Seems I found a (probably rare) decoding bug of Lame: When you decode some MP3s to wav and burn these wav files as an audio CD, your burn shows lots of C2/CU error spikes probably due to DSV problems. Sounds weird, but it really happened to me when decoding some freeformat MP3s bought from allofmp3 with LamedropXPd V.1.3.3 using LAME 3.96.1. The same MP3s decoded with the MAD bundle from rarewares did not cause such a problem.

For a full description of the problem, see my thread at CDFreaks.

Greetings, sTisTi

P.S.: FYI, the files that caused these problems are the original 384 kbps allofmp3 files of the Joni Mitchell album "Ladies of the Canyon". It also happens with the album "Blue". All the other freeformat MP3s i downloaded from allofmp3 did not show this problem.

I think that this is not a problem in decoding by Lame, but in the burner itself.

It seems that you found in your files a pattern causing problem once burned.
Question: is the problem in burning or in reading the cd, ie are the C2 errors the same if you put your birned cd in another reader?

I think that digit-life once found a similar problem with a video file. The problem was then in the reading of the disc that was not correctly handled by some readers.

http://www.digit-life.com/articles2/magia-chisel/index.html

Thanks for the interesting link, Gabriel! Here's a summary of the problem they encountered:

So yes, in the end it boils down to the drive not being able to read back what it wrote. Unfortunately, I don't have another drive at the moment I could use for scanning besides my Plextor 716A. But according to the Digit-life article, the majority of drives should be affected by such an occurence. I am pretty sure my problem is due to the same basic problem described by Digit-life, except that their problematic pattern was caused by a corrupted CD read-out in the first place while mine originated from the decoding output of Lame. And this wasn't just a random error: it happened with a total of 22 files. I'm pretty confident this was triggered by the MP3s I fed to Lame. As I wrote in the CDFreaks thread, they seemed to have a strange ending when viewed in a hex editor. Unfortunately I know way too little about the structure of an MP3 file to interpret this in a meaningful way. Nevertheless, MAD was able to decode the files without corruptive patterns, so maybe Lame has a small bug after all because it does not prevent such a corruptive pattern.
If anyone is interested in having a look at these strange MP3 files, PM me and I will send one or two to you; but maybe my mailer will not let me send these massive (~10MB) files, so if it doesn't work you can also download them directly from allofmp3. I suggest the files "Raniny Night House" or "The Priest" from the Joni Mitchell Album "Ladies of the Canyon", which are both good examples with regard to the strange ending. If you want to encounter C2 errors when burning the decodes, you probably have to burn at least 2 of them in a row as the C2 errors only occured at the track boundaries and not at the beginning or end of the disc.

Lame is not trying to avoid any specific pattern when decoding, simply because there is no forbidden pattern in wav files. In your case, the wav file seems to be perfectly valid, it just happens that when burned the content is disturbing the cd reader.

The output from Mad is different because Mad is using dithering, thus the last significant bit is different.
I think that you are not encountering a problem with Lame decoding, but this is a case of unusual "bad luck" combination.

note: the strange case your are encountering might be interesting to the one who wrote the digit-life article.