Diva Vs. Real Analog

[Gonga]

Funny is, that I love both sounds for what they are. And the raw screaming qualitites of the Omnisphere sound is more appealing to me. Minus the aliasing that is which is very loud at the high notes.

I know exactly what you mean - Omni in this mode is quite an attention-grabber, and a lot of fun to play.

[Gonga]

I used Diva for bass in a demo this week and it sounds gorgeous. I have a theory about analog bass that goes something like this: the best analog bass sounds so good partly because of the attack (minimoog perfect example - explosive attack). When people think a VA's bass is weak, I suspect it's the emotional impact of the attack that's a common factor in most synths.

For me there's only one basic sound area now where Diva is surpassed by any other VA (Zebra), and that's attack. But Diva's attack is still plenty good. I will enjoy Diva a lot this year, but also look forward to future Zebra innovations.

[Syncretia]

Here are some of my fiddlings with Diva:

http://www.kvraudio.com/forum/viewtopic.php?p=4783094

Will it be possible for every copy of a softsynth to have its own sound like an analog synth?

It's funny. I suppose in a sense this is an argument for analog - but's an outdated one. The difference you hear from one analog synth to the next is based on a few variables which create "randomness". These variables include things like heat, out-of-tune-ness, faulty parts or parts that have been replaced and are non-standard, and so on. But, also voltage is just inherently more random than a digital signal. Basically though, all you need to do to attain this variation is to fiddle with a few knobs or change things around. You can subtely LFO parameters to emulate randomness, but I'm guessing there is built in randomness in Diva anyway.

You cannot recreate real voltage topologies ( vco,vcf, vco.)digitally .There are way too variables

Rubbish. Diva may only be 90% of the way there now, but this is a very naive statement. If Diva isn't all the way there now, computers in future will do it with ever increasing accuracy. People ten years ago were saying that digital animation would never look "real", or "organic". That was true ten years ago, but look at the progress. If you have seen Tron Legacy you will have noticed how the creators brought back a rendered image of Jeff Bridge's young face. This animation is a vast improvement over animations made only years before such as Final Fantasy etc. 3D computing animation has only been around for 20 years or so, and it can almost fool us now. In 10 years time, you won't be able to tell the difference between computer generated animation and video photography.

Sound is a lot less complex than this. Anyone who understood what modern computers are capable of these days would not make claims like this.

[sounddesigner]

It's my impression that cpu's are not keeping up. They are adding more cores, but a single core is not that much faster than a few years ago... right?

Actually they are and quite a lot. It is true that much of Moores Law has been used to cram more cores in there but the cores gets faster and faster too.
The 2600 is about 5 times as fast as a Q6600.

That's far from accurate according to ADK benchmarks. The benchmarks can be a bit misleading (but still informative if you know how to read them), this is why i asked Scott and Vinny to not include hyperthreading for one of the processors listed since its still more cores (Virtual) and thus won't show how much true-speed gained. You have to compare current Quads to the QX6700 Quad NOT a 6-core, you can't use hyperthreading or overclocking on current computers when comparing to judge power gains overtime either or it won't be accurate.

The QX6700 can load 54 rxc comps at 32 samples buffersize. The i7 920 can load 130 with hyperthreading but only 105 without. Without HT the 920 is only about 2x the power. The sandy bridge 2600 with HT is 183 RXC's. Sadly Scott no longer benchmark without HT but a educated guess would put Sandybridge Quad at around 150 (it won't be too far from that number) wich is slightly less then 3x the power at 32samples buffer after 5 years time. Keep in mind this slightly less then 3x the power is only at 32 samples buffer-size, when you go to the higher buffersizes where most people work it's about 2x the power (i7 2600 is 212 RXC's with HT and less then that without HT at higher buffersizes, the QX6700 is 87 RXC at higher buffers). Even with HT on your numbers would be off since the sandy bridge I7 2600 quad can only load up to 183 RXC's at 32 samples and 212 at higher latencies. You cannot compare the QX6700 Quad to a 6-core computer with HT enabled and overclocked or you WILL NOT be reading the true-speed increase PER CORE gained overtime with computers that's needed for better plugins and Low-latency performance. Much of what you'll be seeing with the wrong comparisons is'nt true-power gain but substitute-power gain and like everything else in life substitutes are rarely as good as the real deal.

