|
Anybody knows WHY so many softsynths sound THIN compared to hardware?!
|
View this Thread in Original format
| Dance123 |
Hi,
It has been said many times before that most softsynths (with a few exceptions) sound rather thin compared to hardware synths, but WHY exactly is that?
I mean, aren't softsynths and VA hardware synths the same in a way that they both emulate real analog synths with CPU power, so why is it still so that hardware synths sound fatter then software synths which sound rather thin alot of the time?! You can almost always here when a track is made with softsynths, right, but WHY is that??
Anybody has an explanation for this cause it doesn't seem to make sense if you think about it logically or am I missing something..
Thanks! |
|
|
| staticblue |
I don't know much about it, but maybe better algorithms ? Also on a VA hardware synth, all the available "calculation power" is used for making sounds, and this is not the case on a computer.
Just my random guess :p |
|
|
| Derivative |
Softsynths can sound just as good as any hardware synth - Just listen to G-Media ImpOSCar or Minimonsta for proof of that. But softsynth complexity is CPU limited. A hardware synth is basically software - lines of code designed to run off a DSP.
Take a look at Reaktor. You could build a synth on that with similar specification to an Access Virus B. Some aspects you wont really be able to model perfectly (or even anywhere close), like the dual filter. But the number of modules that you would be using (4 oscillators per voice, 2 filters per voice, 2 envelopes and 2 LFOs per voice) would probably grind your PC into dust. And nobody likes running Cubase with 95% CPU load because of 1 synth. Thats why DSP farms exist. Those are for softsynths that are too processor intensive to run on a CPU and do anything else at the same time.
All a digital hardware synth is, is software designed to run off a dedicated chip in a dedicated box.
Now analog...thats a different story - analog is all discrete circuitry. No chips. No lines of code. No software. Its sound made from electricity. 2nd and 3rd generation analog polysynths had some digital components but in most cases the signal path was nearly all analogue. |
|
|
| Dance123 |
1/ But how come that on the Z3ta+ site there are mp3 demos where more then 10 instances at once were used on a recent PC. Check out the Z3ta website to see and hear for yourself over here: http://www.rgcaudio.com/z3ta%2B.htm
How is that possible if you know that Z3ta+ patches can have like 6 oscillators, 2 filters + alot of enveloppes, LFO's, effects etc.. that's even more then alot of hardware VA synths..?? Can you run 10 of those at the same time on a recent CPU?? (I have an older CPU so I haven't tried Z3ta+)
2/ what are "DSP farms". Do you mean those UAD cards or are there also other brands? |
|
|
| thoughtlessjex |
| quote: | Originally posted by Dance123
WHY exactly is that? |
Power of suggestion. People who don't like the idea of their synths being hidden inside of the same box that does the word-processing are bullting reasons to prefer hardware VA synths.
But, like Derivative said, there is a difference with analogue. Essentially, analogue synths are imprecise. They have odd response curves for different parameters because of weird resistances in the circuitry, their amplification methods create distortions that are not tru to the original oscilator, bringing out new harmonics that digital synths can't properly emulate. All this adds to the warmth of the tone.
I wouldn't say that softsynths are necessarily thinner, though. They just take a little more knowhow to program properly. |
|
|
| Derivative |
Dance123. Not every digital filter and every digital oscillator is built the same way in software. Fruity filter takes up hardly any CPU cycles but then it has fixed routing, you dont have the option of adding or subtracting poles and the filter algorythyms arent especially long.
Contrast that with something like PSP Nitro which as a Multi Mode Filter module. You can route 4 filters in sequence,, in series or parallel as well as create feedback loops. OWing to that you can have an insane number of poles in series and it has alot of other built in modules like an LFO and mod matrix. The amount of code for the filter design is also alot longer than fruity filters and to prevent the zipper effect, it rapidly interpolates between CC values. To avoid digital stepping. Thats really processor intensive on its own. But it makes filter sweeping much more seamless.
Zeta's filters are pretty good but they arent anything approaching the complexity of something like PSP Nitro. But that would make sense, as Nitro is a dedicated Multi Mode Filter whereas Zeta's filter is to compliment a fully fledged synthesizer.
Also, it is possible to have sloppy and unneccesary code on DSP. Thats just bloat. Sometimes you get really well programmed softsynths like impOSCar which sound fantastic and dont use up a tonne of CPU cycles. Some VSTs are really bloated - like for instance the VST effect Dirtbag. It sounds and it uses a ridiculous amount of CPU. On my 1.6ghz P4, 1 instance of Dirtbag will take up 50% CPU load. Its just very poorly coded. |
|
|
| Diginerd |
Ok, this is a touchy point of discussion. I'll throw in my 2c.
impOSCAR whilst fattish is no match for the original. An original with a Z80 CPU doing the digital control side beats the snot out of the plug.
Analog synth oscillators fall into a couple of broad categories, those with VCO (Voltage Controlled Oscillator), those with DCO (Digital Controlled Oscillator)
VCOs are unstable, go out of tune every 3 seconds but sound wonderful. DCOs are stable, but thinner sounding.
It is incredably hard programming a VA accurately. It also is very resource hungry. Hardware synths are DSP based, and consequentially have a significant amount of horsepower, coupled usually with some very talented coding to get the most out of it. Measure for measure DSP code is orders of magnitude more powerful doing a repetative task like, well, errr, signal processing!
