Normalising is good?

  • Thread starter Thread starter goldfish
  • Start date Start date
Re: Ok then, BB

Speedy VonTrapp said:
snipp ...On some tracks, I'm getting some of those peaks that I don't want. Perhaps the drummer hits a bit harder on a couple hits, etc., whatever. (Is that even what might cause something like that?

I'm limiter-less. What are my best options? Can the RNC help this situation out if my budget says that no matter what, I've got to work with what I've got?

Or are there limiter plug ins that are worth downloading? Or if the budget says it's ok, but limited, (as most home recc'rs), what should I grab for limiting? Is there an external hardware unit you'd recommend, or is there something software side that you'd recommend?
OR... is it something thath is helped in a different manner altogether, like mic placement, or smacking the drummer, etc?

-Speedy
Click to expand...

I would say that's definiely the case. As playing techniques get smoother, more consistent, stray spikes become less. But limiting is still part of the sound and levels we're often going for.
This can be tackled at the track level, and plugs can work well in the mix stage. UltraFunk's comp/limiter is a good one for example. (Part of sonar now). It's very quick and look-ahead also helps.
The RNC is very fast (it's not brick-wall), but I don't know if it's THAT fast. Especially in super-nice. Doesn't the attack get slowed down also in super-nice? Never really checked that out exactly.
Wayne
 
Re: I need an education...

Speedy VonTrapp said:

As of right now, I can successfully use the RNC as more of a gate device than anything. It works great for this, limiting excess noise under a certain level, and then only activating once that threshold is hit.
...If I'm using a 2:1 ratio, are the levels going to be increased by a factor of 2? Therefore making the signal that runs through it double in db level? It doesn't seem like that's the case, since 25:1 would be just about useless at that point.

-Speedy
Click to expand...
Somehow you got this backwards. The RNC reduces levels by the ratio amount, not expands them.
2:1 means 1db increase at the output for each 2db at the input.
If it was an expander 1:2 would be dynamics expansion, 1:25 would be on-off, a gate.
Wayne
 
Speedy, see if a couple of nights at this web sight doesn't shed some more light on compressors and the 'stacked' compressor approach of the RNC (in super nice mode):

http://www.lanasoft.com/lsamp/manual/strategy.htm

Once you digest that then you'll see why it doesn't take much GR at all to smooth out stuff in 'super nice mode' since there's 3 compressors in there each 'listening' to a different envelope (rms/avg, peak, and a combo). The RNC compressor doesn't have the fastest attack time in the world so some of the 'invisible' transients that BB mentioned can get thru it but there are other limter tools for that. Here' one type, a free VST mastering limiter:

http://www.kjaerhusaudio.com/classic-master-limiter.php

As far as getting used to the controls and what they do you have to work with it - make it sound bad and good - twist every knob you can. Then get more systematic ! :) I usually bring up Cool Edit, generate some pink or white noise and run it thru the device - don't listen to that though ! You can bring up a spectrum analyzer and watch what it does, here's a free VST one:

http://www.elementalaudio.com/products/inspector/index.html

Then once you understand how the compressor 'bends' the noise wave file that has no dynamics put one of your tracks thru it and watch the s.a. and listen.

This is very entertaining - don't forget to make some popcorn first ! :)

Good mixing, compressing, limiting, maximizing and normalizing (in the good sense of the term) !

kylen
 
Blue Bear Sound said:
No... there are often transient peaks that are not noticeable to you, but they defintely impact the amount of normalizing gain that could be applied (sometimes on the order of less than a dB)..... using that same example but using limiting (which will remove transients long before any impact tot he musical dynamics occur) will result in a much larger gain "potential" (on the order of several dB or more).
Click to expand...

I think I'm 99.9% convinced to avoid normalization completely. What's a transient peak? where does it come from and how can we avoid it during tracking?
Also, when you are limiting, how do you know when you've removed the transiet peaks from when you're starting to decrease the dynamic? I guess I am asking how to oper
ate a limiter :D

thanks, Bruce, it's been an educational thread!