Sure computers have gotten faster but it's nickles and dimes here and there and over a long period of time (over 5 years since the QX6700 was released) can add up, but the True-power (More GHZ and better Architecture for the GHZ) is not comming fast enough for many users and developers (Izotope, Madrona Labs, etc all had to make seriouse compromises due to computers not being fast enough with True-power, most of what we get is substitute-power). More physical cores, hyperthreading (virtual-cores that don't add to true-speed of each core thus still don't help in some situations), turbo-boosts (close but no cigar, we had to turn it off for socket 1366) are all substitute-power and mainly what we are getting. We only have about 2X the true-power gain overtime (more then 5 years) at higher buffer-sizes.

I hope Scott and Vinn will start to benchmark computers without HT in additiion to with HT and compare the true-speed/power gained for each individual core of latest computers overtime to the QX6700 wich we know is a Quad that's without HT and loads 54 RXC's at 32samples Buffer-size and 87 at higher latencies. The QX6700 was released over 5 years ago and is the first Quad and best for comparing to deduce true-power gains.

Here is a link to ADK's benchmarks for Audio, it's the second from the top. The i7 920 without HT is no longer listed but i have a copy of the first benchmarks wich have it at 105 RXC Compressors at 32samples buffer-size. Everything else is listed- http://adkproaudio.com/benchmarks.cfm

[goldenanalog]

Well, people will start making parts if there is a demand and maybe Korg with it's Monotrons will help generate a new market.

As for Tempest, they shot themselves in the foot on that one and it's nobody's fault but their own and has nothing to do with the analog nature if it's audio engine (which is hybrid anyway).

First of all, they've yet to put any good audio examples on the site for it. Why? So far I've only heard one single good audio demo from a user. Everything else sounded... well not even as good as software and that good demo came out after some volume bug got correct so I suspect something else was going on with the gain structure that was causing less than stellar audio.

Second, speaking of bugs, they released what was clearly a pre-beta unit and even that was late. How you could release a drum machine in 2011 that didn't sync or respond to MIDI is beyond my powers of comprehension. I got scared away when they missed their first ship date and cancelled my preorder. I don't know much about hardware manufacturing but I can't imagine it was the right move to release it in an obviously unfinished state. Explain to your pre-orders that you ran into issues and you're not going to release a broken product and I think they'll understand. They made it seem like syncing an analog drum machine to a MIDI clock was some ground breaking coding task... yet the Germans seem to have cracked that issue in their $450 MFB-522. Hell, I even have an analog keyboard with a built in sequencer that works really well with MIDI clock and incoming MIDI events. What's that thing called again... oh right! A MoPho!

Anyway, I don't think the analog nature of Tempest is what gave it troubles, it was the software aspects of it. Maybe trying to do too much? Be all things to all people? Not sure but for my money I think I'm going to go for the Vermona DRM1 Mk3 (which sounds fantastic to me) and use Maschine as it's sequencer/digital sound source and still come out cheaper than the cost of the Tempest.

This is a really good post, zerocrossing! Although your Tempest story (in my poor judgement) further supports my theory that analog is indeed at risk of being phased out. Let's use a little logic:

The only reason why a company would release an obviously pre-beta unit is because it was trying to recoup some of it's costs-does that make sense? If not, then Why else would they do it?

I have several Memorymoogs, because they became extremely affordable; because 'Moog Music' was liquidating the synth due to the fact that the costs of the machine had literally buried the company. They released it too soon, and approximately 50% failed, requiring service-center/factory repairs. Memorymoogs were flat-out toxic. So: At a great cost to the manufacturer, there were a series of changes/updates implemented over time fixing virtually every single issue; when I purchased them, they had had ALL of these changes/updates done plus a thyristor (over-voltage protection) installed at nominal additional cost as recommended by Fred McNiff, production manager at the time (the company had also been renamed EJE research, and operated out of Buffalo, NY)

So:

DSI may be doing the same thing--gambling that they can manage the costs of updating the 'sold' Tempests through firmware; let's hope that they gambled correctly, and that there are not significant (and all it takes is one biggie) design/manufacturing flaws that pop up down the road that require that each Tempest goes back to the manufacturer (is there even a service center option?) for repair.

Could they absorb the costs? Let's hope firmware updates do the trick-I'd HATE to see Tempest bury DSI!!!

OK: Like I mentioned earlier:

If someone wanted to take the bull by the horns and design/develop a reimagined CEM-style chipset that employs a good portion of both the analog audio and digital control on board a few solid chips, this would solve a lot of problems.

