rkeeney wrote:
> Pat Farrell;248423 Wrote: 
>> Normalization is a sin. Don't do it.
> 
> Please explain why not.


First, what is normalization?
PCM data are 16 bit values from in the range -32K to +32k

Lets assume you have recorded a signal (doesn't matter if its from a LP 
or a $5000 Neumann mic, etc. that have values from -8K to +8K

This means you are wasting about 12dB of signal, as you could have 
recorded it with more level boost and it would have been from  -32K to +32k.

Now go through all the 16 bit values, and multiply them by 4. This 
"normalizes" the values so they are all nearly dbFS (decibels full scale).

In practice, this sounds nearly the same as cranking your amp's volume 
knob up a bit.

But its not really the same as cranking up the record levels when you 
capture it.

Supposed you had a lot of noise in your signal chain. Say -4K to +4k.
Your signal to noise ratio is 2, since a typical "music" signal is -8K 
to +8K, and the noise is -4K to +8k, divide, and get two.

When you "normalize" you raise the music values, and also raise the 
noise values. The music is louder and so is the noise.

What you want to do is record at a sufficient level that the music 
naturally approaches dbFS. Then the noise will still be at whatever 
level it was, and your signal to noise ratio, which translates into 
quality listening, is better.

Back in the olden days, analog tape would saturate when you over drove 
them, and end up being a slight compression. Digital recording doesn't 
have this nice soft compression, digital clipping is terrible sounding 
and will destroy speakers in a second or two.

The good news is that for most LP playback, the chain is fairly limited, 
so you can setup the levels to almost hit 0dB FS, without going over it.

In a live studio, you typically use a limiter on the channel to handle 
the occasional peaks.

The standard term for this is "gain staging". You want to set the gain 
at each place in the recording chain to have the best signal to noise 
ratio, while also not overdriving any component.

For LP recording, you can play the LP once, record the signal, and then 
use your audio tools to find the max (or min) level. Then adjust the 
levels to that you just kiss up to the max. A record is nice that way, 
it sounds the same every time you play it. Not so with a singer or 
trumpet player.

By doing it twice, you can set the gain exactly right, and optimize the 
signal to noise ratio for each LP.

Purists say that you should only play an LP once per day, as the vinyl 
deforms a little when the stylus plays it. So you could play it, 
measure, and write down the answer, use that setting the next day to 
make the final transfer.


-- 
Pat Farrell         PRC recording studio
http://www.pfarrell.com/PRC

_______________________________________________
ripping mailing list
[email protected]
http://lists.slimdevices.com/lists/listinfo/ripping

Reply via email to