Please take a look at the difference between these two speks.
I use Foobar and I attempted the convert from FLAC to LAME 3.100.1 into 320kbps.
The end result is my MP3 spek should look exactly like my FLAC spek, with the exception of a cut off at 20KHz. But why does the spek in the Mp3 screenshot show a sudden quality deterioration from 16Khz and up?
This doesnt happen when I perform the same conversion with different CDs to Mp3 320 LAME?
Any ideas?
Since frequencies above 16khz are barely noticeable to human ears mp3 uses a polyphase quadrature filter on them, this makes them less accurate, but it leaves more bandwidth for the more important lower frequencies.
Looks like it makes "major" changes to the signal at -100 dB levels ... ? ;)
Lame mp3 could draw an elephant in the spectrum but if it isnt audible it doesnt matter and doesnt count as "quality deterioration".
The codec propably decided it needs more bitrate in the lows because it looks like quite a dynamic track, thats all.
This is by design for mp3 format. Noise shaping is
disabled for last scale factor band in CBR / ABR modes.
Meaning if theres too much demand on lower bands due to
extra precision need for 16k + region, Then the encoder would sacrifice precision
16 + region. You see it in spectograms the lower bitrate its more pronouced.
Welcome to the MP3 and the infamous sfb21 dilemma that psy-models in all MP3 encoders out there encounter this way or another. You didn't see this when ripping your own CDs because the music was different, it has nothing to do with FLAC.
If you want to investigate this a little further, you can try Bladeenc. You will see a truly epic spectrogram there that almost perfectly matches FLAC. The problem is sub-par audio quality that is way inferior to what LAME gives you.
you can try Bladeenc. You will see a truly epic spectrogram there that almost perfectly matches FLAC. The problem is sub-par audio quality that is way inferior to what LAME gives you.
This made me chuckle.
Bernie Grundman really nailed it about this situation: “You can always destroy the sound trying to make it look good on the test equipment.”
You can see a better looking spectogram with lame -q7 or -f as that disables all
noise shaping . But I guess there could be imprecisions below 16 k region if 16 + doesn't
fit properly etc and vice versa. Noise would fall into whatever fits in to the avg bitrate ?
I guess for 320k it won't be too bad as it can fit more.
I think blade enc was doing something similar and even more radical like lame -q9.
Recommended LAME settings for spectrophiles:
https://hydrogenaud.io/index.php/topic,4457.0.html
Many settings are only available on older versions of LAME:
https://www.rarewares.org/rrw/lame.php
Some examples. Read the subsequent posts for passwords.
https://hydrogenaud.io/index.php/topic,111736.msg927918.html#msg927918
Judge from OP's posting history it seems that OP was asking why a track on CD was encoded differently than the same track in the form a flac file. If the CD > mp3 and flac > mp3 conversion were not done within the same software/frontend, the resulting files could be different due to different LAME versions.
Recommended LAME settings for spectrophiles:
https://hydrogenaud.io/index.php/topic,4457.0.html
Spectrophiles???
They should be a Terms of Service 8.b for these people!!! :))