[toddbagwell]

I recently conducted an experiment to see if I was wasting my time downloading (I can only get 1.5Mb DSL in my corner of the sticks) and paying the price premium for HiRez music from HDtracks, iTrax and other websites.
I took a selection of 88.2khz and 96khz songs from a few albums
Metallica Black Album


Sam Cooke Portrait of a Legend


Diana Krall Love Scenes and From This Moment On




REM In Time The Best of REM


Jazz at the Pawn Shop


Santana Supernatural


  My methods involved copying a track or two to a new folder then converting each hi-rez file into a compressed version.
  I used dbpoweramp, MP3tag, and itunes to perform all the necessary manipulations and organization.

  Tracks were converted to MP3s at 32,64,92,128,256,320 kbps and native (roughly 2,500 to 4,000 kbps)

  I made a single album with songs sorted by original album and with an addendum on each track title to identify its bitrate easily with activating the track info on apple’s itunes remote app on the ipad used to select songs. i.e. Enter Sandman by Metallica would be Enter Sandman 32 , Enter Sandman 64, Enter Sandman 128, etc. I organized things this way as to not clutter the rest of my itunes library with duplicate tracks, and to keep all the low-rez files together for easy cleanup after the testing.

  The first part of the test involved my wife and me listening to various tracks in increasing resolution. She is not an avid listener, and was being VERY supportive for this little test. 

  It was very obvious early on as to why low bit rate mp3s are not popular among audiophiles / music lovers. I’ll quickly summarize what we heard (again, this was what WE heard, THIS night, on OUR system, and is not meant to be a definitive treatise on the subject! IMHO, YMMV, etc)

On to Part 2

[toddbagwell]

Part 2: sighted listening impressions.

32kbps : yup, terrible. Very phasey. Almost no soundstage width, and almost like the music was being piped in through a puddle of water. Very uncomfortable. Like noise, and not at all musical. 

64kbps : Meh, still bad, but I think I can hear music in the noise now. Not good, but not so bad that I want to plug my ears like with the 32 samples. Some upper frequencies are more noticeable, and yes, there do seem to be two channels now.

92kbps : better, but not nearly the jump like from 32 to 64. I think if there were no other option, this could do for background listening without much trouble. Highs continue to improve, and the instruments are starting to have their own spaces. 

128 kbps : still better, I think my toes might be tapping! Now I can listen comfortably without any specific problems screaming at me. When I focus on the sound, i.e. critical listening, I can hear the highs and to a lesser extent, the lows, are not what they can be with full-resolution files. There is decent separation between instruments, and the soundstage is adequately wide. Compared to the 32 and 64, this sounds amazing, but still has fairly easily audible artifacts of the compression process. 

256kbps : as we get more data per second, this gets harder to describe, but more air, sense of recording acoustic, more individual instruments, and more frequency extension on the highs, and better control on the lows. For reference this is similar to the itunes download quality (without getting into differences between AAC and MP3 encoding...) More musical, more enjoyable. 

320kbps : EH? did I change the resolution? After a few more notes play, this does indeed seem better. As before, the improvements become still harder to explain. I think I know why some people say "there is more There there"  The sense of separation between instruments continues to flesh out and the frequency spectrum seems complete. Obviously we started with trash at 32 kbps and now with 10x more data per second the sound is transformed into music. I kinda started to wish that I’d not added 256 or 320 to the comparison, because this sounds pretty good. Now I start to wonder if maybe I have wasted some money on needless resolution. 
   Maybe I can remember subtle things that are different in the Hi-rez versions, but some of these songs aren’t played too often (Mixed audience for this test, remember?) and I don’t have the best memory of each and every note. I could definitely listen to this all day, and not really know I was missing anything. Highs are extended, no phasey artifacts are audible to me and the songs definitely get the toes moving. 

