Good source to hear the difference between lossless and mp3?
Oct 8, 2004 at 12:38 AM Post #31 of 47
Quote:

Originally Posted by shimage
edit2: ok, so since I don't want to sound like I'm bragging (since I'm really not that good at it), I'm going to say that it's not only possible, but not difficult to ABX lame -aps from lossless even with pretty cruddy equipment (read: SBLive! and HD490s), particularly with test samples off the lame website. As Publius mentioned though, if you can't hear differences, just be happy and stop asking silly questions.

edit:

Oh.... I see how it is... actually, you generally post stuff as losslessly compressed (or uncompressed) decoded files, so that the testers can't tell which is which. Of course, if they notice that they're bit for bit copies of each other, well...



Where exactly are these test samples on LAME's website? How do I conduct ABX testing with foobar?
 
Oct 8, 2004 at 12:47 AM Post #32 of 47
Quote:

Originally Posted by Publius
Oh, and 16/17 ABX on the two tracks, due to a stereo collapse within the first half second. (I was getting too smug on the one failure.) I strongly suspect that the low-volume highly directional audio on the first few seconds is one of the worst case scenarios for 128kbps and I'm not sure if I could ABX deeper into the track. ER-4S on mobo audio.


Now that you point out the stereo flaw it's easy to pick these two apart ( 7 out of 7 trials in foobar ABX ) and that's with just a chaintech 710, cmoy and KSC-55s.

I've never have been able to ABX lame VBR from lossless though, but haven't tried with my latest amp & headphone upgrades.
 
Oct 8, 2004 at 9:52 AM Post #34 of 47
Quote:

Originally Posted by Distroyed
Is there a better way to abx in foobar besides queuing the same two files up a few times and hitting random?


Yes. Select the files on playlist, right-click and select "ABX two tracks". Just remember to configure the ABX comparator from the Preferences (DSP on/off, replaygain on/off).
 
Oct 8, 2004 at 10:03 AM Post #35 of 47
Quote:

Originally Posted by donovansmith
MP3 suffers from poor transient response, the attack and decay just sound a bit off compared to the original, and its joint stereo modes noticably collapse the soundstage. I generally can pick out whether a track is joint stereo encoded or really encoded in stereo throughout with pretty good accuracy, even though headphones like the Koss KTX-Pro/RadioShack Pro-35s.


Joint-stereo is a good thing when used by an encoder which does it right (ie. no bugs in implementation that ruins stereo image). LAME is one of the encoders and tbh we shouldn't even discuss about any other MP3 encoders. And only the tuned preset modes as someone already pointed out.
 
Oct 8, 2004 at 11:25 AM Post #36 of 47
Quote:

Originally Posted by breez
Yes. Select the files on playlist, right-click and select "ABX two tracks". Just remember to configure the ABX comparator from the Preferences (DSP on/off, replaygain on/off).



Oh, it's a plugin that doesnt come with the special foobar version. You can get it here
 
Oct 8, 2004 at 1:27 PM Post #37 of 47
Quote:

Originally Posted by Distroyed
Oh, it's a plugin that doesnt come with the special foobar version. You can get it here


It DOES come with the special installer. Unfortunately the special installer defaults most things to off, which kinda defeats its "special" purpose.
blink.gif
Just make sure to enable it at install time. Of course you can just copy the dll later on.

As for equipment to distinguish 128/192/WAV, Senn 280 out an Audigy2 (w 48Khz foobar resampling) are more than sufficient.
 
Oct 8, 2004 at 5:11 PM Post #38 of 47
Quote:

Originally Posted by astro
tongue.gif

It's too hard to fake it, it's too easy to tell just by looking at the size of the file.

I guess I do lose a lot of the details from the white noise my computer fans produce, despite them running on such a low RPM to begin with.
My HD497 picks up background noise very easily being open.



2 things that might be worth doing ...

1. reencode them back to wav so they are the same size, and then see if you can tell.

