Skip to main content

Topic: Trying to learn to understand spectral analysis of mp3s (Read 6397 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • bennetng
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #25
M4A file, Bitrate 61 kbps. Cut-off at 20.5 kHz.
[EDIT]Didn't notice I am in the mp3 forum, so it may be off topic? mp3PRO can make the spectrum looks full in low bitrate as well although it requires a compatible decoder.
But M4A/AAC is suppose to preserve more frequencies than MP3
What if I transcode this 61kbps aac to a 320k mp3? Looks like someone (I am not talking about you) will really believe that it is a direct 320k encoding from a lossless source without even listening.

I don't understand why some people (I am not talking about you again) in 2016 still believe that more HF in spectrum = high quality or high bitrate. Lossless codecs can generally reduce a 1411kbps audio file to somewhere about 500-1000kbps, then how can a lossy codec really preserve something up to 20khz at 61kbps? Of course a lot of audio information are generated or synthesized by clever psychoacoustics and DSP experts. Nowadays a lot of people (not you again) pretend they are real experts in these area by merely looking at a spectrogram, how hilarious.

  • pdq
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #26
Hopefully someone will correct me if I'm wrong, aren't the specifications for mp3 and mp4 defined for the decoder?  If this is the case, then the encoder can do whatever the developer(s) decide, provided they create a compliant stream.
What is specified is the format of the data and how all decoders are required to decode it.

The only requirement of an encoder is that it produce data that is compatible with said decoders. There are many examples of encoders that technically meet this requirement, but do so very poorly.

@deathcoreRULES: What you are observing is the default behavior of encoders relating low pass frequency to bitrate. Most encoders follow a behavior roughly what you are describing, but this is not a requirement.
  • Last Edit: 13 May, 2016, 07:17:14 AM by pdq

  • saratoga
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #27
Spectrogram is not a popular program.

AAC is not supposed to have a higher low pass.

That table is not accurate; it says your 160k file is 128k.

People say things like that because they are ignorant. You have acess to Google, you don't need to listen to advice from uninformed people on Reddit.

  • greynol
  • [*][*][*][*][*]
  • Global Moderator
Re: Trying to learn to understand spectral analysis of mp3s
Reply #28
Even if it were a popular program, that would not mean it was built on a solid foundation.

https://en.m.wikipedia.org/wiki/Argumentum_ad_populum
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

Re: Trying to learn to understand spectral analysis of mp3s
Reply #29
M4A file, Bitrate 61 kbps. Cut-off at 20.5 kHz.
[EDIT]Didn't notice I am in the mp3 forum, so it may be off topic? mp3PRO can make the spectrum looks full in low bitrate as well although it requires a compatible decoder.
But M4A/AAC is suppose to preserve more frequencies than MP3
What if I transcode this 61kbps aac to a 320k mp3? Looks like someone (I am not talking about you) will really believe that it is a direct 320k encoding from a lossless source without even listening.

I don't understand why some people (I am not talking about you again) in 2016 still believe that more HF in spectrum = high quality or high bitrate. Lossless codecs can generally reduce a 1411kbps audio file to somewhere about 500-1000kbps, then how can a lossy codec really preserve something up to 20khz at 61kbps? Of course a lot of audio information are generated or synthesized by clever psychoacoustics and DSP experts. Nowadays a lot of people (not you again) pretend they are real experts in these area by merely looking at a spectrogram, how hilarious.
I love and obsess over music, but when it comes to the technical stuff I don't proclaim to be an expert.  I thought I was learning, but  it's now obvious there is much misinformation out there when it comes to comparing bit rates, reading spectrograms, and ripping audio files.

Now I don't know what to believe...  Do high bit rates even matter that much?  Does the format really matter?  Have the encoders improved greatly since the early days, so even low bit rate rips sound good?  Maybe my hearing is going...  No I think it's still good!  Nowadays I always rip my CD's in FLAC because storage is not a huge issue and it allows me to convert to another format later on  Really though, I can admit that with a good portion of my music, I cannot tell the difference between a lot of my FLAC, 320 kbps MP3, and even some lower bit rates files in my library.  My rip of Pink Floyd's Dark Side of the Moon is over 10 years old and it sounds fantastic.  It's a 128 kbps AAC rip that was done with iTunes on default settings and the album has great dynamics.  I still have quite a bit of 128 kbps MP3 files and a lot of it sounds excellent.  I also haven't updated my versions of LAME and FLAC in years and who even knows what version of AAC that was...

