Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Cutoff frequencies of lossy codecs (Read 2472 times) previous topic - next topic
0 Members and 3 Guests are viewing this topic.

Cutoff frequencies of lossy codecs

Hello. Firstly, I want to say that I was wanted to open this topic in the "Lossy audio compression" category but I could not do that because opening a topic in there without selecting a sub category is not possible and this topic is not about a specific audio codec.

I started to see that most lossy encoders are starting to give higher and higher cutoff frequencies for a given bitrate (for example, FhG vs Lame). This may sound attractive, but a higher cutoff frequency means more black space in the frequency area and this sounds horrible to me. (An ABX is not needed per TOS because this is not about quality, this is about "sound color".) For example, a FhG MP3Enc lowest quality encoded 16kbps 11025Hz mono MP3 is pretty good for me, but a FhG Fastencc and especially Lame encoded 16kbps 11025Hz mono MP3 is not very good because of the higher cutoffs. The same thing applies for (for example) FhG MP3Enc encoded 64kbps 44100Hz mono and Lame encoded 64kbps 44100Hz mono. (I can't give an ABX because all of these files are totally untransparent to me. Of course, this does not mean they are all bad.) Of course, very low cutoff points are not good too, but I think the modern encoders are giving too higher cutoff points than the sweet spot. For example, I think 24kbps is the lowest bitrate that makes sense with 11025Hz cutoffless MP3's. Why those modern encoders are giving higher and higher cutoff points every day?



Re: Cutoff frequencies of lossy codecs

Reply #3
Quote
This may sound attractive, but a higher cutoff frequency means more black space in the frequency area and this sounds horrible to me. (An ABX is not needed per TOS because this is not about quality, this is about "sound color".) For example, a FhG MP3Enc lowest quality encoded 16kbps 11025Hz mono MP3 is pretty good for me,
I don't mean to sound rude, but you have a strange idea of what constitutes 'sound color' and an even more strange sense of what constitutes 'good sound'! It must be clear by now that you are in a minority of one. Ultra low bitrate/samplerate does NOT generate anything that sounds good by any metric, certainly not in mp3.

Re: Cutoff frequencies of lossy codecs

Reply #4
I don't know much about low bitrate encoding.    I suspect the LAME developers don't care much about it.    From what I can tell, they are more interested in getting transparent, or near-transparent compression, possibly at the lowest bitrate that's possible.   

Quote
Why those modern encoders are giving higher and higher cutoff points every day?
Maybe because a lot of people judge quality by looking at the spectrum rather by listening?  Of course that's not what we do here.

Quote
I started to see that most lossy encoders are starting to give higher and higher cutoff frequencies for a given bitrate (for example, FhG vs Lame). This may sound attractive, but a higher cutoff frequency means more black space in the frequency area and this sounds
With lossy compression some data must be thrown-away.   There are trade-offs.   If you force it to keep the highest frequencies something else will be thrown away, and what gets thrown-away may be more important than the high frequencies.    Once you are hearing compression artifacts different people may prefer different tradeoffs.    At higher quality settings, I assume the default LAME settings are optimized for most listeners and most program material    OGG or AAC may be able to do better, but I'm pretty sure LAME has been pushed as far as it can go.

Also at higher quality settings, it you hear an artifact it's usually not the loss of high frequencies you hear.   It's usually something else.    The highest audio frequencies (near 20kHz) are weak in normal program material so even if you can hear to 20kHz with loud test-tones in a hearing test, they are usually masked (drowned out) by not-as-high frequencies.     Eliminating "drowned out" sounds is the main way lossy compression gets-away with throwing-away data. 

You are limited to half the sample rate (Nyquist) so at a sample rate of 11,025 the audio can't go above 5512Hz (even with lossless compression or no compression.)

Re: Cutoff frequencies of lossy codecs

Reply #5
Higher cutoff point also looks worse on the spectrum in most cases for me.

Re: Cutoff frequencies of lossy codecs

Reply #6
Higher cutoff point also looks worse on the spectrum in most cases for me.

I'm still confused by what you're describing, but you should listen with your ears, not eyes.

Re: Cutoff frequencies of lossy codecs

Reply #7
@saratoga Looks like you didn't see the word "also". This means I already normally listen with my ears.

