Should I get over my prejudice/ignorance/snobbery for WAV and change to mp3?

[dj matt blaze]

For a quick preview…my results show that I can hear the difference.

:eyeroll: why am I not surprised at that result?

[Xonetacular]

I'll make it even easier: http://forum.djtechtools.com/showthr...126#post474126

I have results for xone, matt blaze, 3heads, and myself. I'd like to get a few more so I can do a different analysis on it.

For a quick preview…my results show that I can hear the difference. Matt blaze already said his results showed he couldn't, but I'll let the others share their own.

I'd like to get some more people to do the test before I comment openly about a few things I noticed. I'm kind of amazed that the scores for one of the songs was so low…I got 10/10 on the one that I thought I did. But after hearing the demos, something jumped out at me that was like a giant neon sign. A really subtle one, but still.

xone suggested posting it on other forums. So, I'll do that and keep a log of the results (i'll have names but won't publish them). DJF, ALDJF, and GearSlutz are on the list. Any other suggestions?

I can't wait to see the gearslutz thread- make the thread title a challenge to get people riled up over it and do it in the electronic music production subforum, I can guarantee it will be a massive thread and full of entertainment, threatened egos, and interesting results.

As for my results, I can't really tell the difference- didn't expect to be able to but I was very surprised with just how little of a difference there was. During the intro I could hear some subtle things when introducing .wav and .mp3 but once that actual test started it all seemed to go out the window.

[mostapha]

:eyeroll: why am I not surprised at that result?

'cuz I'm awesome.

I can't wait to see the gearslutz thread- make the thread title a challenge to get people riled up over it and do it in the electronic music production subforum, I can guarantee it will be a massive thread and full of entertainment, threatened egos, and interesting results.

Good idea. I was wondering where it'd fit in. It's up. If you see it go off the front page, I'd appreciate a bump.

As for my results, I can't really tell the difference- didn't expect to be able to but I was very surprised with just how little of a difference there was. During the intro I could hear some subtle things when introducing .wav and .mp3 but once that actual test started it all seemed to go out the window.

That actually happened to me on a few of them, coincidentally I got a 4 and a couple 6s on the ones I remembered being that way.

The only reason my results look different so far is that I got a 10/10, two 9/10, and an 8/10. The 10/10 one was very obvious to me. One of them has a sound fading basically enough so that the noise floor becomes audible…and the noise floor sounded different between the two files. After that, it was just a matter of remembering which was which. I'm pretty sure the 8 and 9s were reverb tails on hi hats, but I don't remember them as specifically.

I will say that that I've never been able to hear that at a venue. I wouldn't expect anyone to hear it. If nothing else, the earplugs you should be wearing and/or the crowd noise should easily cover up the noise floor in the file…assuming nothing else in the processing chain does.

Anyway, I committed my answers via PM as I said I would. And after he submits, I'll share my google spreadsheet with him so he can check it if he likes (been keeping people's results in different files for convenience). But, my average was 7.3 with a standard deviation of 1.767 for a T-score of 4.12, which is well above the T critical value of 1.833 for a one-tailed test which gives p<.05.

So, in this test, I can hear a difference, though it is very subtle.

Without the ABX structure, I couldn't do it. And it might have colored my results that I was familiar with all of the music. The songs weren't chosen because they showed things that mp3 compression screwed up, but I knew what to listen for in the tracks before the reference passes because I'd heard all of them (as lossless) a dozen times before if not more. I'm not completely convinced that the results would be the same if it was all music I hadn't heard before unless it was a clip that accentuated one of the obvious things (like in the track I got 10/10 on and had no questions about).

Doing deeper analysis of the files was weird, though. The one I randomly chose to look at closer, I did a null test on…and it only nulled to -16dB.

For those that don't know what that means, if you play two identical audio files with the phase reversed in one of them, the result should just be the noise floor of your system (noise in your mixer, converters, amp, etc.).

For that clip, the null test peaked at -16dB…which means either that logic sucks at snapping audio files to a grid or normalizing files or that the things people can't hear are that loud. I'm beginning to wonder if what I was hearing was flaws in Logic's normalizing. The weirder thing was looking at a frequency analyzer of it. The vast majority of what was left over was in the low mids, bass, and sub-bass. There was basically nothing above 400Hz IIRC with a bump around 200 and a lot below 100, which my headphones probably weren't playing much of anyway.

Strangely enough, that also means that it might be more apparent (if you're listening to whatever's in the null file instead of the noise floor) at club volumes. If I'm reading fletcher-munson curves correctly, people are more sensitive to sounds < 400Hz at 100dB SPL than they are at 50dB SPL.

Also, I'm starting to wonder why the normalization was even necessary. Is there any reason why an mp3 (converted back to wav before being imported to Logic) would peak at a different level than the unprocessed wav file? I'm mostly curious about this since i was trying to hear the noise floor in so many of the clips…if it was actually louder in one than the other because of the normalization, that could have given it away without having much to do with the mp3 compression itself.

