Normalising......Good or bad?

[waketek]

This is the explanation from the Masterlink Manual about the fuction on the Masterlink DSP- Normalization

DSP4:NORMALIZER
A Normalizer's function is to scan a Track for the highest peak value, determine the ratio between that
peak value and full-scale, and multiply the Track by that ratio so that the highest peak value of the
Track is equal to full-scale.
The major improvement in this normalizer over others is that the gain multiplication is performed in
real time, instead of rendering the normalized file back to disk. This allows re-normalizing if changes to
the Track Gain, Compression, EQ, Limiting and Track Fade blocks occur after the first pass of
normalization.
The normalizer does not have any parameters per se, other than the current gain multiplier (which is
not user set, but is determined from the Track).

[Timo]

[To Normalise or Not To Normalise]

One of the best threads in the old SOS forums.

[blay]

i had a great time reading through this thread...



cheers

blay

[Wandering Kid]

remix808 is right.

normalising doesnt change the dynamic of the audio at all. it stays exactly the same. what normalising does is scale the entire waveform up so that the peak level of the track hits the ceiling or a ceiling that you predefine (say, normalise to -0.3 dB).

and yes this does mean the noisefloor of the recording is scaled up accordingly. which is why its mostly undesirable to do this.

also. if you have a walloping great big peak in your recording, normalizing will retain that massive peak exactly but scale the entire waveform up so that the peak goes to 0 dB and no further. therefore if the rest of your recording was going at -10 dB roughly and you had a great big spike that shot up to -1dB. normalising it would hardly increase the level of the track at all. if you wanted to get rid of that peak you would have to either notch EQ it out or compress the waveform so it reduces the dynamic range of the audio. the additional benefit of doing this of course is that you can make the overall waveform louder without scaling up the noisefloor too.

i noticed this most when i foolishly had my analog input boost set to 127 on my virus. to those who have never done this it boosts the signal the virus sends to your soundcard and it raises the noisefloor something insane. when i was doing this, the LFO emitted a high pitched peircing sound as it oscillated and the noisefloor was pretty constant in the mid and upper ranges at around -30 dB. thats unbelievably high on its own. anyway i remember writing some dirty hoovers at that point which made it into a tune i have since abandoned - witch engine. that hoover was normalised and it made the low level noise much more prominant. if you listen carefully you can actually hear it on the recording - even with all the other stuff going off.

if i strip the rest of the song away and just play that wav recording it is ear peircingly evident. anyway, whilst i was screwing around with it, i noticed normalising the entire waveform made the ambient noise in the recording even louder. gah.

[Timo]

Don't forget rounding errors.

also. if you have a walloping great big peak in your recording, normalizing will retain that massive peak exactly but scale the entire waveform up so that the peak goes to 0 dB and no further. therefore if the rest of your recording was going at -10 dB roughly and you had a great big spike that shot up to -1dB. normalising it would hardly increase the level of the track at all. if you wanted to get rid of that peak you would have to either notch EQ it out or compress the waveform so it reduces the dynamic range of the audio.

For troublesome spikes, I just import the audio into a wave-editor (wavelab, sound forge, cool edit, etc.), and zoom in and manually pick out the spikes and just pull them down alone. Ie. by highlighting the spike, and then normalising the spike by -10dB (or whatever the average dB level is either side of the peak). With Cool Edit you can simply zoom in and grab the offending spike points and pull them down to any desirable level with your mouse.

Voila, instant headroom, and no need to process the rest of the surrounding waveform. The less haphazard utilitarian processing the better.

[grs]

Nuendo and Cubase have a GAIN setting on each channel, so you would never really need to normalise a recorded wav file. Just adjust the gain of any low level tracks.

Just substitute the word Normalise with Gain and then it is not a mysterious effect.