Topic: Normalized recordings still sound quieter than studio albums.

[optimisticpessimist]

Well... the title pretty much says it all.  I EQ, dynamically compress, and normalize my recordings, yet when I burn them to CD and play them on any system, they consistently are a fair bit quieter than studio CDs.  In my car (which is only a factory stereo system, so obviously not the best guage) I sometimes can max out the volume on the stereo just to enjoy some of these concerts at a loud volume, whereas with almost any studio CD, I could probably never turn the volume that high without damaging my speakers and my hearing.

Is this common?  And btw I'm running an old version of Wavelab.  I believe it's 3.04, but I'm not at my home computer at the moment.

Thanks for the input.

[Javier Cinakowski]

It is common.  Search "loudness wars" on google.


Here is the first link to get you started:
http://en.wikipedia.org/wiki/Loudness_war

[nickee]

Using wavelab you can calculate the average RMS and use it as a comparison of how loud different recordings will be perceived. In wavelab 5 it's under analysis/global analysis (or just press Y) then press analyse and check the results in the loudness tab.

[boojum]

Compression.

[Belexes]

New CD's with all that compression fatigue my ears.  Then I put on DSotM and all is good again.

[Gutbucket]

Then beyond the limiting and loudness processing for studio releases is even more dynamic range compression and peak limiting for broadcast. 

Everyone is familiar with the sound of extreme limiting used for commercial segments where everything is far louder than the program material - commercials screaming at you even with the level turned down.  Yet the peak levels in those commercials are no higher than those of the program material. Broadcasters keep an iron grip on their peak levels (and are forced to by law).

What is a bit more subtle and interesting for me to listen for is the program compression used by broadcasters.  The effect is most evident on the radio. I rarely listen to commercial music radio stations, but occasionally when driving home I'll tune to a station that plays older classic rock - stuff that was not originally super compressed and limited to death but had a relatively wide dynamic range.  The highly advanced program compressors that the station uses to make the dynamics more appropriate for clock radios and car listening dramatically compress the micro dynamics of the music; so much so that it is amazing the music doesn't sound completely smashed like some modern loudness war casualties.  Example- driving home last night I turned on that 'AOR classic rock' station and they were predictably playing a bunch of 70's era Pink Floyd.  I remember spending long hours back in high school listening intensely over headphones for the little background noises and voices that were barely audible little sonic treasures buried within that music.  Those things could never be heard over the radio back then. Yet on my drive home last night I could hear every little sound effect and whisper over the highway speed road noise at nearly the same apparent loudness as the lead vocals!  And although the slam of the dynamic impacts of the music was reduced significantly, it didn't sound overly 'hard' from like severely limited and sometimes hard clipped modern releases.

I prefer the full dynamic range version which have much greater depth, impact and convey more emotion and intrigue with the hidden parts still hidden, but still I sat there impressed that such a major change of dynamics was achievable with enough transparency that most people would never even notice the difference unless it was carefully pointed out to them.  I think that mostly says something about the incredible work that must have gone into the algorithms used in those broadcast compressors and also of the ability of that older material to take such treatment gracefully- material which was not released with RMS levels so high the needles on the VU meters are bent around the pins.

It's an interesting thing to listen for if you're so inclined.

[Javier Cinakowski]

Gutbucket brings up a good point.  A nice thing about this modern compression is that in noisy environments (car, work etc...) the delicate parts of the music make it through.  It would be nice if music was released completely uncompressed, then we had the choice to add it in playback.  I support a standard like "Replay Gain" so I can decide when and where I want smashed waveforms....

[Gutbucket]

Yep.  Would be nice if part of all digital broadcast standards was putting the compression options in the player.  Then the user could choose their compression scheme themselves if they care to delve into the menus and are interested in quality, or let the player choose the best default scheme for them (car, night time movie, portable player, clock radio, full dynamics, etc).  Not sure if that is the case or not.  Unfortunately, probably not.

[newplanet7]

Does anyone here use RMS norm?

[Gutbucket]

The second problem is that consumer players can't really manage the DSP necessary for reasonable "quality" compared with a mastering shop, provided that the mastering compression & limiting are even done digitally . . .

Can't be that much worse than the ultra-compressed stuff and the target for it would be compromised listening situations like noisy cars and tiny clock radios anyway.  People that care about quality could turn the crappy DSP compressor off and enjoy the minimally compressed version without sny in-player processing.  People that don't won't notice or care, probably won't ever know the difference. Everybody wins.

