Subject:Comparing an old CD with a re-mastered version

Posted by: RickD

Date:8/1/2007 3:16:00 PM

I just bought a newly re-mastered edition of one of my favorite CDs. It certainly does sound better than the original. Much cleaner sound, better frequency distribution (not as tinny sounding or distant as before), and much hotter without sounding distorted.

What I want to do is run the same track from each of the discs through some sort of analysis that will tell me what the difference is. For example, the track on the new disc had its peaks rounded so that the overall level could be boosted, or a certain amount of compression was added so that the level could be boosted. Ideally, this comparison would be as comprehensive as possible. Has the stereo separation changed? Has the overall frequency content changed?

I am a programmer who does work similar to this for a living (albeit with data from strain gages and accelerometers rather than audio data) and I have always considered doing it myself. I know that if I try hard enough that I can get a lot of this information from SoundForge but I'm wondering if there is already some easy way of doing it automatically. Any ideas?

Subject:RE: Comparing an old CD with a re-mastered version

Reply by: Geoff_Wood

Date:8/1/2007 3:58:30 PM

Ears and eyes are the best way. Not sure how you could quantify the result otherwise. Watch each version on spectrrum analyser - this should should the difference in frequwncy content.

Looking at the track envelope should indicate if extra compression has been done. Stats (RMS level) has an overall bearing too.

I gues invert a copy of the whole track and adding will give a different 'signal' which would indicate some sort of 'amaount'.


geoff

Subject:RE: Comparing an old CD with a re-mastered version

Reply by: Geoff_Wood

Date:8/1/2007 3:58:51 PM

Ears and eyes are the best way. Not sure how you could quantify the result otherwise. Watch each version on spectrrum analyser - this should should the difference in frequwncy content.

Looking at the track envelope should indicate if extra compression has been done. Stats (RMS level) has an overall bearing too.

I guess invert a copy of the whole track and adding it to the original will give a difference 'signal' which would indicate some sort of 'quantity'.


geoff

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: RickD

Date:8/1/2007 4:27:58 PM

Thanks Geoff,

While I do agree that eyes and ears are one way of assessing the difference, I really do want a quantitative method which you seem to think is difficult to achieve. Excuse me if I sound a bit harsh, but my boss would never accept an answer like that if I told him to trust his eyes when looking at plots of strain and acceleration.

The inverted difference signal is one of the first things I thought of when imagining ways of how I would approach the problem, but there’s more to it than that. You would have to normalize the signals to each other somehow to minimize detecting differences where there really are none other than a level shift. Trivial maybe, but there are other considerations too and I would certainly address them if I were writing the software, but that’s not the point.

To avoid writing custom software myself I would like SoundForge (or some other package) to do it for me. My company’s customers are always amazed at the information we can extract from processing data received from electronic sensors and I’d like to see the same thing in an audio processing product.

Once again, thanks for your input. Any more ideas?

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: jumbuk

Date:8/1/2007 11:29:56 PM

Why not look at HarBal? I think you can download a trial version. This application is able to show you how the EQ profiles of two tracks compare. It is designed for mastering, and does a very good job. I bought version 2.0 and used it for mastering a CD I was producing at the time.

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: Geoff_Wood

Date:8/2/2007 3:30:01 AM

"but my boss would never accept an answer like that if I told him to trust his eyes when looking at plots of strain and acceleration."

- certaily not, but the equivalent would not be eyes and ears, but to trust his sense of touch, if the strain in question was within the normal range of human sensitivity !

"You would have to normalize the signals to each other somehow to minimize detecting differences where there really are none other than a level shift"

- sure normalise to the same peak level, then everything that does cancel is the difference.


Well, if you can somehow quantify the differences, we would all welcome your plugin ;-)

geoff

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: Geoff_Wood

Date:8/2/2007 3:30:02 AM

"but my boss would never accept an answer like that if I told him to trust his eyes when looking at plots of strain and acceleration."

- certaily not, but the equivalent would not be eyes and ears, but to trust his sense of touch, if the strain in question was within the normal range of human sensitivity !

"You would have to normalize the signals to each other somehow to minimize detecting differences where there really are none other than a level shift"

- sure normalise to the same peak level, then everything that does cancel is the difference.


Well, if you can somehow quantify the differences, we would all welcome your plugin ;-)

geoff

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: Geoff_Wood

Date:8/2/2007 3:30:44 AM

How did that happen ?!!

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: Chienworks

Date:8/2/2007 3:48:15 AM

Normalizing to the same peak level wouldn't be correct. You'd have to normalize to the same average level. One file might have a few peaks much louder than average. In that case normalizing to the peaks would make everything else different, even if they were originally identical.

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: RickZ

Date:8/2/2007 12:25:25 PM

I have occasionally "remastered" concert recordings I've made, using entirely different plugins, particularly when I've bought a new plugin. I think the idea of trying to figure out what processes were used on two different CD releases, by means of analysing the two files is not going to be successful, imho.

About the only way to find that out is to talk to the engineer(s), and hope they wrote down what they did. I try to remember to do that, but don't always.

FWIW . .

Interesting topic . .

Rgds,
Rick Z

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: Geoff_Wood

Date:8/2/2007 2:09:59 PM

But the 'average' level is going to be completely different too, which is why I have a conceptual problem with the whole, um, concept.


geoff

Subject:RE: Comparing an old CD with a re-mastered ve

Reply by: Chienworks

Date:8/2/2007 2:44:49 PM

Quite true that. Figuring out what is the "same" in both recordings may be darned near impossible. And that's the easy part of the process.

I suspect that this concept falls within the space of chaos, in that any meaningful representation of the data is going to be at least as complex as the original data. There may not be any simplification possible, which leaves one looking at the original data without any way to interpret it or summarize it.

There's also the possibility that the recording was resampled at some point. Perhaps the original was analog and two different digitizations were done. Maybe the original tape wasn't played at the same speed both times. Maybe the newer version was edited at 96KHz and then resampled back to 44.1KHz. In any of these cases there is no guarantee that the position of the samples in the stream corresponds between the two recordings. If that's true then any sort of mathematical difference comparison will be flawed.

My guess is that frequency distribution and dynamic range are about the only types of information that can be extracted. These will only tell you how the results are different, and won't necessarily tell you how those results came about. Of course, you could make some good guesses with your ears.

Oh wait ... wasn't that the original suggestion anyway?