Fact: There are not that many engineers in the world like Dave Smith, Tom Oberheim that have this skillset:

They understand both digital control, and analog synthesis/audio circuitry, and HOW TO INTEGRATE THEM. Those of us that love and understand analog synthesis need to see a fresh infusion of talent-but (in my feeble mind) the talent can be in large part focused on chip developement for the likes of manufacturers like Korg/Yamaha, Nord, even DSI; so they don't have to worry so much about A/D integration.

When the existing group of A/D specialists retire, who you gonna call?

[zerocrossing]

If you have seen Tron Legacy you will have noticed how the creators brought back a rendered image of Jeff Bridge's young face. This animation is a vast improvement over animations made only years before such as Final Fantasy etc.

Depends on who you're fooling. I work in the video game industry where we work in 3d all the time and there isn't a person I work with that didn't think young Jeff looked good. Horrible in fact. I think Diva gets a lot closer to an analog synth than the best digital human animation. Do a google search for "uncanny valley."

[djshire]

.....did I accidentally go onto Gearslutz and not KVR?

[zerocrossing]

Well, people will start making parts if there is a demand and maybe Korg with it's Monotrons will help generate a new market.

As for Tempest, they shot themselves in the foot on that one and it's nobody's fault but their own and has nothing to do with the analog nature if it's audio engine (which is hybrid anyway).

First of all, they've yet to put any good audio examples on the site for it. Why? So far I've only heard one single good audio demo from a user. Everything else sounded... well not even as good as software and that good demo came out after some volume bug got correct so I suspect something else was going on with the gain structure that was causing less than stellar audio.

Second, speaking of bugs, they released what was clearly a pre-beta unit and even that was late. How you could release a drum machine in 2011 that didn't sync or respond to MIDI is beyond my powers of comprehension. I got scared away when they missed their first ship date and cancelled my preorder. I don't know much about hardware manufacturing but I can't imagine it was the right move to release it in an obviously unfinished state. Explain to your pre-orders that you ran into issues and you're not going to release a broken product and I think they'll understand. They made it seem like syncing an analog drum machine to a MIDI clock was some ground breaking coding task... yet the Germans seem to have cracked that issue in their $450 MFB-522. Hell, I even have an analog keyboard with a built in sequencer that works really well with MIDI clock and incoming MIDI events. What's that thing called again... oh right! A MoPho!

Anyway, I don't think the analog nature of Tempest is what gave it troubles, it was the software aspects of it. Maybe trying to do too much? Be all things to all people? Not sure but for my money I think I'm going to go for the Vermona DRM1 Mk3 (which sounds fantastic to me) and use Maschine as it's sequencer/digital sound source and still come out cheaper than the cost of the Tempest.

This is a really good post, zerocrossing! Although your Tempest story (in my poor judgement) further supports my theory that analog is indeed at risk of being phased out. Let's use a little logic:

The only reason why a company would release an obviously pre-beta unit is because it was trying to recoup some of it's costs-does that make sense? If not, then Why else would they do it?

I have several Memorymoogs, because they became extremely affordable; because 'Moog Music' was liquidating the synth due to the fact that the costs of the machine had literally buried the company. They released it too soon, and approximately 50% failed, requiring service-center/factory repairs. Memorymoogs were flat-out toxic. So: At a great cost to the manufacturer, there were a series of changes/updates implemented over time fixing virtually every single issue; when I purchased them, they had had ALL of these changes/updates done plus a thyristor (over-voltage protection) installed at nominal additional cost as recommended by Fred McNiff, production manager at the time (the company had also been renamed EJE research, and operated out of Buffalo, NY)

So:

DSI may be doing the same thing--gambling that they can manage the costs of updating the 'sold' Tempests through firmware; let's hope that they gambled correctly, and that there are not significant (and all it takes is one biggie) design/manufacturing flaws that pop up down the road that require that each Tempest goes back to the manufacturer (is there even a service center option?) for repair.

Could they absorb the costs? Let's hope firmware updates do the trick-I'd HATE to see Tempest bury DSI!!!

OK: Like I mentioned earlier:

If someone wanted to take the bull by the horns and design/develop a reimagined CEM-style chipset that employs a good portion of both the analog audio and digital control on board a few solid chips, this would solve a lot of problems.

