[PhilS]

Quote:

Originally Posted by hybris /img/forum/go_quote.gif Anyone that can hear "massive differences" between 192kbps and 320kbps (assuming a good quality encoder) . . . .

Why do the differences have to be "massive"? If there's a difference one can hear, isn't that enough? Isn't it up to the individual to determine, assuming arguendo that they hear a difference, what to do about it (e.g., record in lossless and lose space, etc.)?

Quote:

Originally Posted by hybris /img/forum/go_quote.gif Bring home a few friends, encode say 30 seconds of a few tracks in 128, 192 and 320kbps mp3 with the lame encoder. Then you put the same track (but with different bitrate) in a playlist several times, say 15 times (each bitrate repeated five times) in random order. Now the rest of you guys who don't know the playlist order try to rate the sound quality, or even guess which bitrate are playing for each of the 15 tracks. I can tell you right now that your results will be completely random, perhaps you're able to place the 128kbps track a few times.

That test is not very good, IMO. I'll bet lots of people have trouble hearing differences under those conditions. I can see about five different obvious flaws in it -- bascially it doesn't mirror normal listening conditions or the conditions under which people might be able to hear differences.

I think I'll have two songs that I listen to regularly recorded to my MP3 player at 128 (just to start there). Then I'll listen regularly to those two songs over the next week or so in lossless. Then I'll have my wife play the songs at 128 or lossless (her choice) and see if I can tell the difference. Isn't that a fair test -- one that doesn't involve listening to (1) songs you don't know on (2) other people's equipment or equipment you don't regularly use and (3) flipping back and forth rapidly between songs and (4) not trying to determine just differences but being forced to identify which bit rate is which?

 

[ILikeMusic]

Quote:

Originally Posted by PhilS /img/forum/go_quote.gif I think I'll have two songs that I listen to regularly recorded to my MP3 player at 128 (just to start there). Then I'll listen regularly to those two songs over the next week or so in lossless. Then I'll have my wife play the songs at 128 or lossless (her choice) and see if I can tell the difference. Isn't that a fair test -- one that doesn't involve listening to (1) songs you don't know on (2) other people's equipment or equipment you don't regularly use and (3) flipping back and forth rapidly between songs and (4) not trying to determine just differences but being forced to identify which bit rate is which?

I don't see anything wrong with that as an informal test, assuming that you are playing back the lossy and lossless files on the same player and you do the test enough times to eliminate the chance of guessing (and getting it right two or three times in a row isn't enough, that still carries a very high possibility that you were just guessing correctly.)

 

[PhilS]

Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif I don't see anything wrong with that as an informal test, assuming that you are playing back the lossy and lossless files on the same player and you do the test enough times to eliminate the chance of guessing (and getting it right two or three times in a row isn't enough, that still carries a very high possibility that you were just guessing correctly.)

Yes, I'll use the same player and same phones, etc., and keep my phones in my ears so that the only thing that is changing is a different track being selected by my wife or whomever is selecting the track. I also agree about the number of trials.

 

[ChroniCali]

192kbps and 320kbps definitely has a pretty significant difference IMO.

 

[hybris]

Quote:

Originally Posted by PhilS /img/forum/go_quote.gif Why do the differences have to be "massive"? If there's a difference one can hear, isn't that enough? Isn't it up to the individual to determine, assuming arguendo that they hear a difference, what to do about it (e.g., record in lossless and lose space, etc.)?

It doesn't have to be massive, if it's even audible, that's argument enough to go for the higher bitrate. I was commenting on how people (which several have done in this thread) very often exaggerate how much difference there is. If you have to spend a week to prepare in order to be able to hear the difference between 128kbps and lossless, the difference between 192kbps and 320kbps can hardly be classified as "massive", or even significant (as some people claim) ?


Quote:

Originally Posted by PhilS /img/forum/go_quote.gif I think I'll have two songs that I listen to regularly recorded to my MP3 player at 128 (just to start there). Then I'll listen regularly to those two songs over the next week or so in lossless. Then I'll have my wife play the songs at 128 or lossless (her choice) and see if I can tell the difference. Isn't that a fair test -- one that doesn't involve listening to (1) songs you don't know on (2) other people's equipment or equipment you don't regularly use and (3) flipping back and forth rapidly between songs and (4) not trying to determine just differences but being forced to identify which bit rate is which?

Sounds like a fair test, and I'm confident you will be able to hear the difference. But the subject of the thread was 192kbps vs 320kbps, not 128kbps vs lossless.

 

[insyte]

Personally, you have to really concentrate to find the difference, but there is a difference

 

[TubeStack]

I've played the same songs back-to-back, comparing both rates (192 and 320), not knowing which one I was necessarily going to hear, and instantly heard huge differences between the two rates, especially when played over a decent home stereo system (not headphones).

It was most apparent on loud, guitar-heavy material (Metallica's Death Magnetic, ZZ Top's Chrome, Smoke & BBQ set).

Besides an improvement in detail, clarity, bottom-end, and overall warmth that occurred when the songs were encoded at 320, a general annoying sheen (ie. non-warm, brittle) that encased the music encoded at 192 or lower, disappeared when encoded at the higher rate.

Again, this was not over headphones, but a home system, and the difference was clearly apparent.

No need for others to get angry if they don't hear a difference in their own listening experiences and environments, who cares, really.