Re: Cutoff frequencies of lossy codecs

Reply #8
@DVDdoug I think OGG is doing worse. I prefer 22050Hz 32kbps FhG encoded MP3 over 22050Hz 32kbps Aotuv encoded OGG Vorbis, because it has richer spectrum as it has a narrower frequency range.

Re: Cutoff frequencies of lossy codecs

Reply #9
People generally says humans can hear up to 20kHz but I think nobody can hear above 18kHz and no adult can hear above 16kHz.

Re: Cutoff frequencies of lossy codecs

Reply #10
People generally says humans can hear up to 20kHz but I think nobody can hear above 18kHz and no adult can hear above 16kHz.
IIRC, some people have heard both 24 and 27 in trials, but at sound pressure levels you don't want to be exposed to for more than split seconds. That research was on how far up one should measure noise to protect against it, not on what you need to preserve.

Re: Cutoff frequencies of lossy codecs

Reply #11
@Klymins If you seek for quality, go lossless: 44.1KHz/32bfp if possible. Everything else is (almost) useless.
Hybrid Multimedia Production Suite will be a platform-indipendent open source suite for advanced audio/video contents production.
Official git: https://forart.it/HyMPS/


Re: Cutoff frequencies of lossy codecs

Reply #13
klymins is talking about samplerates of 11025 (layer III MPEG 2.5), 16000(?) and 22050, maybe 24000 (layer III MPEG 2)
So, no need to talk about human hearing limits or about methods to losslessly preserve the quality.

We also have to consider the technology limitation ( SFB 21, or SFB 12 on MPEG 2/2.5).

First hand tests with current codecs do not offer a "no-lowpass" option for these frequency ranges, so encoding a 11Khz track gets an almost 5Khz bandwith, rather than a  5.5Khz bandwith, even with the helix encoder.

Academically, it could be tested if not having a filter could be felt as better quality at the expense of slighly more artifacts.
Of course, in that case, the interest here in hydrogenaudio would be exclusively about sound quality, not "graph quality".

Modifying either LAME or Helix would be feasible for this test, but this is not a proposal, just a possibility.
SFB might not be as much a limiting factor here as opposed to 44Khz, since the frequency band is expected to have quite some power.

Re: Cutoff frequencies of lossy codecs

Reply #14
No need for ABX i believe you they sound a certain way, but you could have at least included some examples so we can contribute anything at all and to figure out what "sound color" you are talking about. If we encode this ourselves there is a high risk of confusion.
And so, with digital, computer was put into place, and all the IT that came with it.

Re: Cutoff frequencies of lossy codecs

Reply #15
a higher cutoff frequency means more black space in the frequency area

What?

He means the almost always guaranteed huge emptiness in a spectrogram due to absolutely nothing anyone would listen to having high frequency data.
But it's kind of the reverse, the higher the cutoff, the less pure black. But that depends on your spectrogram's parameters. If you turn the sensitivity down all the blue/purple there would be black too.

And his frequency observations are certainly correct. For almost everyone 16k is a hard limit. Good luck even finding a little child who can hear 20k at _safe_ levels.

Unless we're getting into the realm of possible non-audio high frequency interactions, the typical cutoffs are indeed quite high.

I see he's talking ultra low bitrate which is not of interest to me personally, but I had similar thoughts in higher ranges. Why waste bits on imperceptible frequencies?
So "color" is probably a perceived benefit of more bits representing the target range.

Re: Cutoff frequencies of lossy codecs

Reply #16
a higher cutoff frequency means more black space in the frequency area

What?

He means the almost always guaranteed huge emptiness in a spectrogram due to absolutely nothing anyone would listen to having high frequency data.
But it's kind of the reverse, the higher the cutoff, the less pure black. But that depends on your spectrogram's parameters. If you turn the sensitivity down all the blue/purple there would be black too.

It is the reverse of what he is saying.  Raising the cutoff means more frequencies get through, not less. 

Re: Cutoff frequencies of lossy codecs

Reply #17
a higher cutoff frequency means more black space in the frequency area

What?

He means the almost always guaranteed huge emptiness in a spectrogram due to absolutely nothing anyone would listen to having high frequency data.
But it's kind of the reverse, the higher the cutoff, the less pure black. But that depends on your spectrogram's parameters. If you turn the sensitivity down all the blue/purple there would be black too.

It is the reverse of what he is saying.  Raising the cutoff means more frequencies get through, not less. 