Fact: There are not that many engineers in the world like Dave Smith, Tom Oberheim that have this skillset:

They understand both digital control, and analog synthesis/audio circuitry, and HOW TO INTEGRATE THEM. Those of us that love and understand analog synthesis need to see a fresh infusion of talent-but (in my feeble mind) the talent can be in large part focused on chip developement for the likes of manufacturers like Korg/Yamaha, Nord, even DSI; so they don't have to worry so much about A/D integration.

When the existing group of A/D specialists retire, who you gonna call?

Yeah, you're correct, but I don't think it's hopeless. Like I mentioned, Vermona is making a nice analog drum synth, as is MFB. In fact if you look at all the tiny developers, it's a fairly healthy list. This is just from the companies that Analogueheaven distributes, but I think they get them all.

http://www.analoguehaven.com/what/

So while it's probably a fairly tenuous time for analog synth developers, I think it's also a pretty good time in some ways if you take into account that it's pretty bad for all business right now.

And I think DSI did fix all the major issues with the Tempest, but I think releasing it in the state it was in probably cost them sales and credibility. It wouldn't be the first time Linn released a broken drum machine. My guess is they'll recover but the first thing I'd do if I were them is get some kick ass demos up on their site in high quality mp3 or wav format and even better have them done by someone fairly famous in the electronica world. I know it's a quality bit of kit but they, of all people, should have known the scope of a product like that. The synth engine isn't revolutionary, and neither is the sequencer... so how did it happen? Who knows, but I hope it all works out. The Tempest is the only DSI product I don't own... though I wish they'd release a synth engine only version. I have no need for a sequencer.

[zerocrossing]

.....did I accidentally go onto Gearslutz and not KVR?

Wait, KVR's not the software section of GS?

[pdxindy]

Here are some of my fiddlings with Diva:

http://www.kvraudio.com/forum/viewtopic.php?p=4783094

Nice tune! Thanks for posting

[goldenanalog]

...I don't think it's hopeless. Like I mentioned, Vermona is making a nice analog drum synth, as is MFB. In fact if you look at all the tiny developers, it's a fairly healthy list. This is just from the companies that Analogueheaven distributes, but I think they get them all.

http://www.analoguehaven.com/what/

Very encouraging; thanks, zerocrossing!

So while it's probably a fairly tenuous time for analog synth developers, I think it's also a pretty good time in some ways if you take into account that it's pretty bad for all business right now.

Tough economic times, indeed. Let's hope that the smaller companies mentioned in the AH list are still with us; and survive.

And I think DSI did fix all the major issues with the Tempest, but I think releasing it in the state it was in probably cost them sales and credibility. It wouldn't be the first time Linn released a broken drum machine. My guess is they'll recover but the first thing I'd do if I were them is get some kick ass demos up on their site in high quality mp3 or wav format and even better have them done by someone fairly famous in the electronica world. I know it's a quality bit of kit but they, of all people, should have known the scope of a product like that. The synth engine isn't revolutionary, and neither is the sequencer... so how did it happen? Who knows, but I hope it all works out. The Tempest is the only DSI product I don't own... though I wish they'd release a synth engine only version. I have no need for a sequencer.

I would love a Tempest; but rational thinking precludes me from pursuing one until it's been 'broken in' market-wise. Now: Assuming that it is just a series of firmware revisions that first 'makes it right' and then 'makes it better'; DSI gambled well, and won.

Tech guys have never been strong marketers; and in my experience: Most have a natural distrust of marketing types...who can blame them? Many a marketing promise made that tech is held accountable for, and can't possibly keep.

But of course, then there's our Urs. He builds, sells, supports, and improves, the best soft synths available. And he's a cool guy

[Lotuzia]

Here are some of my fiddlings with Diva:

http://www.kvraudio.com/forum/viewtopic.php?p=4783094

Will it be possible for every copy of a softsynth to have its own sound like an analog synth?

It's funny. I suppose in a sense this is an argument for analog - but's an outdated one. The difference you hear from one analog synth to the next is based on a few variables which create "randomness". These variables include things like heat, out-of-tune-ness, faulty parts or parts that have been replaced and are non-standard, and so on. But, also voltage is just inherently more random than a digital signal. Basically though, all you need to do to attain this variation is to fiddle with a few knobs or change things around. You can subtely LFO parameters to emulate randomness, but I'm guessing there is built in randomness in Diva anyway.

