It finally dawned on me our collection of music CDs could be placed on a USB drive and used in our cars. Both Verbatim and SanDisk make tiny USB-A drives where the memory is placed inside the connector housing itself leaving only a little bit of plastic protruding from the car's socket. This ensures the drive won't damage the socket if bumped. How much memory is needed was the question, and so I looked for reports on the web about the quality of various compression schemes.
Both our cars support MP3 and WMA files. Both also support the later AAC standard but only for the compressors used in Apple's iTunes (this is likely the result of both cars having Apple CarPlay). The cars also support FLAC which is analgous to a ZIP file in that the uncompressed recording is identical to the original. The compression is limited, though, to about 2:1 and so would not hold all the CDs on one drive. The most common compression schemes are MP3 and WMA and both effectively discard portions of the music which most people won't be able to hear. They are analgous to JPEG images. Compressions of 15:1 are available, but will one hear a difference?
The most interesting report was from McGill University, dated 2009 and titled "Subjective evaluation of mp3 compression for different musical genres" by Amandine Pras, et al, and one can find a copy at ResearchGate. They prepared short MP3 clips from 5 musical genre and had a number of listeners from various music backgrounds compare them to the original CD material. Because they tested only a few short clips with one version of compressor, this test was not definitive, but I made my choice based on their results.
Given they've made their paper available publicly, I hope they won't mind if I copy here the two figures I thought most significant, namely Figures 3 and 4:
They found they could categorize the expertise of their listeners as either Musicians or Sound Engineers. They also found they could categorize their music samples as either Acoustic or Electric. The Acoustic included Orchestra, Opera, and Contemporary, so the music samples would have included string and wind instruments and voice. Under Electric was Pop and Metal Rock. All were compressed using the LAME MP3 compressor at bit rates from 96 kbps to 320 kbps.
Their bar charts above show how often the listener picked the CD over the MP3. If they were indistinguishable, the CD should be picked 50% of the time, but because of the sample size, they predict that anything between 46% and 54% are statistically the same.
Note that their Musicians could, with Acoustic recordings, barely distinguish even a 96 k MP3 from the CD original, which is a 15:1 compression. There was no significant preference at all with bit rates 128 k and above. Even the trained Sound Engineers could barely distinguish a 192 k MP3 (7:1 compression) from the Acoustic originals.
My tastes in music run to what they categorized as Acoustic, and I doubt my ear is better than a Musician's. Given a car has low fidelity speakers and a great amount of ambient noice, surely 96 k is enough. However, I might want to listen to the library on headphones on my PC, so I chose 192 k, a compression level at which the Musicians could barely hear a difference in any genre and the Sound Engineers in the Acoustic genre.
One can find generally similar results at SoundExpert. Here, volunteers from the web are presented with music that's been though many different compression schemes. Some of the sources would be what the McGill researchers called Electric, and those forms of music had more audible compression artifacts than the Acoustic sources. So, one might expect higher bit rates to be needed, and this is the case. In their results, MP3 needed 192 k to sound free of compression artifacts to most listeners.
Lastly, I found a news site (which I failed to record) which offered tracks from various genre and asked you to select your favorite from three copies, one each at 128 k and 320 k MP3 plus CD, although you aren't told which is which until you select your favorite. For the one I tried, a Classical track, I picked the 128 k: A golden ear I do not have.