An audio test by me

Petrie. · **on:** April 15, 2008, 10:49:05 AM

This is probably one of the harder audio tests you'll take that you can find online. The goal is to tell which of three files is a 130kbps VBR MP3 file, a 220kbps VBR MP3 file, and finally, an uncompressed PCM file (1,411kbps). There are three different music types represented: pop, orchestral (loud), and orchestral (soft).

Time to give a low down on the file clips. Lossless means how well each clip compressed in a format where no audio quality is lost.

"beat"
Album: Hairspray, Amanda Bynes, 2007
Lossless: 1,170kbps (83%, does not compress well)

"finalstruggle"
Album: Watership Down, Angela Morley, 1978
Lossless: 752kbps (53%, compresses reasonably well)

"worldbeneath"
Album: Dinotopia, Trevor Jones, 2002
Lossless: 495kbps (35%, compresses extremely well)

So what does all this mean? In theory, one song should be relatively easy to pass, or tell which of the three samples is which, one should be relatively hard, and the final should be next to impossible. Mind you, MP3 does not work like a lossless encoding because it guesses what you won't be able to hear and again in theory, will not miss. Its gotten very good as of late, and this test will prove just that.

So remember, you're trying to guess which is uncompressed (or unaltered), which is a high quality mp3 file, and which is a low quality mp3 file with each music sample. Be sure to read "answers.txt" to see how you did. Good luck.

Click here for the files.

PS - If you can't tell the difference between any of them, don't worry...neither can 80% of the world.

When you encode files properly, you'll be amazed how hard it is.

Petrie. · **Reply #1 on:** April 15, 2008, 01:15:15 PM

I'll just post this here so people don't feel too bad. I did a blind ABX test and honestly I failed my own test using headphones.

I only could definitely tell 1/9 files...the rest I couldn't do it or ended up guessing incorrectly. I haven't tested speakers, but I will do it later in the day.

If anyone wants to offer a sample of metal music (say Dragonforce) I'm game for trying it. Next to accoustic instruments, cymbals are very hard to encode.

However, this test shows that if you encode mp3's properly (meaning use variable bit rates, and not a constant one), its very hard to spot a difference or its not large enough that you'd care.

Petrie. · **Reply #2 on:** April 16, 2008, 01:26:15 PM

Last update from me (and I think I scared everyone away from trying once I said I failed

) but I tried with speakers...same result. I could get up to three files, but it was hard as hell to do.

action9000 · **Reply #3 on:** April 17, 2008, 08:20:58 PM

Whew, yeah this is a tough one!

I got the low-quality file on Beat fairly easily...though the rest of the files you uploaded were almost guesswork with slight hunches toward one file or another...got 2 right in total.

Petrie. · **Reply #4 on:** April 18, 2008, 08:13:17 AM

Quote from: action9000,Apr 17 2008 on 07:20 PM

Whew, yeah this is a tough one!

I got the low-quality file on Beat fairly easily...though the rest of the files you uploaded were almost guesswork with slight hunches toward one file or another...got 2 right in total.

Yeah, that's the only one I could pick out. Its been reported that jazz purists won't use mp3 because of trumpets, so that's what I was hoping to see with the finalstruggle clip. The difference didn't appear, even with the trumpet in the far right channel.

Its big for anyone who likes the movie scores and worries about excessive compression....you'll probably never notice it. Its hard for me to accept it myself.