really started from scratch when we built our models, mostly also because we're not EEs. We had to rely on research that others have done before us, so we read papers about transfer curves and effects that people with academic background rated considerable for audio applications.

The actual thing we did was, figure out how SPICE (or, QUCS, which we used) solved non-linear equations and how it does the integration step. Then we applied that method to the equations that were accepted as academic status quo in the field of DSP. In some cases we observed that the outcome was completely wrong, and then we indeed performed our own circuit analysis - until we had a model that behaves well enough.

Once we had the models - which certainly were idealized - we developed "concepts for grit" that we added into those equations to line up our measurements and audible results from the analogue circuits with the results of the computation. That is, we have sat down for months and tweaked 50 parameters that add little offsets here, little offsets there.

Once the difference was below, say, 60dB, the result was within the tolerance level of one unit to another, i.e. the variance between two of those synths were within the error margin of our own solution.

Therefore, we were able to do what SPICE does, in realtime, in a simplified way, yet mathematically correct. The aspects that we had to neglect were modeled by faking them with educated reasoning.

- Urs
Topic:
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump