Quantcast
Why could some digital hardware synths sound better than software? - Page 3 - Gearspace.com
The No.1 Website for Pro Audio
Why could some digital hardware synths sound better than software?
Old 10th September 2012 | Show parent
  #61
Deleted User
Guest
Quote:
Originally Posted by audslu ➡️
I mean what you agree with.
Old 11th September 2012
  #62
Lives for gear
 
synthRodriguez's Avatar
 
🎧 10 years
Hardware synthesizers are designed and built to sound good.

Computers are built to run code with DSP as an afterthought.

Get over it.
Old 11th September 2012 | Show parent
  #63
Gear Guru
 
zerocrossing's Avatar
 
🎧 10 years
Quote:
Originally Posted by scottrod ➡️
Hardware synthesizers are designed and built to sound good.

Computers are built to run code with DSP as an afterthought.

Get over it.
That's just an over simplification. Again, there are plenty of crappy sounding hardware synths out there. Good is good, bad is bad.
Old 11th September 2012 | Show parent
  #64
Lives for gear
 
🎧 5 years
Quote:
Originally Posted by zerocrossing ➡️
That's just an over simplification. Again, there are plenty of crappy sounding hardware synths out there. Good is good, bad is bad.
which ones do sound crappy? cant recall one..
Old 11th September 2012 | Show parent
  #65
Lives for gear
 
volumetrik's Avatar
 
🎧 10 years
Quote:
Originally Posted by scottrod ➡️
Hardware synthesizers are designed and built to sound good.

Computers are built to run code with DSP as an afterthought.

Get over it.
Agreed. We don't really know what the DAC engineers have in mind when they develop soundcards or the way they interact with the operating system/software. They are not going out of their way to make softsynths sound more powerful or whatever. Most people just want to record themselves playing guitar and singing on their PC and it all kind of revolves around that. There hasn't been any real research or advancements in the computer domain as far as improving sound its been the same for the last 5 years no? The whole computer/softsynth thing sounds good but it can't truly beat hardware. Sure Diva sounds good but can you run 5 Divas at the same time with high quality reverb and other effects? CPUs still can't handle it, and in the long term new soundcards/drivers/OS is out and it might be not optimized for it or not be compatible at all...
Old 11th September 2012 | Show parent
  #66
Lives for gear
 
🎧 10 years
Quote:
Originally Posted by bug2342 ➡️
Bad Envelopes are way more serious offenders. Especially since a bunch of the popular plugins today still drag some baggage from the time when things where called CPU hog for doing things right.

Random example: The otherwise beautiful Reaktor still defaults to 400Hz control rate and maxes out at 3.2kHz. Some Waldorf LFOs go to twice that frequency....
That was (hell, for a bunch of stuff still is) quite of an issue for bunch of plugin stuff. Oscilators and filters need exponential function (eats quie a few cycles, but there are low cost workarounds) to calculate coefficients so most coders were like "oh, I'll calculate once per block of 128 (or something like that) samples". Yeah, no wonder it doesn't sound right. But that wasn't only bad part.
For bigger part of VST history most devs implemented envelopes with 4 linear segments (in essence, same way DX envelopes work and very very very not the same way analog envelopes work). And than at one point this or that guy would figure out that analog envelopes have some sort of exponential shape. So they would put that linear env trough some kind of shaper . Yeah right, that didn't hit homerun either.


btw, even in hardware analog land most people dissimis envelopes as trivial nonissue. Well, nope, it is not, so bunch of people bitch about software envelopes in particular.
Old 11th September 2012 | Show parent
  #67
Gear Guru
 
Yoozer's Avatar
 
1 Review written
🎧 10 years
Quote:
Originally Posted by scottrod ➡️
Hardware synthesizers are designed and built to sound good.
...running on industry standard Motorola DSPs which function as well for generating synth sounds as they do for compressing voice for data transmission.

Quote:
Computers are built to run code with DSP as an afterthought.
The algorithms for band-limited oscillators and FIR/IIR filters and physical modeling are identical, what's different is how many cycles it takes to finish the job.

