There's a big difference in the amount and type of distribution of harmonics created if you soft or even hard clip the plugin by going outside the "operating range" of the modeled plugin.

Sure, your test is still valid but it shows how some of these plugins are not optimal as guitar stomp boxes without anti-aliasing measures.

Take for instance a slew of various Acoustica Audio plugins. Some of the older ones almost literally "explode" when driven to exactly 0dBFS and cause much more artifacts than just pure harmonic distortion. Most of them have sensible default settings now and should work at 0dBFS as intended but I just wanted to point out that your logic above just doesn't always cut it. Just as it wouldn't cut in the real world of analogue.

If you'd compare various pieces of analogue equipment the same way, trying to establish which units are "the cleanest" in terms of distortion, you'd be making a complete blunder. The analogue modeled plugin world is not much different in this case either.

The problem arises when it is present misleadingly, be it on purpose or out of ignorance.

.. just saying.

Surely you must be trolling now? You actually raised the test signal past 0dBFS.. it's at +6dBFS in these ones.

@ bmanic

If there is one wheel missing on a car, the air pressure in the wheels can change that fact and make the fourth wheel reappear?

And why are you lying? The volume is the same between the two screenshots.

The volume of the signal is only the variable to enforce a certain level of harmonics, because it are the harmonics that show how effective the anti-aliasing measures are.
A high level of harmonics is used to see it very clearly in the measurements.

How can anyone even believe, that anti-aliasing filters were volume dependent?

Please guys, if you don't understand the technical aspects of a problem, ask, or learn the mathematical basics, but do not make absurd conclusions by looking at a parameter (volume) with absolutely no significance.

There seem to be anti-aliasing measures, but just the loudest 3rd harmonic aliases the most, almost without any reduction:

And also among saturation plugins Fabfilter is setting standards:

I think what bmanic is trying to say is that the lengths to which you push these plugins is in some sense ridiculously outside of what any normal user would ever actually do.

So the counter-argument (I think) bmanic is trying to make is not to deny that the plugins have aliasing, but that there is a line after which the law of diminishing returns kicks in and your effort to eliminate aliasing must start interfering with your effort to mix/master/make music, which defeats the purpose.

This is my thought: If you want to create maximally useful information for the community here, it would be more helpful to test the plugins without cranking them to ultra-hot to show a "real life" picture of how much aliasing buildup normal, sensibly gain-staged plugin use brings forth.

Let's hope @ Fred vom Jupiter never discovers AirWindows.

(Subscribing!)

It is interesting to see and hear stuff like this about digital domain, in contrast to bunch of everyday plugin releases which seem to rely more on graphic/visual innovation and interaction than sound development.

However, FabFilter's video about sample rates shows the differences between oversampling and using higher sample rate in case of heavy serial processing. So heavy processing is rarely used which was stated many times here and the differences were not something I would consider listening again as it's just keeping me from making music.

I am curious though about any improvement that can be done for making software sound (even) better.

@ occultron

Bmanic did not. He was trying to insult me and called the measurements stupid and even accused me of trolling, while he used a silly and totally false technical argument.

You are argueing and therefore I can answer your question:
the argument is not correct, and the technical reason is this:

Measurements only make sense, if they give clear results. Does that make sense?
Harmonic enrichment in normal use cases is taking place in the range of sub 1% THD, or only very few percent. It's very subtle and can barely be seen in an analyzer, because the aliased signal energy is so low.
That makes it such a nasty problem, if ppl do not know how it accumulates with any signal treatment that creates additional harmonics.

For example, brainworx recommends to place their SSL-channelstrips on every track. So from the very beginning aliasing is injected into the signal. Now every every compressor adds to the injected aliased frequencies his own harmincs, every "analog" EQ adds his harmonics to the aliased signal from the SSL and from the previous EQ, then a tape emulation adds his own harmonics to the injected aliased harmonics of pre, compressor, EQ and then this goes to the subgroup where an enhancer may add his own harmonics to the injected aliased harmonics from the strip, EQ, compressor, tape and then the subgroup compressor adds his own harmonics to the aliasing harmonics buildup from the strip, EQ, compressor, tape, subgroup compressor and then it maybe goes into the mixgroup, where each "analog" sounding plugin adds harmonics to the aliased garbage and then it goes into the mixbus a nd gets the final "polishing" by more great harmonics, while some marketing gurus sit in front of their HW-consoles and walls of HW and tell the audience, how great things sound when done with their 100% aliasing plugins ITB.

While everybody becomes furious about the messenger of the bad news?


Someone else attacked me and posted a video from Dan Worrall. Has nobody watched that video? Dan Worrall does the same and has some great examples how aliasing destroys the sound quality even with modest processing chains.

Ofcourse I do not make a video, but I have found a way to show it with just two pictures.
It's up to everyone else what he wants to do with that information about aliasing handling of the plugins I tested and tried.

IMO a rationale behaviour would not be to reject technical facts, but to discuss what everyone could do, if he hears the devastating effects of aliasing, to avoid it.

There really is no point where intentionally eliminating fold-back aliasing from ever occurring should somehow interfere with any other process.

Fred seems to be on a hunt for making that simple by hopefully creating a list of plugins that are good citizens in 48kHz and below projects so that if we intend to work at those sample rates, we can have a list of tools that can do that job without aliasing, thus not needing to have any second thoughts about aliasing interfering in any way with creative decisions.

If that inspires devs to create solutions to make their lower sample rate support more useful like some clearly have, that could only be a good thing.