Spectrogram is not a popular program.

AAC is not supposed to have a higher low pass.

That table is not accurate; it says your 160k file is 128k.

People say things like that because they are ignorant. You have acess to Google, you don't need to listen to advice from uninformed people on Reddit.
The misinformation is ALL over the web and a lot of my findings came from Google, not just reddit.  I mentioned reddit in my posts due to it's popularity and the many subreddits dedicated to audio discussion.  That table I found was in the headphones subreddit and the user talking about spectrograms was posted in /r/audiophile...

  • saratoga
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #30
Do high bit rates even matter that much?  Does the format really matter?  Have the encoders improved greatly since the early days, so even low bit rate rips sound good?

http://wiki.hydrogenaud.io/index.php?title=Hydrogenaudio_Listening_Tests

The misinformation is ALL over the web and a lot of my findings came from Google, not just reddit.  I mentioned reddit in my posts due to it's popularity and the many subreddits dedicated to audio discussion.  That table I found was in the headphones subreddit and the user talking about spectrograms was posted in /r/audiophile...

I don't think you should believe reddit posts just because the person making them posts a lot.  That is probably going to not work out very well for you.

Here is a slightly better source:

https://en.wikipedia.org/wiki/MP3#Encoding_and_decoding

Couple sentences in and you can realize why looking at the lowpass isn't going to work (each encoder can choose a different one, so you'll need one of those tables for every single encoder and/or combination of encoder settings). 

  • greynol
  • [*][*][*][*][*]
  • Global Moderator
Re: Trying to learn to understand spectral analysis of mp3s
Reply #31
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

Re: Trying to learn to understand spectral analysis of mp3s
Reply #32
Thanks for links.  By the greynol, I like your new signature - "Your eyes cannot hear".

It's amazing how much misinformation is out there when it comes to bit rates and spectrogram software.  I am sure Spek has its uses, but now I know it can't distinguish between bit rates and transcodes.  Unfortunately, the misinformation will continue to be spread around by the self-proclaimed audiophile experts.

I want to do these listening tests, but I see the tests have ended sometime ago.  If I take them now, am I able to upload the results to the forum so you guys can see how I did?

  • greynol
  • [*][*][*][*][*]
  • Global Moderator
Re: Trying to learn to understand spectral analysis of mp3s
Reply #33
Thanks.

The links are there so you can conduct your own personal tests in order to determine if your chosen format and file size provide transparent results, or to give you an idea if the results will be acceptable if they are not transparent.

Participation in public tests isn't necessary for you to determine this.
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

Re: Trying to learn to understand spectral analysis of mp3s
Reply #34
Thanks.

The links are there so you can conduct your own personal tests in order to determine if your chosen format and file size provide transparent results, or to give you an idea if the results will be acceptable if they are not transparent.

Participation in public tests isn't necessary for you to determine this.
I know, I just thought I would share my results.  I am going to do a few and get back to you guys.

  • zWore
  • [*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #35
 
I'm trying to learn how to determine the quality of an mp3 file from the audio spectrum.
I'm trying too

The more I analyze, the more I know; by the experience. But sometimes I get some results that I just do not comprehend

[...]
Here's another 320kb track that I find a bit dodgy - am I right to find it dodgy, or is this ok?


It can be that those are transcodes- a clear sign of compression

But it can also be that this compression was there from the beginning- because a musician / producer / mixing person used a compressed sample / recording. There is really no way to tell this unless you have access to the source [in most cases meaning an original CD]

An example: "Running With The Wolves" by Prodigy. At the beginning there is a low quality sample, that is very audible

Finally, if anyone knows of any good tutorials, or sources of further info on how to interpret these kind of spectrums, I'd appreciate it. I'm just starting out as a DJ, so checking through my music to reduce the likelihood of a bad sound over a PA.
This is where I come from: https://forums.mp3tag.de/index.php?showtopic=20777

There is a description of how to fake a good-looking-in-spectrogram file; which led me [among other things] to the conclusion that you can't trust blindly what you see- you have to put it in conjunction with what you hear

