Mastering a CD: average levels?

[Llarion]

I have a question:

When you are taking a bunch of songs, that may have been recorded at different time, with different sonic palettes, and different final mastering techniques on a per-song basis; what is a basic rule of thumb in temrs of normalizing/leveling so that the songs all have a seamless flow from an average volume perspective? I'm fine with mastering single songs; gettign peak volumes up without clipping,a nd compressing enough where the average volume is solid, but when I did my first CD; I think I could have done better in the song-to-song relationships.

Is there a rule of thumb as to where to have RMS levels? Is there a particular tool or plugin that is good to help in this regard?

(I'm an Audition user, with Waves and T-Racks, I'm fond of using the L1 Maximizer, the C4 multiband compressor, and the 10 band paragraphic EQ)

 

[SouthSIDE Glen]

Is there a rule of thumb as to where to have RMS levels?

The only very general rule of thumb is that the optimal RMS level for any given song is directly linked to the "density" of the song itself.

Let me use a very extreme hypothetical to explain what I mean. Take two seperate raw ( by raw I mean no extra processing like EQ, compression, limiting, etc. of any type) recordings; one of an electric guitar power chord with heavy distortion and sustained into amp feedback, and another of an a cappella vocal rendition of "Amazing Grace", and set them both to peak at 0dBFS. Analyze the RMS level on both of them and you'll find that the guitar will intrinsically have a much higher RMS level than the a capella vocal.

The main reason for this is that RMS is an average level calculated over a period of time and not a snapshot of volime in a single instant. The guitar waveform averages a much higher volume because there is always sound there; there is very little decay in volume, the volume is at a relatively constant level and the vaveform never approaches zero. It is an extremely "dense" waveform over time. The vocal, OTOH, not only has variability in volume in the vocal itself (vowel sounds and percussive starts of some hard consonants hitting the peaks harder than the soft consonants and recessive syllables, etc.), but there are a lot of pauses in between lyrical lines, words and even syllables. This is a much less dense recording that's all over the place. The guitar looks more like a seismograph during an earthquake whereas the vocal more closely resembles a complex heartbeat.

You can compress and limit the vocal until you've squashed the life out of it, but you'll still have a smaller RMS because of the quieter sections of the recording. The vocal will sound like crapola with no dynamics, and you still won't have a matching RMS to the power chord with little or no processing done to it.

The moral of this story is that if you have a CD that has a mix of power anthems and emotive ballads, or a mix of denser and lighter instrument arrangements or anything else like that, it would be a mistake to try and master the CD to have equivalent RMS levels for each song. I know this phrase has become a tired cliche on this board, but in this case more an ever it's the accurate truth to day it's best to master the songs by ear and not by the numbers. In this case what that means is to arrange and master the compliation in such a way where the apparent volume between songs seems to match appropriately, and not in a way where you're forcing the numbers the equate.

And don't be afraid to let "Amazing Grace" even sound lower than the rest, regardless of the density of the arrangement; it can bring out better emotion if it's at a different listening level than "I Want To Rock And Roll All Night"

G.

 

[Llarion]

The only very general rule of thumb is that the optimal RMS level for any given song is directly linked to the "density" of the song itself.

Let me use a very extreme hypothetical to explain what I mean. Take two seperate raw ( by raw I mean no extra processing like EQ, compression, limiting, etc. of any type) recordings; one of an electric guitar power chord with heavy distortion and sustained into amp feedback, and another of an a cappella vocal rendition of "Amazing Grace", and set them both to peak at 0dBFS. Analyze the RMS level on both of them and you'll find that the guitar will intrinsically have a much higher RMS level than the a capella vocal.

The main reason for this is that RMS is an average level calculated over a period of time and not a snapshot of volime in a single instant. The guitar waveform averages a much higher volume because there is always sound there; there is very little decay in volume, the volume is at a relatively constant level and the vaveform never approaches zero. It is an extremely "dense" waveform over time. The vocal, OTOH, not only has variability in volume in the vocal itself (vowel sounds and percussive starts of some hard consonants hitting the peaks harder than the soft consonants and recessive syllables, etc.), but there are a lot of pauses in between lyrical lines, words and even syllables. This is a much less dense recording that's all over the place. The guitar looks more like a seismograph during an earthquake whereas the vocal more closely resembles a complex heartbeat.

You can compress and limit the vocal until you've squashed the life out of it, but you'll still have a smaller RMS because of the quieter sections of the recording. The vocal will sound like crapola with no dynamics, and you still won't have a matching RMS to the power chord with little or no processing done to it.

