[Music Alchemist]

  What you're describing doesn't reflect how lossy works. There isn't an overall distortion at medium bitrates. It's momentary artifacting in specific parts of the music caused by too little bandwidth available to render the sound. Either the sound can be rendered or it goes splat for a second. As you raise the bitrate, the artifacts become fewer and fewer until they disappear entirely.
 
Overall coloration of the sound or higher distortion overall at high bitrates doesn't sound like lossy artifacting. It sounds like expectation bias, problems with level matching, too much time between samples (auditory memory), or some sort of distortion being added by the equipment itself.
 
It is much better to match levels using measurements. You can get a ballpark idea by balancing by ear, but if the differences seem subtle to you, it's entirely possible that the levels aren't balanced perfectly and there really isn't a difference at all. That's how bias works. Everyone is subject to it.
 
I don't know about other encoding programs, but iTunes drops the volume a dB or two when it encodes. I think it's trying to prevent clipping in hot mastered recordings as it encodes.

 
That's partially what I meant. Sorry if my wording wasn't specific enough. Differences were perceived in specific parts, as well as overall for some songs with repetitive sections. Dull/muddy/harsh versus clear and punchy (in comparison), for example. I don't see how that would be caused by the things you listed.
 
However, if you go to much lower bit rates, there is more noticeable distortion everywhere.
 
I use dBpoweramp. (Forget iTunes!)

 

[Chris J]

And you have just demonstrated the blindspot by defending science with that statement. Because audio equipment contributes to an aesthetic experience, useful research would typically need to be comprised of both quantitative and qualitative, where the qualitative would take that into account. And to design qualitative research methods and do the resulting analysis of the data, one would have to understand theories from social science and the humanities, which a lot of researchers in the hard science would not have a clue about, and thus why they feel it can be "deliberately eliminated." To rely only on the quantitative is sort of like thinking one can analyze cake and compare it to pie by analyzing it's chemical composition and other physical characteristics without doing research into what makes it taste good and having people taste it.

But I guess it depends on whether your end goal for research is how the equipment measures or which people will like better

NVH anyone?

 

[Music Alchemist]

NVH anyone?

 
I really hope resonance/vibration control doesn't get added to the debate...

 

[Roly1650]

Music Alchemist, if you don't feel comfortable in being able to level match properly you can always run both files through Audio Diff Maker. You'll have to bounce the lossy files up to the same bit depth/sample rate as the original file, but that's pretty simple. Then you'll get the difference channel and you'll be able to hear, (or not) what went missing. Run a test with the original files on both sides first, this will tell you how good your sound card/system is, you should get a match way way down there!

 

[Music Alchemist]

Music Alchemist, if you don't feel comfortable in being able to level match properly you can always run both files through Audio Diff Maker. You'll have to bounce the lossy files up to the same bit depth/sample rate as the original file, but that's pretty simple. Then you'll get the difference channel and you'll be able to hear, (or not) what went missing. Run a test with the original files on both sides first, this will tell you how good your sound card/system is, you should get a match way way down there!

 
Hmm, Audio Alchemist. Sounds catchy!
 
I've used that software before. However, I don't have the equipment necessary to do a test with it and the iPod, and I won't be doing any more tests in the near future regardless. I'll be sure to consult with you and other experienced individuals when the time comes.

 

[Roly1650]

Hmm, Audio Alchemist. Sounds catchy!

I've used that software before. However, I don't have the equipment necessary to do a test with it and the iPod, and I won't be doing any more tests in the near future regardless. I'll be sure to consult with you and other experienced individuals when the time comes.

Ah, spotted that did you? I'll blame on senility.

 

[Chris J]

I really hope resonance/vibration control doesn't get added to the debate...

Threw it in there because it encompasses subjective and objective measurement.

 

[bigshot]

I'm not familiar with it, but it may be a problem with your encoder or sound card. Compression artifacts aren't particularly subtle. They don't sound like muffling or harshness above 192. They are usually big digital splats or weird outer space gurgling that just happen here and there for a short moment, and then the sound is perfect again. They stick out like a sore thumb over acoustic instruments. The higher the bitrate, the rarer they become until the entire track is artifact free. The best codec I've found is AAC. At 256 it is completely transparent. MP3 LAME is good too, but it usually needs to go to 320 to be completely artifact free. Fraunhofer (plain vanilla) MP3 isn't as good. It still has problems with some problematic tracks at 320. Rare, but it does happen sometimes with that codec.
 
