What plugin synth has the least aliasing (especially in the oscs)?
psytendo
Started Topics :
1
Posts :
73
Posted : Mar 2, 2008 17:51
Yeah, makes sense, altho i'd think that ANY soundcard brickwall filter would be better than no internal brickwall at all in a plugin...
BTW. found an interesting synthy on the internet:
http://www.audjoo.com/Helix.html its free!!
its just vsti so i need some convertor to make it work under logic. but the audio demos are promising.
Listen to some of the demos, it has a very nice high end, not perfect but much better than most stuff i have heared from other expensive plugs.
It seems to be based on wavetables because i hear some artifacts but at least its harmonic .
Started Topics :
9
Posts :
1448
Posted : Mar 2, 2008 17:52
Quote:
On 2008-03-02 16:59, psytendo wrote:
@UnderTow:
Interesting feature of rapture!
But as i understand interpolation, it doesnt fix aliasing.
This is Sync-Interpolate. This is also what is used in sample rate converters. Try it, there is no aliasing when this option is turned on and the synth is rendered/bounced/frozen.
Quote:
One option could be an internal brickwall filter just underneath the main nyquist frequency (samplerate / 2).
But you'd need to oversample the oscs (make them run at a higher samplerate) and filter that down.
This is probably what Predator uses but it won't (and can't) remove aliasing completely. It just moves the problem further away. You need a completely different approach like what Rapture (and probably Massive) uses.
Quote:
Also a linear phase FIR brickwall has some latency, not sure how much of a problem that is.
I doubt using a linear phase filter would make that much difference. After all, nearly all these synths already have (non-phase linear) filters included.
It makes sense to use linear phase filters for some applications when you don't want to affect the phase of the original material but with synths, there is no original material. Whatever comes out of the synth's output is the original material.
Quote:
Judging by Massive i'd say that its pretty heavy on the cpu.
I am guessing that Massive uses Sync-Interpolate in realtime.
Quote:
Another solution could be to create the waves from some sort of additive synthesis and just never add patials that are higher than the niquist frequency.
AFAIK Predator (and all Rob Papen's soft Synths) use some kind of additive method to generate the original wave forms. (But that doesn't mean it used realtime to produce the different pitches). Looking at the output of Predator, things are very dynamic but a bit messy.
UnderTow
heretical
Started Topics :
1
Posts :
77
Posted : Mar 2, 2008 18:07
I really don't know why everone cares about aliasing so much. IMO aliased oscilators sound much more "analog" at the lower registers than anti aliased. It also really screws up alot of FM sounds, you can't get the harshness like you can with older FM synths before anti aliasing became this huge marketting nonsense.
Reaktor 4 and 5 might work for you if you want anti aliased. I would be very suprised Massive has aliasing for this reason.
Early versions of absynth had no anti aliasing, current version lets you turn it on and off.
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Mar 2, 2008 18:18
sound with no aliasing are more "clear" to me and need less eq but maybe i m wrong ,i call aliasing all the bad freq in oscillators
UnderTow
Started Topics :
9
Posts :
1448
Posted : Mar 2, 2008 18:18
Quote:
On 2008-03-02 18:07, heretical wrote:
I really don't know why everone cares about aliasing so much. IMO aliased oscilators sound much more "analog" at the lower registers than anti aliased.
Do you have examples of this? Are you sure you are not just associating a particular implementation of non-aliased oscillators with a digital sound? After all, analogue gear does not alias.
UnderTow
heretical
Started Topics :
1
Posts :
77
Posted : Mar 2, 2008 18:52
Quote:
Do you have examples of this? Are you sure you are not just associating a particular implementation of non-aliased oscillators with a digital sound? After all, analogue gear does not alias.
UnderTow
Well only purely qualitative examples obviously, nothing quantitative.
One example is the sh101 knockoff in reaktor. After using it for a few years in reaktor 4 or 5 which only has anti aliased oscilators, I installed reaktor 3 just to mess around. Reaktor 3 would be considered a very dirty sounding synth by modern standards but to my ears the sh101 sounded far closer to the original than the 4 or 5 versions.
I also noticed this from listening to some kid pound on his fathers moog on youtube. First thing that came to mind was "wow that sounds like a really dirty aliased signal path".