There are several options:

1. Raise the project sample rate to 192k. Problem: CPU; still not enough if high harmonics are used somewhere ITB.
2. Put aliasing plugins into a plugin-wrapper with oversampling (workflow nightmare).
3. Replace aliasing plugins with plugins who offer protection from aliasing.

The best for us as customers in the long run is to create pressure on developers who destroy signal integrity with aliasing.

Wow, sorry that didn’t work out.

Although the plugin does not come close to the absolutely ****ing incredible sound of the HW, this is a plugin where aliasing finally is decently handled by Brainworx:

Some people obsess endlessly in circles about measurements.
In the meantime, other people with ears are releasing records (and soundtracks, and commercials, etc) heard by millions around the world.

I think most people are just posting on forums.

Actually, they're very close.

Well, it is, but it is not relevant to the research at all. Some of that “people with ears” made “Death Magnetic”.... a lot of “people with ears” are great artists (even engineers are), but it does not mean they don’t use bad technology, or just plain “flawless”. Appeal to authority is a logical fallacy. We all making mistakes. We are all not ideal.

About ears: my own hunting for aliasing actually begun after i was hearing that’s something is wrong with most nonlinear processing ITB. When i stopped using not so great saturation plugins - it’s all improved tenfold.

I also want to point out that this is only one part of the problem since as far as I know every DAW might capture aliasing through different transformative operations, and I’m not just talking about the included plugins (which usually suffer the same fate), but menu-based operations, especially regarding pitch-shift. So we also need oversampling or aliasing suppression options within basic transformative DAW operations too.

If anyone knows of a DAW that has this covered already, please let me know.

I know Logic doesn’t for sure.

Thanks for posting it, though!

Can you do some UAD ones? I've never seen an aliasing test for any of theirs.

Hey I understand about aliasing and I agree, if you hear it and it bothers you, the you should look for ways of getting rid of it.
If you are doing an amazing Jazz or Orchestral music that needs the airy full sound of every bit and you are hearing it, yes, by all means do what'cha gotta do.

Aliasing in my world is a non void. I don't think about it at all and it has never caused me to lose a mix session nor song placement etc.
I think my mixes sound great as I do Hip Hop, R&B, POP etc.

To me, it's like hiss was to us in the 70's, 80's and 90's. Hiss when compressed could be up to -40db and more and when you pressed stop and you didn't have gates on, we were like wow but we still got it done.

I've had this opinion from the beginning after seeing all the pictures and where it is on the db spectrum but what confirmed it for me was the Dan Worrall video from Fabfilter as he has went deep into it and came up with the same opinion as me. If you haven't seen it, I would highly recommend.

https://www.youtube.com/watch?v=-jCwIsT0X8M&t=827s

Marc

You seem to also have some issues with reading comprehension. Look at what I wrote.

Anyhow, looks like you aren't trolling but actually serious about all of this. I'm truly sorry you have to be so bullheaded and refuse to listen to the criticism about the method you are using and refuse to understand where you are making mistakes and thus creating all this worthless data.

Well.. carry on.

There WAS a complaint against the sound of digital... using the same exact descriptions: cold, hard, harsh, etc.

So the awareness of aliasing has always been there. As with all things math, the truth is in there is you know how to identify, quantify and improve it.

BTW, working at higher sample rates creates as many problems as it supposedly addresses.

Yes one would think this is his aim. Unfortunately he is creating a bunch of false data and doesn't even seem to understand why it is false and worthless.

Heck, his "mission of getting rid of aliasing" is something I can very much sympathize with (just check my post history over the years)..

But I'm also a hard advocate for fair and actually useful data, especially if it is presented aggressively "for a cause". Misleading data, as has been presented in this thread, does nobody any good.

There's no such thing as "lying for the greater good".. it's just lying and usually for selfish gain. I think it's pretty clear Fred isn't here for any public service cause considering the super aggressive and 100% willfully ignorant stance.

His attitude and stance is so ignorant and aggressive that I suspect none of the people giving good feedback at first, pointing out some flaws in the method, stuck around. Not actually sure why I wasted these minutes writing this either.. so I'll just bugger off.

Better let Fred do what Fred does.

Yes, I understand. I guess you and bmanic just disagree about methodology here. His point is that the amount of aliasing in a plugin does not scale linearly with input, and because you drive the plugins so hard you end up with results that show a higher _ratio_ of input signal vs aliasing noise, which is not necessarily interesting information if few people actually ever use the plugins at such settings.

I'd be interested in seeing a chain of, say, 8 various Waves/PA/etc. mixing plugins at normal settings and see how much that aliases, because that's information that's easier for me to compare to my own real life working conditions.

Of course it's also useful to know that some plugins handle aliasing so well that you really _can_ crank them if you want to, so it's been interesting to look at the screenshots so far, too.

Well, seems he could (should?) have discovered the previous thread on plugin quality that included talk about aliasing... you know, like someone suggested almost immediately;

thus..

See above post/link and do a search and I'm convinced you'll find something.

Are there any technical analysis of the (possibly detrimental) effect of extensive oversampling on sound/tone/signal integrity?

Curious, since Chris from Airwindows seem to prefer (low level) aliasing to any kind of oversampling whatsoever - though he always seems to be on his own esoteric dsp path (and always runs at 96k) so hard to know what is applicable in normal conditions.

I’d like to read more about that as I’ve only seen upsides except for computing resources.

Let me just say that Fred Vom Jupiter has golden ears and I’m not even joking.