The challenge for a VA or digital synth is to squeeze out as many voices of polyphony as possible out of 'm - as you can see at CEM (Curtis) SSM Hardware Chips used in these Synthesizers - welcher dsp wo? the Novation needs several while the Virus does the job with just one.

When the Alesis Ion was released, its 8 voices were kind of meh compared to the 24 the Virus and Novations were pushing. Instead of going for quantity, they went for quality, which was considered to be pretty revolutionary.

Last but not least:

The Oasys runs on a general purpose computer (P4). The Kronos runs on a general purpose computer (Atom). The Hartmann Neuron runs on a general purpose computer (P3-800).

Good luck trying to call the sound quality on those an "afterthought".

Quote:
Originally Posted by volumetrik ➡️
Agreed. We don't really know what the DAC engineers have in mind when they develop soundcards or the way they interact with the operating system/software.
Their worries consist of getting the buffer filled in time. The only task of the OS is to send that data to the hardware.

Quote:
They are not going out of their way to make softsynths sound more powerful or whatever.
No, of course not; they can't. The bits that are sent to the interface are what they are. What they aim for is transparency and neutrality.

Quote:
Sure Diva sounds good but can you run 5 Divas at the same time with high quality reverb and other effects?
Let's put it this way: which hardware can run comparable algorithms to Diva? All of 'm use a computationally cheaper filter algorithm except for perhaps the Solaris, which tops out at 12 voices max.

Furthermore, does every VA have high quality reverb? The GS answer is - "no, get a Bricasti or an H8000 because those are loads better at reverb than the built-in crap".


Here you go - 22 instances playing on an i7.

Quote:
CPUs still can't handle it, and in the long term new soundcards/drivers/OS is out and it might be not optimized for it or not be compatible at all...
You have some strange ideas about what "optimized" means. The OS plays a completely different role in this case.

Windows is not optimized for software synthesizers. A completely stripped Linux distribution minus all the code except for audio and perhaps a bit of I/O is also not optimized; but because there are so few services running, it means that there's more computational power available to run softsynths. The Kronos hits 140 notes of polyphony with sample playback; it'd probably score less than half of that if all kinds of services were running in the meantime because they take up processor time (worse: because those services are scheduled at completely different intervals, you might have 60 voices at one point, and 40 the next, because something would take up too much CPU time). That's all there is to optimization. What kind of bits roll out of the plugin, they're not going to change because you're not checking your mail at that moment.

The DAC receives data. Where that data comes from doesn't matter. If the DAC makes anything sound better, that's great - but that's the same as putting a smiley EQ curve on your amplifier; you're lying to yourself that the source is more pleasant than it actually is.
Old 11th September 2012 | Show parent
  #68
Lives for gear
 
kilon's Avatar
 
🎧 10 years
Quote:
Originally Posted by audioconsult ➡️
which ones do sound crappy? cant recall one..
well my andromeda for one, and blofeld can easily go to the crap zone as well. Some synths are not that "bread and butter". Some synths are made to annoy the hell out of you. Andy is definitely one of them. I cant begin to describe how many times I want it to throw it out of the window. But it also showed me that if I want great sounds I need to **** my lazy ass down and make them oscilation by oscilation.

I have no doubt it takes a single dedicated sound designer to make some presets in the most crapist of crapist synths of all time to send the same synth to Hypedom Nirvana.

Of course it depend whats the definition of sounding crap, there are a million factors.

At least for me its a moot point, soft synths lately do sound as great. I use Alchemy , no problems making the sounds I want.But I still prefer hardware :D
Old 11th September 2012 | Show parent
  #69
Lives for gear
 
volumetrik's Avatar
 
🎧 10 years
Quote:
Originally Posted by Yoozer ➡️
Their worries consist of getting the buffer filled in time. The only task of the OS is to send that data to the hardware.

