forum

Bitrate Test - Can you tell the difference?

posted
Total Posts
40
Topic Starter
kikones34
*I don't feel like writing kind-of-long posts in english, so I apologize if I make huge grammar mistakes*

Yesterday I was wondering: It's really necessary to waste disk space with 320Kbps mp3 files? Why not use 192, or 128Kbps? Is there any real difference?

I could not answer my question.

So I've made a little app to test if the human ear can really tell the difference between 320, 192 and 128Kbps files.

The test consists of 7 segments of random songs that I had in 320Kbps. Every segment has three versions, recoded at 320, 192 and 128Kbps with Sound Forge 10.
Three notes will appear in the screen for every segment. A bitrate is assigned randomly to each one. You'll have to listen and distinguish them.

When you feel comfortable with your decisions, click the text that will appear at the bottom of the screen to check the bitrates.

After you finish all the 7 songs, you'll go to the results page.


The numbers between brackets are the real right and wrong answers. The numbers with no brackets are the "fair" right and wrong answers.
Click the spoiler below to read a boring explanation.

SPOILER
There are only 3 possibilities for each song:

3 good
1 good, 2 wrong
3 wrong

It's unfair to count the right and wrong answers this way, because there are more wrong than good answers (5 vs. 4)

So I've made the counting system like this:

3 good = 2 good (because if you've made 2 good the other one will be good)
1 good, 2 wrong = 1 good, 1 wrong (because if you've made 1 good and 1 wrong, the other one will be wrong)
3 wrong = 2 wrong (there's no real reason more than convenience here)

And this way there are the same wrong and good answers (3 & 3).

But I'm not totally convinced with this system so I kept the normal counting between brackets.
The final mark is calculated based on the "fair" right answers.
You will get and osu! rank depending on how well you've done :D

TL;DR: Here's the download link: Dropbox

Note that it MAY have some bugs, because I'm the only person who tested it, but if you don't make strange things all should be good.

Oh, and don't forget to post your results and your opinion about bitrates ;)!
Kanye West
Interesting idea, looks worth trying.
Nyquill
I always only get 320 right lmao
Stefan

Damn, I suck. I think the D is a bit buggy, but idk.
Aurele
lolok



Pretty well done :3
IppE
Mozaik Role is the only one where MP3 encoding has real troubles making it sound transparent at bitrates lower than 320kbps, the rest is pretty bass heavy or just not that hard on the encoder to begin with...
Frizz
lolaudiophiles

This is also worth a try though http://www.klippel.de/listeningtest/lt/
zertap
Depator


Not so much difference between 192 and 128 thought.
Corin
It's pretty easy.

Just listen to the highs freq shit, anything lower than 320Kbps gets snubbed slightly and sounds more... flat I guess.

Can't think of the word.
Jarby
I'll post here later when my house is quieter. My dad is using power tools and my washing machine is loud as fuck too.

Corin wrote:

It's pretty easy.

Just listen to the highs freq shit, anything lower than 320Kbps gets snubbed slightly and sounds more... flat I guess.

Can't think of the word.
Humans generally can hear up to 20 kHz. With modern encoders, higher frequencies are cut first. You can see this by looking at spectrograms and it's a good way of checking for transcodes because you can easily see if the frequency range is too low for the bitrate.



Noooo my audiophile cred
XPJ38
I got a D but honestly I couldn't tell the difference between the bitrates at all :?
Ephemeral

IppE wrote:

Mozaik Role is the only one where MP3 encoding has real troubles making it sound transparent at bitrates lower than 320kbps, the rest is pretty bass heavy or just not that hard on the encoder to begin with...
pretty much this

do this with orchestral/instrumental music and you will absolutely hear a difference between all three with ease even through laptop speakers

good little app though
Jarby
Also, I'm wondering if you encoded all of these from a lossless source or did them all from the 320 you grabbed.
Elysion


Bangarang HURRRRRRRR ;_;
Suzu_old_1
Topic Starter
kikones34
Wow, I'm not used to big forums at all o.o
Thank you for all your responses :D!

And now, let's start..

Frizz925 wrote:

lolaudiophiles

This is also worth a try though http://www.klippel.de/listeningtest/lt/
Interesting

Ephemeral wrote:

IppE wrote:

Mozaik Role is the only one where MP3 encoding has real troubles making it sound transparent at bitrates lower than 320kbps, the rest is pretty bass heavy or just not that hard on the encoder to begin with...
pretty much this

do this with orchestral/instrumental music and you will absolutely hear a difference between all three with ease even through laptop speakers

good little app though
Thank you :oops:
To be honest, I don't know anything very much about mp3 encoding, so I've chosen the songs randomly :o
Maybe I'll upload a version with other kinds of music to see the difference between the results... maybe..

