During the conversion of the CD, some data is obviously lost.

My question is: Why do some recordings lose more data than others?

It all reached a bizarre level today when I was listening to a minidisc on which I had Jurassic Park 2 and Jurassic Park 3. The JP2 minidisc recording sounded like a typical compressed ATRAC/minidisc 66kbps data file, with some distortion in the higher frequencies, some loss in the music (the violins become squeeky and muddy for example). But JP3 did not seem to have lost anything. Most of the parts were clear, with no distortion.

I guess I can understand why pop songs lose less than film scores or classical music, since there is much less to be lost in the conversion process. But what I cannot figure out is why some big orchestral albums lose more of their original sound quality than other big orchestral albums.

Both JP2 and JP3 are huge orchestral extravanganzas. But the effects of conversion are very different.

This difference also occurs with the two William Alwyn albums from Chandos. Volume 1 sounds thinner and squeekier than on CD, while volume 2 remains bold and clear.

When copied to minidisc, Silva's new rerecording of Nino Rota's Romeo and Juliet gets muddy strings whenever the strings are in the background, but the previous Silva album of original tracks does not seem affected as much as the rerecording.

I’m not a sound recording expert, but I have been dealing with the same thing, since I want to compress my recorded compositions (They are usually huge files, 4 min = 46 Mgs)

The loss of quality mainly depends on the recording and sampling rate. The better recording process, the less sound quality loss at the moment of compression.

For sampling rate, frequency may be from 11025 Hz to 44100 Hz. 11025 Hz is suitable for voice recording, 22050 Hz is suitable for tape quality recording, and 44100 Hz is suitable for CD quality recording.

As you said, if the song is very orchestral, it might sound with less quality when compressed. This is because it is using more frequencies.

66 kbps is a very low rate for orchestral pieces. I would recommend 90 kbps to enjoy them (It will never be like a 320 kbps)

I could think of two reasons of why some recordings lose more data than others:
1.- Because at the moment of recording, it was recorded with a very different rate than the one used for compression, or it was not very well recorded (ha ha, that’s why the recording staff is so important).

2.- Because the software you are using for compression losses data during the process.

The best compression I have been able to do (less size/ bigger quality) was a 90kbos MP3, using “Creative wave studio”

I hope this had helped you to clarify your doubt.

Stereo separation will affect the encoding quality as well. Can pieces of the music be encoded as a 66kbps mono chunk or do we need 33kbps for each channel of a simple stereo chunk? (This is just a simple example. I think most stereo algorithms are better than this.)

Loud recordings will require a higher bitrate than soft ones. Noise and crap that may not be audible on the soft and quiet recording will need to be encoded into the signal for the loud recording.