There are also some links in that thread that will take you further down the rabbit hole



just convert lossless file to mp3 few times with different bitrates and notice the difference. that's the best way to learn.
Yes, that is a good method for learning

I also recommend comparing few files [with different kind of music] encoded to MP3 with best quality settings, but done in two different programs; and also comparing them with the uncompressed source

I also recommend checking your software that you use for encoding to lossless formats- it may be that you have some weird settings. [Just compare your WAVs / FLACs / WHATEVERs in spectrogram and see if the look exactly the same, so that you will be sure you are encoding them good]



[...]
More recently I've sourced tracks of Youtube that weren't available anywhere else, which often results in a file with at best 160kb, and again I've noticed some of these sound horrible at certain points.
[...]
This varies and changes. I remember I was checking few years ago, what audio quality 720p videos on YouTube are suppose to have. And the answers were inconclusive; in that sense that policy of YouTube was changing in time



Check this out: look for "Portal: No Escape" short video on both YouTube and Vimeo [https://www.youtube.com/watch?v=4drucg1A6Xk, https://vimeo.com/30064863]

Then skip to the end titles and save the music by record-what-you-hear method in your operating system [in 44 kHz, 16-bit]

Then download this video from these both sources via http://www.clipconverter.cc

And then compare the files in spectrogram: you will get different results [different presence of frequencies]



An so this example shows that it is worth to check different sources [if they are available]; and that sometimes it's better to record than to download



I thought I would get some intelligent answers on here, but it seems that my post has also attracted a number of trolls.

If you can't usefully contribute to a discussion, please don't bother to post, particularly when it's clear that you haven't actually read and understood my OP. Particularly those commenting on blind hearing tests, since that is completely irrelevant to my question which had NOTHING to do with what is or is not removed from a compressed audio track.
[...]
This went somewhat off topic- but that is a relevant knowledge, if you want to get the big picture




I'm trying to learn how to determine the quality of an mp3 file from the audio spectrum.
Answer: You really can't.
[...]
That's half true

You can't; but you can see transcodes and frequency cut off lines [indicating low / bad quality]

And also: seeing a single track can be misleading. But analyzing whole album, asuuming it came from the same source [and assuming it is not some OST by various artists], will give you more accurate info



https://hydrogenaud.io/index.php?action=dlattach;topic=111736.0;attach=9930
M4A file, Bitrate 61 kbps. Cut-off at 20.5 kHz.
[...]
If you look only for the cut-of line, then you will be misled

But if you actually take more time and see how this looks and compare it to other files that go way up but also sound good, the a little red light may flash in your head. For me this file looks somewhat little strange [= suspicious]; and therefore after playing it I have no doubt that it is of very low quality. But it sound so bad, that I would not even think of analyzing it



[...]
Why would people spread all this misinformation about bit rates in correlation to spectrograms?  I mean Spectro, another popular spectrogram program says this right on the front page of their website... 
Quote
Spectro lets you view vital data about compressed audio files and creates a spectrogram of the wave data. This allows you to quickly and easily spot quality issues with a file and also look for transcodes.
Spectro will soon support an automatic transcode detection feature and will be able to scan your music library for suspect files.
The key to understanding the usability of such software are the last two words from that quote:
"suspect files"

Not "plain wrong" or "absolutely of low quality", but only suspicious; but at the same time "quickly and easily" to be spotted



[...]
I don't understand why some people (I am not talking about you again) in 2016 still believe that more HF in spectrum = high quality or high bitrate
[...]
Because in more cases it is true?



Nowadays a lot of people (not you again) pretend they are real experts in these area by merely looking at a spectrogram, how hilarious.
I don't consider myself to be an expert

And that 61 kbps M4A sample surely gave me something to think about



Now I don't know what to believe...
I had the same problem when started to use spectrograms

Do high bit rates even matter that much?
If you are looking for a very good quality [or at least the presence of high frequencies], then anything below 320 kbps you can throw out without even listening / analyzing. At least in MP3 format

Does the format really matter?
In what sense?