No, of course not; they can't. The bits that are sent to the interface are what they are. What they aim for is transparency and neutrality.
Someone might think about implementing features in the hardware that the software can utilize in order to sound different or better but they dont bother. So if an engineer has softsynths in mind he could think about making them sound better.

Quote:
Originally Posted by Yoozer ➡️
Let's put it this way: which hardware can run comparable algorithms to Diva? All of 'm use a computationally cheaper filter algorithm except for perhaps the Solaris, which tops out at 12 voices max.
Diva might have extremely complex algorithms but they don't mean anything once they hit the DAC. Why run such algorithms in the first place? What is so special about Diva anyway, can it emulate hardware well or something?

Quote:
Originally Posted by Yoozer ➡️
Furthermore, does every VA have high quality reverb? The GS answer is - "no, get a Bricasti or an H8000 because those are loads better at reverb than the built-in crap".
I wasn't referring to built-in FX.


Quote:
Originally Posted by Yoozer ➡️
You have some strange ideas about what "optimized" means. The OS plays a completely different role in this case.
Optimized might mean that when a new CPU is out it has features and in order to utilize it for speed the code needs to be implemented in the software in order to utilize thoes features, furthermore a future OS might have a different architecture that no longer functions with the way the software was written in the past.
Old 11th September 2012
  #70
Gear Addict
 
🎧 10 years
CPU performance used to be a factor in 2003 (single-core Pentium 4 or Athlon XP), but not any more. General purpose CPUs have by far surpassed the power of dedicated DSP's. Not only that, but modern CPUs have 4, 6 or even 8 cores to distribute the load.

Something like DIVA would not be possible on a simple Freescale DSP or 10.
Old 11th September 2012 | Show parent
  #71
Lives for gear
 
🎧 5 years
Quote:
Originally Posted by volumetrik ➡️
Optimized might mean that when a new CPU is out it has features and in order to utilize it for speed the code needs to be implemented in the software in order to utilize thoes features, furthermore a future OS might have a different architecture that no longer functions with the way the software was written in the past.
"New CPUs" don't just implement features that make them in any way incompatible to older ones. With every new Intel CPU, the old command set of all predecessors is maintained for compatibility. This is why even the newest i7 can still run old Win98 installations, all those different Linux distributions or OSX Versions. The only major change that demanded new code was when Intel went from 32 bit to 64 bit - but still that's no problem: if you keep your old 32 Bit OS, you can still run all the ancient 32 Bit code on the new CPU.
As for new OSses: Modern software isn't designed as a monolithic block, that just accesses internal OS features without different interface layers inbetween. So if something needs to be adapted to newer OS versions, it's mostly the drivers or the system libs that are affected, and those are adapted by the OS and/or hardware provider not the plugin developer. And last but not least: Even if the DAW vendor makes changes to it's plugin interface that demand a recoding, it implements them as a NEW version while maintaining the older versions for compatibility (like with steinbergs VST, VST2 and VST3).
Old 11th September 2012
  #72
Lives for gear
 
🎧 15 years
That Diva demo sounds horrible to my ears.
Old 11th September 2012 | Show parent
  #73
Lives for gear
 
volumetrik's Avatar
 
🎧 10 years
Quote:
Originally Posted by chk23 ➡️
"New CPUs" don't just implement features that make them in any way incompatible to older ones.
Ok but there have been cases where old software just refused to work on new CPU/OS no?
Old 11th September 2012
  #74
Lives for gear
 
lysander's Avatar
 
🎧 5 years
I wish I hadn't clicked on this thread, but I must be slightly masochistic.
It seems Gearslutz has a never ending stream of people who really want to share their strong opinions about things they understand absolutely nothing about.

Thank god there are people like Yoozer who have enough patience to bring facts and rationality back in the discussion - we need the moderators to create a Gearslutz Saint tag for you.
Old 11th September 2012
  #75
Bio
Lives for gear
 
🎧 10 years
Quote:
Gearslutz has a never ending stream of people who really want to share their strong opinions about things they understand absolutely nothing about.

