192khz, verdict

General feedback, questions and feature requests.
Post Reply
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

192khz, verdict

Post by MadGonzo »

I realized something today, which made alot of sense, and im ashamed to say it never occured to me (the whitenoise)

The whitenoise you usually hear in the upperranges when trying to make a supersaw, which i have diligently tried to do for some time, is because of low (unacceptable) khz sample quality, such as 48khz, or even 96.. you have to go all the way to 192 to really make it inaudible on a (digital) VSTi.

In theory the sound should be clearer and more "desirable",
and MY GOD, it was. the sound changed COMPLETELY, in a good way
i could render out the difference, whereas render1 in 192khz and render2 in 48khz, but whats the point ->

However, cpu usage increases exponentially, and so youre back to the good olde render to listen era :P
i bought my quad core ~10ghz (FL8.5 supports multicore, 2.8ghz per core) for this same reason, to not be annoying when making music

needless to say, it was a shock to see cpu make a hyperjump
i also learned that 11 unison voices is a good maxima for supersaws in general, 10 should suffice

I dont know if theres anything you (karmafx) can do to optimize for this kind of thing, because its natural that exponentially more work means exponentially more cpu usage :P
but you should take a look, if youre able to produce 192khz
in any case, it was a real pleasure listening to my instruments with this highest quality available (my soundcard supports it too so!)

Other than that, is there any way to overcome this? such as hardware related

Verdict: It's serious business.
If available, don't get intimidated by having to render previews.
At least you should be able to play one instrument at a time and freeze the others.

I suppose that analog synths (sometimes) are of an un-resonant nature because of inherent cycle "flaws" caused by (such as) copper wires,
which makes the cycles not interfere as often, if ever, generating unique waves pretty much always. Prevents whitenoise? Probably
(I also believe you can get 192khz without any mentionable cpu loss from them either way)
karmafx
Site Admin
Posts: 1194
Joined: 07 Mar 2005 16:37

Re: 192khz, verdict

Post by karmafx »

It sounds to me like the "whitenoise" you're referring to is actually aliasing.
Can you please explain what exactly you mean by "supersaw" and how you go about creating it?
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

Re: 192khz, verdict

Post by MadGonzo »

karmafx wrote:It sounds to me like the "whitenoise" you're referring to is actually aliasing.
Can you please explain what exactly you mean by "supersaw" and how you go about creating it?
pick any way you want... 6 oscilliators (with different tuning and/or octaves), or just a regular osc2 + heavy unison (more noise)

try the Bagpipes pad, with 8 voices in 192khz
sounds different :P
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

Re: 192khz, verdict

Post by MadGonzo »

http://commandante.fwsnet.net/192

_3.wav (192khz)
_4.wav (48khz)
_4.mp3 (48khz with percussion, bass and sublead)

difference is enormous (3 vs 4), including the filesize :(
i blundered on the instrument though.. seeing as that isnt a supersaw and you cant really (well) hear whitenoise differences

anyways ill post some stuff below:

KVRAudio http://www.kvraudio.com/forum/viewtopic ... 09#3556309

My own study on this reveals that 1 saw is cool, 2 saws is cooler, 3 saws is great, 4 saws is fantastic, 5 saws is extreme 6 saws is holy crap!, 7 saws is omg! and 8 saws is terrific, and 9 saws is yeah uber terrific, 10 saws is terrific, 11 saws is extreme, 12 saws is fantastic, 13 saws is "is there a whitenoise osc running?" 14 saws is "it sounds like the last one." 15 saws is "hmmm..."

The reason for this is because many VSTi modules have oscillators that are too perfect. The oscillator reads in a digital wave form and begins playing it from the start of the wave. Now because of this, each time the wave "cycles" it is exactly the same every time. which means that when you stack them and detune them, all the overtones are exactly alike and produce a somewhat digital hiss in the upper frequency range.

Now if you're smart you'll have picked a virtual analog VSTi where there is a simulation of an acual oscillator, the wave form will not be exactly the same each time. This is not the same as a random addition or subtraction from the slope in the oscillator's algorithm. Real hardware oscillators are imperfect because of the "disinhibitions" in the hardware, that because the signal is usually running through real copper wire, and various components that have strange physical properties that can't always be predicted, but follow a rational pattern that can be imitated, but not simulated precisely.

A computer has no components so far that can be utilized to produce a real analog signal, so the virtual analog VSTi is the substitute.

Now your sound buffer on your soundcard can as of late take floating point values, which is a huge leap in sound quality. But that does nothing for people in the time domain. The biggest reason that the sounds end up sounding noisy is because running your audio at 44.1 or 48 khz still isn't enough to facilitate a broadband saw wave because the saw wave is a mixture of sine waves using a successive logarithmic mix of all harmonics. the problem is when the harmonics go past 11 khz. There just aren't enough samples per second at 44.1/48khz to form a true sine wave at frequencies above 12 khz. I suggest to anyone who wants to experience a supersaw in the best possible way, to run your audio buffer at 32*f/192khz this will allow anyone to form a precise sine wave at frequencies all the way up to 22khz.