You want to preserve all the data- always stick to lossless formats like FLAC and WV

You expect max quality from lossless formats like FLAC and WV found on the Internet- rarely you can get a fake [which you will be probably able to detect on spectrogram]

You only found what you were looking for in some lossy format- first use spectrogram and then your ears

Have the encoders improved greatly since the early days, so even low bit rate rips sound good?
For me all MP3s below 256 sound bad, and around ~128 sound terrible

[...]
Nowadays I always rip my CD's in FLAC because storage is not a huge issue and it allows me to convert to another format later on  Really though, I can admit that with a good portion of my music, I cannot tell the difference between a lot of my FLAC, 320 kbps MP3, and even some lower bit rates files in my library.
I have the same "problem"

I recently replaced few albums [of different kind of music] stored in maxed out MP3s for FLACs. To have them in better quality

And every time I thought that in those FLACs I heard something that was not present in MP3s, I went back to the old version [in MP3 format] and played that fragment. And every single time I had to admit, that it was only my imagination / wishful thinking: that new sound noticed in FLAC was already present in MP3 for all those years. I simply could not tell the difference

And may I add: I'm over 30 years old, but in the last two years I started to hear the compression in [low quality] files. If my collection was not kept in an orderly matter and if I had not implemented from the beginning a reliable info system based on tags and dates, I would think that I was wrong on that; that this audible compression was something done by me recently. But my files do not lie- with time I gained ability to hear [consciously notice] more. Based on what I see in spectrogram, I correctly detect with me ears low quality files something like 9 times out of 10. And I am now simply stunned how I could not hear 3-4 years ago what I'm able to hear now. And yes: my computer / audio hardware and settings are the same for around 10 years now and very similar for 15. But my friend who publishes a journal abuut cognitive science says, that this is not something that extraordinary, in terms of how a human brain works



And basically that is was why I started to use spectrograms- first I started to suspect some things and only then I have looked for proper tools and info about it



So all you doubters: do not let anyone tell you, that using spectral analyze for determining quality, is always a waste of time

My rip of Pink Floyd's Dark Side of the Moon is over 10 years old and it sounds fantastic.  It's a 128 kbps AAC rip that was done with iTunes on default settings and the album has great dynamics.  I still have quite a bit of 128 kbps MP3 files and a lot of it sounds excellent.  I also haven't updated my versions of LAME and FLAC in years and who even knows what version of AAC that was...
I also sometimes stumble upon in my collection of some very old MP3s that to my surprise sound very good [and also look good on spectrogram]



Now I don't know what to believe...
Just your ears, and I do mean only your ears.
[...]
By that logic audio, all editors are bad, because with your ears you cannot hear lines painting a waveform



[...]
I want to do these listening tests, but I see the tests have ended sometime ago.  If I take them now, am I able to upload the results to the forum so you guys can see how I did?

[...]
The links are there so you can conduct your own personal tests in order to determine if your chosen format and file size provide transparent results, or to give you an idea if the results will be acceptable if they are not transparent.
[...]
Why would anyone in 2016 want to listen to music in for example a 128 kbps MP3?

Are you using Compact Disc for storage?

Do you have an Internet connection bandwidth counted in single kB digits?



I appreciate the time all you guys took in 2008 to conduct those [MP3] tests. But that's like talking about today video formats resolutions [and codecs] while evoking empirical data from the beginning of the DVD area. It gives context but practically is irrelevant

[Of course I know that most people, especially younger ones, do not know what's that all fuss about 128 vs. 320 vs. FLAC, because they go to YouTube and listen most of the time to some real crappy encoding without a thought; as they know nothing about audio formats. But I suppose that we here, at least in this topic, are after high quality solutions]


##################################################################################################


My overall experience is that:
- rarely some MP3s little below 320 kbps still can sound quite good
- rarely even some FLACs can sound bad, even when looking good on spectrogram
- I can't hear the difference between maxed out 320 kbps MP3 and a FLAC
- older MP3s tend to have a clear frequency cutoff line somewhere below 20 kHz, indicating a poor quality at the spot
- multiple frequency cutoff lines below 16 kHz [at different points of the track] usually indicate lower quality, even if in other fragments of that track frequencies goes up to ~20 kHz
- some official / legite optical and purely digital releases can look A-OK in spectrogram but still sound terrible
- music ripped from vinyls often have noise and all kinds of glitches


