Help with loudness normalization

I have folders with over 100 tracks each, and I need some advice on using the best settings to achieve a fixed volume throughout each folder?
MP3 Gain is no good because it requires a lot of clipping.

And I need to know the best settings using Loudness Equalization?

And want to set a volume to 0db the norm or just under, how do I achieve that also I don’t understand LUFS?

Should I use Perceived Loudness or RMS also.

If you have tried this and found out what works best, any help would be appreciated?

Also does it apply some form of hard limiting to very quiet audio?

Many Thanks,

Karl.

If you have tried this and found out what works best, any help would be appreciated?

Here is a procedure that will work without any limiting or compression.

  1. Normalize (or Amplify) all of the tracks for 0dB “maximized” peaks.

  2. Listen or run a loudness scanner* to check the LUFS level of each track (and keep notes).

  3. Choose the quietest track as your reference.

  4. Use the Loudness Normalization effect to match everything to your reference track.

MP3 Gain is no good because it requires a lot of clipping.

No, it doesn’t “require clipping”. There is an option to allow clipping. If don’t allow clipping (which is normal) some quiet-sounding songs won’t be boosted to the full target volume.

Setting a lower target volume** gives MPGain (or other procedure/algorithm) more “room to work”.

If you allow clipping the MP3 won’t necessarily be clipped because MP3 can go over 0dB. But if you play it at “full digital volume” your digital-to-analog converter will be clipped (if the MP3 goes over 0dB) so it’s usually best not to do that.

And I need to know the best settings using Loudness Equalization?

There is no “best” setting. There are broadcast standards and there is a standard (RMS level) for audiobooks. Typically, you need to use compression/limiting to hit those standards without clipping.

There are standards for CDs or MP3s. It’s up to the producer (or mastering engineer) to decide on loudness and dynamics (and the overall sound).

And want to set a volume to 0db the norm or just under, how do I achieve that…?

That won’t give you volume-matching. There are many quiet-sounding files with 0dB peaks. Perceived loudness is more-related to the short-term average (or RMS) and the frequency content. LUFS tries to measure perceived loudness but it’s complicated and two people might not agree when two different songs are “equally loud”, especially if one is highly-compressed (constant volume) and the other is highly dynamic (very-loud and very-quiet parts).

Since there are many 0dB normalized songs that sound quiet, you can’t simply make the quiet tracks louder to match the louder tracks. You have to make the loud tracks quieter. It’s a compromise… If you match all of your tracks to your quietest track you’ll probably find that most songs are too quiet.

IMO, the default MP3Gain and ReplayGain settings are a good compromise, but some people complain that the result is 'too quiet". (If you have enough analog gain you can just turn it up and it’s not a problem.)

… also I don’t understand LUFS?

…Should I use Perceived Loudness or RMS also.

[u]RMS[/u] (a special kind of average) is better than peak normalization. LUFS is even better. LUFS takes the [u]human frequency-sensitivity[/u] into account. (Mid-frequency tones sound louder than low-frequency or high-frequency tones of the same intensity.)

Also does it apply some form of hard limiting to very quiet audio?

Normalization does not use limiting. It’s simply a linear volume adjustment (like turning-up or down the volume control before the track starts) so it does not change the character/quality of the sound. (However, if you push peaks into clipping that’s a “bad kind” of limiting.)

You can use limiting (and or dynamic compression) with make-up gain to bring-up the volume. But, since limiting “pushes down” the peaks it changes the character/quality of the sound. You might like it and you might not.


\

** MP3Gain and ReplayGain use a dB SPL (acoustic loudness) reference (**+**89dB) which is different from the digital LUFS reference (such as **-**23dB). There are reasons for that but it gets confusing since we are working with digital levels (dBFS) and the actual SPL level depends on the speakers, amplifier, volume control, etc…

I have folders with over 100 tracks each

All that and it can depend on where the tracks came from. A very common complaint is from people trying to match LP transfers with the exact same song mastered in the last couple of years. That effort is pretty much doomed. The older work is going to have very gentle if any volume management where the modern version is likely to have very heavy volume compression to appear louder (and thus more successful) than everybody else.

Then there’s the magic of increasing the loudness without increasing the volume. This may be what you ran into. Matching loudness perception without causing the tops of the blue waves to crash. This is also where LUFS starts to be important.

There was a recent post from someone whose client insisted that the delivered waveforms hit 100%. Are you in that camp?

Koz

Thanks for the help still a little confusing, but it explains a lot. Yes there is a few vinyl rips in there that are very quiet. I will try the first suggestion, and maybe try and get the very quiet tracks from another source.

Thanks,

Karl.

How do I run a loudness scanner in audacity?

Karl.

An important question that, unless I missed it, hasn’t been stated yet: How will you be playing these files?

On a Windows 7 Ultimate SP1 64 Bit PC, Using Realtek HD Audio Card (24 Bits 48000 HZ, No sound effects or EQ), Windows Media Player 12, With manual EQ in the player.

Either with HyperX HX-HSCL-SR Cloud Silver Pro Gaming Headset, or Creative GigaWorks T40 Series II (2.0) Multimedia Speakers with MTM Audiophile Configuration and BasXPort Technology.

Or on android phone using Google Play Music manual EQ and Sennheiser HD 100 On-Ear Headphones.

How do I find what LUFS level my base track I’m going to use is?

Thanks,

Karl.