The thing about compressed audio is that it's not all created equal. There are codecs, and there are bitrates, and there are workarounds to distribute the bitrate where it's needed most (VBR). All of those can make a big difference. If you find the sweet spot, it can be audibly identical to lossless.

 

[Music Alchemist]

  I'm not familiar with it, but it may be a problem with your encoder or sound card. Compression artifacts aren't particularly subtle. They don't sound like muffling or harshness above 192. They are usually big digital splats or weird outer space gurgling that just happen here and there for a short moment, and then the sound is perfect again. They stick out like a sore thumb over acoustic instruments. The higher the bitrate, the rarer they become until the entire track is artifact free. The best codec I've found is AAC. At 256 it is completely transparent. MP3 LAME is good too, but it usually needs to go to 320 to be completely artifact free. Fraunhofer (plain vanilla) MP3 isn't as good. It still has problems with some problematic tracks at 320. Rare, but it does happen sometimes with that codec.
 
The thing about compressed audio is that it's not all created equal. There are codecs, and there are bitrates, and there are workarounds to distribute the bitrate where it's needed most (VBR). All of those can make a big difference. If you find the sweet spot, it can be audibly identical to lossless.

 
Many consider dBpoweramp to be the best software for converting files, ripping CDs, etc. (Not that I'm endorsing it. I'm just saying there is not an encoder problem.)
 
I'm not so sure about some of that. I will have to conduct further tests in the future.
 
For reference, if you didn't see before in my past posts, my private tests were done on an iPod Classic with lossless, QuickTime tVBR AAC, and MP3 files encoded with the highest quality settings.
 
Come to think of it, though, I did compare an old 128 kbps CD rip to lossless and that was one with especially noticeable differences.
 
I also have some downloaded 320 kbps MP3 files that are very distorted compared to the lossless versions, but they may just be faulty encodings.

 

[bigshot]

There really isn't any reason to use VBR if you are encoding at 320.
 
It's like anything else... if you know how it works, you can get the best out of it.

 

[SilverEars]

Is AAC the most compact compression scheme that retains the most information right now?

 

[bigshot]

From my comparisons, yes. Others are better at lower bit rates, but if you are looking for perfect transparency, AAC can do that at 256 VBR, where LAME requires 320 CBR.

 

[cjl]

  There really isn't any reason to use VBR if you are encoding at 320.
 
It's like anything else... if you know how it works, you can get the best out of it.

For MP3, assuming I remember how the format works, there is no such thing as 320 VBR. 320 is the max supported bit rate, so 320 is always CBR. 240(ish) VBR can be basically identical from a perceptual standpoint though.

 

[bigshot]

Yes, exactly. I read somewhere there are some encoders that allow you to select 320 VBR, but I think that can only reduce bit rate, not increase it.

 

[Music Alchemist]

  There really isn't any reason to use VBR if you are encoding at 320.
 
It's like anything else... if you know how it works, you can get the best out of it.

 

  For MP3, assuming I remember how the format works, there is no such thing as 320 VBR. 320 is the max supported bit rate, so 320 is always CBR. 240(ish) VBR can be basically identical from a perceptual standpoint though.

 

  Yes, exactly. I read somewhere there are some encoders that allow you to select 320 VBR, but I think that can only reduce bit rate, not increase it.

 
I used a custom command line encoder for QuickTime AAC that someone made for me, but basically, tVBR AAC creates a variable bit rate depending on the content in the music. You don't set a bit rate with VBR; you set a quality setting. Off the top of my head, I think it can convert files with a bit rate anywhere from 200 to 400 kbps when using the highest quality setting of Q127. I don't remember which setting I used for MP3 other than that it was 320 kbps. MP3 can be VBR, ABR, and CBR. If you use the highest MP3 VBR setting of -V 0, the file can be a lower bit rate than normal, estimated at 240 kbps. I just tested it and the converted file is 251 kbps. 320 kbps can be both ABR and CBR.