  Hi-Rez : [disclaimer, some of you will cry foul, and hate me for this, but in my system I have quicktime and the ASUS soundcard driver set to upsample all songs to 92/24 as my equipment won’t decode 88.2, and I don’t notice enough difference with upsampled redbook CD’s to switch these settings back and forth. As I said earlier, IMO, and just my experience] 

  With uncompressed Hi-rez, the room I was in was gone, and we were at the recording session! The 320 that I’d just thought was plenty good enough appears not to be the limit of what my ears or equipment can discern. 
 All the preceding descriptions still applied in regards to the frequency extremes, soundstaging and separation of individual instruments and voice(s). When I stopped listening critically, I noticed that I could relax and just enjoy the spectacular performances for what they were. This was the most “real” the least fatiguing and the best reproduction of all of the sample rates. It truely felt as if the space around us had been transformed, we were there at the pawnshop, or standing next to Lars Ulrich. 

on to part 3

[toddbagwell]

Conclusions:

OK, so much for what we all probably know, how did we do when testing each other blind?

For any scientific minds out there, yes this was a single blind test. if I had hidden the bitrate in the comment tag, and then recorded my guesses as to each track's bitrate, then pulled up my recently played song list and looked at each comment field to check the accuaracy of my guesses it could have been double blinded. I didn't chose to do this, as we were just having fun. You are free to interpret the results however you wish. 

  We took turns playing a few tracks at random sample rates for one another and the person holding the ipad got to choose the bitrate and order in which things were played. The other person tried to guess what sample rate was being played, based on our memory from the sighted listening. If we didn’t know, a ballpark guess was allowed with a higher or lower bitrate then played for comparison. Again, very artificial arrangement when compared to normal listening. We’d both had a drink or two with dinner, so we didn't worry too much about being super secretive.

  My wife did very well, guessing correct on about 60-75% of our tests. She claims to have one bad ear and doesn’t read my issues of stereophile, so we can consider her to be the “average person on the street” (but only for this test, she is otherwise amazing!)

 I did a bit better, about 70-85% correct. After about 10 minutes my brain turned to mush, and I lost all sense of reference. This was during the Metallica track, which sounded much more compressed than most of the others, making distinctions harder to nail down as the bitrates climbed.

  This was a fun little experiment, and I’m glad I did it. I learned that I'm not throwing away money on high resolution, and I also learned that if I want to buy a song or two from the Itunes store on a whim, I won't be buying utter crap (say what you will about pop music, some I find catchy)

  If you’re still reading, thanks. I hope that this novel of a post might inspire one or two people to try it out for themselves. It was fun to experience all the compression artifacts in short order and to see what all the fuss is about with ABX testing and how listener fatigue can confuse what is an otherwise well set up test.

Later,
Todd

[ted_b]

Todd, there is nothing in your part 2 or conclusions.  Also, why did you do just Mp3's?  Why not redbook at least.  Some of these selections are simply upsampled redbook (i.e I wouldn't buy them as hirez, including REM) and therefore are not great hirez examples. Some, like Pawnshop, are better.

Edit:  I see you've edited/filled in part 2.

[toddbagwell]

sorry ted

All files were hi rez. I just linked to amazon pics quickly.

the metallica, REM and Dianna Krall love scenes were 2 channel rips from DVD-A discs that had been gathering dust for a few years after i went to a computer based set up.

as to why no 16/44.1, I did all of this in two hours before dinner and wanted to make sure there were some easily audible differences for more normal listeners.

I have all the files and could probably make 16/44.1 copies for a more stringent analysis.

todd


edit= as to the choices, yup, not the cream of the crop, and not intended to be. It was more of just a fun little thing to spend a few hours trying to involve my wife in the audio hobby. I learned a bit as well.

[ted_b]

No problem.  You gained a hirez fan (your wife) so all is good. 

[Geardaddy]

nice experiment Todd.  I too would want to hear your impressions of Redbook vs higher resolution formats since that is where much of the debate lies....

[toddbagwell]

yup. having typed and posted it all, I realize that it isn't really a comparison of "standard" resolution 16/44.1 vs the higher stuff, if the patterns hold true to the prior comparisons, I still think it would favor hirez. [ does this win any sort of "Obvious Statement of the Year" award? ]

I'll look around in my software settings to see if I can generate redbook resolutions from hirez files. I'll be careful not to just mix two different file sources (i.e. cd track vs hirez download) to ensure that the track volumes are equal so as to not favor a lower res file that just happens to be louder.


thanks
todd

[Geardaddy]

I am pretty sure hi rez would win too, but it may also depend on the CD.  Use some CDs that you think are exemplary....