I think what has got lost with modern synth programming is that because analogs don't alias that anti aliased oscilators must sound more like analog. Complete nonsense IMO. In the 90s when the debate was how "phat" or "analog" a synth sounded there had to be a reason none of those synths used anti aliased oscilators. I doubt it was a matter that the programmers just didn't know hot to impliment the algorithm. More like the market would have never accepted the pussy/softness of anti aliased oscilators.
I also wouldn't doubt that anti aliasing is a huge reason why virtually no softsynth has a character of its own. Nord modular 1 has the nord character, G2 has the same character as every other anti aliased soft synth.
psytendo
Started Topics :
1
Posts :
73
Posted : Mar 2, 2008 18:57
UnderTow,
First of all, it is sinc interpolation, not sync Big difference and easy mistake to make, so i'm not going to hold it against you
Playing a sample at a different samplerate than the 'main' samplerate IS a form of sample rate conversion, so here interpolation is useful.
I'm not sure if interpolation is a solution to aliasing as it implies the calculation of EXTRA samples between the original ones. But in this case we actually want to reduce the bandwidth of the oscilators so we need LESS samples than the osc is trying to produce.
The only use of interpolation in an anti-aliasing algorithm i can think of is if you internaly reduce the samplerate in the oscs to well below the niquist freq and then apply interpolation to smooth the jaggies. But this would mean loss of the high frequencies (and, indeed, no aliasing).
BTW. the Sinc function IS useful in anti aliasing tho.
Sinc is a bit of a magic thing, popping up everywere in signal processing.
Linear phase brickwalling is a good thing in the way that it doesnt interact with the original waveform. It alows you to take away the unwanted audio (above niquist) without changing the hearable part. I agree its not nessesary to have a linear phase but it is easy to implement as a FIR.
The problem with a brickwall filter is that it is very very sharp. If you do not implement it as linear phase then the phase related to frequency would be altered dramaticaly.
Not saying it is good or bad, but it WILL alter the wave that the oscilator is trying to produce.
You also say that most synths already have linear phase filters but i'm not sure what you mean by that. If you can see aliasing then at least you can say that the oscs are not brickwall filtered, linear phase or otherwise.
It could be that there is some filtering at the end of the synths chain but the problem is generated by the oscilators and cannot be filtered out afterwards. So the oscilators need their own sharp filters to cut of all freqs above niquist for no aliasing.
psytendo
Started Topics :
1
Posts :
73
Posted : Mar 2, 2008 19:00
@heretical
I realy agree that there is no problem with aliasing in the lower registers. I like a bit of grit myself btw, i'm not sure if you realy mean aliasing. is is it that 'digital' sound? then its just the low samplerate of the waveform (which has a niice sound, especially filtered )
Any VA synth can sound good in the low end anyway.
My problem is getting nice transparent high freqs.
Different story altogether...
psytendo
Started Topics :
1
Posts :
73
Posted : Mar 2, 2008 19:01
@PoM
What you mean is ARTIFACTS Aliasing is just one of the possible artifacts, and a particulary nasty one since its not harmonic...
heretical
Started Topics :
1
Posts :
77
Posted : Mar 2, 2008 19:25
psytendo,
I don't think its the sample rate though. Reaktor has been able to do extremely high sample rates since 1999.
If you listen to the Moog Voyager
You can hear that "bite" to the sound in the higher registers that vanishes with band limited waveforms. Sure modern softsynths doesn't alias in the same way an analog doesn't alias, but you have thrown the baby out with the bathwater. I believe the way older digital implimented that analog "bite" is what gave synths their characteristic sound. A simple subtractive patch on a nord lead 2 sounds nothing like an older roland synth with the exact same settings.
With anti aliased softsynths a simple subtractive patch sounds virtually the same on any synth.
They all sound "good" but they all sound "good" in the exact same way.
heretical
Started Topics :
1
Posts :
77
Posted : Mar 2, 2008 19:30
that Sylenth1 is a good example. Extremely clear quality, alot of snap to the envelopes. However, if you listen to the sound samples in the player they all basically sound exactly the same.
When the filter opens up on a patch, you hear that exact same old same old, anti aliased soft mush type sound.
psytendo
Started Topics :
1
Posts :
73
Posted : Mar 2, 2008 19:54
heretical,
With samplerate i mean the samplerate of the waveform, not the samplerate of the plugin itself.
Imagine a sampler that generates 96000 samples per second.