Jarby wrote:

I'll post here later when my house is quieter. My dad is using power tools and my washing machine is loud as fuck too.

Corin wrote:

It's pretty easy.

Just listen to the highs freq shit, anything lower than 320Kbps gets snubbed slightly and sounds more... flat I guess.

Can't think of the word.
Humans generally can hear up to 20 kHz. With modern encoders, higher frequencies are cut first. You can see this by looking at spectrograms and it's a good way of checking for transcodes because you can easily see if the frequency range is too low for the bitrate.
Does this mean that songs with lower frequency have less audible loss than songs with higher frequency? Or just the opposite thing? Or neither? (I dunno :? )

Jarby wrote:

Also, I'm wondering if you encoded all of these from a lossless source or did them all from the 320 you grabbed.
The point of this test is to decide if I should re-encode my 320Kbps files to a lower bitrate or not. So I encoded the other files from the 320 one ;)

-------------------------------------------------------------------------------------------

Alright, and now a genius conclusion: It depends on the ear and the song if the bitrate makes any difference.
(So I still don't know what do do with my files :) !)


EDIT: Oh, there are really a lot of smiles in the post, It's fine?
JFRN
interesting, i will test it on my computer.

and, good coding on Game Maker :D!
Topic Starter
kikones34

JFRN wrote:

interesting, i will test it on my computer.

and, good coding on Game Maker :D!
YOU! Y U every time D:?

.
.
.

..Hi :3
HakuNoKaemi
I, somehow, perfect'd 3 songs, did around 58-60 or so.
Well, Depend mostly on the volume you want to hear them, with headphones and a normal volume it's difficult to see the difference from 196 onward in all song.
Differently if you must hear them with speaker at high volume (example: you're a DJ and play them in some kind of Disco, or your music should be played by all), then an higher bitrate is unavoidable.
Remi_Scarlet
128 is easy to tell but between 192 and 320... Eh. I can't hear enough of a difference for me to care. Besides, I'll be way too focused on hitting all the notes while playing than just listening to the music.
Jarby

kikones34 wrote:

Does this mean that songs with lower frequency have less audible loss than songs with higher frequency? Or just the opposite thing? Or neither? (I dunno :? )
Details in the treble end are lost as bitrate declines, yes. This usually doesn't matter as much for songs with unimportant, simple and quiet parts in that area.
Kitsunemimi
Oops, silly me I got one of the tests completely wrong~ >3<


I can tell the differences between the clips pretty easily, but sometimes it's hard to decide which one is actually 320 and which one is 128~.

SPOILER
Oh wait a second...

kikones34 wrote:

if the human ear can really tell the difference
...Ah well whatever~
Flanster
funny how i managed to spot oppan gangnam style best

first and only try!


but yea i suck at this, my headphones are cheap too barely tells difference :v
IppE

Jarby wrote:

Humans generally can hear up to 20 kHz. With modern encoders, higher frequencies are cut first. You can see this by looking at spectrograms and it's a good way of checking for transcodes because you can easily see if the frequency range is too low for the bitrate.
LAME has the default frequency cutoff at 19.5 kHz so it can be heard with good enough equipment depending on ~stuff~ like how the song was mastered etc.

And of course if your ear hears it to begin with.
Jarby

IppE wrote:

Jarby wrote:

Humans generally can hear up to 20 kHz. With modern encoders, higher frequencies are cut first. You can see this by looking at spectrograms and it's a good way of checking for transcodes because you can easily see if the frequency range is too low for the bitrate.
LAME has the default frequency cutoff at 19.5 kHz so it can be heard with good enough equipment depending on ~stuff~ like how the song was mastered etc.

And of course if your ear hears it to begin with.
What? Cutoff depends on bitrate.
peppy
A lot of the songs used for testing were noticeably compressed (as in the frequency range) to the point that encoding at 192kbit (or even 128kbit to a lesser extent) is transparent – that is, indistinguishable to a human. I had trouble distinguishing most of them apart as a result. In fact, a lot of the samples felt like they were encoded from a low quality original encode or rip (was this the case, possibly?). It would make more sense if you substantiated your original rip sources as being from original CDs you own.
ztrot
IppE just blew my mind, not really :P
Jarby

peppy wrote:

