I've two mp3 files (the same song, a jazz live concert)



Both mp3 were not encoded by me

To my taste the 294vbr mp3 sounds better (mainly the voices in the song) but the applauses sucks, sound noisy like something being fried. On the other hand the 192cbr sounds a little more flat in general but the applauses sounds great.
So I'm wondering if any of the features described in the lame tags can explain why the mp3 with best quality has this defect

thanks in advance for any help

Unless you have the source (= reference), it's hard to decide which one is closer to the original.
I'd think it's safe to say that the VBR one is most likely the better one.

who did encode the songs for you? what settings did the person(s) use? or did you 'obtain' them somewhere?
I may be wrong about that, but doesn't the min.bitrate of 160 in the vbr-file indicate that custom settings have been used? if that is the case, it may be that the vbr-file does NOT have to be most likly better...

I agree, any user-sided modification/interaction seems very probable. Nethertheless it may be encoded with the extreme preset of Lame 3.91 and a manually set alterations like this:
Lame.exe --alt-preset extreme -q0 -V0 -F -b160 input.wav output.mp3
The characteristics (as far as verifiably) of such a file are fully according to the second Mp3 file mentioned above.

The '-V0' switch would explain why the average kbps rate is so big, improving the quality generally - even if this Mp3 tendentiously seems to be a 320 kbps CBR file with 'exceptions'.
The '-F' would enforce the encoder to take a minimum bitrate of 160 kbps also for (digital) silence. This is not very efficent but applied on 'normal' music track this doesn't really matter - it has no notable influence on file size and no consequences for the acoustic quality (correct me please if i would be wrong here).
More problematic is the '-q0' switch specially on older Lame 3.9x versions - as far as i remember it was rather experimental.

Of course this is a very 'post-theoretic' approach to estimate the roughness of Mp3 encoding.
I mean... a friend of mine encodes live concerts captured from analog TV with an extreme preset...

____
Edit:

Just now I have had a suspect on this:
I have once encoded 'AC/DC Live' with a similar command line with Lame 3.92 and the applause/audience has sounded like gone down in an acoustic soup.
On older Lame versions this maybe was a problem of pre-echo and the psycho-acustic model 'nspsytune'. - I really hope it was a difficulty that no longer exists in newer Lame versions.

Apparently, -m s was used as well, which doesn't improve things.

Different Noise Shaping contributing to the harsh sounds in the applause?

(Damn) You are right! - Well, sorry! - I fully othersaw this only looking at 'safe joint stereo: yes '

Thanks too in another sense: Now the first time i have discovered the test tool i use called 'encspot' (console version) reports 'safe joint stereo: yes' on a full stereo file.
That seems not very logical indicating the use of a joint-stereo procedure on an Mp3 that was encoded as stereo!?
Untill now I never was aware before of this strange behavior...

now, there couldn't be any safer joint stereo than plain stereo...
you might want to bring that to the attention of the author who showed up here again after a long time... if he's still up and working, 'long time no see'...

I was really shocked a little bit about myself this morning... Of course i never had the idea to ckeck the 'safe joint stereo' entry if one of the first entry tells me it is a (full) stereo Mp3 file.


But what are you telling? - The encspot's author is back on work to improve that clever program? - Where?
I mean an update would be fine - specially because I had two little wishes:
- Make the smart tool show us also the LR-MS frames allocation/percentage!
- Please report the full Lame string if there is any.
If encspot is really based on 'mp3guessenc' from Naoki Shibata this should be possible i hope.