I have an idea that might or might not work.

The problem is that a lot of old music (80's etc) are mastered with slightly less bass and treble than they would have been today. If it is possible to measure the amount of bass/treble used today and create a reference-point, it might be possible to create a EQ curve that should be applied to the old content to match it up with most modern records.

This EQ setting could easily be defined in tags, and applied by a EQ DSP during playback.

I think the algorithm for this would be much like the replaygain algorithm, just using multiple bands, but I have no technical knownledge to back this up. I would like to hear any positive thoughts on this: Would anyone benefit from this, is it a good idea at all, and how would it be possible to implement?

Edit: I did some editing of this initial post, since some tests showed that... VL are crap nonetheless Still I think this might work, and like your oppinions on it.

This is based on previous experiences I have from using a discontinued DSP called "VolumeLogic".

VolumeLogic is a multiband compressor/limiter that uses a special algorithm to "remaster" the music while playing. Now, many of you would probably not care or start a flamewar that such DSP's are fake, crap or just useless - Generally I share the same oppinion, but for once I actually think it made a pretty good job much of the time, although I think It's design was flawed.

This plugin is great especially on older songs that lacked bass and treble, while it didn't do much to most newer songs that were already well mastered. However, I dislike that it compresses the music, and also that this process is done on-the-fly, instead of statically on the entire song. Also it seems that this plugin introduces some sort of a saturation filter that gain a lot of noise on many tracks.

I for myself don't have the need or urge to use such an algorithm, but maybe because I never experienced the mentioned VolumeLogic.

As for implementing: Perform an Fourier transform (spectrum analyzer) on the signal, measure the intensities of the "high" and "low" frequencies (boundaries have to be specified), and after that according to measurement write some metadata-info about the measured levels. Then use a plugin which reads this info and EQs the levels up to the desired level. But since I'm no programmer I wouldn't know how do that, or if it is the right idea, even.

VolumeLogic doesn't saturate filter at all, it doesn't even have a special algorithm. It's just AGC > Multiband Comp > Limiter and that's all the processing that happens.

The real problem with VolumeLogic is that you can't fine tune the settings, it only allows you to use presets. It's basically exactly the same stuff that radio's use in their broadcasting chain (except in software form).

This online radio uses the exact same software as VolumeLogic except with configurable settings, so they've made it sound a lot better than any of the presets that come with the commercial version.

We've just done a few tests with VL, and it seems it applies +20db to low and high bands by default (I thought that was variable).

Still, I'd like to know if something like this is doable.

Hmm, I think I'd be more interested in making new music, with its overcompression and the overused smiley-face EQ sound more like older, more naturally processed and EQ'd music. Well, different strokes for different folks, of course

Time to bump this thread, as it's still bugging me



Actually my idea would also be able to do this. If you have a reference point that are flat (say non-smiley), it would revert the smiley-made music to a more flat norm.

I've done some test, by looking at the frequency response in Audition, and trying to equal it to the frequency response of another track (say a reference track) - just for fun Of course it didn't turn out very good. The bass alone were significantly overboosted, compared to the amount I would have wanted.

Maybe I just didn't do it right, or is missing some kind of logic here.

Someone with remastering experience might have some knownlegde to if it's possible to do some analysis and automatically apply a proper EQ?

With Har-Bal you can analyse a track (in your case the "old" track) and match its frequency response to another track (the "new" one).

You could do that with some "reference" albums to get a general "80s-to-now" EQ filter.

Then the easiest way to use that filter in apps like foobar2000 is creating an impulse response with the EQ filter and use the impulse in the convolver.

This method is non-adaptive, of course.


example:

Underworld - 1996 - Second Toughest in the Infants
to
Underworld - 2007 - Oblivion with Bells

...gives this EQ filter, so Oblivion with Bells has more bass but less highs. This leads to this impulse (created with Sound Forge Acoustic Mirror), which can be used with foobar2000's convolver.


Do you have any "reference albums" in mind?

It looks very promising. Can't wait to try it (currently at work)!

I played a bit with this, and just for fun I tried "correcting" a non-remastered track with a remaster with remarkably more bass (Enya: Ebudæ - From the original album and the one from 'Paint the sky with stars')

I would expect it to automatically adjust them to be equal, but although it seem to adjust the mids fairly well, it does close to nothing with the lower bands.

I may have to do a little more testing, but unfortunately it doesn't seem to handle this very well by it's own

After you have opened a track and a reference track select the IntuitMatch cursor. It looks like this:

+-+
M

Then click somewhere on the left of the spectrum, hold down, and draw to the far right, then let go.