Audio CD distortion

[Firewave]

I use ImgBurn from time to time to burn audio CDs and once in a while I have a CD, that is seriously distorted. The sounds files are actually okay. Unfortunately I couldn't really emember, which files were cuasing it. I ran into this issue again recently and started investigating. The files are once again okay. I tested them with Winamp, MPC-HC and GraphStudio - the two latter are using the installed ffdshow filters like ImgBurn does. I first suspected it's the file since it is just 22kHz, but even resampling the file with ffdshow to 44kHz didn't lead to this distortion. And I am quite positive it happen with 44kHz files in the past. So is there a way for me to reproduce how ImgBurn converts with the file with a sample application, so I can find out what is actually at fault?

 

I attached a screenshot with the waveforms of the original MP3 and a WAV, that I ripped from the Audio CD using CDex.

[Firewave]

I just tried it with a 22kHz WAV file created from the initial MP3, which was playing fine and I get the same distortion on the Audio CD. So there's either a bug or it is a limitation (I am not familiar enough with audio to say). If it is the latter, maybe the "Create CUE File..." could warn about those files.

 

I forgot to mention that I am using 2.5.2.0 on Windows XP x64.

[Firewave]

I found one of the earlier cases. This time it's a file, that is in Mono. It doesn't cause such a bad distortion as the other case, but it still results in lower quality.

 

See the attached waveforms:

[LIGHTNING UK!]

It's probably the 32bit float stuff that's messing it up.

 

When you burn to disc it has to be 16 bit.

 

If there's a problem going from 32 bit to 16 bit then the people that make your filters need to be told.

[Firewave]

These waveforms are comparing a 44kHz Stereo MP3 against the WAV I extracted from the burned disc. It looks almost identical to me, so I doubt it is the 32-bit float stuff.

[LIGHTNING UK!]

So humour me and just try feeding it the same song twice (as track 1 and track 2) - one that's that using 32bit float and one that's 16 bit PCM.

 

Obviously you'll need to check that both sound ok before burning

[Firewave]

I have to find a way to convert it to 16-bit first. Audacity lets you change it and you can say it as 16-bit WAV, but when you open it, it still says 32-bit float.

But I don't understand how the filter can be at fault when it also happens with a 32-bit float WAV file? And why it only seems to happen with lower quality files (mono, 22kHz)?

[LIGHTNING UK!]

Now you've got me confused!

 

Anything that's not 44kHz, 16 Bit, Stereo will be subject to some form of conversion by one or more filters that DirectShow pulls into the graph.

 

All ImgBurn does is say 'Here's a file, I want the output in 44kHz, 16 Bit, Stereo, do it!'. What happens after that is all down to DirectShow.

[Firewave]

I am confused myself, but to sum it up:

 

MP3, 22kHz, stereo -> serious distortion

WAV, 22kHz, stereo -> serious distortion

MP3, 44kHz, mono -> decreased quality

MP3, 44kHz, stereo -> no difference

 

I don't know what to make out of the 32-bit float information from Audacity, so I didn't list it.

 

All files are playing fine with GraphStudio, which appears to be just an enhanced version of GraphEdit and is also just using the DirectShow filters. And ffdshow on my system, which is used to decode the MP3s, is even configured to output 16-bit PCM by default.

 

Could you provide some sample code or program, that performs the conversion, so I could see if the output is already distorted?

[LIGHTNING UK!]

Check your PMs