You cannot recreate real voltage topologies ( vco,vcf, vco.)digitally .There are way too variables

Rubbish. Diva may only be 90% of the way there now, but this is a very naive statement. If Diva isn't all the way there now, computers in future will do it with ever increasing accuracy. People ten years ago were saying that digital animation would never look "real", or "organic". That was true ten years ago, but look at the progress. If you have seen Tron Legacy you will have noticed how the creators brought back a rendered image of Jeff Bridge's young face. This animation is a vast improvement over animations made only years before such as Final Fantasy etc. 3D computing animation has only been around for 20 years or so, and it can almost fool us now. In 10 years time, you won't be able to tell the difference between computer generated animation and video photography.

Sound is a lot less complex than this. Anyone who understood what modern computers are capable of these days would not make claims like this.

To answer the first part, I'll try to answer the second one : Yes its currently impossible to recreate 100% all the interactions that you can find in a real analogue synthesizer, because there are too much interactions. The way the physical components are placed relative to each other implies interactions, the action of the components might vary with heat, and the way voltage is handled. An analog synthesizer is a piece of art, and implies multiple solutions that manufacturers used to fix, more or less, problems due to electronic/electric components, or their interaction.

Not counting models and circuit revisions ( like changing an important IC) or just changing slightly the model or the place of this or that component because a new model, more resistant or "wich does the same job but cheaper" or "wich has more tolerance with heat" or " Wich vary less with heat" or " the new board is more easy to repair" , has appeared.

For the first part of the answer, if its between tow models with close serial numbers, of the same synth, the question is valid. If its different analog synths from different brands, then of course everyone who has tried at least two different analog synths in his/her lifa has the answer.

Note than randomness, either with pure random, or LFO modulation is very different though. Differences between two incarnations of the same synth same revision will not have random modulation. They will have a sound differing slightly from another model but always the same difference, so a kind of static one, wich btw includes all the non static differences due to analog ( If I've lost only 95% of readers at this time, I'd consider it as an optimistic % ).

Lfo are cyclic modulators, and therefore are not suited to simulate this kind of differences. Random, or pushed to the extreme chaotic functions like on the Synthix, are not suited either, because they will modulate the targets in a continuous unpredictable way, wich is not the goal to reach here.

To achieve this two possible ideas are : A general button wich will handle several parameters at once to give fundamental parameters some different values ( like the one found on the SYnthix ) or specific functions wich would adress fundamental parameters one by one like an EV segment lenght, or the way voices are summed up, or the drive-> Filter, etc etc ie changing the parameters of functions who emulate the real electronic components implementation.

The drawback is that handling all these parameters one by one can become quickly very complex, and that people might tweak all these parameters endlessly while not beeing able to decipher with ears the changes they bring.

AN example of this can be seen on one of the chorus modes of the Syntix, where a toggle can change the -emulated- behavious of two of the inner components of the original Synthex chorus. But of course it would be difficult to offer such variants for each parameter. Not only could it be very time consuming for us, but it would also probably be seen as very counter productive by musicians.

So the cursor to add this kind of features has to be put at the right place : What has musical consequences, what IS musical, or lees, or not : Not so easy

[hakey]

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

[jupiter8]

It's my impression that cpu's are not keeping up. They are adding more cores, but a single core is not that much faster than a few years ago... right?

Actually they are and quite a lot. It is true that much of Moores Law has been used to cram more cores in there but the cores gets faster and faster too.
The 2600 is about 5 times as fast as a Q6600.

That's far from accurate according to ADK benchmarks. The benchmarks can be a bit misleading (but still informative if you know how to read them), this is why i asked Scott and Vinny to not include hyperthreading for one of the processors listed since its still more cores (Virtual) and thus won't show how much true-speed gained. You have to compare current Quads to the QX6700 Quad NOT a 6-core, you can't use hyperthreading or overclocking on current computers when comparing to judge power gains overtime either or it won't be accurate.

Damn, don't where i got that info from. I remember reading it and was all like Wow the 2600 is 5 times as fast as a Q6600 which i still think is a decent CPU.

But if you look at the 32 buffer test there's an OCd 2600 that runs 261 plugins compared to the 54 of the non OCd 6700. Maybe that's what i remember though that isn't a fair fight. Ah well Moores law will make me right soon anyways. Just wait for the Ivy Bridge.

Regarding HT: well yeah and maybe no. That subject gets a really complex really fast. You do have a point there.

[Lotuzia]

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

While the probabily of Hackey bashing a Xils-Lab product in such a thread is one from the beginning ....

Do your maths, and see who has the most elegant behaviour