Thank god there are people like Yoozer who have enough patience to bring facts and rationality back in the discussion
This +1000
Old 11th September 2012 | Show parent
  #76
Gear Guru
 
Yoozer's Avatar
 
1 Review written
🎧 10 years
Quote:
Originally Posted by volumetrik ➡️
Ok but there have been cases where old software just refused to work on new CPU/OS no?
The switch from Power PC architecture to Intel is an (in)famous example, but in that case "old" means "hasn't been updated at all in the past 5 years" because Apple gave more than enough warning shots. If a developer didn't do anything about it, they're just not interested in it anymore.

Quote:
Originally Posted by volumetrik ➡️
Someone might think about implementing features in the hardware that the software can utilize in order to sound different or better but they dont bother.
They don't, because it'd tie the hardware to the software. We've had enough of that in the past, and the software-turned-hardware builders product list is littered with "works only on X" stuff. Steinberg Midex, AMT8, etc.

There's the UAD Apollo which allows you to use hardware features so that the software can sound better. Ditto with SonicCore's boxes. Thing is, then you're locked to yet another plugin format of yet another developer where plugins that won't run natively will yet again be unsupported in a decade or so, with no way to emulate anything.

The problem is not non-nativeness. The problem is a guaranteed lifetime of support, which would do a lot of good for software anyway, because you're now at the mercy of a small developer team.

Even bigger companies have a hard time dealing with this.

Quote:
Diva might have extremely complex algorithms but they don't mean anything once they hit the DAC. Why run such algorithms in the first place? What is so special about Diva anyway, can it emulate hardware well or something?
For that I have to defer to Urs Heckmann himself, but one of the big issues is in the filters; they're zero-delay feedback filters. These take up quite some computational power.

Quote:
Optimized might mean that when a new CPU is out it has features and in order to utilize it for speed the code needs to be implemented in the software in order to utilize thoes features
Well yes, that's why the new version of Diva uses way less CPU on an i7 or i5 as opposed to a Core 2 Duo or Quad. It's done by using extensions; software builders have to recompile their code so that the compiler can use these, and it happens. Besides that I know of NI Massive that it shows which multimedia extensions are detected, so I'd imagine that certain library calls or code paths are changed when those features are available.

Even without those extensions an i5 or i7 is faster than my old Core 2 Quad, so if I decide to buy a new computer I'll get less CPU usage.

Quote:
furthermore a future OS might have a different architecture that no longer functions with the way the software was written in the past.
If it changes that radically the entire application has to be redeveloped anyway and you're SOL.

As it stands, 10-year old plugins still work on Windows. OS X is more of a moving target (forcing developers to stay up to date more) but that's a matter of philosophy, and if your plugin developer has experience with OS X they keep that up to date as well. It's the big honking legacy apps like Photoshop that have to be entirely rewritten from Carbon to Cocoa that are dragging their feet.

Cross-platform frameworks like JUCE should mean that most of the work is in the algorithm and in making it talk to RTAS and AU. Of course, I'm not saying it's easy - it's just doable. Meanwhile, the OS keeps on trucking along happily.

Of course, there's no telling what radical changes will happen in next OSes. Even with massive core parallelization you could still emulate a single core in software and let the plugin talk to just that, and you'd have something sub-optimal, but it'd work. In the end, the OS has as a task to distribute resource usage, not the plugin.

My story however doesn't fix this:

Quote:
Originally Posted by living sounds ➡️
That Diva demo sounds horrible to my ears.
so use your ears, get the demo, IMHO it's definitely wortwhile enough to keep around - saying this as an owner of a Juno-60 and Model D. They're great machines that I'll never let go because they're dear to me in various other ways that have nothing to do with music.

But if you're looking for that Moog sound, no amount of substitutes or plugins will help anyway: buy one, or give up while you still have money.

It's between your ears; no amount of trickery can convince you that you're playing the actual machine while you know you're not, and you'll keep doubting all the time, and wondering if you're missing anything.

