AC3 degradation? POLL - Lend me your ear (but I wont sing you a song)

[mats.hogberg]

In another thread, we're debating how much an MP3 is degraded when converting to AC3.
I'll let you be the judge.
What I've done:
I took a random MP3, (192 KBPS, 44.1 kHz) and decoded to wav.
Took the wav, and encoded to 256 kbps AC3 with ffmpeggui, and decoded to wav again.

....and please don't cheat by looking at wave forms and stuff - let your ears decide.
A.wav
B.wav


/Mats

[raquete]

I took a random MP3, (192 KBPS, 44.1 kHz)

is one valid test but the user from "the other thread" never posted the bitrate of his mp3 sources to encode to ac3: https://forum.videohelp.com/viewtopic.php?t=314527
and the worse,he posted

The MP3s are from Audio CDs but the trouble is that some of them are bit old and scratched.

and we can imagine how bad will be his results.

downloading your samples...

[raquete]

from A.wav and B.wav,
B.wav is worse!..and both have littles distortions.

and please don't cheat by looking at wave forms and stuff

is not needed,B.wav have stranges variations in the volume!!!

regards!

[Scorpion King]

B sounds the best here. I think. Whichever the top one is.

A, the second file, sounds muted, dull... not full sounding like the first file.

Why didn't you put A 1st?. I keep getting confused which is B.

Nice Test. I enjoyed it.

[mats.hogberg]

I've re-ordered the links (but a is still a, b is still b)
50/50 so far!

/Mats

[GTRBudda]

To my ears (hearing the clips after my band had a practice tonight. We play stuff like that ), A sounds a little cleaner, and at a very slightly lower volume. Perhaps some low end loss, too, compared to B. A is the ac3 to wav? B sounds a little fuller. I'd probably want A coming through my tv speakers and B in my car stereo. For over all sound though, I think B is better.

[raquete]

I've re-ordered the links (but a is still a, b is still b)
50/50 so far!

/Mats

too bad...but it's ok!
post number 5 edited after 2 votes.
you are editing your posted in the exact moment that i quoted...was wroten:
"1 correct,1 wrong!"
because only had 2 votes in the moment that you wrote in the first time and i still don't vote...

you are counting with Scorpion King's vote and you own vote,right?

[raquete]

can you post the MP3 (192 KBPS, 44.1 kHz) used to the work please?
(short sample like the .waves posted)

thanks!

[mats.hogberg]

I'll PM it to you, not to disclose too much. But it'll be in about 10 hours, as I'm at work right now.
Just 2 votes sounded a little lame, so I changed it to 50/50 instead And no, I haven't voted. That'd be cheating, right?

/Mats

[raquete]

I'll PM it to you, not to disclose too much. But it'll be in about 10 hours, as I'm at work right now.

don't need faster,thank you so much!

That'd be cheating, right?

all is clever!

regards.

[gadgetguy]

Where's the option for "I can't tell the difference"?
(I only listened with the crappy speakers in my monitor.)

[mats.hogberg]

Options added on popular demand

/Mats

[3dsnar]

I've listened to both. IMHO they differ only (by my subjective evaluation) in terms of slow amplitude variations, as someone noticed. I.e. sample B is louder at the beginning but the rest of the clip seems similar to A.

The other issue is that mp3 performs fairly well for rock tracks, where very fast transients are not present (as in some electronic stuff), and also there are usually no long strings (classical) where it is easy to catch the quantisation noise.

I have used near field monitors Yamaha MSP5 and headphones AKG K401.

I cannot evaluate which is better for me, because in this sort of paired comparison test it would be nice to know the original (in case if the differences are not so obvious).
With reference to original it would be possible to determine which one is more similar to the source.
Can you upload the original?
Cheers, 3d

[mats.hogberg]

All tru, no doubt. But not the issue here. The issue is:
How much would an MP3 degrade when reencoded as AC3?

/Mats

[Scorpion King]

[edit]

( I had to delete my 'wave form' photo I just put up)

[src2206]

Its good to see that my question has started a very healthy debate

Frankly I do not think that I should vote- but I'm going to keep a very close watch of this thread for the user's comments.