And I use Spek, because its window can be maximized and it gives really good look at the data. Sometimes, when in doubt, I also use Audacity. And I tend not to use Spectro, because it has window with data way to small. I've also tried some other software which I can't recall, but although it's spectrogram was probably exceeding capabilities of that from Audacity, it didn't have a quick access / interface that I require in my workflow. And I've also contacted Brian D’Alessandro, the co-author of paper entitled "MP3 Bit Rate Quality Detection Through Frequency Spectrum Analysis", and learned from him, that the algorithm they used for atomated analyze of thousands of files was used via a command line input [wich is also no suitable to my needs]

  • saratoga
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #36
128k mp3 these days is remarkably good.  If you think 256k sounds bad it means you probably know what you're talking about.

  • bennetng
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #37
https://hydrogenaud.io/index.php?action=dlattach;topic=111736.0;attach=9930
M4A file, Bitrate 61 kbps. Cut-off at 20.5 kHz.
[...]
If you look only for the cut-of line, then you will be misled

But if you actually take more time and see how this looks and compare it to other files that go way up but also sound good, the a little red light may flash in your head. For me this file looks somewhat little strange [= suspicious]; and therefore after playing it I have no doubt that it is of very low quality. But it sound so bad, that I would not even think of analyzing it
Do you mean you can determine a file is lossy/transcoded from a lossy source by merely look at the spectrogram, even without the original lossless file's spectrogram as a reference?

  • greynol
  • [*][*][*][*][*]
  • Global Moderator
Re: Trying to learn to understand spectral analysis of mp3s
Reply #38
If you think 256k sounds bad it means you probably know what you're talking about.
I think you missed the word don't.

Clearly he doesn't.
  • Last Edit: 31 May, 2016, 11:54:54 AM by greynol
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

  • mjb2006
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #39
@zWore -

You mustn't make public statements here along the lines of "anything less than 320 usually sounds bad / I can hear the difference" without proof ... this is a violation of TOS#8.

Some food for thought:

Cutoffs at 16 kHz do not indicate low/bad quality. They indicate that in those portions of the song, precious bits did not need to be wasted on the difficult-to-efficiently-encode, noisy, sporadic, quiet or inaudible harmonics and hiss at the extreme upper end of human hearing. Rather, the encoder realized that there was nothing to gain by keeping whatever was up there (if anything), and the available space was much better utilized in the lower, far more audible bands. So if anything, the parts your ear is most sensitive to may well be higher quality than the parts where scalefactor band 21* was hogging space.

Frequency cutoffs happening below 16 kHz in a high-bitrate file do suggest a possible transcode from a lower bitrate, if you know the source material has audible content above the cutoff.  However even when there is audible content above the cutoff, the exclusion of those frequencies can be a positive change if what's being removed is mostly noise. Consider the sound of very old recordings or instruments which do not have very high harmonics, and a lot of tape hiss plus vinyl or shellac background noise... cutting out the high frequencies can make some people say it sounds better because it's less noisy. Others will say they think it sounds better with the noise, because they think the music sounds brighter when the treble-y noise is present, even though the musical parts are confined to the lower frequency bands.

* sfb21 in high-bitrate mp3s is the 16+ kHz range, and is unique in that accuracy in this range is inversely proportional to the efficiency in all the other bands. IOW, if this range is preserved, it will either be unusually noisy, or will make the 0-to-16-kHz range (the parts you can hear) take up an unusually large amount of space, possibly enough to adversely impact perceived quality. You can usually trust the encoder to make good decisions about when and how to encode sfb21.
  • Last Edit: 31 May, 2016, 11:50:04 AM by mjb2006

  • greynol
  • [*][*][*][*][*]
  • Global Moderator
Re: Trying to learn to understand spectral analysis of mp3s
Reply #40
Now I don't know what to believe...
Just your ears, and I do mean only your ears.
[...]
By that logic audio, all editors are bad, because with your ears you cannot hear lines painting a waveform
What a completely asinine thing to say.  Have you any clue how wave editors are used and why they often come with tools which allow the user to see what is going on in the frequency domain?

