BuzzCut (The Erosion Clipper)

[LusiD_Music_UK]

Come on, the dev backpedaled on that one thing, fairly. That doesn't change the fact the critique was based on a combination of two things:
A somewhat half-baked understanding of the nyquist theorem.
A pedantic interpretation of the copy text.

"Bullying" is a bit strong maybe, but it is pretty hostile in my opinion.

Again, the devs claims *are* true. Replacing (yes, replacing. And yes, even without oversampling) the reflected harmonics with noise is in fact what happens. It's what the nyquist theorem predicts should happen, and it does! Try the plugin yourself to find out.

Is resulting sound is pretty darn aggressively noisy? Yup! Is it "better quality" than just oversampling the clipper. Absolutely not. But if you try the thing for yourself it'll immediately become obvious to you that a clean high-quality sound isn't the goal. I personally won't buy it though I did try it and think it's good. I won't have much use for it for the music I make. But the reception the dev got is undeserved. It's a good product and does do what it says on the tin (unless you are committed to a pedantic read of the "eliminate aliasing wordings).

Thanks for the backup here. I will say that it is absolutely designed to ALLOW for a prestine sound. (try the "$120" preset) 16x oversampling with just a touch of erosion, the -100dB aliasing is replaced with "analog noise" arguably the "cleanest clip that is possible period". So its for both dirty and clean! But thank you <3

[Andreya_Autumn]

Right! I admittedly didn't try that setting.

[martinvankoek]

First of all, best of luck with your product and further development.

I'd like to know if you'd take issue with the following since the concept is the same:

Dithering replaces truncation distortion with superior "analog noise".
Finally, you have the freedom to choose: "digital aliasing" or "analog air".

That seems 100% accurate and fair to me.
(I acknowledge "analog air" could sterilized and described as "noise" but obviously it's designed to be marketing copy, not a DSP thesis)

There is definitely a similarity to dithering since both methods add randomness before a nonlinear/discontinuous operation. I am not denying that analogy entirely. What I take issue with -as I already said above- is the strong "problem/solution" framing (which I believe you justify based on that similarity?)

With your threshold modulation method, aliasing still happens exactly as before and any energy that lands above Nyquist will fold back, I believe we already agree on this point. And yes, with dither, quantization also still happens exactly as before; it just changes the statistical character of the error. BUT... there is a big practical distinction which I believe matters.

In the case of dither; "replaces distortion with noise" is a lot more defensible because there is a known standard where the distortion-like components are reliabily replaced by a controlled noise floor, independently of the signal. The bad artifact is really converted to a controlled noise floor. With your method there is no single and most importantly "modest" amount of noise that "solves" aliasing across signals in the same way dither addresses quantization issues, is there? If the claim is replacement here then a good question would be "by how much under which conditions?" and at "what noise penalty"? In many cases, the most offending alias products are so strong that the amount of broadband noise required to mask them is impractical for most use cases, is it not?

So yes, I admit that there are similarities. But no, I wouldn't say it is a 100% accurate and fair comparison.

[LusiD_Music_UK]

.......

[LusiD_Music_UK]

If the claim is replacement here then a good question would be "by how much under which conditions?" and at "what noise penalty"? In many cases, the most offending alias products are so strong that the amount of broadband noise required to mask them is impractical for most use cases, is it not?

So yes, I admit that there are similarities. But no, I wouldn't say it is a 100% accurate and fair comparison.

So I had a question today along those lines that I feel answers this pretty well so I'm going to quote:

Hey man so I have do a question actually about the clipper plugin

So pretty much how do I set it up so it functions as a regular clipper plugin for really loud mastering but without using oversampling? Relying only on the erosion clipping and noise for anti aliasing purposes so it makes the aliasing inaudible (but of course still showing up on something like plugindoctor but not actually significant or audible in the end result)

Thank you!

Hi,

Here’s the results with a fresh analysis.
This is assuming 1 or 2dB of clipping and no soft clipping. (Soft clipping means you need to drag the buzz threshold down a bit further)
The harder you clip the more erosion is required. The same goes for more high frequency content.

Masking “ALL” aliasing under normal conditions

At 1x oversampling (OFF),
it’s actually impossible to “fully” mask all aliasing (meaning NO audible inharmonic frequencies) no matter how much erosion is used.

At 2x it’s like 8dB of erosion for full masking

At 4x oversampling you need around 4dB

At 8x you need around 2dB

At 16x you need around 0.2dB

So if we’re aiming for “analog” clipping, that’s what you’re looking at!

Hope that is useful!
....................................................

For clarification, when I say you need "x dB of Erosion" I mean the BUZZ threshold's DISTANCE from the hard clipping threshold. The "Gap" between them where the erosion "gradient" starts at the hard clipping threshold and works it's way down.

The sine tone I was using was around 10khz.

I'm going to make some content demonstrating this, but those are the results I observed this morning with regards to how much (ballpark) is required.