Curious about the phase rotation techniques.  Is that a phase shift for a particular frequency band? Isn't that what occurs with typical minimum phase EQ? Constant phase shift across all frequencies?  - the kind of thing that makes matrixed surround encoding work? That would do weird things for people listening with home surround decoders so I guess not.  Thanks for any light you can shine on this for me.

[echo1434]

First of all, Yes... All modern CDs are MEGA LOUD due to the loudness war, but I'm glad that most audience recordings are NOT like this. 

But aside from that, I've never understood why people rely on normalizing to make their recordings louder — because all it takes is one stray peak to screw up the entire normalizing measurements.

Just to make a simple example, let's say that somebody clapped really loud and it was recorded at -1 dB RMS, but the music only peaks around -6 dB RMS on average. So you go to normalize, and it finds that -1 dB peak and therefore only amplifies your recording by 1 dB. Really not a very "intelligent" measurement there.

That's just a really generic example, but most music has unimportant peaks that will inevitably limit the amount you can increase the volume with simple normalization.

My solution has always been to use the hard limiting feature Adobe Audition (or Cool Edit Pro). I basically find the loudest song in the performance, and see how much I can boost it without significantly compromising the dynamic range. There's a bit of an art to this, but it's not very difficult if you know what you're doing. Anyway, once I find this number I apply this amount of boost to the entire recording. And believe me, the results are ALWAYS better than what I would get with normalizing!

This is basically what they did to albums in the early '90s, before they started getting crazy with the volume. Most of those recordings were louder than stuff from the '80s, but they sounded great.

There are other good limiters out there, like Waves L3 (which will work as a plug-in with most editing programs). I'm sure there are others, but those are the ones I'm most familiar with.

[EarlyMorningRain]

My solution has always been to use the hard limiting feature Adobe Audition (or Cool Edit Pro). I basically find the loudest song in the performance, and see how much I can boost it without significantly compromising the dynamic range. There's a bit of an art to this, but it's not very difficult if you know what you're doing. Anyway, once I find this number I apply this amount of boost to the entire recording.

this is more or less the method I do, however I accomplish this in a slightly different manner than you describe.

Just to make a simple example, let's say that somebody clapped really loud and it was recorded at -1 dB RMS, but the music only peaks around -6 dB RMS on average. So you go to normalize, and it finds that -1 dB peak and therefore only amplifies your recording by 1 dB. Really not a very "intelligent" measurement there.

but in this example, if you used your approach I quoted above (hard limiting), then you would only be amplifying the recording by 1 db, so you'd end up with the same results as if you normalized (vs hard limiting), no? Or are you lowering that clap that peaked at -1db?

[newplanet7]

I usually run conservatively at about -6 to-4.
If I get a huge peak above that I just hit them up manually
with a pencil tool and minimize them by redraw.
Then use peak norm to .1.
Pretty easy.

[optimisticpessimist]

Interesting info regarding commercial recordings... Of course I knew there is a lot of compression used in making them, but I didn't realize the full implication of that process.

And I guess I technically misspoke when I mentioned "normalizing."  I use the Waves RCL plugin on Wavelab to apply the dynamic compression and gain to the entire recording.  I never use a ratio higher than 2.5 or 3, and just scan through the entire recording, playing with the threshold and gain until I feel comfortable with it.  I'd say the loudest parts of most of my recordings still fluctuate between 0 to -3 dB.

Now that I think about it, though, this conversation leads me to another question.  In the master section of Wavelab, I always apply EQ first, then sometimes a stereoimager, then the dynamic compression, and finally a plugin called ultramaximizer, which I mostly use to make sure I haven't applied too much gain and my levels aren't clipping.  What I'm wondering is, should I be applying EQ and processing the file in one step, and THEN working on the dynamic compression and gain in a second step, or is it ok to apply all of these plugins in one processing step?  I feel like I'm accomplishing what I want to accomplish effectively, but I want to make sure I'm not overlooking something, or misusing my DAW.  Thanks again.

[Dede2002]

I usually run conservatively at about -6 to-4.
If I get a huge peak above that I just hit them up manually
with a pencil tool and minimize them by redraw.
Then use peak norm to .1.
Pretty easy.

That's what I do.