So I noticed a new AGC feature in v3. Is this the same as VolAdj, just built into the player now? I tried searching for this on the forums but didn't find much on it.

Messing around with it, I noticed AGC has more of an effect on the signal than VolAdj does.

Yeah, it's dynamic compression. It's a different algorithm than is used in Hijack, so it's possible people will prefer one or the other. I personally think that once v3 is stable enough to be the player of choice for everybody, Mark should dump the VolAdj stuff from Hijack. Unless there's a real marked difference between the two algorithms... I personally don't like the dynamic compression approach, and would prefer a fixed per-track gain adjustment applied based on the RMS or peak level of the track. Unfortunately the one time this was tried, it caused problems, and it never got a chance to see the light of day in a finished form. I'm crossing my fingers that it might make it in some day, though.

The bulk of the code for a static per-track volume adjustment is already in hijack, in the form of the volboost_dsp parameter. It shouldn't be too difficult to provide an ioctl to control this from userland.

But unlike when the player did it and calculated it the first play, userland approach would need to precompute it.

Certainly true. I wonder if there's enough information in the dynamic data partition to do this...

Is there any reason why replaygain (mp3gain, vorbisgain, &c.) shouldn't be used for this? It's pretty well established in other parts of the digital music community, deals with many of the issues involved (e.g., heavy compression), is being worked on by the empeg crew for the karma? The website is a little out of date, but the background and rational are there. The biggest problem is the complete clusterf--- of mp3 tagging standards.

Once the issue of how to best tag an mp3 file is figured out (last word was that it worked for newer lame encoded files) it should be rather simple to write a userland program that amalyzes and tags the tune file that could be accessed via a hijack menu. It would of course take a while to analyze and tag thousands of files, but can easily be done on the pc side before syncing files to the player.

--Nathan

yn0t_ looked at the dynamic data partition and discovered the adjustment info wasn't being set anymore, fwiw.

I personally think that once v3 is stable enough to be the player of choice for everybody, Mark should dump the VolAdj stuff from Hijack. Unless there's a real marked difference between the two algorithms...

Currently there is a marked difference between the two algorithms. Both work on a similar principle, but take slightly different approaches and result in different kinds of "feel" to the compression. Some people will prefer one over the other. Also note that Hijack's parameters are adjustable.

Yeah, the normalization field is gone, but the peak data (for Info:Seek) is still there. I don't know nearly enough about the seek tool data to be able to figure out if it could be accurately munged into a single number that represents the overall volume of the track, though.

If people were willing to deal with using a program on their PC to tag all of their files with a number that represents the amount of gain that ought to be applied, it would be pretty easy for me to cook something up in emphatic that raises or lowers the volume when the track is played back. The thing is, that's a large ordeal for people with large music collections. Even if we had a way to do it while the song is being played back on the empeg, we can't write to the files unless we mount the drives read-write, which is a bad idea. The thing about the original approach that was tried in earlier player releases is that it was done automatically on the first pass through the track, and stored on the dynamic data partition for us.