Critical mp3 encoder quality bug needs to be fixed

[LeoWattenberg]

Fwiw: I’d suggest formally forking lame if you believe that lame has a bug in it that they’re unwilling to fix. A patch that gets passed around is likely to lose context eventually, while a standalone project is going to have some sort of web presence (and if it’s just the reader.md on the repo), and a maintainer/contact person. This isn’t a technical thing necessarily, more a marketing thing which makes it easier to convince people to adopt the changes you’re making.

Also, I’m not convinced there’s an issue in the first place. As people in hydrogenaudio thread have repeatedly stated, placebo is a strong factor in audio quality. Without a proper scientific experiment (ie, blind ABX test at sufficient repetitions), it’s virtually impossible to evaluate. Given that you’re trying to get people to believe you, I’d suggest running this experiment yourself to get some data to point to if people tell you they’re not convinced it’s an issue in the first place.

[jjchiarella]

I appreciate the feedback.

I am nowhere near technically savvy enough to fork and run a Github or whatever is necessary. I should get back on ABX, which I have not done or set up in several years. I am confident that I can catch the regression (or whatever you will call it) from the cymbals in stereo music. These are real samples and not artificial “killer” samples.

This was not the best format, as I had to randomize a list and play them back, then check, but getting eight out of ten was good enough.

Part of the problem was that I could not find an original sample, only one recorded at 320 kbps CBR, probably with LAME at the defaults. The sample already had that telltalle distortion on cymbals. I saved it one more time at LAME CBR 320 kbps with the default (-q3) to have more cymbal distortion. Instead of distinguishing from that distortion being present or not, I had to differentiate that effect being more pronounced or less pronounced.