Fun listening test - MP3 vs. CD

I put this online so I have to write e-mails about MP3 quality less often...


I've set up a little listening test, using a 30-second excerpt from one of our new recordings (Schubert's Piano Trio No. 1, Fournier - Janigro - Badura-Skoda, Westminster 1953). This is downloadable in three files:

1 - The CD-quality 16-bit 44.1k WAV file

2 - The LAME-extreme encoded MP3 file

3 - A WAV file which includes both of the above - with the MP3 cut into the middle of the WAV file with seamless joins. Again this is at full CD resolution.

The point of the test is to download the latter, play it back on your most analytical audio equipment, and see if you can hear the change from CD to MP3 and back. It is instant (so no clever cross-fading). The two source-files for this final file are online for reference purposes, though if you have the facility you can of course A/B these as well.

I'm quite happy to state that I cannot hear the joins. I can see them on-screen, but only by using visual means can I identify where the sound switches from CD-quality to MP3-quality and back to CD-quality.

Have a go and see how you get on:

http://www.pristineaudiodirect.com/More/HowTo.html


The idea of this is that it's just for fun - I've no desire to get into heated debates with the audiophile community...

Directly comparing the mp3 sample to the wav sample I can notice a difference in the high end with the mp3 sample having a less clear and fine high end than the wav sample.  However listening to the 3rd sample I can't tell where it switches.

I've done a sample like this in the past where the file switches back and forth between wav and mp3 several times.  In that sample I used todays commercial pop music and it was easier to tell than this sample was.  The mp3 was also encoded using lame, with vbr settings.

I find pop easier to use for MP3 discrimination as well, because the most audible signs IMO are in the highs - that "swishiness" on cymbals & such.  There's usually more of that in pop, jazz, etc.

It also depends on bitrate; the "extreme" setting is pretty high, with many frames at MP3's max - 320k.

I cannot hear it switch in the third file. Sometimes you think you hear it, but then, no, I cannot be certain and say "here it is".

1st violin motif at start is MP3 as soon as it is repeated a few bars in - yes/no?

Not sure when it returns to non-MP3 as it seems quite cloudy to me from there all the way to the end. I'd need to listen to the latter section a bit more. BTW, I only listened to the mixed track, not to the others at all.

The idea behind the exercise is to prove to potential purchasers of our historic classical music that they're not missing out on anything they can really hear if they buy downloads instead of CDs. I'm getting bored of answering e-mails which start "I've been told that MP3 really degrades the sound quality..."

If you open the WAV file in Audition and switch to SFE the difference is clear. I'd rather offer these files at maximum quality if I'm going to try and persuade people to jump into the relative unknown of classical MP3s, though in general the music restored from 78s is offered at lower bitrates. I'll probably add a demo of this in the next few days, but expect much the same result.

and now for an MP4 (AC3) test?  
I think sometimes the mp4 sounds better.  
Maybe I like the distortions added.

Why not flac? It would lower the bandwidth and put doubts to rest.

and now for an MP4 (AC3) test?  
I think sometimes the mp4 sounds better.  
Maybe I like the distortions added.

That's a distinct possibility and an astute observation.  And it speaks to why ABX testing is key; one may "sound better" but still be less accurate.

As an aside, I thought AC3 was earlier (and not the same) as MP4.  AAC is an MP4 flavor, IIRC.

Why not flac? It would lower the bandwidth and put doubts to rest.

Or WMA lossless.  EVERYONE can play that, right?

2 - The LAME-extreme encoded MP3 file

This is not telling much, as you have not elaborated on what lame version you have used.  There are various audio quality differences between the versions.

The recommended version is Lame 3.97 beta 2 found here:

http://www.rarewares.org/mp3.html

and has been tuned to be transparent an estimated 98% of the time (using the "-V 2 --vbr-new" profile)

it would be more interesting to try this profile before going to the extreme profile on this version.

Why not flac? It would lower the bandwidth and put doubts to rest.

Or WMA lossless.  EVERYONE can play that, right?

Too big for me on a lowly 64k ISDN link running the business! If you can't hear the joins in the file posted, what's the need? Some of our customers are confused enough as it is!

Why not flac? It would lower the bandwidth and put doubts to rest.

Or WMA lossless.  EVERYONE can play that, right?

Too big for me on a lowly 64k ISDN link running the business! If you can't hear the joins in the file posted, what's the need? Some of our customers are confused enough as it is!

Point taken.  I forget, whilst I'm coveting others' 2Mb connections, that I have a speedy connection in comparison to ISDN or dialup!

Or WMA lossless. EVERYONE can play that, right?

Well, maybe they can until the next update of whatever MS fancies.

Or WMA lossless. EVERYONE can play that, right?

Well, maybe they can until the next update of whatever MS fancies.

AFAIK, no media player update has ever broken playability of older WMA files.  That's no guarantee it will never happen, of course, but I don't think that's too likely.  No more so than any other compression format, anyway.