2. encode from wav -> mp3 -> wav -> mp3 for several generations and see how bad they get. This might convince you that, while one generation of compression is below your hearing threshold, quality loss is real, and others night not be hearing only placebo effects.
 
Oct 8, 2004 at 7:09 PM Post #39 of 47
I abx'd the two files 10/10 on the system in my signature and I am pretty sure I could do so on less expensive equipment. As for 192 vs lossless, I tend to be able to abx those as well but it takes longer. With the two given samples I performed ten trials with no failures in about 90 seconds.
 
Oct 8, 2004 at 7:58 PM Post #40 of 47
Quote:

Originally Posted by Distroyed
So you'd consider the Dac1 to be cruddy equipment?


When did I say that?

astro: I thought that there would be some samples on the lame website, but apparently I'm full of it. There are other places you can go to get them, though; you just have to follow your nose. I know that ff123 hosts a large collection of test samples. They tend to be samples from real music that people listen to that also tend to difficust to encode well for whatever reason. In the case of castanets, this reason is that mp3 has a serious problem with pre-echo which is totally absent in mpc.
 
Oct 8, 2004 at 9:04 PM Post #41 of 47
A question about the high frequencies...
The TV in my house makes a high frequency ringing sound whenever they're on; however my mom, dad and 12-year-old sister say they can't hear it. But I can hear it and know that the TV is on even before I walk into the living room. Of course, the speakers volume is turned all the way down.
Around what frequency would that be?
Do I have a better chance at hearing the mp3 differences than my family members?
(regardless, I guess I don't know what to listen to but I did abchr-java test on my mac and I have a hard time telling 112k CBR mp3 from original... on a Senn HD600 / firewire audiophile... 96k is obvious because the pitch changed)

Quote:

Originally Posted by dip16dac
It is easier to see with cool edit frequency analysis that the 128 one only goes up to about 16 khz were as the 192 one has 18 khz stuff. It depends on how good your hearing is to be able to tell the difference since most equipment can handle both.


 
Oct 8, 2004 at 9:41 PM Post #43 of 47
Quote:

Originally Posted by es2mac
A question about the high frequencies...
The TV in my house makes a high frequency ringing sound whenever they're on; however my mom, dad and 12-year-old sister say they can't hear it. But I can hear it and know that the TV is on even before I walk into the living room. Of course, the speakers volume is turned all the way down.
Around what frequency would that be?



About 15.625KHz. Older people have declined high frequency hearing, but most 12-year-olds would probably be able to hear that frequency.

You can test your hearing with your computer and a tone generator program (foobar2000 has one, "Add location" and enter tone://<insert here the freq. in Hz>). Of course you are limited by the headphones' frequency response (a decent sound card should have flat response).
 
Oct 8, 2004 at 9:48 PM Post #44 of 47
I just did ABX with castanet.wav that I found by digging through the lame site, which brought me to another site that had it.
I did the test 3 times, 128kbps vs wav, and I got it right all three times! I had to use waveout, because abx wouldn't work with kernel streaming. But I can hear the difference. There isn't a distinct difference, but the echoes on the wav seems a tad more present.

I tried ABX with a few other 192/128kbps mp3s, and I failed miserably.
 
Oct 8, 2004 at 10:08 PM Post #45 of 47
Quote:

Originally Posted by Mr.Radar
For ABX'ing an Etymotic ER-4x paired with an EMU 1212m and a very detailed amp would probably be the best.


dac1 and ER-4S is pretty good, I can hear up to 17.5khz too so I could probably be decent at abx'ing if I cared. What I want is a simple abx program for the karma (ie I just want it to give me right/wrong and I'll do the statistics myself), since what I can abx with the dac1 is meaningless in terms of what I should put on my karma. I guess I could assume that the analog out on the chaintech is in the same ballpark as the karma but I don't actually have any idea if that is valid.
 

Users who are viewing this thread

Back
Top