TEST: MP3 vs WAV, can you hear the difference?

C

charger

New member
This is a little test I made up because of a discussion that came up in this thread:
http://www.homerecording.com/bbs/showthread.php?s=&threadid=82429

http://www.astrondelta.com/charger/audio/mp3test/

I ripped 19 seconds of the song "mon enfant" from the Ralph towner album Diary (ECM, 1974). I chose this album for its dynamic source material, high frequency content, and liberal tape hiss (one of the most obvious places where MP3 artifacts show up). Using Pro Tools LE on Windows, I created 2 sessions, one @ 16-bit and one @ 24 bit. I imported copies of the stereo file into each session separately. In the 16-bit session, I used Waves' L1+ Ultramaximizer to boost the level by 3.4dB, while limiting to -0.1dB. In the 24 bit session, I used Waves' L2 to boost the level 3.4db while limiting to -0.1dB. In both sessions, I dithered the output with Dither Type 2 on the Ultra setting (you still have to dither from Pro Tools in a 16-bit session, because the output is always 24 bit from the Pro Tools mixer). I then bounced these files down to separate stereo wav files @ 16 bit.

Then I compressed these wav files to MP3, using LAME encoder version 1.30, engine 3.92 MMX. I set the VBR to level 3, with a minimum of 32 kbps, and a maximum of 320 kbps. I converted these MP3 files back to wav using Winamp. I then encoded two more MP3 files, using the same settings, from these wav files. These result files were also converted to wav files. For the 4 MP3 files I created, the sizes ranged between 384-392 k. The wav files all come out to 3.25 MB.

MP3 test: challenge

See if you can identify them correctly... of the 6 wav files listed below:

1 was a 24-bit wav, dithered to 16 using L2
1 was a 16-bit wav, dithered using L1
2 were MP3s of the above (wav > MP3)
2 were 2nd-generation MP3s of the above (wav > MP3 > wav > MP3)

test1
test2
test3
test4
test5
test6
 
#'s 4 and 5 are the WAV files, but i can't tell which is dithered from 24. i won't even try to guess which are second generation MP3s.
 
Well I tell ya.... I hear no difference or very little difference on my headphones but I can hear a diff on the monitors... I am afraid to try and identify them cause I could look really stupid so I will let the more seasoned ears take as stab..
 
i won't listen to them myself, due to lack of good speakers on this computer, but i will say this: listen for a lack of super-high frequency information, if you have good ears. LAME automatically does a roll-off at 18Khz. You can't really tell by listening to a song by itself, but when you compare it to the original WAV file, the difference in high-frequency content can sometimes be noted.

other than that, LAME is such a good encoder, that when you're dealing with really high bitrates or VBR, it's basically impossible for anyone to spot actual compression artifacts (in first-generation mp3) on anything but the most expensive headphones and monitors in the world.
 
wave 4 sounds like the original wav (my quess the L2)
wave 5 idem (maybe the L1)

the others are hard...
I've always been using VBR high quality (like you did) mp3 encoding b'coz it comes damn close to the original
I noticed the difference in the high-frq ranges for wave 4 and 5 (they were more present) and could discover more hiss.
MP3 tends to attenuate these frequencies (atleast the xing and lame encoders)

ever heard about mp+?? where you have more presets, also concerning attenuation of frequencies...

I could not hear a big difference in the other waves, which must be the mp3's. although my first 'quess' was that wave6 could be one of the double encoded ones...

nice test by the way ;)
 
Great Idea! I'll take a shot.

#4 24bit wave
#5 16bit wave
#3 & #6 First Gen MP3s
#1 & #2 Second Gen MP3s
 
not many people are trying, oh well, 'ers my go. :D

#1 2nd gen mp3
#2 16 bit mp3
#3 24 bit wav
#4 16 bit wav
#5 24 bit mp3
#6 2nd gen mp3
 
1 - MP3 1st generation
2 - MP3 2nd Generation
3 - MP3 1st Generation
4 - L2 WAV
5 - L1 WAV
6 - MP3 2nd Generation
 
The only problem with this testing method is the answers are given which will influence others while listening or some might just plain ole copy...
 
Sorry, just had a new baby! But now I'm at work, and my "answer" sheet is at home. Now that there are a few responses, I'll post the answers when I get home tonight.
 
here's my shot at this:

4 - 24 bit L2
1 - 2nd generation 24 bit
3 - 2nd generation 16 bit
6 - 16 bit L1
2 - 1st generation 24 bit
5 - 1st generation 16 bit
 
test1 L2/24 bit 2nd generation MP3
test2 L1/16bit 2nd generation MP3
test3 L1/16bit 1st generation MP3
test4 L2/24bit wav
test5 L1/16bit wav
test6 L2/24bit 1st generation MP3


too tired to check your answers... did anyone get it right?
 
I think this would have been easier had there been a full band like rock music or something that is busy with cymbals and covers the whole frequency spectrum.
 
yeah, the cymbals would have made it much easier.
but hell, atleast I could discover the waves... gheh.

oh, was it recorded/edited with pro-tools by the way? had that pro-tools sound I've heard so much.
 
It was recorded in 1974, so not much chance they used Pro Tools... Since there are no fader moves, the only sound you should hear is the sound of the Waves processors...
 
:( I cheated, but I didn't mean too :)

I wanted to do the best comparison I could, so I used "save target" and burned the tracks to CD. I wanted to use my O2Rs DA converters with my best pair of AKG headphones.
As I drug the files into my CD layout I noticed 3 pairs of file sizes, 4&5 were the smallest, 1&2 the largest.
Since jrosenstein and roman had both chosen 4&5, I was pretty sure what was what before I listened.
I didn't exspect to be able to hear a difference between 16 and 24bit. I thought that because the files both came from a 16bit source, they would sound the same.
But I was wrong, you can hear the difference in the reverb tails.

I feel I could have picked 4&5 without knowing the file sizes, but I don't think I could have sorted the MP3s.

Nice test charger, I learned stuff:)
 
You must log in or register to reply here.
Back
Top