[john970]

In preparation for the 1er, I did another listening test last night to help determine if I would get any benefit out of losslessly encoding my audio collection (I encode at 256Kbps vbr aac and occasionaly buy music at either 128 or more often 256 from the iTunes music store).

Tester:
I consider myself to be a fairly discriminating listener, although I am realistic about it and revile undefined audiophile terms like "more musical" that people revert to using when they are trying to "describe" an imperceptible difference.

Bitrates tested:
I tested AAC only for compression at 96Kbps VBR, 128Kbps VBR, 256Kbps VBR and ALC (Apple Lossless) as a control, which was usually around 900kbps.

Audio Equipment:
Denon 2807 processor and amplifier with B&W DM 604 tower speakers using MIT cables. No separate subwoofer was used for the test. I also used Shure e2c headphones, switching between headphones and the speakers during the test.

Source:
I used some tom petty tracks, batman begins soundtrack, and some 2pac. They were encoded from the CDs using my Mac Mini and iTunes encoders and sent to my Denon over an optical out (digital), so the Denon was used for the DAC and amplifier stages. No EQ was enabled (DIRECT mode on the denon). My center and surround speakers we not used.

Methodology:
This was a blind test, which is absolutely key in this type of testing. I setup the 4 tracks on the screen (per song), started playing them all simultaneously, and switched between them and attempted to order them from best to worst quality for a selection of tracks. Sometimes I would play a section over on different recordings until I felt I heard a difference.

Results:
This was HARD. I expected to have trouble between 160&256, but identifying 96 vs. lossless in a controlled environment listening to the same track was suprisingly difficult, and not obvious. I believe Variable Bitrate Encoding (VBR) is instrumental in this: complex sections are given enough attention to be indistinguishable while silent or simple passages are compressed heavily. This makes my job very hard...

96kbps: I identified all of these tracks, but just barely. The tracks had a general sense of reduced range - the high shrills, cymbals, etc. sounded more muffled. I can see getting less out of your music at this bitrate.

128kbps: I identified all but 1 of these which I confused with 256. Very difficult: listening to certain passages it was possible to tell the difference. Depending on the material, I can see getting less out of your music at this bitrate on a good system.

256Kbps: I confused 1 of these tracks with 128 and two more with ALC. This sounds great - hard to tell between this and lossless. Certain music and certain tracks for certain people on high end equipment might be able to spot the difference but: You get full enjoyment out of your music at this bitrate. There are some differences, but they are academic.

ALC: I confused two of these with 256 and got the other 3 right. This is not much better than luck. On some tracks, I feel comfortable saying an audiophile would be throwing darts guessing between the two (256 and ALC) on a good audio system, maybe getting 2/3 on a CD he/she had heard 100 times.

Conclusion:
Lossless codecs are needed to ensure when the next hot codec comes along you have a way to reencode. And someone out there has a proper audiophile listening room and the right ears to spot differences between 256kbps and the ~900kbps data contained in a PCM audio stream. This should be regarded as the "gold standard", and 256kbps as the "consumer standard", as I believe you get every bit of enjoyment out of your audio using 256kbps audio.

Thus, I will use 256kbps VBR AAC as my standard audio format.

Caveat:
I didn't do a scientific test on this, but 128 CBR MP3 sounds substantially worse than anything I tested above. CBR is just way worse for a given bitrate, it is very easy to pick a complex section and spot the difference. I wouldn't pay for music at this bit rate/format.

Other notes:
HD Radio stations usually push a 96Kbps HE-AAC stream on FM, which is similar sounding to a 96kbps VBR AAC (read: good) from a polk audio hd-radio reciever and my shure e2c headphones.

Satellite radio uses a similar, slightly more efficient codec, at 40-50kbps on average for music stations. It is completely obvious that this is not a CD to even the least discriminating listener, to the point where I do not enjoy music on it. I use their online feed (128kbps wma) and record it, and transfer it to my iPod after converting it to 160kbps vbr mp3 using LAME).

Makes it obvious to me why SACD and DVD-Audio failed.

PHEW.