Hi folks --

Sorry to ask yet another normalization question, but. . .

I'm using SF to do some mild pre-processing of pre-recorded music for
an audio cd, and I'd like to normalize the tracks so that they have a
similar apparent volume across the cd. My understanding is that RMS
normalization is a good tool for this.

I'm used to using my ears and eyes to check for clipping and distortion
and such, but I've run into some confusion over SF's "dynamic
compression." Basically, if SF is compressing the peaks so they don't
distort, how should I be determining if I've set normalization too
high?

I think that if I set it too high and then ask SF to "dynamically
compress" the peaks, then I'm essentially flattening the overall
dynamic range of the clip, but that's just an inexperienced guess, and
I'm not sure how to tell how high is too high or when is not enough.

Just wondering the standard by which the more experienced among you
decide "higher. . .higher. . .nope that's too high" when it comes to
RMS normalization. And understand, I'm am interested in theory here,
but I also need practicality - tips, tricks, what you look for, etc.
Thanks very much!

-- Dave

" wrote in message
oups.com...

Just wondering the standard by which the more experienced among you
decide "higher. . .higher. . .nope that's too high" when it comes to
RMS normalization. And understand, I'm am interested in theory here,
but I also need practicality - tips, tricks, what you look for, etc.
Thanks very much!

I would venture to say that very few of us use RMS normalization; we adjust
the relative level of our tracks by ear, which is the only effective way to
do it. I don't usually make statements quite as absolute-sounding as that,
but in this case I am.

Peace,
Paul

" wrote in message
oups.com
Hi folks --

Sorry to ask yet another normalization question, but. . .

I'm using SF to do some mild pre-processing of pre-recorded music for
an audio cd, and I'd like to normalize the tracks so that they have a
similar apparent volume across the cd. My understanding is that RMS
normalization is a good tool for this.

The perception of loudness depends on a lot more than RMS amplitude.
Peak-to-averge ratio matters, and so does the distribution of frequencies in
the music. For example, a song with a lot of energy below 100 Hz and above
10 KHz can have the same RMS value as a song with a lot of energy in the 2 -
4 KHz range, and the latter would sound a whole lot louder. A song that is
consistently moderately loud generally sounds louder than one that has a few
loud short peaks that kick up the RMS value of something that is mostly
softer.

I'm used to using my ears and eyes to check for clipping and
distortion and such, but I've run into some confusion over SF's
"dynamic compression." Basically, if SF is compressing the peaks so
they don't distort, how should I be determining if I've set
normalization too high?

What sounds good to you.

I think that if I set it too high and then ask SF to "dynamically
compress" the peaks, then I'm essentially flattening the overall
dynamic range of the clip, but that's just an inexperienced guess, and
I'm not sure how to tell how high is too high or when is not enough.

Try several different things and compare them to determine which you prefer.

" wrote in message
oups.com...
Hi folks --

Sorry to ask yet another normalization question, but. . .

I'm using SF to do some mild pre-processing of pre-recorded music for
an audio cd, and I'd like to normalize the tracks so that they have a
similar apparent volume across the cd. My understanding is that RMS
normalization is a good tool for this.

Yes. But SF's Wavehammer is better still.

I'm used to using my ears and eyes to check for clipping and distortion
and such, but I've run into some confusion over SF's "dynamic
compression." Basically, if SF is compressing the peaks so they don't
distort, how should I be determining if I've set normalization too
high?

If you are using the straight normalisation with an average RMS weighting,
you can Scan Levels and/or select If clipping occurs = "Normalise peak
value to 0dB" .


I think that if I set it too high and then ask SF to "dynamically
compress" the peaks, then I'm essentially flattening the overall
dynamic range of the clip, but that's just an inexperienced guess, and
I'm not sure how to tell how high is too high or when is not enough.

Yes, flattening the dynamic range is exactly what you are trying to acheive.
You have to flip between all the different tracks to compare.

Just wondering the standard by which the more experienced among you
decide "higher. . .higher. . .nope that's too high" when it comes to
RMS normalization. And understand, I'm am interested in theory here,
but I also need practicality - tips, tricks, what you look for, etc.
Thanks very much!

Purely down to personal taste.

geoff

Paul, Arny, Geoff, thanks for the great answers guys. I've shipped the
project (not critical, a cd for friends, but I did want it to sound
okay) and I've learned a lot both from your responses and fiddling
around. Especially the very interesting differences in perceived
volume between tracks of different types that have a similar RMS value.


I'm not sure if it makes any sense algorythmically (sp?) but I was
noticing that tracks with light music and a spoken voice sounded much
louder than a guitar-heavy rock track at the same scanned RMS value.
Now that my deadline is over I'll have a little more time to dive into
the practicalities of what the whole root/mean/square thing actually
means just to understand it going forward. Then I'll use my ears like
always :-)

I'll also look into Wavehammer, because it's never too soon to get
another piece of software!
Again, thanks for taking the time to respond.

-- Dave