IK White 2A: lots of aliasing.

[alteregoxxx]

I noticed that IK White 2A has "a bucket" of aliasing, almost seems IK didn't use upsampling on it...this fact alone is strange, but it becomes even more strange when you see that their Black76 doesn't suffer of this problem, or at least, the aliasing is very low.

Here the two compared:





I've highlighted in red the aliasing caused by the 3th harmonic (8.1Khz) but there are a lot of other "alias" frequencies as you can see...
sample rate is 44.1 Khz and the stimulus signal is a pure sine at 12 Khz.

Ik guys?

[Compyfox]

Did you do a sine sweep?
Does it still occour at the same places?

Did you try to oversample "manually" (either with a samplerate of 96kHz or a 3rd party 2x oversampler)?
Is it still there?

Have you messed with SPAN's resolution or higher accuracy?

Do you keep in mind that a compressor/limiter is nonlinear by nature, so this could happen due to other reasons?


Did you maybe try to post on IKM's board as well?


A couple of weeks ago, there was a thread on KVR regarding program dependent release with both Black and White, now it's probably aliasing that is wrong.

I'm just saying...

[penguinfromdeep]

Is that aliasing or something else? LIke just harmonics generated by the emulation, I dunno?

[sunny_j]

at 44.1khz, use 993hz input tone for better results

[bezusheist]

i think that's just harmonics,,,i get very similar results with COMPassion in classic opto and british type 1 mods...(44.1k w/oversampling on...12k tone)
sounds good whatever it is...


edit: hmm, it's not there @96k for 12k tone, but for 11k it is...maybe it is aliasing?...or magic?...

[alteregoxxx]

Did you do a sine sweep?
Does it still occour at the same places?

Did you try to oversample "manually" (either with a samplerate of 96kHz or a 3rd party 2x oversampler)?
Is it still there?

Have you messed with SPAN's resolution or higher accuracy?

Do you keep in mind that a compressor/limiter is nonlinear by nature, so this could happen due to other reasons?


Did you maybe try to post on IKM's board as well?


A couple of weeks ago, there was a thread on KVR regarding program dependent release with both Black and White, now it's probably aliasing that is wrong.

I'm just saying...

Short answer:

Black 76 doesn't have aliasing in the same measure (not even comparable) of White2A...so.. (..and we can say all but an 1176 is a "linear device")

Long answer:

Sine sweep is no useful to easily measure the aliasing

In regards to the non intrinsec non-linearity of compressors and limiters,well...i know...but "good non-linearities" has nothing to do with what we are discussing here
"good non-linearities", referring with this term to "saturation" and similar concepts, must have an "harmonic relation" with the stimulus signal, that my case is a 12Khz pure sine: so 24 Khz, 36Khz, 48 Khz maybe attribuited to "good non-linearities" (read saturation pr similar concepts), NOT 8.1 Khz

Maybe intermodulation distortion, that is a typical phenomenon in limiters and compressors (and a "not-good non-linearities"), is also happening,for sure, producing non-harmonic frequencies...but, BUT 8.1Khz is exactly

44.1Khz - (12Khz x 3)

or, in other words, the alias of the third harmonic (36 Khz)
This thesis is confirmed by the fact that knowing the frequency of the stimulus signal i am capable to predict where the alias frequency will be...

also,this same predicted alias frequency is there even changing the analyzed plugin (Density MKII rather than Molot rather than Ferric -really really a lot of aliasing for this last- etc etc..)
Moreover, i can refer you of a compressor that shows ZERO alias: stripBUS from Sknote. Measure for yourself.

Furthermore...changing the SR to 96 Khz doesn't have sense at all in this contest: we all know that if you pull up the SR at 96 Khz, the usable bandwidth become greater, 0 to 48Khz, so is obvious that aliasing appears only for frequencies (including generated harmonics frequencies) greater than 48 Khz.. Upsampling was "invented" for this,no?

...remember however to don't get fooled if trying with a stimulus at 12 Khz and SR at 96 Khz,analyzing a device capable to produce 5th order harmonics, you don't see alias: it's there but you can't see it because is the choice of the stimulus frequency that is wrong for the case; the stimulus frequency in fact doesn't have to be an integer divider of the SR to correctly see the aliasing phenomenon

Last, Voxengo Span settings has no relations with this facts...changing them doesn't affect the measurement at all, as it should be...

[alteregoxxx]

i think that's just harmonics,,,i get very similar results with COMPassion in classic opto and british type 1 mods...(44.1k w/oversampling on...12k tone)
sounds good whatever it is...

My friend, that is aliasing ...i bet the big spike after 12Khz is precisely at 15.9 Khz (5th harmonic aliasing), no? zoom and check it
magic? ...no...aliasing...

edit: hmm, it's not there @96k for 12k tone, but for 11k it is...maybe it is aliasing?...or magic?...

No magic, as i said...i explained this in my previous post...

at 44.1khz, use 993hz input tone for better results

no no...993 Hz is really a low frequency value to see aliasing...even at 44.1Khz in fact you would need a device capable of (at least) a "significant" 21th order harmonic (!!!!) to see aliasing

[sunny_j]

I was merely quoting andy from cytomic, he had mentioned this in the GS thread. don't know much tho..

[macmurphy]

does any of this make the White 2A or COMPassion sound rubbish?

no. thought not.

[vata44]

+1 macmurphy

some people with all their science... only trust your ears

[Compyfox]

Gee, now this is a lot of yadda-yadda on a simple question.