The links are there so you can conduct your own personal tests in order to determine if your chosen format and file size provide transparent results, or to give you an idea if the results will be acceptable if they are not transparent.
Why would anyone in 2016 want to listen to music in for example a 128 kbps MP3?
If you can't hear a difference (NB: I said, hear, which is how we consume audio; rather than "see" which is the entire content of your overly-leavened naive post of saccharine), then I don't see the problem; not that I ever cited 128kbps MP3 as a general threshold for transparency.

  • Last Edit: 31 May, 2016, 06:38:44 PM by greynol
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

  • greynol
  • [*][*][*][*][*]
  • Global Moderator
Re: Trying to learn to understand spectral analysis of mp3s
Reply #41
There is a description of how to fake a good-looking-in-spectrogram file; which led me [among other things] to the conclusion that you can't trust blindly what you see- you have to put it in conjunction with what you hear
You've completely missed the target: when it comes to what you hear you are not doing yourself any favors attempting to include what it is that you see.

https://www.youtube.com/watch?v=G-lN8vWm3m0

Until you can understand why seeing is not involved in the development in lossy perceptual audio codecs you really shouldn't be posting in these types of discussions.
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

  • zWore
  • [*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #42
If you think 256k sounds bad it means you probably know what you're talking about.
I think you missed the word don't.

Clearly he doesn't.
Yes...

I know all about that Good Quality coming out of high compression, be it audio or video

I also know people, who believe in UFO, read horoscopes and fear of witchcraft. Would you like to make friends with them?

Do you mean you can determine a file is lossy/transcoded from a lossy source by merely look at the spectrogram, even without the original lossless file's spectrogram as a reference?
Rarely

More often I get suspicions- and then I use my ears

And to answer another questions:

a] most of the time [in what I do] I do not have access to original lossless from the get-go

b] most of the time I work from the other side- first I hear, then I look at the spectrogram

c] I still [although very rarely] get some results that do not make sense or my knowledge / experience is not enough for valid evaluation of them; as I mentioned it in my first post

@zWore -

You mustn't make public statements here along the lines of "anything less than 320 usually sounds bad / I can hear the difference" without proof ... this is a violation of TOS#8.

I deducted you are refering to the Terms Of Service

>>
8. All members that put forth a statement concerning subjective sound quality, must -- to the best of their ability -- provide objective support for their claims.  Acceptable means of support are double blind listening tests (ABX or ABC/HR) demonstrating that the member can discern a difference perceptually, together with a test sample to allow others to reproduce their findings.  Graphs, non-blind listening tests, waveform difference comparisons, and so on, are not acceptable means of providing support.
[
<<

All I wrote was my overall experience, as I called it

But I see how it violates [I must say] this good rule

So I'm sorry for that


Some food for thought:

Cutoffs at 16 kHz do not indicate low/bad quality. They indicate that in those portions of the song, precious bits did not need to be wasted on the difficult-to-efficiently-encode, noisy, sporadic, quiet or inaudible harmonics and hiss at the extreme upper end of human hearing. Rather, the encoder realized that there was nothing to gain by keeping whatever was up there (if anything), and the available space was much better utilized in the lower, far more audible bands. So if anything, the parts your ear is most sensitive to may well be higher quality than the parts where scalefactor band 21* was hogging space.
Well, speaking from my OVERALL EXPERIENCE based on thousands of "spectrogram & ears" evaluations [followed if needed / possible by acquisition of questioned music in lossless or at least higher quality], in most cases cutoffs at 16 kHz did indicate that

Some of those that did not, most likely concur with what you are saying; but it some cased it could also be, that my hardware quipment wasn't just able to present the difference

Frequency cutoffs happening below 16 kHz in a high-bitrate file do suggest a possible transcode from a lower bitrate, if you know the source material has audible content above the cutoff.  However even when there is audible content above the cutoff, the exclusion of those frequencies can be a positive change if what's being removed is mostly noise. Consider the sound of very old recordings or instruments which do not have very high harmonics, and a lot of tape hiss plus vinyl or shellac background noise... cutting out the high frequencies can make some people say it sounds better because it's less noisy. Others will say they think it sounds better with the noise, because they think the music sounds brighter when the treble-y noise is present, even though the musical parts are confined to the lower frequency bands.
[...]

