YT vid 'explains' why analog emulations don't sound as good as they should? Any truth in this?

[Dr Huge Longjohns]

This chap reckons that plugins emulating analog gear fail for all sorts of aliasing/oversampling/nyquist reasons I don't understand. Is any of this true, please, digital boffins out there?

https://www.youtube.com/watch?v=4taroKS_N6Q

[Martin Walker]

Hi Dr. Huge,

Yes, it's all true, although as Monsieur Snake Oil admits, he does exaggerate a bit for effect

Personally I think it's slightly unfair to use Decapitator as his example plug-in to demonstrate aliasing issues, since it's by definition designed to produce evil distortion and lots of upper harmonics

Most other plug-ins will have far less high frequency energy being generated, so any aliasing should be at correspondingly lower levels - after all, how many signals are you likely to be using with perhaps 7kHz or higher at near clipping levels?

Moreover, many other plug-ins where clarity is of tantamount performance will use internal oversampling where required (or even provide such options as user-defined settings), so that aliasing doesn't happen, or at the least is minimised to be at inaudible levels. It seems to be generally accepted that any lines that appear in the audio spectrum at less than -125dBFS can be consider inaudible.

Martin

[Dr Huge Longjohns]

since it's by definition designed to produce evil distortion

Not necessarily, to be fair. It models desk channels so you can choose to have a tiny amount of it. But Herr Snakeoil basically says it's always going to impart nastiness you shouldn't be getting because of its non-aliasing etc design? Does this mean one should always check the specs of an allegedly analog-modelled plugin very careful?

[Martin Walker]

since it's by definition designed to produce evil distortion

Not necessarily, to be fair. It models desk channels so you can choose to have a tiny amount of it. But Herr Snakeoil basically says it's always going to impart nastiness you shouldn't be getting because of its non-aliasing etc design? Does this mean one should always check the specs of an allegedly analog-modelled plugin very careful?

To some extent yes, but personally I think you have to judge with your ears in context.

Here for example I've just run up the wonderful Decapitator with a more subtle preset that emulates 'Neve Fizz'

With an 10kHz sinewave input at a level of -5dBFS, yes, you can see aliasing apart from the wanted 10kHz and 20kHz lines, but as you can see from the THD (Total Harmonic Distortion) figure of around -60dB, the distortion products are already at low 0.1% levels in total, while the highest aliased spike is at about -87dB (that's equivalent to 0.0044%) - not very significant in the overall context to my mind.

As always your likeage may vary from plug-in to plug-in, and it's well worth doing a few tests at higher frequencies to check for aliasing problems, but in practice I wouldn't rule out using Decapitator based on these figures

Martin

[Dr Huge Longjohns]

BTW, I forked out for the Metric AB plug on your recommendation. Excellent as you said. Can't find a way to activate the help balloons at the moment, though. The manual says click on the settings icon to get a dropdown, but doesn't work for me!

[Martin Walker]

BTW, I forked out for the Metric AB plug on your recommendation. Excellent as you said. Can't find a way to activate the help balloons at the moment, though. The manual says click on the settings icon to get a dropdown, but doesn't work for me!

Glad you like it!

Help mode does work for me, although for some reason when I first tried it nothing seemed to happen. Click on Settings and select the drop-down option Enable tooltips then hover over a function - if you move the mouse about then the help balloon won't appear, but as soon as you maintain your position over a particular function it will.

Good luck!

Martin

[Dr Huge Longjohns]

I don't get the dropdown menu, that's the problem. Are you on 1.1?

[Dr Huge Longjohns]

Dropped a note to PA support and they can't get it to work their end either...think it might be a bug!

[Martin Walker]

Aha - couldn't find a version number on mine anywhere, and nor could I find it on the PA website. Finally found my original ZIP download, which has the filename adptr_metricab_win_x64_1_0.zip

Now updating to the latest version 1.1 to get all the improvements, so I'll probably lose the help function as well

EDIT: Yep, no help function any more - it's been tromped upon by the new enhanced Settings page

2nd EDIT: No, I've just found it - in the new enhanced Settings page under General Settings there's a tick box labeled Tooltips. Click on this and your help functions all return

Martin

[Dr Huge Longjohns]

DOH!

[CS70]

It's very hard to gauge someone's argument when it's presented like that.

Confirmation bias is very strong. If they guy could show that he consistently distinguishes digital vs. analog distortions in a blind setting, it would be worth listening to his explanation.

As it is, it is not really worth the time.

[Tomás Mulcahy]

...all sorts of aliasing/oversampling/nyquist reasons I don't understand.

That's because he doesn't explain it very well. There's a really great, long running thread on Gearslutz about this. Several plugin developers chip in, and are great at explaining this stuff! And they're honest about their own products that are getting tested by users in the thread.

As it is, it is not really worth the time.

Agreed. The GS thread is far more useful, even though it'll take up more time LOL:
https://www.gearslutz.com/board/music-c ... hread.html

[Dr Huge Longjohns]

116 pages of Gearslutz! It better be good, Tomas!

[Dr Huge Longjohns]

So got as far as page 20 in this very interesting GS thread. And whilst the conclusion was broadly aliasing:broadly bad but often inaudible, this pic stopped me in my tracks. This was the frequency analysis of, get this, a single 910hz sine wave at 44.1k subjected to one Waves modelled EQ and 2 modelled compressors. Eek is the word!

[Martin Walker]

Well done for getting to page 20 thus far Dr. Huge!