I have tested this, and I have found that the sound comes out to be a lot more clean, and less dirty from aliasing. if your VSTi doesn't support those high modes, then I guess you'll need to get yourself something better.

I believe that 64-bit audio will also make it possible to give digital signals the same subtle dynamics that an analog signal possesses. No more gritty chalky sound yay!

For the future in sound, I want to have one of those Intel i7 processors, and a 64-bit OS and 8 TB of onboard RAM Very Happy I'll be unstoppable xD

KVRAudio reply http://www.kvraudio.com/forum/viewtopic ... 09#3556309

that reply wasnt really ontopic but seeing as im not a master of the arts hopefully someone knows, or have any ideas

btw this thread altogether was just me trying to gather some information on such as, should i continue using 48khz sound
is there any way to improve the overall sound of karmafx synth
and so on
i guess it always will be, at the cost of cpu :P
right now the synth does a bangup job as cpu usage is at extreme lows
but imo theres no harm in seeing what can be done
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

Re: 192khz, verdict

Post by MadGonzo »

im sweating
i just rendered a demo song using 96khz
its sounding mighty good, and i didnt really do much except the usual insert instrument, add beat

anyways, this is better than porn
ill upload later, its 40MB after all :(
karmafx
Site Admin
Posts: 1194
Joined: 07 Mar 2005 16:37

Re: 192khz, verdict

Post by karmafx »

...or it could just be accumulated aliasing.
Is the problem just as bad with hardsync *off* ?
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

Re: 192khz, verdict

Post by MadGonzo »

ill have to do some extensive testing to find out if hardsync has any say
turning hardsync off is a no go on the instrument i was testing with so ill have to make another, preferrably supersaw which spreads alot saws :P

anyways here is the 96khz render:
http://commandante.fwsnet.net/192/ghosts2k9_2.wav
thats a 96khz version

http://commandante.fwsnet.net/192/ghosts2k9.mp3

btw dont misunderstand, i dont know if its just the fact that theres a huge "quality difference" between the two
but if 96khz or more is required for ungodly sound, then thats how it is
regardless, what i posted above which doesnt come from me should make sense to you, i dont understand it fully :P

aliasing, if youre talking about frequencies morphing together because there isnt high enough samplerate? something like that
i dont know, but if theres any room for improvement you should tell :P

btw which filters do you use? n-point? sinc? linear? it would be cool to be able to choose :P does filter apply for aliasing? :|
karmafx
Site Admin
Posts: 1194
Joined: 07 Mar 2005 16:37

Re: 192khz, verdict

Post by karmafx »

Aliasing can be a problem (even in digital hardware synthesizers), but we need to establish wether the "whitenoise" your'e experiencing is aliasing or something else and wether it is justifiable or a bug. Aliasing does not normally sound like whitenoise, though.

I'll listen to your samples, and be happy to discuss the samplerate issue in depth...To get aqauinted with aliasing, you can take a look in the manual under "Analog vs. Digital".
karmafx
Site Admin
Posts: 1194
Joined: 07 Mar 2005 16:37

Re: 192khz, verdict

Post by karmafx »

the second link doesn't work ?!?:
http://commandante.fwsnet.net/192/ghosts2k9.mp3
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

Re: 192khz, verdict

Post by MadGonzo »

oops, http://commandante.fwsnet.net/192/ghosts2k9_2.mp3
was version _2

well after reading your manual, which was extremely revealing about karmafx (well written btw)
i can tell you right now its probably not a sound issue :P
come to think of it, distortion would be completely different for "exotic" samplerates

btw what interpolation does the synth use internally?
you should have a 512 point sinc option :P
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

Re: 192khz, verdict

Post by MadGonzo »

turning off internal reverb, using reverb vst version = 1000% sound improvement
just like i said earlier, internal reverb leaves alot to be desired
might be related, who knows
karmafx
Site Admin
Posts: 1194
Joined: 07 Mar 2005 16:37

Re: 192khz, verdict

Post by karmafx »

MadGonzo wrote:turning off internal reverb, using reverb vst version = 1000% sound improvement just like i said earlier, internal reverb leaves alot to be desired might be related, who knows
oh, no... :oops: (btw: Havn't we been over this before? :lol: )...but weird?! :shock: The code they use is 99% the same.
Are you sure it is not the settings. Remember, "size" in the synth equals "diffusion" in the reverb.
I'll look into it, but if possible please post or pm me an example.
MadGonzo
Posts: 296
Joined: 07 Jun 2008 23:14

Re: 192khz, verdict

Post by MadGonzo »

well for one, i can never seem to recreate the same unique sound your VST has
i also posted a new demo in my bank thread, which is awesome sounding, and (wait for it...) it uses the VST reverb :P

ah well, i could investigate more i suppose, but im pretty sure something is "funny" here...
ill look into it

btw. i checked out the internal delay, and its pretty solid, so its NOT the issue
Post Reply