The moral of this story is that if you have a CD that has a mix of power anthems and emotive ballads, or a mix of denser and lighter instrument arrangements or anything else like that, it would be a mistake to try and master the CD to have equivalent RMS levels for each song. I know this phrase has become a tired cliche on this board, but in this case more an ever it's the accurate truth to day it's best to master the songs by ear and not by the numbers. In this case what that means is to arrange and master the compliation in such a way where the apparent volume between songs seems to match appropriately, and not in a way where you're forcing the numbers the equate.

And don't be afraid to let "Amazing Grace" even sound lower than the rest, regardless of the density of the arrangement; it can bring out better emotion if it's at a different listening level than "I Want To Rock And Roll All Night"

G.

That's pretty well what I thought. Thanks! In my toolbox, though, I cannot seem to find a tool to analyze RMS level over time... What do you suggest? I probably have six of them and don't even know it

 

[SouthSIDE Glen]

That's pretty well what I thought. Thanks! In my toolbox, though, I cannot seem to find a tool to analyze RMS level over time... What do you suggest? I probably have six of them and don't even know it

By definition, RMS is a reading that's measured over time. Any RMS meters or measuring devices you have are automatically doing that. The length of time over which they are measuring depends upon the tool. Many RMS level meters are showing the moving average over the last few milliseconds of playback. OTOH some RMS calculators that give you a single number are analyzing the entire track (or highlighted section of a track) and giving the average for that entire period of time.

G.

 

[Llarion]

Any RMS meters or measuring devices you have

AHA!!! We found the key! I honestly don't know if I indeed have one. Example names, please?

 

[SouthSIDE Glen]

AHA!!! We found the key! I honestly don't know if I indeed have one. Example names, please?

Oops, sorry, I misunderstood .

Audition has a good RMS caclulator built into it. In the editor window, highlight all or part of your track. Then pull down the "Analyze" menu from the menu bar and select "Statistics" from that menu. It will analyze the highlighted section and report "Minimum RMS", "Maximum RMS", "Average RMS" and "Total RMS". To be honest I have not chased down just what the difference is between "Average" and "Total" RMS, but the "Average" figure seems to agree the most with the values reported by other tools (e.g. Sonic Foundry's RMS calculator), and is the one I refer to the most.

G.

 

[Llarion]

Perfect! That's EXACTLY what I was hoping for! Woo hoo!!! Tanks man, I'll experiment with it!!

 

[scrubs]

The audition tool is good. Voxengo has a plugin called SPAN and Elemental Audio has one called Inspector that are both good (& free).

 

[$a1Ty]

yeah audition is mad, the noise reduction is the bomb among other things

 

[mshilarious]

To be honest I have not chased down just what the difference is between "Average" and "Total" RMS, but the "Average" figure seems to agree the most with the values reported by other tools (e.g. Sonic Foundry's RMS calculator), and is the one I refer to the most.

I'll take an absolutely wild guess that average is the mean of the RMS samples (50ms or whatever), whereas total is the RMS of the entire selection. My algebra is no longer good enough to prove it, but they aren't mathematically equivalent. There is probably some psychoacoustic reason that average RMS is a superior measurement, I imagine.

 

[SouthSIDE Glen]

I'll take an absolutely wild guess that average is the mean of the RMS samples (50ms or whatever), whereas total is the RMS of the entire selection. My algebra is no longer good enough to prove it, but they aren't mathematically equivalent. There is probably some psychoacoustic reason that average RMS is a superior measurement, I imagine.

Yeah, I figure it's something like that myself. I haven't bothered chasing it down because 1) I have a life and 2) the "Average RMS" reading is almost always exactly what what is reported by the RMS calculators in other tools, so for the sake of continuity, I figure that's the one to watch.

Not that I really watch RMS levels - especially in Audition - all that much. I occasionally look more as a curiosity and mental reference than anything else during my premastering, which I usually do in Sound Forge.

G.

 

[Llarion]

yeah audition is mad, the noise reduction is the bomb among other things

Amen to that...

I have a preset filter in there; I recorded my empty room with all the machinery running, and sampled that noise and saved it as a profile, and now I apply it to any open mic recordgin I do, and bam, room noise is gone!!

 

[Llarion]

I'll take an absolutely wild guess that average is the mean of the RMS samples (50ms or whatever), whereas total is the RMS of the entire selection. My algebra is no longer good enough to prove it, but they aren't mathematically equivalent. There is probably some psychoacoustic reason that average RMS is a superior measurement, I imagine.

Certainly seems reasonable to me. Can't wait to try this stuff out. That's what I love about this joint; you all always re-enegize me when I'm in a rut...