But now imagine you load a flute sample and play it 4x too slow.
The sampler will still produce the 96000 samples per second!
But, since there are not enough samples in the flute sample the sampler has to insert extra samples to fill up the 96000 samples it needs to produce. Still with me? If it would just repeat the previous sample value you would get the effect you are talking about, the gritty digital sound. If you would use an interpolation algorithm then you would lose the digital artifacts and get a smooth waveform.
But -none of this has anything to do with aliasing-.
What you hear in the moog demo is not aliasing!! (well, there is aliasing from the low quality of the movie! lol)
Notice how the 'extra' digital atrifacts move together with the pitch that is played.
This is just the result of replaying a waveform at a speed slower than the samplerate (which can be considered infinite in the analog domain) without interpolation.
And again, i dont mind a bit of digital grit, but i already have so many ways of producing this kind of sound with my nord modular ( how about using sequencers as oscilator! hah! That will get your PPG blood pumping ).
"They all sound "good" but they all sound "good" in the exact same way"
There are more ways to manipulate the sound than by introducing aliasing...
And aliasing also always has the same character, so no real improvement there, aliasing is just another sound.
-==-
Anyway, this is not a discussion about the pro's or cons of aliasing.
I'm just looking specifically for plugs that have no aliasing or as little as possible.
psytendo
Started Topics :
1
Posts :
73
Posted : Mar 2, 2008 19:56
Quote:
On 2008-03-02 19:30, heretical wrote:
that Sylenth1 is a good example. Extremely clear quality, alot of snap to the envelopes. However, if you listen to the sound samples in the player they all basically sound exactly the same.
When the filter opens up on a patch, you hear that exact same old same old, anti aliased soft mush type sound.
Still this is something i want to use in my music. Having everything sounding digital is just as boring as having everything sound analog. So i want to combine both. Whats the problem with that?
UnderTow
Started Topics :
9
Posts :
1448
Posted : Mar 2, 2008 19:56
Quote:
On 2008-03-02 18:57, psytendo wrote:
UnderTow,
First of all, it is sinc interpolation, not sync Big difference and easy mistake to make, so i'm not going to hold it against you
DUH! I should proof read what I type.
Quote:
Playing a sample at a different samplerate than the 'main' samplerate IS a form of sample rate conversion, so here interpolation is useful.
Agreed.
Quote:
I'm not sure if interpolation is a solution to aliasing as it implies the calculation of EXTRA samples between the original ones. But in this case we actually want to reduce the bandwidth of the oscilators so we need LESS samples than the osc is trying to produce.
I think what happens is that each sample value in the OSCs gets replaced by a (windowed) sinc function centered at and scaled to the sample value. Then the output of all the sinc functions are summed and you have an alias free output signal.
Quote:
Linear phase brickwalling is a good thing in the way that it doesnt interact with the original waveform. It alows you to take away the unwanted audio (above niquist) without changing the hearable part.
Well it does to a certain extent as all linear phase filters have pre and post echo. (ringing). That is one reason why minimal phase filters are often used for various purposes (including sample rate conversion).
Also I am not sure you can really use a linear phase filter for this particular purpose as AFAIK, you need a minimal length of material to be able to calculate the filter. I don't think this is practical for synthesis. Not to mention the latency...
Quote:
If you do not implement it as linear phase then the phase related to frequency would be altered dramaticaly.
I believe you can get good results with minimum phase filters.
Quote:
You also say that most synths already have linear phase filters but i'm not sure what you mean by that.
No I meant non-linear phase filters. In other words the phase of the oscillators are already altered by theses (resonant) filters and no one is complaining about that AFAIAA.
Quote:
So the oscilators need their own sharp filters to cut of all freqs above niquist for no aliasing.
My point was that you can not really use filtering. As soon as the signal is created, it needs to be alias free. It can not be filtered after the fact. (Well maybe if you use insane oversampling but that to me seems like the wrong solution to the problem).
UnderTow
psytendo
Started Topics :
1
Posts :
73
Posted : Mar 2, 2008 19:59
heretical wrote:
"In the 90s when the debate was how "phat" or "analog" a synth sounded there had to be a reason none of those synths used anti aliased oscilators."
YES!. antialiasing is computationaly heavy!
Processing power in the 90's was barely enough to run decent main filters so noone was worried about aliasing in the oscilators yet!
.
).