I want to know how good AAC compared to MP2
For Example 192 AAC stereo = how much from MP2?
I can not comment on old MP2 encoders.
If you want sub-band-type encoders, can I suggest Musepack (MPC) to look at?
Because MPC is excellent (and virtually transparent) at 192kbps.
From this book (https://link.springer.com/chapter/10.1007/978-1-4419-6184-6_6), a figure showing the results of a late 1990s formal listening test, with MP2 and MP3 called "LII" and "LIII", respectively:
You can see that AAC-LC is as good as MPEG-1 Layer II (MP2) at roughly half the bit-rate.
But I'm pretty sure AAC-LC has likely improved more than MP2 has over the last 20+ years, not to mention stuff like HE-AAC and so on (though that's not quite so useful at high bitrates)
EDIT: just going off of software encoder development over that timespan. AAC (even LC) has had significant development, has MP2 had any development at all?
MP2 is, technically, much much simpler than AAC-LC, HE-AAC, and xHE-AAC, and as far as I know (never worked with it), it is inherently CBR because it doesn't do fancy lossless parameter packing and requires fixed-length frames. So once you get a working encoder implemented, there's not much more you can do. Also, MP2 standard approval (committee draft) was 30 years ago.
I assume that, by the time the above listening test was done, AAC-LC was already in good shape performance wise since MPEG-2 AAC standard approval was 25 years ago. But you're right, AAC-LC probably has improved some more since then, and other codecs (incl. xHE-AAC which I worked on) outperform MP2 even more clearly. I don't have numbers for such comparisons, but compared to modern AAC-type encoders, MP2 probably requires close to 3 times the bit-rate at the same audio quality.
I think both AAC and MP3 have probably improved a lot since then. At least few people in listening tests posted here report "slightly annoying" or "annoying" artifacts at 128k, IIRC some tests have found both formats nearly transparent at those bitrates for a great many samples.
Keep in mind, though, that those formal tests were, and are, usually done with CBR settings, where it's harder to improve the encoding quality. Also, those tests were usually done with developers as listeners, i.e., experts. And if I look at, e.g., Kamedo2's 2013 AAC-LC vs. LAME test (with encoders which have had very subtle quality tuning since then), I see, at 128 kbit/s, roughly the same quality verdict for AAC-LC (qaac, scoring in the upper quarter as AAC-LC in the 1990s test) and MP3 (LAME, scoring slightly above the middle of the quality scale as in the 1990s test):
As pointed out in #4 above, MP2 is much "simpler" - and that is because it set out to serve a different purpose. MP2 was designed to be broadcast and streamed and for the receiver to easily pick up decoding after an error. MP3 was for a decoder that would have access to the entire file.
If you are asking for AAC vs MP2 in a broadcast setup, like DAB+ vs DAB, you must also take into account what "AAC+" (/HE-AAC) are actually used by the industry. If you are streaming, you have to compare what you actually receive.
thanks all for your comments