I used to encode movie soundtracks to mp3 using Lame 3.90.3's quality level 6, which gave an mp3 around 124 kbps and a decent sound quality. Further versions gave bad results, same goes to new recommended Lame 3.97.2. Here's a quality example:

Lame 3.90.3 -V 6 -h MP3 [259 kb]
Lame 3.97.2 -V 5 -h MP3 [241 kb]
Original saved into FLAC [1,05 mb]

So, what's up with modern Lame? It seems that I can't trust it to encode any sound, unless the quality level is between 0 and 4! Those between 6 and 9 I find useless. The minimum is quality level 5, but you can be almost sure it'll plant ringing artifacts somewhere in the track.

What should I be looking for?

(And where? Those samples are reasonably wrong and nothing jumps out at me)

The hight frequencies of sample encoded with Lame 3.97.2 are messed up, while encode made with Lame 3.90.3 is almost clean. I hear it very well in my headphones, while it's almost undistinguishable in my speakers.

This high frequency distortion issue is common for all latest versions of Lame (quality levels 5-9). As soon as your headphones are on it's very audible.

OK, downloaded just the FLAC and -V 5 mp3. I don't hear anything out of the usual. I don't have the best audio setup but I used my Bose TriPort headphones with my Sigmatel C-Major sound card.

I did experience popping noises in both tracks, the popping noises were effectively carried over to the -V 5 mp3.

My ABX results aren't worth posting as I was unable effectively distinguish between the mp3 and FLAC (had a 15% chance that I was "guessing" so I discarded the results).

You too will also want to conduct a blind ABX test, that way we/you can narrow things down. No offense or anything but, you can make all the comments about audio quality that you want. Those comments wont really be taken seriously until you conduct a blind ABX test (or if you have conducted a test, post your results).

Isn't -h redundant with VBR in LAME 3.97? Have you also tried --vbr-new?

Good to see I'm not the only one who didn't notice anything

Biont:

Did you try and ABX it?

given which day today is, maybe there is nothing to ABX

-andy-

I haven't tested these files, but I recall that 3.90 -V settings were NOT tuned (hence the existence of --alt-preset). Therefore, it would be very strange if a non-tuned and four-years encoder will generally provide better quality than current lame versions which are benefiting from several tunings and enhancement. It would be even more strange to consider 3.97 -Vx as "useless" compared to the old and untuned one. Furthemore, -V5 seems to be more than "the minimal [...] quality level" considering the rank it obained on recent listening tests (tied with aoTuV, iTunes AAC and WMAPro at ~130 kbps).


Anyway, Biont is maybe complaining about ringing issue occuring with low volume content and LAME encoder. I noticed them, and experienced myself that -V4 is by far better than -V5 & inferior VBR steps on such content. I recall that --athaa-sensitivity switch really help on some situation (at least with 3.97b1, but probably less with 3.97b2). Biont: try ABR instead VBR with 3.97. If you're experiencing ringing issue with low volume content, it should be the best settings for 130 kbps and less.

What's destination of mp3 encoding?

If compability for standalone players ( MPEG4-ASP) has priority then vbr is no go. Mp3 vbr in avi container is quite problematic.

high frequencies? there didnt seem to be any high frequencies in your sample, there wasn't anything above 5khz worth while IMHO. Do you have a better sample?

I don't know how ABXing is done. I'm a user, and I care about what I hear, and not what I see on some charts. So, as I said, those samples I provided show very well the difference between VBR aproaches of old Lame and new Lame. If you can't hear the big difference even with your headphones on, then what's the point of ABXing?

Perhaps guruboolez is right about my complain being all about "ringing issue occuring with low volume content". Because when I normalize that sample and encode it again with Lame 3.97 VBR 5 - the ringing is gone. Dynamic music is full of low volume content and therefore any Lame settings, that distort the sound pretty bad, I count useless. And that's VBR quality levels 5-9. The tip about using ABR instead of VBR helped. ABR treats audio differently and ringing is not produced. I'd say the quality of 3.97 ABR 128 beats the quality of 3.90 VBR 6 a bit on my sample.

And again about the distortion I hear. I got Yamaha monitor headphones RH-5Ma. The distortion happens to hiss (I meant hiss, saying "high frequencies"). If you listen to FLAC, the hiss is heard in the background throughout the whole track. But if you listen to Lame 3.97 VBR 5, the hiss appears at the very start, and when piano starts, the hiss turns to ringing right away. That doesn't happen to 3.90 VBR 6 and 3.97 ABR 128.

Igor, my destination are regular computers, I don't aim for hardware players, but I know, that XviD + Lame VBR (ABR) in AVI play well on some hardware players I tried (like those Xoro players popular in Russia).