Nyquist theorem, fold-over, aliasing and all that (when downsampling to Redbook).

 

QUE beatboxing

 

Perhaps we are talking too much about MP3. Lossy codecs aside, it's true that most people can't hear above 18khz. It's a stat. My brother in law can almost hear dog whistlers (overstatement) and I probably can't hear above 16Khz. 24bit vs 16 makes a bigger difference, specially with tracks with big dynamic ranges (or if you go wild with your speakers which makes a kind of dynamic range increase but with your ears bleeding)

I may be wrong, but wasn't VP9 Google's bet on the browser video codecs war? And they "adopted" the good oldie Ogg Vorbis as its audio codec.
I mean, every codec that Google/Youtube backs up ends almost fully supported for obvious reasons, including mp4-aac which before winning the browser codec war was a mess: quicktime version, videoconsoles version, and so on...

RIP Theora

 

Last edited: May 5, 2017

It is amazing how many BULLSH*T are written every time there is a subject like this one ;)
It was the first reason i left KVR. And may be i'm going to do the same here...

Simple answer : you CAN cut freqs...or not. And if you don't, someone else is going to do it FOR YOU, LIKE A BUTCHER !

CD players WILL cut above 20kHz, to avoid aliasing. Everything above 20kHz to 22.05kHz is considered as "aliasing" by Redbook specs.
And most CD players filters sucks...
side note : some even cut UNDER 20Hz !

Audio compression (Youtube, mp3...) can cut highs and/or lows, according to some settings.
Example : 320kbps MP3 (best quality) cut at 20kHz anyway...like Redbook CD

When you export to CD/mp3, you use 16bits 44.1kHz. So anyway, your signal WILL be cut by export, at least, at 22.05kHz.

So, NO, there is nothing mandatory. But most ppl will not hear their signal IS ALREADY cut when exporting to 44.1kHz. And start talking bullsh*t about freqs above 20kHz, whereas they don't even hear above 8kHz, because they already shot their ears by listening to loud music

Sorry, i had to say it...

PS : if you can avoid transforming this website in KVR...do it

 

there is a factual difference between people hearing limitations in the upper 20khz register, and the effect on the rest of the signal when cutting the upper frequencies in audio.

there is a difference between a human beings hearing range , and the effect had on the entire audio signal , by chopping out highs ( double in your case saying it already is, why would you do it gain if nothing is there? just to destroy the transients below cutoff?)


Just want to say regarding this entire topic of cutting out upper highs of signal,
what goal is the outcome of your mix? to have average mediocre masted songs to listen to?
if your goal is to compete with best mixes in music best in genre, to have your songs sound like the very best pros, then the way you go about dealing with your mix matters.

load up a drum loop , bring up spectrum analyzer on master fader like fabfilter pro q 2,
next cut everything above 18khz, and even do it again twice if you want, now bounce to your fav shit codec like mpthree-deez nuts, is there any difference of the original loop? tiny bit. ;D
the thread question is "does professionals..?)
answer no, they do not, but you can do anything you want, even the opposite of what they do if that result gives you something you love the sound of

 

Last edited: May 5, 2017

This may be a dumb question. Anyway, when you mention this effect, are you talking about human hearing perception or "technical" side effects of filtering?

 

I am guilty of doing the exact same thing (matching mp3 cuts). I was ripping the audio from 1080 youtube vids for references. Now I got a folder full of .flac files for referencing.

My thread, lol
https://audiosex.pro/threads/what-the-heck-is-this-weird-dip-in-eq-curves.29879/

 

@Baxter and @Pipotron3000 already said it pretty clear. Still, to generate more awareness:

Nyquist means that a signal must consist of lower frequencies than Nyquist to be able to reconstruct the signal. In a digital domain there's only bits, which is a fixed size (there's no such thing as half a bit, etc.). And therefore when reading in a signal it must be done in a rate that is at least double as high as the highest frequency in the original material. So, Nyquist means 22.05 kHz in a 44.1 kHZ domain, 48kHz in a 96 kHz domain, 11 kHz in a 22 kHz domain, etc.