Last edited by Yoozer; 11th September 2012 at 03:37 PM.. Reason: zero-latency -> zero-delay!
Old 11th September 2012 | Show parent
  #77
Lives for gear
 
🎧 10 years
Quote:
Originally Posted by Yoozer ➡️
For that I have to defer to Urs Heckmann himself, but one of the big issues is in the filters; they're zero-latency feedback filters. These take up quite some computational power.
Actually, zero-delay feedback is a bit tougher on CPU but not that tough. Thing with Diva is this: you have equations for various components (like, Ic=Is*exp(Vbe/Vt) for collector current of BJT and I = C*(dV/dt) for capacitor current). Now you have these software packages that do circuit simulation (like SPICE). You define circuit as network of nodes and with equations like above you end up with system of differential equations that describe circuit. To calculate some voltage or current in circuit you have to solve that system of differential equations numericaly and voila. Except, voila part comes with one hell of a CPU tax. What Diva does is exactly that, it runs circuit simulation in real time (just with simplified models, that is only most significant stuff is taken into account). Not like real thing but it's much closer than usual attempts.
Old 11th September 2012 | Show parent
  #78
Lives for gear
 
audslu's Avatar
 
🎧 5 years
Quote:
Originally Posted by lysander ➡️
I wish I hadn't clicked on this thread, but I must be slightly masochistic.
It seems Gearslutz has a never ending stream of people who really want to share their strong opinions about things they understand absolutely nothing about.

Thank god there are people like Yoozer who have enough patience to bring facts and rationality back in the discussion - we need the moderators to create a Gearslutz Saint tag for you.
Yea page 1... 2... 3 *BAM* , BS detector exploded.
Old 11th September 2012 | Show parent
  #79
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by Yoozer ➡️
so use your ears, get the demo, IMHO it's definitely wortwhile enough to keep around - saying this as an owner of a Juno-60 and Model D. They're great machines that I'll never let go because they're dear to me in various other ways that have nothing to do with music.

But if you're looking for that Moog sound, no amount of substitutes or plugins will help anyway: buy one, or give up while you still have money.

It's between your ears; no amount of trickery can convince you that you're playing the actual machine while you know you're not, and you'll keep doubting all the time, and wondering if you're missing anything.

I have tried it and don't get what people are raving about.
Old 11th September 2012
  #80
Gear Guru
 
Yoozer's Avatar
 
1 Review written
🎧 10 years
Sorry if I wasn't clear - that was intended as a reply to volumetrik, not you No amount of arguing on my side can convince you when your own ears don't.
Old 11th September 2012 | Show parent
  #81
Lives for gear
 
🎧 5 years
Quote:
Originally Posted by volumetrik ➡️
Someone might think about implementing features in the hardware that the software can utilize in order to sound different or better but they dont bother. So if an engineer has softsynths in mind he could think about making them sound better.
Hmmmm.......
Any digital synth does the following:

- read user input from a buffer
- calculate some stuff
- write audio data to a buffer

The only thing hardware can do to improve this chain is optimize input and output, so you can have short buffers and improve the speed of "calculate some stuff".

Thats where all the funny letter combinations of your CPU come in.
Old 11th September 2012 | Show parent
  #82
Lives for gear
 
🎧 5 years
Quote:
Originally Posted by recnsci ➡️
And than at one point this or that guy would figure out that analog envelopes have some sort of exponential shape. So they would put that linear env trough some kind of shaper . Yeah right, that didn't hit homerun either.
Don´t disrespect the shaper, the shaper is your friend.

Actually there is absolutely nothing wrong with using shapers, if your shaper has a reasonable resolution and your readout happens frequent enough.

Most FM synths (hardware) do that for sines (except for the high resolution part...) and most CPUs do it for the more interesting functions.
Old 11th September 2012 | Show parent
  #83
Lives for gear
 
🎧 15 years
Here's a quick demo with no processing whatsoever of the ancient Linux port "ZynaddsubFX". Every bit as big and crisp sounding as digital hardware synths. Freeware, too. I don't hear this kind of quality in most plugins.

