file conversions

Hi everybody,

I have a question about converting from mp3 to flac and back again.

The problem is that some mp3's don't play on my car player, although they work on other players, so I tried converting the mp3 to flac and then converting the flac back to mp3 again, using the same bit-rate as the original, and now they play ok.  Perhaps some bad data in the original which got dropped during my conversions...??

But I'm wondering if I lost any data through the conversions. I know flac is lossless, so I don't think I lost anything during the first conversion, but then when I convert back to mp3, even though I used the original bit-rate, would the conversion have dropped any data?

Thanks for any info...

Yes, loads I'm afraid. Every time you re-save as mp3 you loose another lot of audio info.

Hi,

Could you please elaborate a bit?  I had read something that when going to mp3 it drops some data, but only in some frequencies that may or may not be heard by the listener.  So my thinking is, if those particular frequencies were dropped during the first pass, then they probably don't exist on my original mp3, so therefore there might not be anything left to drop on my second pass.  Is this a too-simplified theory that doesn't work in practice?

How about if I simply convert mp3 to mp3, specifying a slightly higher bitrate for the second mp3, and leave out the intervening flac step?  Obviously the file would be larger, but would I still lose any data?

Thanks.

[MP3 is a distribution-only format - not one for playing about with at all.]

How about if I simply convert mp3 to mp3, specifying a slightly higher bitrate for the second mp3, and leave out the intervening flac step?  Obviously the file would be larger, but would I still lose any data?

I'm afraid you would - and 'simple' is nowhere in it. The first thing you have to realise about MP3 files is how they work in the first place - and the real explanation's not the usual rubbish you get to hear about 'throwing away' signals either. What they have to achieve is something called a 'coding gain'. This is basically finding a way not to store anything that is determined to be redundant, certainly - but this is not the actual loud signals you can hear. The coder uses a complex theoretical argument about the 'masking' effect of human hearing to determine what can be modified without apparent loss. Unfortunately the use of the word 'masking' is somewhat misunderstood... the whole explanation is rather complex, but it involves the establishment of 'critical bands', and it's the signals surrounding these that are the target of masking.

MP3 files achieve their coding gains by reduction in the bit depth of the 'masked' signals. So, contrary to popular belief, the masked signals are not  'thrown away' as such - the compression achieved is based only on a bit-depth reduction. That's how it's possible to have a number of different coding rates - the smaller the bit depth for the masked signals, the more you save. What does the bit-depth reduction do in practice? Reduces the available dynamic range somewhat, and increases the noise of the supposedly masked signals. What's worse is that not all signals are treated the same, even in the same frequency band; steady-state signals get more masking treatment than transient or moving ones, so there's a temporal dimension in this too. See what I mean about not simple?

So what happens when you open an MP3 in Audition is that it's decoded, worked on as a WAV file and then re-encoded? Since, even if you did nothing to the file, the re-encoding would be working on the same critical bands, the masked signals would end up being even noisier than they were, and because of the peculiar nature of some of the rest of the processing to re-establish the whole signal upon replay, the results sound even worse than they did before too - inevitably. If you change the bitrate, you are in an even worse position, because the way the critical bands are interpreted changes at different rates. Where this generally leaves you can only really be described as an appalling mess, quite frankly - don't go there.

Bottom line - MP3 is a distribution-only format - not one for playing about with at all. If you have problems with your original MP3s, it's probably because the headers are corrupted. I think that there may well be ways of repairing these without touching the audio if you look around on google, although I haven't checked recently.

The most likely reason for mp3's not playing back on certain systems is that they are encoded in a way that the system doesn't recognise. Joint stereo is a problem with older decoders so try encoding as full stereo. Joint stereo sounds better for a given bit rate because it assumes a certain amount of correlation between channels but our old (2001 vintage) DVD player won't play joint stereo files. You should also check the bit rate as some older decoders don't work at certain bit rates.

Cheers

James.

Thanks for your detailed information Steve... makes things a bit clearer for me.   And many thanks for your idea about fixing corrupted headers... I suspected there was corruption somewhere, which is why I went down the re-encoding path, but now I see there are better ways.  In fact, last night I discovered that one of the files I re-encoded still doesn't work.

I just googled and there's a ton of programs out there to fix corrupted files...

Also thanks James, I'll keep your point in mind...

Possibly not. You cannot lose any data after conversion. You have kept the bit rate to remain same. So, there is no chance of losing any data. And keep converting you'll get the things done in your favor.

If you're going to post garbage, please do it on another forum.

I've been having problems with playing MP3s in a program called TLC (a radio program used by Scott Studios).  We get a lot of commercials with a 48k bitrate, and we need to convert them to 44.1 as that's what our automation system uses.

When I'd convert the files in Audition, (actually Cool Edit pro 2.0), the "new" mp3 wouldn't play in TLC. 

We've been struggling with this problem for a while, but have recently discovered what the problem was.  When we were saving the mp3s - "File - Save as" there was a box checked that should not have been.  "Save extra non-audio information" was checked, and when we unchecked the box it has been smooth sailing ever since.

I'm not exactly sure what "extra audio information" means, but it was giving us problems.  Maybe Steve can enlighten me on what exactly that is, and why it was causing the problems it was. 

Hopefully your problem is just as easy to fix.

Randy.

***EDIT***  I just realized this was an old post.  Steve, if you could still answer my question that'd be awesome.  Thanks.
 

I'm not exactly sure what "extra audio information" means, but it was giving us problems.  Maybe Steve can enlighten me on what exactly that is, and why it was causing the problems it was. 

I think what happens might be this:

When you open a file - any file, regardless of what format it is - in Audition, it's converted internally to a wav file, and in that format it's possible to add things like cue marks, and have then saved at the end of the file as an additional 'chunk' of non-audio information. If you try to save a file like this as an MP3 it appears to try to covert this information to audio, which it isn't, and this makes the MP3 data file size information wrong, which is probably why the file won't play.

And if this is what really happens, then it's a bug, because it's preventable. But since it's in CEP2, then, er, don't hold your breath for a fix anytime ever!

I always have that flag checked, mainly because I want the que marker saved for other work, but I do prepare some files for mp3 encoding. Lots of cue marks are not uncommon. However, I use LAME to do the mp3 encoding, via a simple GUI front-end, and the extra audio content has never caused any problem there.

I like RazorLame but foobar2000 is more popular.