If you only produce for yourself and only have players with a rate of 96 kHz, you don't need to cut at 22 kHz. But if you share your music, chances are it will be listened to on devices with way lower rates. Those devices will take care of converting and cutting to match their technical specs, but those automatisms are no audio engineers. That's why you should treat your songs in such a way that those automatic routines get nothing to work on (which could ruin your mix).

Also, everything above Nyquist will (some more, some less) fold back into the audible range, which is called Aliasing. Once it occurs you can't filter it out. Aliasing occurs way before mastering! Even some plugins already introduce Aliasing if they don't properly up/downsample. Take care of aliasing while mixing, not while mastering. Mastering is to prevent future aliasing, it can't eliminate already existing aliasing.

 

yes sir. there is a difference between those two. both have to be accounted for (thought about independently) not a dumb question at all.


"
tulamide
@Baxter and @Pipotron3000 already said it pretty clear. Still, to generate more awareness:

Nyquist means that a signal must consist of lower frequencies than Nyquist to be able to reconstruct the signal. In a digital domain there's only bits, which is a fixed size (there's no such thing as half a bit, etc.). And therefore when reading in a signal it must be done in a rate that is at least double as high as the highest frequency in the original material. So, Nyquist means 22.05 kHz in a 44.1 kHZ domain, 48kHz in a 96 kHz domain, 11 kHz in a 22 kHz domain, etc.

If you only produce for yourself and only have players with a rate of 96 kHz, you don't need to cut at 22 kHz. But if you share your music, chances are it will be listened to on devices with way lower rates. Those devices will take care of converting and cutting to match their technical specs, but those automatisms are no audio engineers. That's why you should treat your songs in such a way that those automatic routines get nothing to work on (which could ruin your mix).

Also, everything above Nyquist will (some more, some less) fold back into the audible range, which is called Aliasing. Once it occurs you can't filter it out. Aliasing occurs way before mastering! Even some plugins already introduce Aliasing if they don't properly up/downsample. Take care of aliasing while mixing, not while mastering. Mastering is to prevent future aliasing, it can't eliminate already existing aliasing.

"
this.

 

stuff under 20 khz sounds shite if it post 70s, cut it all!

 

This discussion made me think about the advantages of analogue and in this day and age analogue hardware is made so much better specs wise than 20 or more years ago. Just imagine: no nyquist frq problems - no aliasing, in analogue there is no limit to high frequencies, no worries about dithering, just smooth sound. On the other hand you do get noise, and making analogue copies does introduce degradation and it's much harder to do everything that digital editing does. Hmmm. It makes me think, have we lost our way? I mean going *all digital*. We should take advantage of the advantages of both somehow for best quality audio.

p.s. on the other hand, nobody cares. Just rare people who still have ears, or what's left of them, but still appreciating quality sound.

 

Last edited: May 7, 2017

lol thats what i was implying

 

personally I cut out 16khz and above, because if i slide that lpf up and down from 16k-22k or whatever the higest is allowable, I notice that the sound is hardly changed ata ll and is usually a tiny bit better IMO at 16k hz. i do the same thing at 30hz and below with a hipass filter.. ive never heard this theory of it affecting harmonics in the audible range.. but i have 5 different sets of speakers to test my sound on and some can actually play down in the 20hz range(with a powered subwoofer box) but none can do anything about above 16khz.. anybody got a link to more concrete information on the harmonics being negatively affected by cutting out inaudibles?

 

As far as I know only a sharp brickwall LP filter can cause ringing at the slope and aliasing in the lower frequencies. I don't think there's anything wrong with 6 or 12dB LP filters or HP, for that matter, although that video about phase shifting is interesting. I think correcting DC offset is more important.

Somebody correct me if I'm wrong.

 

It's simple and I'm gonna resolve your question with my usual flawless logic:
1. We need to record the original sound in analog media
2. Then we sample it 192Khz/64bit to digital
3. We put the analog copy in a safe
4. Two years laters we open the safe
5. Then we sample it at double the frequency and bit depth: 384Khz/128bit
6. Go back to step 4

Hell yeah, I'm good...

PS: It's already happening with classic movies like say Twelve Angry Men
Recorded in analog gear for theater purposes.
First digital release: 480p (DVD)
Second one: 720p (HD Ready)
Third iteration: 1080p (Full HD)
Current iteration: 2k
Next: 4k