• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

MP3 & It's limition workarounds?

2M2B

Active Member
Joined
Mar 9, 2021
Messages
118
Likes
73
Been messing around with a beta Lame 3.99 encoder after finding out about the --allshort switch. I have no idea why it not used in the finalised encoders since with 256kbps CBR with Allshort on Eig.wav goes from having pre echo artifacts to being fully transparent?. Turns out that switch forces MP3 to act like a Subband codec from a thread on HA in 2002.
 
U

UKPI

Guest
Pre-echo happens because quantization noise spreads throughout the entire block in time domain. Using a short block has the advantage of containing pre-echoes caused by attacks in a shorter time frame (This results in attacks masking pre-echoes more.) but has a disadvantage of having less frequency domain resolution (This hurts codec performance because of less effective utilization of frequency domain masking. Less resolution, less control.) and using more blocks when encoding the same amount of PCM samples (More blocks mean more overhead. That leads to using more bits.). Because of this, only using short blocks would be a bad idea since it isn't suitable for encoding parts of music that doesn't have strong attacks. AAC has increased the length of its long blocks and shortened its short blocks so that the encoder can fare better for both circumstances.

These threads of hydrogenaudio would be helpful:
https://hydrogenaud.io/index.php?topic=7296.0
https://hydrogenaud.io/index.php?topic=4878.0
 
OP
2

2M2B

Active Member
Joined
Mar 9, 2021
Messages
118
Likes
73
Also, another thing I don't get with Lame. The MP3 codec supports 8 ~ 640kbps and has a subband mode, If It MDCT mode is weak why didn't they switch to subband for higher bit rates(160 ~ 640kbps)?. There was already few decoders that could read 640kbps MP3 files like MAD or patch older decoders to see frames at 384 ~ 640kbps.

The fact --Allshort helps with attacks/transients even on 224kbps CBR despite being a dirty method. Musepack Is pure Subband codec like MP2, HA's own tests at 192kbps VBR on par with Opus but in reality it robust at 160kbps on electronic killer samples.
 
U

UKPI

Guest
The MP3 codec supports 8 ~ 640kbps and has a subband mode,
There was already few decoders that could read 640kbps MP3 files like MAD or patch older decoders to see frames at 384 ~ 640kbps.
Bitrates above 320kbps are non-standard. In the header of MP3 stream/file, all the supported bitrates are explicitly written while everything else are clumped into a single classification called freeformat. It is a hack. Some decoders do support it, but the majority of devices/software with MP3 playback capability does not. It wouldn't be that good of an idea to put effort into nonstandard bitrates that only a few people would use.

If It MDCT mode is weak why didn't they switch to subband for higher bit rates(160 ~ 640kbps)?
MP3 is a hybrid codec which first uses the PQMF filter (subband) of MP1/2 and then further divides the signal with MDCT. There is no subband mode or MDCT mode. Those filters must be used together.
 
Last edited by a moderator:
Top Bottom