A lot of the songs used for testing were noticeably compressed (as in the frequency range) to the point that encoding at 192kbit (or even 128kbit to a lesser extent) is transparent – that is, indistinguishable to a human. I had trouble distinguishing most of them apart as a result. In fact, a lot of the samples felt like they were encoded from a low quality original encode or rip (was this the case, possibly?). It would make more sense if you substantiated your original rip sources as being from original CDs you own.
I checked at least a couple of the 320s and the spectrograms seem appropriate for what it's worth.
Topic Starter
kikones34

peppy wrote:

A lot of the songs used for testing were noticeably compressed (as in the frequency range) to the point that encoding at 192kbit (or even 128kbit to a lesser extent) is transparent – that is, indistinguishable to a human. I had trouble distinguishing most of them apart as a result. In fact, a lot of the samples felt like they were encoded from a low quality original encode or rip (was this the case, possibly?). It would make more sense if you substantiated your original rip sources as being from original CDs you own.
The purpose of this test was not to test the quality difference when encoding from a lossless source, but to test the quality difference when re-encoding from a 320Kbps MP3 source. (I should've specified this in the original post, maybe :roll: )

So, if you're right, the conclusion should be: There's no audible difference.

But then, there is some people that says:

Nyquill wrote:

I always only get 320 right lmao
And:

Remi_Scarlet wrote:

128 is easy to tell but between 192 and 320... Eh. I can't hear enough of a difference for me to care. Besides, I'll be way too focused on hitting all the notes while playing than just listening to the music.
So.. Does this really make any sense?

I don't know, I think I'm just going to re-encode the 320s to 192s
peppy
You're missing the point, though. Even if the source was "320kbit", it could have been encoded to 192 or 128 before it got to the 320kbit. This is common for illegal music. For this reason, you should rip it yourself if you are to create something like this, so you can ensure a clean entrance path which doesn't hinder the testing environment.

I have done double-blind tests myself when considering switching to AAC. For a lot of the music I listen to – especially high female vocals – the different from lossless to AAC 128kbit was as obvious as night and day, but AAC 256kbit was such low confidence I have since switched all my music to this format. Not looking back, either.

When I say AAC 256 I am talking about the iTune settings, which is actually VBR with a max of 256kbit.
Kitsunemimi
When I was doing the test, I could hear the differences between the clips. There were some songs however that were much harder to distinguish because of the nature of the songs themselves. For example, one of the songs (I can't remember which one) just sounds really crummy and it's not apparent that the high frequency ranges are there, but they do exist (otherwise I wouldn't have been able to get the results that I had).

Besides, didn't Jarby say that he checked the mp3s, confirming that they were indeed 320 kbps, and not just upscaled?
bagnz0r
<DarkScenario>
I bet this guy pre-fucked up every song at 64kbps and re-encoded it at higher bitrates so that you fail. :)
</DarkScenario>

Also:
http://www.foobar2000.org/components/view/foo_abx

I keep my music at similar level peppy is.
All my music is 240 kbps AAC at this point for the obvious reason - SSD.
Kitsunemimi
I'm using an SSD too but I stick with lossless FLAC :3

(of course I wouldn't say I have a ridiculous amount of music as of this point but... meh :P)
mnh48
LOL..
Topic Starter
kikones34
Some final mostly useless conclusions:

·My test is definitively not made accurately, thus the results are definitively not accurate.
·We don't know if the 320s are in fact 320s (although Jarby said that "the spectrograms seem appropriate for what it's worth").
·Despite all that, there's people that hears a difference and can get decent results in the test, so it seems there's at least some variance.
·The kind of song really matters when encoding.
·Illegal music is illegal.

Thank you for your cooperation

And...

bagnz0r wrote:

<DarkScenario>
I bet this guy pre-fucked up every song at 64kbps and re-encoded it at higher bitrates so that you fail. :)
</DarkScenario>
You caught me D;
bagnz0r

Kitsunemimi wrote:

I'm using an SSD too but I stick with lossless FLAC :3

(of course I wouldn't say I have a ridiculous amount of music as of this point but... meh :P)
256 GB is hardly enough to maintain my 400+ collection of albums and still be able to keep tremendous amount of projects.
I might get an external drive or NAS though.
asl97
i can heard a high flute like sound in Mozaik Role, it sound more clearly in the higher bit-rate, did anyone hear it too?
something high in unbreakable and the one with higher bit-rate in express emotion is clearer.
Frizz

asl97 wrote:

i can heard a high flute like sound in Mozaik Role, it sound more clearly in the higher bit-rate, did anyone hear it too?
something high in unbreakable and the one with higher bit-rate in express emotion is clearer.
It's already stated that MP3 compression causes frequency cutoff on highest frequencies in previous posts.
show more
Please sign in to reply.

New reply