[3dsnar]

All tru, no doubt. But not the issue here. The issue is:
How much would an MP3 degrade when reencoded as AC3?

/Mats

OK, can you upload the original? (so it will be possible to compare A and B to it and chose more similar -> less degraded)
Thanx

[mats.hogberg]

Use your ears - that's what counts. One of the wavs above is the original. One is the original reencoded to AC3.
The original discussion was wether an MP3 reencoded to AC3 would sound "horrible" or not.
We all agree that the better your source material is, the better the end product, but if your source is MP3, how much worse would it sound when reencoded as AC3 (for Video DVD use).

/Mats

[3dsnar]

OK, I gave my vote for option 3.

----

I think that more reliably find out what you are looking for,
you would have to do the test in the following way (for example)

1a) Take the original wave (not encoded) -> reference
2a) Encode it to mp3 (lets say high bitrate -> high quality)
3a) Re-encode the mp3 to AC3.
This would make SET 1
---
1b) Take the original wave (not encoded) -> reference
2b) Encode it to mp3 (lets say low bitrate -> with noticable artifacts)
3b) Re-encode the mp3 to AC3.
This would make SET 2

Having these two sets (3 sounds in each set: A, B and REF). The question could be staded:

Which of the two sounds in each set (A or B) has better quality, with regard to the reference sounds?

Results should be published after all votes, so people would not get suggested by previous votes

If listeners would strongly prefer one of the options, we would find out that
mp3 is better (or mp3->AC3 is better, which could be theorerically also possible).
Otherwise (around 50% for A and 50% for B) both are the same, hence AC3 256 kbps does not affect mp3 encoded signal.

Above results could be considered separately for high quality mp3 as well as low quality mp3.


Just my thought
Cheers, 3d

[mats.hogberg]

I'll keep this in mind - I think it'd be interesting too, once this is over.
As far as this site goes, all poll standings are immediately available at any time - And you're even allowed to comment it. Very unscientific, but fun!

/Mats

[Scorpion King]

I'm surprised we have only 10 votes so far. Where is everybody....

[TheFamilyMan]

My two cents: Given just this music clip, it's a pretty limited test of how much fidelity is lost when encoding an MP3 to AC3. I imagine the encoders (mp3 and ac3) employed also have an impact on the results. I voted for B as the AC3 since it's volume fluctuates so much. Beyond that, I can't really tell which sounds "better".

[raquete]

Given just this music clip, it's a pretty limited test of how much fidelity is lost when encoding an MP3 to AC3.

i agree with you(completely)!!!! for this reason i don't will vote.
we need the source .wav and the mp3,this is the minimum and,of course:the description of how the mp3 and the ac3 was done.
i'm sure that the mp3 here was done with more care than the mp3 of the other post where the user call for help!
again the link and the reason of this poll: https://forum.videohelp.com/viewtopic.php?t=314527
here and there are completely differents cases.
mates.hogberg know what he do and how to do but src2206(user of the other thread) is not sure or don't how his mp3 was done.

I voted for B as the AC3 since it's volume fluctuates so much.

i bet that you're right but i'm not sure because,as you wrote,the sample is short and more: was cutted in differents places!

regards!

[vhelp]

(o) Wav A is MP3, Wav B is AC3
( ) Wav A is AC3, Wav B is MP3

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

My first impression was that *both* sounded good to my ears.

Then.. I gave a little more attention to the challenge and evaluation.

Now, I'm not much for audio, and have been known to do 192k and be
happy with it, but I have to say, that my audio is screwy at the moment.
I have a bad connection sound card (Turtle Beach) and it has sure signs
of ware. Sometimes Rt chan is louder than left most of the time.., else
the center (stereo) is not even, and more so towards to Rt chan.

Oh well. Anyways.

This is sure a tricky one. You don't know which direction to vote, but..

** SOUND: B=loudest, A=lower
** QUALITY: B=slight noisy, A=more/less quiet'er

So, I'm thinking that you encoded the A.mp3 -> B.ac3,
** because you said that you took a 192k -> 256k, which means to me,
** louder or more distortion, which to my ears, B did sound bit tadd loud'er
** and noisy (maybe distorted a little) to me. I find that when you convert
** to a higher bitrate, things tend to get noisy/distorted a little.