Has anyone saved the OSX version of this plugin when it was still availible, by any chance?
Attached Files

Zynadd.mp3 (4.16 MB, 810 views)

Old 11th September 2012 | Show parent
  #84
Lives for gear
 
Ossicle's Avatar
 
🎧 10 years
I wish I'd still have my Wavestation EX to show how much ballsier it sounds to its sw counterpart.

The reason why modern computers cannot match the musicality of 20 year digital technology old is a mystery to me (and I was too lazy to read the long answers in this thread ).
Old 11th September 2012 | Show parent
  #85
Gear Addict
 
🎧 10 years
Quote:
Originally Posted by Ossicle ➡️
I wish I'd still have my Wavestation EX to show how much ballsier it sounds to its sw counterpart.

The reason why modern computers cannot match the musicality of 20 year digital technology old is a mystery to me (and I was too lazy to read the long answers in this thread ).
This thread hasn't conclusively answered that, although it's provided a lot of interesting opinions.
Old 11th September 2012 | Show parent
  #86
Lives for gear
 
🎧 15 years
Quote:
Originally Posted by [email protected] ➡️
This thread hasn't conclusively answered that, although it's provided a lot of interesting opinions.
I think modern computers can "match the musicality" (see my example above), the question is why most plugins don't do it.
I've asked designers these questions five years ago and they couldn't tell me. So I went the hardware route.
Old 11th September 2012
  #87
Lives for gear
 
🎧 10 years
because they're hardware and hardware is for real men.
Old 11th September 2012 | Show parent
  #88
Lives for gear
 
volumetrik's Avatar
 
🎧 10 years
Quote:
Originally Posted by Yoozer ➡️
so use your ears, get the demo, IMHO it's definitely wortwhile enough to keep around - saying this as an owner of a Juno-60 and Model D. They're great machines that I'll never let go because they're dear to me in various other ways that have nothing to do with music.

But if you're looking for that Moog sound, no amount of substitutes or plugins will help anyway: buy one, or give up while you still have money.

It's between your ears; no amount of trickery can convince you that you're playing the actual machine while you know you're not, and you'll keep doubting all the time, and wondering if you're missing anything.
what are you trying to say here?

Yeah I probably should test out Diva on a high end computer and hear it live, but I'm not looking for the Moog sound anyway, now if you're saying that Diva can come close to sound as fat as a real Model D well thats amazing.
Old 12th September 2012 | Show parent
  #89
Lives for gear
 
🎧 10 years
Quote:
Originally Posted by bug2342 ➡️
Don´t disrespect the shaper, the shaper is your friend.
Oh, don't get me wrong, I love shapers on envelopes, IMveryHO there is no env that can't benefit from DC coupled saturator/softclipper (I actually emulate sorta' frequency dependant saturator with Diva mod matrix to get particular CV response of some old analog synths {there is response of CV path especially with VCFs but that's long story on another topic}). My point was that most of software guys got whole process wrong for quite a long time, linear envelope can be great, I looove DX liine, but for emulation of analog it's bad starting point.
Old 12th September 2012 | Show parent
  #90
Gear Addict
 
🎧 10 years
Quote:
Originally Posted by living sounds ➡️
Here's a quick demo with no processing whatsoever of the ancient Linux port "ZynaddsubFX". Every bit as big and crisp sounding as digital hardware synths. Freeware, too. I don't hear this kind of quality in most plugins.

Has anyone saved the OSX version of this plugin when it was still availible, by any chance?

I spent ages hunting for an OS X version. No luck though.

There's something special about ZynaddsubFX for sure. Sad that it's not available.
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 627 views: 104769
Avatar for damoog
damoog 27th April 2019
replies: 835 views: 177322
Avatar for jlgrimes11
jlgrimes11 30th August 2020
replies: 518 views: 41619
Avatar for audslu
audslu 28th December 2015
replies: 498 views: 60861
Avatar for GlenGood
GlenGood 4 weeks ago
Topic:
Post Reply

Welcome to the Gearspace Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…

Forum Jump
Forum Jump