Hi Fur! Welcome to the board.
The conclusions I draw about MiniDisc are from blind-testing ATRAC against other audio codecs, using well-recorded CD's and several hard-to-encode samples that are known to pose difficulty for transform encoders like ATRAC and MP3. Back when I owned a MDS-JE530 deck and MZ-N505 NetMD player, I encoded clips from about 20 songs and samples, including the following (not an all-inclusive list): Stravinsky's
The Rite of Spring, second movement; Dave Bruebeck's
Time Out; Pink Floyd's
Keep Talking; The Police's
Walking on The Moon; Return to Forever's
Romantic Warrior (jazz); Mile's Davis
Bitches Brew; a door-creaking sample; a sample of a tolling bell in the end of Pink Floyd's
High Hopes; the opening of AC/DC's
Hell's Bells, and a sample of applause (plus a few more that can't recall off the top of my head).
I recorded each sample to ATRAC digitally through the TOSLINK input of the deck, and recorded the resulting MD through the digital outs using a stand-alone Philips CD-RW deck. To test LP2 @132kbps, I used my NetMD player for the same samples and used Total Recorder to grab the wavs while playing the samples in OMG Jukebox 2.2. The wav files from the deck were grabbed from the CD-R using EAC's secure mode, as were the orginal wav samples to be compared to. I then used a blind-testing program to listen to each sample, to see if I could tell the sample apart from the original (called ABX'ing around these parts). Short story: on almost every sample tested, my ability to tell the ATRAC encodings apart from the CD source was significant, p<.05. For 132kbps, I was able to ABX every single sample from the original, and the level of confidence was even greater, p<.01. Thus, I objectively tested ATRAC/ATRAC3 against the orginals and found them to be different than the CD source. Was this done using expensive components? No. The grade of your euipment has little effect on one's ability to ABX samples, for the type of artifacts I listened for (pre-echo, ringing, etc.) can be heard just as well on cheap heaphones as they can with expensive ones, etc.
Artifacts I noticed with ATRAC1 Type-R: ringing and harshness on high-frequency notes such as brass and percussion, pre-echo, crackling sounds on bell ringing, "whooshing" sounds on some electronic notes, and occasional popping/dropouts on loud material. And this was using Type-R compression....the early iterations of ATRAC were so bad that I actually liked metal cassettes better than MD when it first came out. Needless to say, I don't own that deck or the portable anymore....I sold them on eBay last year.
As far as ATRAC3 @ 132kbps sounding like a 192kbps MP3, I would suggest that you try encoding some material using LAME 3.90.3 and the --preset 192 setting (or --preset standard, which results in an average file of around 192kbps). Preset standard is transparent on 99% of material, and that is due to the type of testing and tuning that I described. It is common knowledge around here that even TYPE-R doesn't do as well as --preset standard.....read the "other audio formats" forum and you will see this mentioned several times. Thus, ATRAC3 (LP2 setting) doesn't even come close to alt-preset standard; in
another thread, a member actually prefered LAME @ 128kbps to ATRAC3 at 132. For one, the 17.5Khz cutoff on LP2 recordings is noticeable on certain recordings, and there is a loss of detail on orchestral music (as guruboolez points out in that thread I cited), as well as pre-echo. Listen to the door-creaking samples on the page as well, and you will see how ATRAC3 and ATRAC3plus artifacts...when it runs out of bits, you get strange sounds that even MP3 does not create.
In short: I have nothing against MD, but have tested it enough to know that better formats exist. If it works for you, great! Just make sure to do comparison testing to some LAME encodings and orginals before assuming that 132kbps ATRAC3=192kbps MP3. I think you'll be glad you did.