Head-Fi.org › Forums › Equipment Forums › Sound Science › I found a good test track for MP3 codecs
New Posts  All Forums:Forum Nav:

I found a good test track for MP3 codecs

post #1 of 14
Thread Starter 
From : Time on earth: Crowded House: Track One: Nobody Wants to

For a while I have thought that my MP3 version (Lame 3.98, VBR 0) sounded a bit rough (distorted) epecially on my WD HDTV. So today I took the MP3(Lame 3.98, VBR 0) and a fresh Wav rip. I used the same Software and PC that I used to rip/encode the track previously.

First I ran both through Cool Edit Pro, the Wav file shows 8 possibly clipped samples, the MP3 shows over 400 !. The track is recorded with a generally high average level it peaks at 0db a lot on the Wav file.

Then I did a DBT in Foobar I homed in one a particular small segment and I got 16/20 i.e P < 0.006, it was quite tough , my ears are not the best but it is the first time I have ever been able to positively DBT between Wav and a VBR 0 mp3 file. I did not use replay gain since the overall levels were within 1/100th of a db.

What is more puzzling is why my WD HDTV/optical/Entech sounded more distorted than my PC/Musiland 01 USD USB to Optical/Entech

I have done PC/HDTV comparisons before and always come up null, but here the difference is gross. I even swapped digital cables

Okay, I took the optical feed from my WD HDTV and recorded it in Audacity from a USB ADC with optical input. It sounds just as bad through the PC now, but no worse, but in Cool Edit pro it now shows 54,000 possibly clipped samples. It looks like the WD is cranking up the digital output which being already quite often at 0db appears to add a substantial amount of clipping. The overall level is upped by about 1db which is not a lot unless you are bang on max

Well, there you go then....
post #2 of 14
OMG!!! This coming from the formidable Nick Charles.

Have you tried the same test with constant 320 as opposed to variable. From my undestanding variable peaks at 320 and on paper constant is better. The last time I did a blind test with 320 constant and lossless I truly failed to pick a difference.
post #3 of 14
Thread Starter 
Quote:
Originally Posted by SP Wild View Post
OMG!!! This coming from the formidable Nick Charles.

Have you tried the same test with constant 320 as opposed to variable. From my undestanding variable peaks at 320 and on paper constant is better. The last time I did a blind test with 320 constant and lossless I truly failed to pick a difference.
Not yet, good idea, I will try that next weekend. I think the problem is that the track is just really hot and it gives Lame problems.

Normally I fail dismally at DBTing "good" MP3 and wav, and I always do high bit-rate encodes. Since this is 1 track out of 12,000 I will not be claiming golden ears
post #4 of 14
There are guys here that believe 128 cannot be differentiated from lossless. I'm afraid you already have golden ears.
post #5 of 14
Thread Starter 
Quote:
Originally Posted by SP Wild View Post
There are guys here that believe 128 cannot be differentiated from lossless. I'm afraid you already have golden ears.
I am a huge fan of "good" MP3 and normally consider lossless to be a waste of hard disk space, but clearly it is not always transparent even to my old ears. This is, I think, a pretty extreme example though, the track is slightly distorted in itself and very hot but the encode does add levels of crappy to it, perhaps 320 will be better.

The problem now is that having identified added distortion I will be inclined to hear it everywhere, my Streamer does support FLAC, but I don't want to have to re-rip my 800 CDs

What is more annoying is how the WD HDTV adds even more crappy on playback, that was unexpected, I will need to test this with a wav file and see if it does the same thing...
post #6 of 14
Yeah, there you go...
LAME V0 (and MP3 in general) may be transparent in most cases, but certainly not all the time. Another reason to go lossless.
post #7 of 14
Thread Starter 
Quote:
Originally Posted by krmathis View Post
Yeah, there you go...
LAME V0 (and MP3 in general) may be transparent in most cases, but certainly not all the time. Another reason to go lossless.
The vast majority of my music collection is either classical or older "popular" music which both tend to be rather less prone to the hotness disease of mastering which we see so much of today, the problem track is however from a very modern recording, however I have very very few very modern CDs so cannot say for sure if hotness in and of itself is a problem for MP3 codecs...

I will re-encode this CD to FLAC but will only re-encode CDs which I think are problematical as and when...
post #8 of 14
Quote:
Originally Posted by krmathis View Post
Yeah, there you go...
LAME V0 (and MP3 in general) may be transparent in most cases, but certainly not all the time. Another reason to go lossless.
Because he is able to barely discern one small section out of one track out of thousands?
post #9 of 14
Yep, you've stumbled upon the rare and elusive killer track.

Hydrogenaudio has them pop up ever so often. Most of the time MP3 should work fine for the purpose (even 128cbr can work with certain recording, honestly!) . . . but once in a while it's not enough.

Personally, I don't stumble on them enough to really care myself~
post #10 of 14
Quote:
Originally Posted by ILikeMusic View Post
Because he is able to barely discern one small section out of one track out of thousands?
Correct!
I doubt the OP have ABX tested thousands of tracks, hence hard to say how many of them are transparent or not.
post #11 of 14
Thread Starter 
Quote:
Originally Posted by krmathis View Post
Correct!
I doubt the OP have ABX tested thousands of tracks, hence hard to say how many of them are transparent or not.
MY DBTs run into the dozens rather than the thousands, so on the one hand you are correct I do not *know* how many I could DBT correctly.

On the other hand this is the first time a track has really alerted me to it's crapness in a way that caused me to suspect a limit in the encoding. Maybe I could pick out a few more with concentrated listening, dunno. To re-encode all my tracks to FLAC seems a bit like a juggernaut solution for an occaisional problem. If I were starting from scratch then yes I would probably go the FLAC route as hard disk space is so cheap now.
post #12 of 14
Quote:
Originally Posted by nick_charles View Post
If I were starting from scratch then yes I would probably go the FLAC route as hard disk space is so cheap now.
That's true regarding PCs but not most portable players. My current mobile library comes to about 28 gb at LAME -V0, and that is compatible with a wide range of portable players. If I were to use FLAC I would be limited to a very small subset of what's available and I can't see any logical reason for accepting that limitation in that I doubt I could reliably A-B one in a thousand tracks. And the 'even one instance is too much' philosophy is simply impractical in the real world, in this or just about any other realm.

Nick, would you mind pointing out the time point where you hear the problem/artifact that gives it away? I'd like to try this track myself at various bitrates.
post #13 of 14
Thread Starter 
Quote:
Originally Posted by ILikeMusic View Post
Nick, would you mind pointing out the time point where you hear the problem/artifact that gives it away? I'd like to try this track myself at various bitrates.
Sure, I chose from ~1:02 to 1:11 to DBT including the lyrics "They make it go away, pretending that it's all ok". It took a lot of concentration and many repeats. I would be interested to see how different bit-rates cope.

However, it sounded far worse on my WD HDTV which made me investigate, but on my PC the MP3 sounded much better and the difference was a lot less.

Good Luck !
post #14 of 14
Quote:
Originally Posted by nick_charles View Post
Good Luck !
Thanks! I have a lossless copy of that track, going to give it a go...
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › I found a good test track for MP3 codecs