I'm pretty sure I've eventually waded through all the pages, and can reveal that the best choice of sinewave signal when sampling at 44.1kHz to most easily reveal any aliasing is 993Hz.

I believe it was Andy of Cytomic in that thread who eventually came up with that figure, and it works well, since any aliased lines appear neatly between the other desired lines of the harmonic series

"Ok, I've worked these to be as close to 1000 hz as possible, be integer values frequencies, and spaced the aliased harmonics at close to 400 hz up and down from the main harmonic peaks for easy detection:

44100 -> 993 hz for easiest detection of aliased harmonics
48000 -> 992 hz
88200 -> 998 hz
96000 -> 996 hz
176400 -> 989 hz
192000 -> 998 hz"

Martin

[Martin Walker]

...and here's a very interesting discussion about aliasing by the very expert (Chris Johnson of Airwindows fame) that Monsieur Snake Oil consulted before making his video. It makes an awful lot of sense to me:

http://www.airwindows.com/unbox/

Here's a very telling paragraph that reinforces my initial reply in this thread:

"many of my plugins have gentle enough distortion curves that they throw a limited number of harmonics. If you are only generating harmonics within the range of digital audio’s frequencies, you’re fine and there will be no problem until you feed the system a frequency that’s too high. You’re not automatically feeding superhigh frequencies all the time if you’re working with natural recordings: not all sounds contain that kind of high frequency content."

...before he goes on to explain the philosophy used in his Unbox plug-in that only distorts those frequencies that WON'T generate aliasing

Martin

[Dr Huge Longjohns]

Yes, very interesting. Makes you wonder why all the top end plugin developers don't do something similar as a matter of course?

[Dr Huge Longjohns]

I've started putting some of my own plugs through the GS test! Very interesting. Logic's own plugs are on the whole remarkably clean and alias free at all frequencies, certainly the EQs and Compressors, and whilst their basic limiter is great their Adaptive Limiter is shocking!

[Folderol]

Maybe I'm missing something obvious, but surely anything that is going to generate distortion should also include post-effect low-pass filtering.

[Hugh Robjohns]

yes...

[Martin Walker]

And there was I expecting Hugh to write a highly readable essay on anti-aliasing, and he managed to chip in the perfect response with one word

Bravo Hugh!

Martin

[Hugh Robjohns]

Sorry... been away... too tired to write essays this morning...

But basically this bloke is saying that using a plugin designed to sound nasty makes it sound nasty... I'm not really going to get very exited about that.

Any well-designed plugin will ensure it complies with the Nyquist requirements. Most decent plugin these days employ oversampling and proper anti-alias filtering of the output specifically to ensure that. Some less well-engineered plugins might not -- in which case they are, technically, BROKEN!

Sometimes broken things make noises that some people like -- and I so know some people that really like the sound of aliasing.

But aliasing isn't rocket science. Both the reasons for it occurring, and the solutions to prevent it, are very well known.

If you have plugins that sound nasty, and you don't want nasty, don't use them. Use something that works properly.

Or, process your audio through them at a much higher sample rate and then use a decent sample rate conversion tool to apply the required anti-alias filtering back to the base-band sample rate of the rest of your project.

H

[Dr Huge Longjohns]

Most decent plugin these days employ oversampling and proper anti-alias filtering of the output specifically to ensure that.

How can you tell? Can you give us some examples of what you consider 'decent' plugins? I'd have considered Soundtoys 'decent'?

[Elephone]

I don't understand why no one is actually seeing whether the emulations do the job.

Has absolutely nobody got the hardware and (at least a demo) VST of the emulation (or a preset on a non-specific analogue model that is suppose to emulate a real unit) to compare the two? Then, simply put the exact same digitally-recorded audio file through both and do a spectral analysis and provide audio examples.

I understand that different analogue hardware units (of the same brand and model) will produce different results, but some kind of consensus as to what to expect should be reached. If not, we need to compare different hardware units and then compare an average to the VST

Then, we wouldn't need to rely on our ears as much for an opinion. I say 'ears', but we can't separate the brain from the hearing system because we are prone to all kinds of biases that make us actually hear stuff differently.

In actual use, of course use your ears, but for these comparisons, we should really be using the scientific method to negate opinion.

[muzines]

I don't understand why no one is actually seeing whether the emulations do the job.

Has absolutely nobody got the hardware and (at least a demo) VST of the emulation (or a preset on a non-specific analogue model that is suppose to emulate a real unit) to compare the two? Then, simply put the exact same digitally-recorded audio file through both and do a spectral analysis and provide audio examples.

Well, I don't know where you have been, but if you frequent the common audio forums such as Gearslutz, KVR, and others (inc YouTube), you'll find people are *always* doing these comparisons and tests.

They mostly become meaningless, for a variety of reasons. People just use the results to convince themselves of the outcome they wanted. If the outcome shows otherwise, then they just poke holes in the tests. It's a no-win situation - and there's very little objectivity - regardless of what the facts and figures demonstrate.

I can go into more details, but there are lots of variables in these things. Let's face it, with analog synths for example, very few sound identical from one unit to another - so what's the "reference" point anyway? Put up a recording of a Minimoog, and ten people go "that doesn't sound like mine". If you can't even get people to agree on a reference point, then comparisons are never going to draw general concensus. It's like if you ask 20 different people what the sound of "tape" is, you'll get twenty different answers, tastes, biases and preferences.

In actual use, of course use your ears, but for these comparisons, we should really be using the scientific method to negate opinion.

Welcome to the very flawed and biased world of human opinion...

For some people, the very mention of using *science* turns them off straight away....