Al
 
Transient peaks are sharp spikes of short duration... they are created by percussive sounds, or even a hard pick attack on a clean guitar part (the start of a harmonic)

They can also be produced by an overzealous guitarist causing the string to hit the pickup.

There are many ways to produce transients....

The limiter WILL reduce transients, because of the nature of limiting (the signal hits a brickwall and goes no higher) - but there may be an area between the peak of the transient and the let's say "average peaks" of the rest of the signal that you can aim for without affecting any of the musical dynamics at all.

The only way to know what that area is, is to look at the waveform. But you should be using your ears anyways - again, go extreme with limiter settings to hear what it does to your sound, then back off gradually noting again the difference in sound. Eventually, you should be able to see some gain reduction on the limiter's meters, yet not hear the dynamics being squashed.

Note that this is all very context-sensitive -- it depends completely on the signal involved.
 
So, then, riddle me this:

What's the best way for me to experiment? I run all of my sounds through my mixer, into a yamaha MD4S. Once I've got them down, I dump them to Cakewalk PA9. (4 unbalanced 1/4" inputs on a Lexicon CORE 2 sound card.)

Will it do much good if I run a track out of PA9, and into my RNC, then into my Yamaha? It seems like it should work, but I don't know if it will or not.

Other than that, I'll have to do it while tracking. I'd hate to waste someone else's time having them play something over and over so I can play with my RNC while they hit the snare. I understand that when it comes time to lay down a good track, I'll have to fiddle with the the RNC in this manner anyway, but for now, we're just talking about experimenting, and me learning how to use what I've got.

Also, Kylen...

I'm very new to VST. Is there a trick to make those plugins work with PA9?

I just feel a bit ill at ease with what I have. I have the feeling that the RNC is only going to work with things as I track them. Is that an ignorant statement? Can I (or better yet, should I ) run audio that I've already recording through the RNC, then back to the computer?

I seem to be quite a bit more comfortable with tracking since first coming to this site. Now, it seems that working on some sort of processing of the sounds, after or during tracking, is one area that I'm very very new to.

I hope that I'm not annoying people with questions here that might very well belong in the newbie forum.

Thanks for the help from everyone!

-Speedy

(And, apologies to goldfish for his thread being hijacked, although, still somewhate on topic, right? ;))
 
Song to Song level adjustment

NORAMLIZING:
the way i understand it, correct me if i'm wrong please...

i hate analogys but.....
if 5 people are standing side by side on a noisy sidewalk in NY.
one of the 5, is 8ft tall and very skinny, the others are all 3ft tall and fat.
8ft tall skinny guy is our "spike""transient peak".
3ft tall group is the bulk of the song.

so........in DB terms........
if the 8ft tall guy = -5db(loud, spike, trasinet peak)
if the 3ft tall people= -8db (quietter)
if sidewalk Noise= -12db
(there in the background..ssshshhshshsshshs, click,pop, shshshssss..taxiiiii... which may not even hear it on plasticshtcomputer speakers..its there)

so then if we NORMALIZE to 0db......

Only the -5db spike is increased, brought to the desired 0db.
the -8db increases to only -3db
the -12bd Noise increases to -7db (hey i can start top hear the SSHSHSHSHSHHSHSSS,CLCLCICCCKK,POP,,SHSHSSSHSHSZZZZZ...taxxiiii!!!)

so GOOD TRACKING takes priority over NORMALIZING?
1) GOOD TRACKING the noise is greatly reduced.
2)GOOD TRACKING the spike isn't there to begin with so need to normalize.

NEXT SCENARIO:
???seems song to song level adjustments would still be an issue, if say you were putting together a cohesive group of songs...recorded at different locations, differnent dates for ex..
is that a mastering house issue or Alesis Masterlink type thing?
how would the SONG to SONG levels be addressed if not normalized?

i hope i got this right....fhkn boss#5 keeps interruppting me wanting me to go build the pyramids...!!need more bricks!!! go build the king a pyramid you peasent mthrfhkr!!

feel free to "pounce on me" if i'm wrong...
 
