Normalize then Compress, or visa versa???

S

SamJamaica

Guest
if i had a track that i recorded, would it be better to first normalize the track, and then compressed it, or first compress it then normalize it? i'm actually going to try it out, so i can hear for myself the difference, but i would appreciate other people's opinions, and maybe if some one knows why it would sound better, please let me know
 
I would normalize it first. You could do both in either order and end up with the same or different results depending on your compression settings.

The reason I say normalize first is because normalize doesn't alter anything but the volume of the track. Compression on the other hand while it does make the sound louder, it also affects the dynamic range of the audio.

So, to me, it makes more sense to do the 'non-destructive' transformation first. You can always normalize again after you compress if you need to.
 
Usually but not strictly Normalize then compress.

Compression squeezes the dynamic range by lowering louder things into quieter ones. so if you don't normalize before , the compressor might not catch all the inputs with his current settings.


 
NFX has some good points.
but it depends on what you want to do, (using the same settings for both things either way) if you normalize then compress, you will essentially end up with a wider dynamic range, but if you compress first, it will seem quite a bit louder (this is what they do with commercials, they compress the sh*t out of it, then normalize and it seems way louder than the tv show you're watching) --sorry kinda went off-topic :rolleyes:
anyways, if you compress first, your noise floor will be a lot higher and it may sound worse. but that all depends on how you want it to sound :P
 
I would
1. normalize (note: to peak, not rms),

2. then I would compress,

3. and following that, I would normalize to peak again.

step 1, to get the clip to maximum volume without clipping.

step 2, for the desired compression.

step 3 because, depending on the attack settings on your compressor, the transient peaks in your clip would have been reduced (if using a fast attack setting), and you will probably be able to get more volume out of your clip when you normalize to peak, as now the loudest peak of the clip will no longer be touching the ceiling that you set when you first normalize the clip in step 1.

hope this helps...
 
I never normize. Its an irreversible process and I learned to never apply that to a recorded signal -- unless you have a backup of the original.
 
Normalizing Kills the dynamic range... even more so than compression. ..
 
Rayne said:
Normalizing Kills the dynamic range... even more so than compression. ..

Normalizing does not kill dynamic range. It just takes the loudest peak in your song and sets it to zero. All pro mixers and mastering engineers never normalize, so I suggest you don't do it either.
 
Ummm What FOE said. Normalize should be a last resort trick to get something that was recorded too quiet raised in volume, but then if it was that quiet it was recorded wrong in the first place and should have been done over.

Normalize will never affect dynamic range but it can severly raise the noise floor that wasn't heard before. So I guess the percieved dynamic range could then be affected.
 
ok then....

With all said above, take a step back. Should I EQ first then Compress? or the other way around. I feel that I should EQ first, don't know why it just seems logical. Someone correct me if I'm wrong! It just sounds better in the end to me.
 
There are 2 schools of thought on that. One would be to eq then compress because you want to control the dynamics that would be increased from a boost.

The other is you compress and then slightly boost what you want accented.

Either one works, and I think it really depends on what end result you are trying to achieve. Just be carefull with eq after compression if the original intent was to control transients.
 
Tim20 said:
if it was that quiet it was recorded wrong in the first place

could you shed some light on how to record to avoid normalizing? sorry, i'm a total newbie
 
The conventional way using a mixing console is to set ur faders at o (unity gain) and then raise the input gain (on desk) until it reaches the max point without overloading (clipping) The better the signal recorded the less noise.

Try closer miking or if its a line in signal, try maxing the output of the line signal and then boost at the desk if required.

to FOE and TIM20:
if you normaize it raises the highest peak to the max output right? so it does this propotionally right? so... does it actually make a difference wether you normalize or just raise the fader? because by raising the fader ur surely just increasing noise too, plus adding extra noise from the fader? (if analog desk).
 
thanks

Thanks I'll try that.