A higher cutoff frequency means more black space below the cutoff frequency, and black spaces those below the cutoff frequency are much easier to detect than black spaces those above the cutoff frequency.

Re: Cutoff frequencies of lossy codecs

Reply #18
A higher cutoff frequency means more black space below the cutoff frequency, and black spaces those below the cutoff frequency are much easier to detect than black spaces those above the cutoff frequency.
Do you mean "black spaces" the lowest amplitude range for the spectrogram (which its appearance depends on color scheme, the lowest amplitude range would be black, white, or any other color)?

Obviously, unless the difference between audio files either encoded using any lossy codecs or downconverted to lower samplerate/bitdepth and original WAV/FLAC files is so significant to the point it is even audible and doesn't require ABX to hear these differences, we shouldn't post spectrogram images (from any sources like foobar2000, Audacity, Spek, and even my own spectrum analyzer and spectrogram project) per ToS #8

Re: Cutoff frequencies of lossy codecs

Reply #19
@DVDdoug I think OGG is doing worse. I prefer 22050Hz 32kbps FhG encoded MP3 over 22050Hz 32kbps Aotuv encoded OGG Vorbis, because it has richer spectrum as it has a narrower frequency range.

Narrower frequency range gives richer spectrum? Errrrr. what?
Sorry, dude, but your takes on pretty much anything don't make any sense, and i've seen you spreading such kind of misinformation across multiple threads. What is your point?
Random digital audio nerd girl

Re: Cutoff frequencies of lossy codecs

Reply #20
@DVDdoug I think OGG is doing worse. I prefer 22050Hz 32kbps FhG encoded MP3 over 22050Hz 32kbps Aotuv encoded OGG Vorbis, because it has richer spectrum as it has a narrower frequency range.

Narrower frequency range gives richer spectrum? Errrrr. what?
Sorry, dude, but your takes on pretty much anything don't make any sense, and i've seen you spreading such kind of misinformation across multiple threads. What is your point?
I can only agree. All the similar comments suggest a pronounced hearing defect, not wishing to sound(!!) unkind.

Re: Cutoff frequencies of lossy codecs

Reply #21
@DVDdoug I think OGG is doing worse. I prefer 22050Hz 32kbps FhG encoded MP3 over 22050Hz 32kbps Aotuv encoded OGG Vorbis, because it has richer spectrum as it has a narrower frequency range.

Narrower frequency range gives richer spectrum? Errrrr. what?
Sorry, dude, but your takes on pretty much anything don't make any sense, and i've seen you spreading such kind of misinformation across multiple threads. What is your point?

@DVDdoug I think OGG is doing worse. I prefer 22050Hz 32kbps FhG encoded MP3 over 22050Hz 32kbps Aotuv encoded OGG Vorbis, because it has richer spectrum as it has a narrower frequency range.

Narrower frequency range gives richer spectrum? Errrrr. what?
Sorry, dude, but your takes on pretty much anything don't make any sense, and i've seen you spreading such kind of misinformation across multiple threads. What is your point?
I can only agree. All the similar comments suggest a pronounced hearing defect, not wishing to sound(!!) unkind.

It's your opinion, and that's my opinion. We must respect to different opinions. Please be more polite.

Re: Cutoff frequencies of lossy codecs

Reply #22
It's your opinion, and that's my opinion. We must respect to different opinions. Please be more polite.
Everyone is entitled to their own opinion but there is no requirement that it should be respected. You continually profer statements and opinions that are completely contrary to the perceived wisdom on this and most other forums without any evidence that makes any kind of sense to the majority. Of course you are entitled to your opinion but please don't continue to try to convince others that your opinion is more valid than that of the majority.

Re: Cutoff frequencies of lossy codecs

Reply #23
It's your opinion, and that's my opinion. We must respect to different opinions. Please be more polite.
Everyone is entitled to their own opinion but there is no requirement that it should be respected. You continually profer statements and opinions that are completely contrary to the perceived wisdom on this and most other forums without any evidence that makes any kind of sense to the majority. Of course you are entitled to your opinion but please don't continue to try to convince others that your opinion is more valid than that of the majority.

I'm not trying to convince others, thank you.

Re: Cutoff frequencies of lossy codecs

Reply #24
I'm not trying to convince others, thank you.
Well, you constantly try to promote settings as "good" that make my toe-nails curl up. My humble opinion