How Much Am I Losing?

[timsilva]

Okay, on a previous post Spoon said we always lose some quality during conversion when using dBpoweramp.

What I would very much appreciate from you folks is some clarity so I have a better understanding on what to--and what not to--expect from my purchase.

First, how much am I losing? The only reason I bought the program was to "clean" some old MP3s. If I apply DSP actions, say: Fade-in/Fade-out/Volume Normalize on a 128kbps song what do I have at the end? a 112kbps? 96kbps? Even though the target was 128kbps CBR? So that my new "cleaned" 128kbps song is in fact a 96kbps? It would be nice to have some idea, some transparency.

second, is this losing of quality compounded? The more Effect/Action applied on a single section the more quality is lost?

Once again thanks.

Tim

[Spoon]

Re: How Much Am I Losing?

There is no concise answer on the internet, even if you convert a 128kbps mp3 to 320 kbps you still loose a little quality.

Quality is lost when doing the conversion, so running it multiple times looses quality. Fade and volume normalize if done correctly would not loose quality, only the conversion.

[Porcus]

Re: How Much Am I Losing?

Okay, on a previous post Spoon said we always lose some quality during conversion when using dBpoweramp.

"when using dBpoweramp". Beware that quality loss when transcoding will happen regardless of which program you use.

If you were using lossless files -- WAV, FLAC etc -- you could transcode to another lossless format without losing quality. If you are using lossy formats -- like mp3 -- then what will happen, is (with exceptions noted by Spoon) the following:
- it will unpack the file to the bitstream you hear (i.e. like to a lossless format except that losses will be preserved as well)
- it will then be processed
- then when saving to mp3, the encoder will remove part of the audio information just like when it was encoded the first time.


Every mp3 encoder does this. You can object with "why does it have to remove information, it has already been done once?", and the answer is that there is no known way to avoid it, much like there is no known way of compressing a strongly encrypted file: it is only possible in theory. (In theory it is even simple -- just try all possible mp3 files and pick the smallest which decodes to your signal. Needless to say you will be waiting til doomsday, just like you would if trying to brute-force hack a million-character password.)


Some operations might be performed with negligible quality loss though, as Spoon indicates.

[Porcus]

Re: How Much Am I Losing?

By the way, suggested solutions:

- If you can avoid transcoding, avoid transcoding. Rather configure your player to do the processing while playing. For a (Windows-based) computer, Foobar2000 has a fair bit of capabilities. It will take you a bit of time to learn it though, but Foobar2000 and dBpoweramp are the reasons this computer is still running MS-Windows.

- If you are looking for volume normalization: Foobar2000 can also do that by tags (which does not alter the audio itself). This is however (as far as I know) a non-standard way which will not be recognized by other players, but it is harmless.

- If you want to do conversion: Keep the original. In a write-protected folder. If you ever need to do a second conversion, convert from the original and not from the copy.

[GeorgeButel]

Re: How Much Am I Losing?

MP3DirectCut is claimed to do certain simple mp3 edits losslessly--splitting the file, deleting parts of it, or normalizing it. It supposedly does not do any recompressing, which is where your quality loss comes in--the same thing that happens when you re-save mp3s or jpegs. The interface is awkward, and the normalizing only has discrete settings if you are not happy with its default. There is no fine tuning on the normalize function. If, for instance, it finds a peak of -1.4 db, it won't normalize it, but you can tell it to, but you can only select a setting of +1.5 dB (it goes in multiples of 1.5). The wave form displayed in the interface is warned about when you open it up: take heed, and open the file in something else, like Sound Forge, if you want to delete what you might think is "silence" and make sure you aren't messing anything up.
If you want to do any other audio editing, my suggestion would be to convert to a lossless format such as wav or flac, do all your editing in that, and then convert back to whatever lossy format you desire. The other programs don't give me much choice about mp3 format, so, when necessary, I will not save the edited file in the other program except in the lossless format I used to begin with, then, when I'm through, I'll use dB Poweramp to convert it back to mp3. An example of where I would have to do more complex editing would be if, say, I had recorded off of an old tape or record, and had forgotten to do something like dc offset, or needed to do some kind of complex cleaning (Sound Forge keeps getting better and better hen it comes to that), and didn't want to re-record it. Obviously, saving the original file in a lossless manner would eliminate the problem.