That said, comparing a JP8080 (VA) to an MKS-80 (Analog that the 8080 is closely based on) clearly shows the differences. The MKS-80 doesn't do a supersaw pad, but it does 8 voice unison with DCOs and squishes the 8080 like a 400lb man sitting on Peewee Herman. No contest. Both are from the same well resourced manufactuer with a massive budget to do the VA, but still not the same thing..
Nothing sounds as gritty as a Microwave II / ppg wave. Even though the former is an all digital synth it's a combination of factors that makes a difference. Serious DSP Code, serious architecture, and a lot of love
FM-7 does a very good job of mimicing the DX family, but again a TG-77 is sonically much more satisfying(and that's before getting up to an SY-99) , even though next year the TG-77 would be legal to drive in the USA.
Same goes for samplers, an Emu EIV Ultra series sampler beats practically any software based sampler hands down in terms of raw sonic quality and flexability. You can even have 16 digital outs, 128 voice poly and 32 part multi timbral in one box, each with wicked amounts of serious filtering, modulation matrix to die for and a bunch of other tricks. Why does it sound better? Because it was built to a specification not a price, it also is very pure in its design.
It has dedicated hardware that only does the task at hand, it consequentailly isn't mucking about wondering whether that email is coming in, if that other instrument needs some time or any of those other tasks your computer has to deal with just to stay running.
It also means that it can use some pretty wild algorithms which translates as computationally intensive filtering, and amazing transpostion abilites. Transposing digitally whilst maintaining accuracy is tough. The nailled it.
What's interesting is that it was based around severalASICs (Application Specific ICs). These are built to do one thing and one thing well, and are even more specialized than DSP chips. There's one for Filtering, one for general dutys (Transpostion, modulation) there was even a plug in upgrade for FX (The R chip as sound in the RFX) each uns in parallel as part of a whole.
I have to wonder (and this is pure speculation) if clock jitter and cpu jitter produced by swapping tasks is part of the issue on general purpose computers. They're just trying to do too much at once.
On the otherhand I'm pretty sure it will be just a matter of time before the software beats the current crop of hardware VAs in terms of quality. As for the other instruments (Be they digial, analog or hybrid) it's touch because they have real character and in many cases soul. That's really hard to copy...
Not my most incisive post, but hey. |
|
|
| Dance123 |
Hi,
1/ Is there any difference between a "DSP" (which hardware VA synths seem to use) and the CPU in your computer. I mean, aren't those the same thing or are DSP's" something different?!
2/ what are "DSP farms". Do you mean those UAD cards? Are there also other brands?
Thanks! :) |
|
|
| Eldritch |
| quote: | Originally posted by Diginerd
...As for the other instruments (Be they digial, analog or hybrid) it's touch because they have real character and in many cases soul.
|
Oh, please. :haha:
I don't mean to laugh at you or your opinion, but I just find that extremely funny. :D
| quote: | Originally posted by Dance123
Hi,
1/ Is there any difference between a "DSP" (which hardware VA synths seem to use) and the CPU in your computer. I mean, aren't those the same thing or are DSP's" something different?!
2/ what are "DSP farms". Do you mean those UAD cards? Are there also other brands?
Thanks! :) |
They're basically the same.
You can do alot more with DSP, because you have dedicated CPU power.
You can use almost 100% of the DSP's power while on a softsynth, you really don't want to use more than 20% CPU, because there wouldn't be enough to use for other plugins, and the sequencer itself.
This was actually explained earlier in the post. :p |
|
|
| Eldritch |
| double post, sorry |
|
|
| Pjotr G |
| I think in a lot of cases, the bigger synth manufacturers can afford to have more time put into the development of algorithms. More sophisticated algorithms = better sound. Of course there are also softsynths that sound sublime, and I'd be hard-pressed to tell what is what, beside the obvious character of some synths. But that way of telling apart is beside the issue, as I can also tell certain hardware synths apart from each other. |
|
|
| Diginerd |
There is another side to this too. DSPs are generally directly coded using a low level programming language which reguires a significant degree of skill. Where it gets interesting is if you have "DSP Farms" with several chips per card strapped together you get a massive amount of parallel processing all synced in lockstep. This system will then perform right up to 99.9999999% load with practically zero impact on the host computer.
So coming back to soft synths (And whilst we're at it let's looks at ITB FX, compressors EQs etc) why do they tend to sound thin.
Coding for DSP is harder than general purpose computers. It also produces more efficiant code, which means that more complex routines can be squeezed out of a given amount of power.
Now, given that people pay a premium for these systems (This can be extended to VAs too, the Virus TI ain't cheap) the best and most experienced talent is writing code for them, and some of the best hardware is setup for them.
This is one of the reasons those big expensive pro Tools rigs are big and expensive.
Coming back, some VST(i)s do sound bloody marvellous, but come with a high computational overhead. It's a never ending circle though, as more power is aailable coders get sloppier (less efficient code), and complexity of the code increases, which then chews more power.
So, my position is that the only reason they sound thinner is down to the talent of the codeer and the computational resources available.
As for the Soul comment above, each Minimoog is different, there are at least two personalities of MKS-80, and with the additional time that has passed the sound of individual units will be changing. These have character and lack the repeatability of a pure digital model.
Look at the amount of CPU used by something like Phosycon to emulate a cheaop old synth. It's certainly ery good and very close, but still not quite there. Scary huh? the simplest of synths, the extreme amount of CPU.
We have a way to go yet before general purpose software replaces hardware (Outside of VAs) |
|
|
|
|