Short answer:

Black 76 doesn't have aliasing in the same measure (not even comparable) of White2A...so.. (..and we can say all but an 1176 is a "linear device")

From the looks of your screens, it has - and god these screens are tiny (and I hate image shack with passion, since the zoom doesn't work for me on either machine I have).

Sine sweep is no useful to easily measure the aliasing

But it shows you if you're "wrong" on that behalf. If it IS aliasing, then it should move along the fundamental.

This thesis is confirmed by the fact that knowing the frequency of the stimulus signal i am capable to predict where the alias frequency will be...

In theory!

also,this same predicted alias frequency is there even changing the analyzed plugin (Density MKII rather than Molot rather than Ferric -really really a lot of aliasing for this last- etc etc..)
Moreover, i can refer you of a compressor that shows ZERO alias: stripBUS from Sknote. Measure for yourself.

First and foremost, your "analysing" showed a form of compression going on ALREADY. I think 2-3dB if I read that right on the White. This means, the signal is getting "non-linearic changed" right from the start.

I'd say you run into either intermodulation, or a simplified waveshaper. Why? Density and Ferric are INTERNAL OVERSAMPLED, StripBus (IIRC) can be changed on that behalf. I can't say anything to MOLOT.

Furthermore...changing the SR to 96 Khz doesn't have sense at all in this contest: we all know that if you pull up the SR at 96 Khz, the usable bandwidth become greater, 0 to 48Khz, so is obvious that aliasing appears only for frequencies (including generated harmonics frequencies) greater than 48 Khz.. Upsampling was "invented" for this,no?

Upsampling or more common "oversampling" is created just for that purpose, yes. To have less distortion in a signal. The thing is... if the plugin was sloppy programmed, it even shows up in higher frequency ranges. So to see if this "aliasing" is gone, you either use a higher samplerate or oversample it.

...remember however to don't get fooled if trying with a stimulus at 12 Khz and SR at 96 Khz,analyzing a device capable to produce 5th order harmonics, you don't see alias: it's there but you can't see it because is the choice of the stimulus frequency that is wrong for the case; the stimulus frequency in fact doesn't have to be an integer divider of the SR to correctly see the aliasing phenomenon

This is why you do a sine sweep and don't stick to ONE single frequency.

Last, Voxengo Span settings has no relations with this facts...changing them doesn't affect the measurement at all, as it should be...

But it shows you more detail if you raise the sample buffer. Furthermore, have you cross checked with ANOTHER analyser, or maybe even C.Budde's analyser just to be sure? I don't trust SPAN alone on that behalf. It showed me more than once that there was signal or content there that other meters didn't show me.


Whatever it is, only the programmer from IKM can clear that. And to clear that, open a ticket at IKM's board. Else I get the feeling this turns into another slugfest here on KVR yet again.


Oh on regards of reading out compressors and the likings, contact bootsie. Especially since you said his Density and FERRIC does have aliasing. I'm pretty sure he'll love to discuss that out with you to iron out bugs.

does any of this make the White 2A or COMPassion sound rubbish?

no. thought not.

I'll throw in another one: Sknote GTS-39 - 9 harmonics, though a bit on the agressive side (soundwise) that thing. Then again, use either White or GTS-39 for 2-4dB gain reduction and your ears start to sweat since it (might) melt away from what it hears.

Just my 2c.

[audiovet]

DELETED

[alteregoxxx]

Compyfox...to make it short...in the meanwhile i talked about this with the Molot developer and he confirmed that this IS ALIASING (obviously)..so...no bla-bla-bla...i would not seem presumptuous but, as electronic engineer, i know a little about the phisics behind sampling theory and a little about inter modulation distortion, non-linearity of electronic circuits etc

In regards to Bootsie plugs, try for yourself and you'll see that BootEQMKII saturation doesn't suffer of the aliasing problem at all, Ferric instead, even without dynamics or limiter knobs engaged (so intermodulation is not the problem here)shows a lot of aliasing...

so if Ferric is oversampled (i don't think so because there isn't a word about oversampling in the Ferric pdf manual) is bad oversampled, or at least BootEQMKII (in the manual Bootsie says explicitly it is 4X oversampled...) is better oversampled, no?

Density MKII maybe 2X oversampled because there is aliasing, more than that of BootEQMKII saturation with the saturation knob at max, but a lot (and i mean really a lot) less than Ferric.

However i wrote a post on VarietyOfSound blog to invite him to this discussion

Now...On the question "does any of this make the White 2A or COMPassion sound rubbish? " the answer is: yes and no; no, if you use one instance; but for sure a mix with 50 instances of White2A, it will sound worse than a mix with 50 instances of oversampled White2A, because their aliasing will build up, creating audible non-harmonic frequencies; audibly worse or really really really slightly audibly worse are the two extremes, depends on the audio material)

as an extreme approximate similarity...imagine 1 violin on 100 violins that play a slightly detuned C (not a G, a 5th interval,to keep alive the similarity with the non-aliased good harmonics) while the others 99 are in tune...no problems....imagine now 50 detuned violins on 100...

[Fritze]

So if this aliasing does it make sound bad it should be AUDIBLE, right? Would you please make an audio comparisson, alteregoxxx? Once at 44k and once at 96k should do the trick. Only a simple mix with 50 tracks with a Whitie, just as you said.
Thank you.
You know dancing about music is like swimming about talking. Or something like that.

[@midnight]

I know the initial version of FerricTDS had 4x oversampling and then he revised it to remove the oversampling and made it zero latency, probably why there is aliasing now.