But, then I recall from my own experience, that when you transconvert
audio to another format (assuming w/out filter cleanup) the audio is
usually a little lower in volume, hence my evaluation of A=lower.

( Its hard for me to tell (and test) becuase my audio system is messed
up as I noted earlier, but I gave it my best shot. )

But, either way, you are snagged by this challege
( I voted, A )

-vhelp 4154

[mats.hogberg]

we need the source .wav and the mp3,this is the minimum and,of course:the description of how the mp3 and the ac3 was done.

Once more: The purpose of this test, is not to determine if AC3 or MP3 is better. The purpose is to determine if your average MP3 will sound horrible or not, after being reencoded to AC3. What it sounded like before it was encoded as MP3 is irrelevant. You already have the MP3 (decoded as WAV) - it's one of a.wav or b.wav

i'm sure that the mp3 here was done with more care than the mp3 of the other post where the user call for help!

What makes you think that now? You've already found that

both have littles distortions

Like I said, it's a straight CBR 192kbps encode, using Lame. Nothing fancy at all - Just what anyone equipped with CDEx would do.

/Mats

[raquete]

The purpose is to determine if your average MP3 will sound horrible or not, after being reencoded to AC3.

but "we all know" that depend of the quality of the source--->always!!!
and to be very clear: to encode the mp3 was used the source .wav but to encode the ac3 was used the mp3...seems just? i wrote in the begining; encode the ac3 from the sources and not from lossy to lossy.if src2206 have only mp3 as source we can't "force" one good result...never will be!!!

What makes you think that now?

what? that you know what you're doing.i'm sure that you know how to encode mp3,you're not a newby! or i misunderstood your question?

Just what anyone equipped with CDEx would do.

not anyone mats(let me call you mats as a friend)you know how to use and lots of people too but not everyone.again i repeat for you:we don't know how the mp3 of src2206 was done...he is not secure about it too.he have some 128,192 and 320 and to convert the know bitrates form unknow convertions desrve deep tests.this is what i wrote in the other thread:

src2206,
try few samples first and tell us about the quality of the results.
...or post one mp3 and the resulting ac3 if possible! can you?

i'm trying to help and not to "ilude" the user writing that his results will be cool or reasonables.
i repeat what i wrote here again:

here and there are completely differents cases.
mates.hogberg know what he do and how to do but src2206(user of the other thread) is not sure or don't how his mp3 was done.

[mats.hogberg]

we all know" that depend of the quality of the source--->always!!!

Quite. And here, the MP3 is the source. And the question is, again, how much would an MP3 degrade when reencoded to AC3. To stay true to the original thread I should have asked src for a sample MP3 - But that didn't occur to me at the time; I just took one I had at hand in my library.
...and please call me Mats, by all means - I try to be a friend to all here!

/Mats

[raquete]

...and please call me Mats

all right,Mats with big "M" for big friend!

And the question is, again, how much would an MP3 degrade when reencoded to AC3.

how much?
(again)worse than mp3 source!

[mats.hogberg]

Of course - every encoding to a not lossless format will degrade quality some. But is it a significant quality loss, or just of academic interest?

/Mats

[gadgetguy]

I think Mats has given us a perfectly designed and executed challenge.
The original source for the mp3 is (as Mats has stated repeatedly) irrelevant to THIS test. We all know that converting the orginal to mp3 degraded the audio, and that converting it again to ac3 will further degrade it, but his challenge was whether we as individuals, using just our ears to judge, could discerne which was which. This is not rocket science. Either you can hear a difference or not, and if you can hear the difference, which is the better (and presumably the source mp3)?
It's a pretty small sampling, but the responses would indicate that most people can't tell which is the source, which means that for most people it would probably be OK to convert again for their own use, but each would have to judge each individual circumstance for themselves. I don't think Mats even hinted that this is the proper or best way to handle the audio for any given clip. It is simply a challenge to anyone who thinks that just because the audio is technically further degraded, it will be unlistenable. If you are one who can hear the difference and can identify the source, then you clearly don't want to do this to any of your clips. But if you can't tell, then it might be a viable solution to your situation.
(Just don't try to sell it to someone who can hear the diference.)