 

[PhilS]

Quote:

Originally Posted by hybris /img/forum/go_quote.gif Sounds like a fair test, and I'm confident you will be able to hear the difference. But the subject of the thread was 192kbps vs 320kbps, not 128kbps vs lossless.

Yes, but there are suggestions, if not outright assertions, in this thread that people cannot tell the difference between any of the formats, and I thought I would start with what is most likely to be the largest difference, and then go from there if it's worthwhile. Moreover, part of my interest is not just to test the difference, but also to explore some issues about testing itself.

 

[hybris]

Quote:

Originally Posted by TubeStack /img/forum/go_quote.gif I've played the same songs back-to-back, comparing both rates (192 and 320), not knowing which one I was necessarily going to hear, and instantly heard huge differences between the two rates, especially when played over a decent home stereo system (not headphones). It was most apparent on loud, guitar-heavy material (Metallica's Death Magnetic, ZZ Top's Chrome, Smoke & BBQ set). Besides an improvement in detail, clarity, bottom-end, and overall warmth that occurred when the songs were encoded at 320, a general annoying sheen (ie. non-warm, brittle) that encased the music encoded at 192 or lower, disappeared when encoded at the higher rate. Again, this was not over headphones, but a home system, and the difference was clearly apparent.

I'm impressed you found any detail, clarity or overall warmth whatever the bitrate on Death Magnetic. The sound is compressed so hard it's actually distorting. If you found a clear difference on 192 and 320kbps on that album I think you need to get a new mp3 encoder.

 

[hybris]

Quote:

Originally Posted by PhilS /img/forum/go_quote.gif Yes, but there are suggestions, if not outright assertions, in this thread that people cannot tell the difference between any of the formats, and I thought I would start with what is most likely to be the largest difference, and then go from there if it's worthwhile. Moreover, part of my interest is not just to test the difference, but also to explore some issues about testing itself.

Go ahead, we're always interested in tests and its results.

 

[Eleazar]

I think the best thing to do would be to hear the artifacts that compressed audio introduces. It's important to train your ears so you know what you are listening for. I would encode a song that you are very familiar with and that preferably has a lot of frequency range, high highs and low lows. I trance song with someone like Sarah McLachlan's voice would be good. You could also try two different well known songs. If you have any choral music or opera you listen to, pick a part with women's soprano. Another kind of music that seems to very audibly produce sound compression artifacts for me personally is Gregorian chant, whether male or female, especially when they sing in unison loudly and the solemn echoing begins.

I listen on Pandora.com which does 128 AAC and it is surprisingly awful sometimes.

Once you get an ear for what poorly compressed sound is, you won't have trouble picking it out from higher kbps encodings. Do realize that on some music it will be more prominent than others, it can actually be difficult to tell a difference with certain music. For mp3's I use hydrogenaudio's suggested VBR settings and the latest stable release of LAME.

 

[PhilS]

Quote:

Originally Posted by Eleazar /img/forum/go_quote.gif I think the best thing to do would be to hear the artifacts that compressed audio introduces. It's important to train your ears so you know what you are listening for.

Yes, this is exactly what I'm trying to do. I haven't completed everything yet, as I got interrupted, but my very preliminary reactions were as follows:

1. At first, the differences between 128 and lossless were not as great as I anticipated.

2. There really is a "training" issue. As you listen more and more, certain differences between the formats appear, and it's as if each one has a signature that is recognizable once you become familiar with it.

3. The ability to listen and train sighted seems to me to be a crucial element. Having trained myself knowing what I was listening to has made it much easier to identify the differences blind. Although I haven't completed all the blind testing yet, I think this it is quite apparent that this makes a difference. I seem to recall that this phenomenon was pointed out in the "Null Hypothesis" thread, and I would suggest that this is one of the deficiencies in some of the DBT's that have been undertaken.

I'll report back more when I have a chance to do some more blind testing.

 

[ILikeMusic]

Quote:

Originally Posted by PhilS /img/forum/go_quote.gif 3. The ability to listen and train sighted seems to me to be a crucial element. Having trained myself knowing what I was listening to has made it much easier to identify the differences blind. Although I haven't completed all the blind testing yet, I think this it is quite apparent that this makes a difference. I seem to recall that this phenomenon was pointed out in the "Null Hypothesis" thread, and I would suggest that this is one of the deficiencies in some of the DBT's that have been undertaken.

All of the ABX software I've seen does account for that factor, i.e. you can listen to an identified 'A' or 'B' as many times as you like before listening to the unknown 'X'.

 

[PhilS]

Quote:

Originally Posted by ILikeMusic /img/forum/go_quote.gif All of the ABX software I've seen does account for that factor, i.e. you can listen to an identified 'A' or 'B' as many times as you like before listening to the unknown 'X'.

The software may do this, but not all of the DBT's I've heard or read about do this (not just talking about DBT's in this area).

 

[AtomikPi]

I have ABX'ed in Foobar all the way from 128 CBR to 320 CBR vs. flac. Honestly, although I could successfully ABX them all (3/3 each - cymbal crashes in "For Whom the Bell Tolls" made this relatively simple) I honestly think even 128 CBR, except for a few minor digital artifacts sounds 95% as good as FLAC. The successively higher bit rates (128, V2, V0) each sounded slightly closer to flac but each sounded extremely similar to each other. I'm still planning on ripping to V0 to be safe, but I wouldn't say you're missing too much with even 128.