Of course I differentiate between [let's say] modern pop music, blues recorded in 1920's, rip of 60 minutes score from a 1997 game that was somehow filled on a single CD with cinematics cut scenes

Of course most of [at least my] music is of [supposed] good / excellent quality [if we are talking about the source material]

Now I don't know what to believe...
Just your ears, and I do mean only your ears.
[...]
By that logic audio, all editors are bad, because with your ears you cannot hear lines painting a waveform
What a completely asinine thing to say.  Have you any clue how wave editors are used and why they often come with tools which allow the user to see what is going on in the frequency domain?
Please enlighten me

Or just skip to the part, when you can explain, why exactly I should use ears and only my ears; and not spectrogram

The links are there so you can conduct your own personal tests in order to determine if your chosen format and file size provide transparent results, or to give you an idea if the results will be acceptable if they are not transparent.
Why would anyone in 2016 want to listen to music in for example a 128 kbps MP3?
If you can't hear a difference (NB: I said, hear, which is how we consume audio; rather than "see" which is the entire content of your overly-leavened naive post of saccharine), then I don't see the problem; not that I ever cited 128kbps MP3 as a general threshold for transparency.
There are two poles. One is the Pono-mindset pole. The other is the 128 kbps-is-good pole

And although humans can survive on both poles, they aren't really habitable

There is a description of how to fake a good-looking-in-spectrogram file; which led me [among other things] to the conclusion that you can't trust blindly what you see- you have to put it in conjunction with what you hear
You've completely missed the target: when it comes to what you hear you are not doing yourself any favors attempting to include what it is that you see.

https://www.youtube.com/watch?v=G-lN8vWm3m0

Until you can understand why seeing is not involved in the development in lossy perceptual audio codecs you really shouldn't be posting in these types of discussions.
Thank you for that very informative video; a video that has nothing to do with how I use spectrogram

That McGurk Effect would affect my evaluations if I somehow could remember that 4328th song that I checked on a Spectrogram looked BAD and that it is this exact file that I am listening to right now. Or if I was first and always checking spectrogram and only then playing a file

Because as I explained, I work [most of the time] like this: first I hear BAD quality among my collection, then I check that specific file in spectrogram, and then [in most cases] I get better quality of that track; the better quality that I actually hear [and can compare with the old file / version]. And so I feel very sorry for all those people that do not comprehend such modus operandi and because of which they devalue spectrogram analyze for checking quality

Thanks for links.  By the greynol, I like your new signature - "Your eyes cannot hear".
[...]
And on the side note: there are some substances, that allow you to turn music into multiple sensory experience. And there are people [like myself and my close relative] that can get nice feeling shudders when listening to music while being completely sober. Also, anyone who was not exposed to the wonders of music while in lucid dream state, should not give in to narrow minded statements abut how music is perceived

  • saratoga
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #43
If you think 256k sounds bad it means you probably know what you're talking about.
I think you missed the word don't.

Clearly he doesn't.
Yes...

I know all about that Good Quality coming out of high compression, be it audio or video

I think you massively overestimate your own understanding.  Have you actually compared 128k/256k on a modern encoder or are you just assuming?  Because it sounds a lot like you haven't actually tried. 


  • bennetng
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #44
May I propose a quiz? We use several different formats or same format with different encoders/settings to encode/transcode a source file, decode them to float 32 wavpack (to avoid clipping if the decoder supports), attach the encoder details and original file in an encrypted 7z file and let him sort the quality out by using any methods he likes. For example:

Quiz archive:
encoder a.wv
encoder b.wv
encoder c.wv
encoder d.wv
encoder e.wv
...

Answer archive (encrypted)
original.wv
encoder setting.txt

That could be quite interesting even if each of us only contribute one 30-second clip. We can even comment/guess other members' files (in an encrypted way in order not to affect other members' judgment). Not to disclose the password until he answered the quiz.