I know that this test shows that I'm hearing something that apparently other people aren't. But I'm starting to wonder exactly what it was.

It's weird, because I'm one of the ones who can hear a difference. And actually doing the test has just about convinced me that the difference doesn't matter that much.

Did I really put in that much work only to end up proving that I'm right but concluding that it doesn't matter? Weird. I need another drink.

Edit: oh yeah…if you've done the test or are going to, please let me know whether or not you think you know what to listen for (regardless of whether or not you think you did well) and what you did the test on (what headphones/monitors/etc.) so I can play with statistics some more.

I like statistics.

[deevey]

I'll give this a go when the file finally downloads and I have the 45 mins to spare...you need a faster host for the file :P

[djproben]

Here's someone else who did null tests - he was testing FLAC vs WAV (as expected, the null test result was null) but he also tested FLAC vs mp3 and while there is clearly a signal in the null test, it doesn't look very loud: http://www.computeraudiophile.com/bl...experiment-94/

(note that his m4a part of the experiment is invalid as it used different source files, but the WAV/mp3 comparison used the same files).

If you want to listen to the results of a null test, here's one: http://www.noiseaddicts.com/2010/04/...en-wav-vs-mp3/ ... the file definitely has content though it's not much. There are differences but based on this information it is minimal, and the most noticeable differences will be at frequencies that many of us can't even hear (16k here).

Downloading the soundtest file now to see what results I get. I think I agree with the above though; if there is a difference it is *very* subtle, and certainly wouldn't be noticeable in a large noisy venue (and probably not even at a more intimate setting on a nice home system without almost cartoonishly careful listening). I mean seriously, if you're listening for a slight smudge on the attack of a snare sound when you're playing "Last Night a DJ Saved My Life", you're doin' it wrong...

[3heads]

I took the test. Im the guy who said you can't hear a difference and yes there was instances where I was like, "ok this is the mp3" and then the next trial comes on and I'm like "wait, no this is the mp3 and that last 1 was the wav" Then all those years of test taking comes into play with...I just said wav 3 times in a row, this 1 has to be an mp3 and the mind plays tricks. Im not surprised at the difficulty.

Haha, the same phenomena happened to me. I even sometimes changed some stuff around while listening to the next trial - might not have been the smartest move.

Overall the results show I can not reliably tell them apart (didn't expect myself to be able to, though). As for your question mostapha, I don't have a clear idea what to listen for, mostly I tried observing the low end (as that's the thing where the difference between 320kbps and 192kbps mp3s becomes really apparent - I just adapted that observation). For some of the songs I had the impression I could hear a difference during demo-mode, but telling them apart during trials was still mostly guesswork. Other songs sounded virtually the same even when demoed in the beginning.

I did the test on Yamaha Hs80Ms (the rest of the signal chain shouldn't matter I guess, but for completeness' sake: Audio8 -> Rodec MX-180 mk2 -> Hs80M)

[mostapha]

I'll give this a go when the file finally downloads and I have the 45 mins to spare...you need a faster host for the file :P

It's on Georgia Tech's webspace <3 gigabit lan hops away from a top-level router. If it's slow, either it's your downstream or they're throttling it……which means I might have a nasty email waiting for me.

@djproben, I'll look at those. Should be interesting.

seriously, if you're listening for a slight smudge on the attack of a snare sound when you're playing "Last Night a DJ Saved My Life", you're doin' it wrong...

That might be the single most intelligent sentence in this entire thread.

[deevey]

It's on Georgia Tech's webspace <3 gigabit lan hops away from a top-level router. If it's slow, either it's your downstream or they're throttling it……which means I might have a nasty email waiting for me.

Deffo not my side ... although they might throttle due to my location I guess (head-fi blocks me completely!) hopefully you don't have a nasty mail hehe.

[sarasin]

Every sound engineer I have every talked to has a tale of tweaking a song to perfection...only to find that the EQ/effects/whatever were bypassed the whole time. If you find someone who will not admit to that they are "new" or a "liar." It took me about 5 years of tweaking sound systems to catch myself doing that exact thing. I adjusted the !@#% out of a system...only to find out that the whole rack I had been working with was completely out of the loop.

HAHAHA..yep!

This is too true also!

I even catch myself doing it at home....when mixing. Bring in a track...decide its tight as fuck!
Then notice the X Fader is all the way to one side. Thus only 1 track is playing....DOH!

Stoners...

But I have also done it PLENTY while producing...PLENTY!

[3heads]

I even catch myself doing it at home....when mixing. Bring in a track...decide its tight as fuck!
Then notice the X Fader is all the way to one side. Thus only 1 track is playing....DOH!

Stoners...

Haha, happens to absolutely everybody. Was mixing in a tune during a gig (at least I thought I was) until I noticed (after 30seconds or so of careful mixing) that I was pushing up a line-fader on a channel I wasn't even using