I will def. stay away from Normalizing from now on, I've got my output and levels set nicely in Soundforge, and everything records beuatifully, even with my 16 bit soundblaster 5.1 card :victory:

I used soundforges normalize on one of my mixes a few days ago, and it was less than desirable! I ended up with a louder, yet flater sound, all the lower end freq.s seemed to diminish, and I was left with merely the dominate Highs, mid and bass. Perhaps I did something wrong, but I wasn't impressed. Thankfully I got a nice recording set up now so I won't have to stretch and use it.

appreciate all the help!

GTM
 
I ended up with a louder, yet flater sound, all the lower end freq.s seemed to diminish, and I was left with merely the dominate Highs, mid and bass.

I dont think a 'normalize' function would have caused all that. Probably, by doing normalize you have made the track somewhat louder, and humans do not perceive the same frequency proportions at different levels. This explains why a dull recording would sound better when played out loud, cause us humans perceived higher freq better on higher intensity.

for stan Chic: yeah..u're logic made much more sense to me than FOE's. I think the prof never normalise cause they are already confident with the level of the mix in the first place. He doesn't need to bring anything up ...coz it's already at 0db. So, normalizing SHOULD be avoided, just for the sake if you can make a thing right in the first place, why bother to fix it later. agreed?

:cheers:
 
I don't think it is neccessarily of a case the the "pros don't do it" but more a case that if all else was done correctly in the recording chain when capturing the track, then normailize doesn't have a usefull function.

I can see a case where normalize would be bad. Example: Some instruments have noisy amps and cranking up the volume produces audible hiss. If an attempt was made to hide the hiss by reducing the fader on the mixer and then normalize was applied that noise would likely return.

If several tracks were recorded that way and then the mixed stereo track was normalized most likely all of the noise would be raised.

If dither has been applied to a mixdown and then it is normalized, it only stands to reason that it will sound worse, since dither applies an inaudible low noise floor to maintain dynamic range during the process.

The best technique I have found is to adjust the instrument being recorded to its hottest output without getting distortion or hiss and then set the mixer fader to unity gain, then dial in the preamp to get the hottest signal. It should not need to be normalized and will have the highest bit depth.

Oh and make sure you always have the inputs of the soundcard set to zero for maximum resolution of bit depth. Don't control the input signal from here.

:cheers:
 
hot....

ok.. i understand...recording it hot captures all the bits....
.
lets say iam just sampling from the turntable to mixer into my mpc2000xl... i got the mpc record gain... all the way to the bottom and my mixers output at its max....

lets say i sample the sample i want from the record as hot as it can be ...lets say -.05db .....

the i sample some drums(kick snare, and hat) from that record..... once again as hot as can be -.05db

obviously if i start lacing the beat.... its gonna be distortion...
.
.
so some of you guys are saying i should risk losing that surge/adrenaline/rush of creativity to be burdened with the process of compressing those samples and drums back to reasonable levels....??
 
Well... though I think there's some good thinking and communicating going on -- there's also some missed communication (and differing definitions).

When a lot of us are talking about "normalization" we're talking about a straight multiplication of the levels... but some software, like Sound Forge offers "enhanced" normalization that involves compression as well.

This is the diff between what's sometimes (as in SF) called "Peak Level" normalizing and the technique of using compression ("If clipping occurs - use dynamic compression" valuebox) and gain makeup to hit a taget "Average RMS power" (you can also use "equal loudness compensation" to try to attempt to compensate for the Fletcher-Munson law ( the ear hears high and low freqs easeier at high volumes than lower -- so you can attempt to boost bass and treble in the lower level sections [obviously, if that's not done right you can end up with serious problems]). SF has all these functions clustered in one Normalize dialog -- which is convenient but helps confuse already ill-defined contemporary terms of art.


I don't think you lose much by straight normalizing. Obviously, unless you're doing an even double, you'll induce basic rounding error -- but working at 24 bit rez -- which has 256 times better resolution than our CD standard 16 bit audio -- it's hard for me to imagine that being much of an issue.
 
Last edited:
Back
Top