  • sven_Bent
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #45
@bennqu
we don't have to make it that difficult there is free ABX software. but im guessing that Zwore is one of those kind of people that like to make statements of his abilities but rarely want to do a real test to show em.
They exist everywhere, and always full of statements and half truths, but when you offer them to prove themselves through a simple ABX they have all excused to suddenly not have time or the right equipment for it.

zWore if you really want to prove you point that rarely anything below 320kbits sounds good, would you mind doing a simple ABX test ? I will provide the necessary files for you and even provide a file that has been double encode into mp3 ?
You can even provide your own test track  of your "GOOD qualtiy source"
Sven Bent - Denmark

  • bennetng
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #46
Of course we can simply use ABX, but the problem is that zWore also likes to look at the spetrograms. If he cannot differentiate two or more files by listening he can judge the quality by his eyes as he claimed. The test is aimed at his claim about judging quality by looking at spectrograms. I am interested to know his guessing accuracy.
  • Last Edit: 03 June, 2016, 05:14:09 PM by bennetng

  • sven_Bent
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #47
@BenetnG

oh I see now.  i was more focused on his ability to hear the difference.  You are absolutely right then
then agian instead of comparing orignal vs mp3. it would be fun to see mp3 vs dlb mp3. since both with have the same aproimatley cut off", but once should sound worse than the other.
Sven Bent - Denmark

  • bennetng
  • [*][*][*][*][*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #48
One of my old attempt:
https://hydrogenaud.io/index.php/topic,49288.0.html

Using one format/one encoder with default settings could be easy, but with multiple formats (not limited to mp3) + different encoder + transcode it can be extremely difficult. Of course some may think that it is unfair to artificially create some hard-to-guess scenarios but it is still a valid method to prove a concept, and who knows such cases will not happen in real life? For example, if someone believes in the "good looking spectrogram" concept (not only limited to cutoff), he may deliberately change the default encoder settings and distribute his files.

  • zWore
  • [*]
Re: Trying to learn to understand spectral analysis of mp3s
Reply #49
And on the side note: there are some substances, that allow you to turn music into multiple sensory experience. And there are people [like myself and my close relative] that can get nice feeling shudders when listening to music while being completely sober. Also, anyone who was not exposed to the wonders of music while in lucid dream state, should not give in to narrow minded statements abut how music is perceived
I forgot to mention synesthesia


And for everyone still bearing a "Your eyes cannot hear" tattoo on their foreheads I prepared some links:

http://www.medicaldaily.com/your-brain-music-how-our-brains-process-melodies-pull-our-heartstrings-271007

http://www.cog.brown.edu/courses/cg195/pdf_files/fall07/Gaver-howdowehear.pdf

https://www.mcgill.ca/mpcl/research


As for the quality, all I said was
Do high bit rates even matter that much?
If you are looking for a very good quality [or at least the presence of high frequencies], then anything below 320 kbps you can throw out without even listening / analyzing. At least in MP3 format

and also

rarely some MP3s little below 320 kbps still can sound quite good
I operate with [live by] the first statement when acquiring new music. Which I think cannot be argued with, in regard to high frequencies

The second one is based on what music I've already have and the new that happens to have cut-off lines below 19-19.5 kHz


I think you massively overestimate your own understanding.
I think you're probably right
Have you actually compared 128k/256k on a modern encoder or are you just assuming?  Because it sounds a lot like you haven't actually tried.
Having tried new software for making 224-256 MP3's I'm having difficulty to tell the difference between them and 320's

Of course we can simply use ABX, but the problem is that zWore also likes to look at the spetrograms. If he cannot differentiate two or more files by listening he can judge the quality by his eyes as he claimed. The test is aimed at his claim about judging quality by looking at spectrograms.
Yes. First I listen to, then maybe suspect, and only then check it out. Around 90% of the time I suspect correctly; which is in regard to that old music / codecs. I have a vast collection of music accumulated throughout the years, but only recently started to nice bad and terrible quality of some of it. And spectrograms are giving me a proof of that

I am interested to know his guessing accuracy.
I will be glad to participate in such evaluation of my skills and claims: that I can tell if MP3 is of lower quality than ~256 [as I have never claimed that I can hear the difference between lossless and MP3 320]

Am I to provide some song that I know, but in lossleess format; that will be encoded by you to lower-quality-but-made-to-look-in-spectrogram-like-320 so that I cannot cheat? Should / can I provide a rock / pop song, a classical piece and maybe something electronic?

I can also provide you with a sample of a lossless file that looks A-OK in spectro, buts sounds terrible; which is rare case of something I do not comprehend yet