But I will say, yes, dithering is more of an exact science than how I'm handling this. It's more of a "use your ears" type deal with BuzzCut but I don't think that invalidates the concept. Also, this is part of the reason I included a test tone that sweeps all the way to nyquest, so you can pull up and FFT and see for yourself.

[martinvankoek]

So I had a question today along those lines that I feel answers this pretty well so I'm going to quote:

Hey man so I have do a question actually about the clipper plugin

So pretty much how do I set it up so it functions as a regular clipper plugin for really loud mastering but without using oversampling? Relying only on the erosion clipping and noise for anti aliasing purposes so it makes the aliasing inaudible (but of course still showing up on something like plugindoctor but not actually significant or audible in the end result)

Thank you!

Hi,

Here’s the results with a fresh analysis.
This is assuming 1 or 2dB of clipping and no soft clipping. (Soft clipping means you need to drag the buzz threshold down a bit further)
The harder you clip the more erosion is required. The same goes for more high frequency content.

Masking “ALL” aliasing under normal conditions

At 1x oversampling (OFF),
it’s actually impossible to “fully” mask all aliasing (meaning NO audible inharmonic frequencies) no matter how much erosion is used.

At 2x it’s like 8dB of erosion for full masking

At 4x oversampling you need around 4dB

At 8x you need around 2dB

At 16x you need around 0.2dB

So if we’re aiming for “analog” clipping, that’s what you’re looking at!

Hope that is useful!

1 - There is no mention of where the Noise Amplitude parameter was set in any of those tests. Why is that? It is a very important variable and greatly contributes to the noise penalty,

2 - That user you've quoted is asking a great question. They specifically want to handle the most offensive bits of aliasing which are the loudest spikes. But applying a "practically acceptable" level of noise will only help to mask the least offending aliasing, right? The loudest tonal spikes will poke right through the noise floor.

3 - Remember Zeisner's first point?

Ask yourself: Why does it feature oversampling if this 'stochastic erosion' would reduce aliasing? And what's the point of adding even more noise?

This is a very relevant question. Oversampling is doing the heavy lifting here. So heavy that it renders whatever Erosion is doing as an "anti-aliasing" tool essentially negligible.

Which brings me back to my main objection:

"Standard clippers have a problem(aliasing). We have a solution(Erosion)"

Let's forget about "completely eliminate/replace" for a second (which is why the Dither analogy failed). Let's just talk about "meaningfully reducing the most offensive aliasing artifacts". Can you in all honesty say Erosion does that? Or at least helps with that? This is exactly what your quoted user is asking about. So the honest answer I can think of here is:

"Unfortunately, no. Erosion isn't practically helpful with those offensive tonal spikes at any reasonably acceptable level of application. It does not help reduce them at all"

Unless you find an SNR of around 20-30dB acceptable.

By the way, I am not claiming this is a bad plugin or it has no uses or whatever. I am sure it can have great creative applications! The only issue I have is with the specific marketing bits, which I find dishonest. (Well at least the ultra-misleading graph comparison was already addressed)

[LusiD_Music_UK]

There is no mention of where the Noise Amplitude parameter

The noise amplitude was set to the default value.

But applying a "practically acceptable" level of noise will only help to mask the least offending aliasing, right?

The answers given were to mask "all aliasing". Every single aliasing harmonic.

Oversampling is doing the heavy lifting here. So heavy that it renders whatever Erosion is doing as an "anti-aliasing" tool essentially negligible.

It's not negligible. The difference is very audible especially at lower oversampling rates and as "a creative effect".

Which brings me back to my main objection:
"Standard clippers have a problem(aliasing). We have a solution(Erosion)"

Replacing aliasing harmonics with noise is a solution.
Is it providing you with options... an alternative vibe.

What sounds cleaner?
A - 8x oversampling with low level aliasing harmonics
B - 8x oversampling with random noise

The answer I believe is B.
You're absolutely free to disagree.

I really do appreciate the pushback and the chance to defend my claims. Have you actually tried the plugin? What do you think about how it SOUNDS? Do you think erosion clipping sounds better than normal clipping? I'd find that a lot more interesting

[LusiD_Music_UK]

Out of interest, how would you market this erosion clipping plugin martinvankoek? Genuinely, the erosion clipper concept is in fact novel (seemingly) and it does in fact mask aliasing and provide a more "analog" profile. Do you not think that is worth a mention? How would you frame it if you were writing the marketing copy? Not a gotcha, genuine question.

To clarify the mechanism, there's no noise added to the signal path. The erosion modulates the clipping threshold itself, so the noise-like character only appears at clipping moments.
Even if erosion clipping does have the downside of increasing the SNR for a few microseconds (with flat noise that the brain is more likely to ignore as irrelevant). I think Problem, Solution still applies as perfectly reasonable marketing.

"Hey bro! You smell, here's some deodorant.
That's not a proper solution to the underlying problem!"

