Hello, this is my first post. In LAME, the lowpass filter used in encoding can be disabled by using -k. As a result, mp3's can have a frequency response almost identical to that of the original CD audio files. Try this when encoding >192 VBR files. Together with -m s for stereo (not Joint Stereo) can produce excellent mp3 files at 210-230 kbps.
The whole point of lossy encoding is throwing away inaudible information to save space. So a (high enough) lowpass is *good* as well es -m j (= lame is allowed to decide if a frame is encoded as left/right (= "true" stereo) or mid/side (= what's referred to as "joint stereo" mostly). Lame does this in a decent way). Using -m s and -k you'll run into audible problems caused by the 320kbps upper limit - the bits wasted on inaudible frequencies and useleless stereo information can't be used to encode important audible information. If you're a human use the --alt-presets with lame (3.90.3 or 3.93.1). E.g. at bitrates arround 210-230kbps --alt-preset extreme will give you best possible quality (see sticky thread in this forum). If you're a cat, a dog or some other animal capable of hearing higher frequencies than us humans, you better not use any lossy psychoaccoustic encoder at all, as most likely the psychoaccoustic model isn't tuned very well for your needs anyway.
The low-pass filter used by default in Lame cuts at about 18 Khz. Using a wav editor and an uncompressed wav originated from cd, one can cut these frequencies and hear the lack of them even if being a human. Of course, if Lame is forced to keep freqs. above 18Khz in a constant bitrate, less information will be available for the rest of the signal. When using variable bitrate, though, the result is just a higher bitrate. So the question is: Is it worthy to raise the bitrate 10-20 kbps in order to keep these frequencies? If you want to make files for a portable player, NO. If you want to make high quality files for a personal database and you own a modern harddisk, why not?
With majority of music the difference between 18kHz lowpassed and original isn't noticable. If you don't believe this, try yourself. To avoid immagination having influence on the results, perform the test in a double-blind way (aka ABX), e.g. using foobar2000's ABX tool. Anyway, the lowpass of --alt-preset extreme (~ 220kbps VBR) is 19.5kHz, the lowpass of --alt-preset standard (~200kbps VBR) 19.0kHz which is (both) inaudible on real music for sure for 99.9xxx % of ppl / 99.9xxx % of music. The problem here: Even in VBR mode, mp3 can't use higher bitrates than 320kbps. At places where that many bits are needed the behaviour is identical to CBR. Keeping high inaudible frequencies won't cause audible differences in most cases, only a raised bitrate - true. In some cases, when 320kbps is necessary to avoid audible artifacts even with lowpass, removing the lowpass will make these artifacts noticable or make already audible artifacts sound worse. The --alt-presets for lame (--alt-preset standard, extreme & insane) have been developed and tuned using lots of double-blind listeining tests performed by well-trained listeners. If you find some lame setting that gives better quality at the same bitrate (and you've verified this by doing ABX tests) feel free to post your results at http://www.hydrogenaudio.org - ppl there will be interested for sure. Lots of background information about what I've said you'll find in the FAQ there and using the search feature. BTW: For high quality PC archiving/listening you might want to use something else then mp3 to avoid the limitations of mp3 as a format (pre-echo artifacts, gapless playback problems, slow encoding at high quality etc.), e.g. musepack (mpc), wavpack lossy or some lossless format (flac, monkey's audio, wavpack).