Yes. But it is still "A" solution. And a seemingly new one at that. (I don't see any other clippers that offer this). Even if you'd say it is just a "colorbox/dirty-vibe" tool. Fair enough, I think for getting LOUD. It's a contender for being the best clipper on the market. Which was my goal. ABUSE. DESTRUCTION. With regards to aliasing. I think having "THE ABILITY" to mask it if you so desire, is a real benefit that deserves to be mentioned.

[LusiD_Music_UK]

After some thought, I want to clarify something I assumed went without saying that I think is part of the misunderstanding here.

Like any tool, erosion clipping requires correct application to achieve the intended result.
(assuming the goal is masking of aliasing which is more precise than just "VIBE")

Saying "erosion masks aliasing" implicitly assumes you've set the "buzz threshold" and oversampling appropriately for your signal, just like saying "a compressor controls dynamics" assumes you've set the ratio and threshold correctly. I'm not claiming it works regardless of how you use it. It's very versatile.

The one honest caveat I'll acknowledge: at 1x oversampling, FULL masking is impossible regardless of settings, so if you're running zero oversampling, erosion alone won't mask all of it (but it still is arguably better sounding).

Above 1x OS, with appropriate settings, the (replacement, masking, covering up, hiding) is real and audible.

[LusiD_Music_UK]

Alright guys, I have tweaked the product description in a way that is still marketable and now hopefully avoids this particular semantics debate coming up again. (I'm sure there will be plenty of others lol)

The marketing now reads as follows:

The Problem with Standard Hard Clippers
Standard hard clipper VSTs have a fatal flaw: Math. When you drive your signal into a hard ceiling, traditional plugins introduce digital aliasing, crush your high-end, and turn your mix into a harsh metallic mess.

A New Approach ...... (No longer claiming to be a "Solution")
BuzzCut is the world's first "Erosion Clipper". It modulates the clipping threshold with shaped noise. This masks inharmonic foldback aliasing with superior "analog noise" (much like how dithering masks truncation).

Finally, you have the freedom to choose: "digital aliasing" or "analog air".

[El°HYM]

I wont get into this 'fight' & seeing the developer is really trying hard to satisfy the all 'claims' in this thread. Yet, I dont see 'analog noise' here, it should be digital, or emulated as there is nothing analog in a vst - plugin. I might been thinking way too far 'out of the box' here, who knows. This plugins is definitely an interesting one and should work in the proposed environment.

[billinder33]

The marketing now reads as follows:

I wouldn't worry too much about this pedantic feedback you're getting on your marketing descriptors. Acustica and Pulsar Modular both spew pure BS every plugin release and their cult followings never push back. Not sure why as an independent dev you're under such microscopic scrutiny. For the folks using clippers to drive loudness in bass music, dubstep, D&B, etc., the finer details of aliasing is the least of their concerns.

IMO the things you should be focusing on in your marketing is 1) "can get louder than your average clipper because of X" (as noted by others, I've not tested that specifically) and 2) "has a unique transient response unlike any other clipper on the market". These are your actual product differentiators and benefits. Don't get lost in the weeds, stay laser focused on the real value you are providing to your customers.

Then, since it's not dead obvious how to actually use the plugin, on the website I'd have a very brief, 3 or 4 step animated walkthrough of how to achieve these results right above the download links, so people have a general idea of how to get these results before they demo.

[martinvankoek]

I really do appreciate the pushback and the chance to defend my claims. Have you actually tried the plugin? What do you think about how it SOUNDS? Do you think erosion clipping sounds better than normal clipping? I'd find that a lot more interesting

I am a bit busy this weekend but I promise to give it a proper test and tell you what I think when I have time Will also respond to your other questions

[LusiD_Music_UK]

I wont get into this 'fight' & seeing the developer is really trying hard to satisfy the all 'claims' in this thread. Yet, I dont see 'analog noise' here, it should be digital, or emulated as there is nothing analog in a vst - plugin. I might been thinking way too far 'out of the box' here, who knows. This plugins is definitely an interesting one and should work in the proposed environment.

Let me just check the thread.... and..... More semantics?

Of course the noise is not "literally" analog - it is of course "digital" - because of course it is.
It's a VST plugin.
The idea is that erosion clipping should SOUND more analog/natural than normal clipping.

Thank you for saying it's interesting! I'm happy people are talking about it at least. Very positive reception for the most part

[LusiD_Music_UK]

I wouldn't worry too much about this pedantic feedback

I know right? I've seen so much really crazy marketing from other companies and see so little pushback if any. (Although I am new to KVR threads lol)

I think it's actually a good thing... The extra scrutiny is (in my view) because it's a "new idea" that seemingly nobody else has done in a clipper thus far. Which is why I built it.
"Extraordinary claims require extraordinary evidence" - and apparently scrutiny (fair enough).
I'm happy people find it interesting enough to debate with me about it at least! On a macro level, it's almost like an antibody reaction to a new idea and seeing it it can survive. That's pretty cool.

I appreciate the advice. I'm currently working on a manual and some video content that will explain everything in detail. As the dev, I naively assumed that everything is fairly self explanatory... Apparently I was mistaken haha