Re: Song to Song level adjustment

COOLCAT said:
NORAMLIZING:
the way i understand it, correct me if i'm wrong please...

i hate analogys but.....
if 5 people are standing side by side on a noisy sidewalk in NY.
one of the 5, is 8ft tall and very skinny, the others are all 3ft tall and fat.
8ft tall skinny guy is our "spike""transient peak".
3ft tall group is the bulk of the song.

so........in DB terms........
if the 8ft tall guy = -5db(loud, spike, trasinet peak)
if the 3ft tall people= -8db (quietter)
if sidewalk Noise= -12db
(there in the background..ssshshhshshsshshs, click,pop, shshshssss..taxiiiii... which may not even hear it on plasticshtcomputer speakers..its there)

so then if we NORMALIZE to 0db......

Only the -5db spike is increased, brought to the desired 0db.
the -8db increases to only -3db
the -12bd Noise increases to -7db (hey i can start top hear the SSHSHSHSHSHHSHSSS,CLCLCICCCKK,POP,,SHSHSSSHSHSZZZZZ...taxxiiii!!!)

so GOOD TRACKING takes priority over NORMALIZING?
1) GOOD TRACKING the noise is greatly reduced.
2)GOOD TRACKING the spike isn't there to begin with so {I think you meant - no} need to normalize.
Click to expand...
Good analogy - pretty much spot-on - except the 8ft spike is probably more like 12ft in many songs, so the normalized increase would be only 1dB....!


COOLCAT said:
NEXT SCENARIO:
???seems song to song level adjustments would still be an issue, if say you were putting together a cohesive group of songs...recorded at different locations, differnent dates for ex..
is that a mastering house issue or Alesis Masterlink type thing?
how would the SONG to SONG levels be addressed if not normalized?
Click to expand...
Song to song levels are typically handled by either simple gain changes, or limiting, or compressor, or all of the above depending on the nature of the tracks! Normalizing in a mastering situation pretty much doesn't occur, according to Bob Katz.
 
No Normilizing required

i get it then...thanks for the response.

yes, i meant to include "NO" normailizing is needed with good tracking technique.
 
Hi Speedy,
VST - I've only begun using a VST Adapter (Cakewalk VST Adpater v4.3.1 mostly) beginning with Sonar2.2 and Cool Edit Pro 2.1 so I'm kinda new at it myself. I don't know if VST (with an adapter) works in Pro Audio 9 or not.

Tracking vs Re-amping - I think it's common to use whatever tools you like to shape and create the sound while recording or during mixdown or post-mixdown like pre-mastering a completed mix. Obviously if you add something to the sound while recording you want to be pretty good at it since it's harder to fix a 'mistake' You said that's where you hang mostly ! If you know that sending a dynamically wild track out to your RNC and back into the computer will let it sit better in the mix then you can do that during mixdown - kind of reamping I think they call it. You can also re-amp a whole completed mix if you like. That's what I'm doing now - I mentioned something about it on the Saturation thread here in this forum using some outboard gear.

If you check out 'The Mastering Engineers Handbook' (Bobby Owsinski) there's info about how the pro's re-amp and when. I'm sure the word normalize is in there too - many times !

http://www.amazon.com/exec/obidos/tg/detail/-/0872887413/002-0655101-3493655?v=glance

Good Tracking and Re-amping !
kylen
 
Blue Bear Sound said:
Transient peaks are sharp spikes of short duration... they are created by percussive sounds, or even a hard pick attack on a clean guitar part (the start of a harmonic)

They can also be produced by an overzealous guitarist causing the string to hit the pickup.

There are many ways to produce transients....

The limiter WILL reduce transients, because of the nature of limiting (the signal hits a brickwall and goes no higher) - but there may be an area between the peak of the transient and the let's say "average peaks" of the rest of the signal that you can aim for without affecting any of the musical dynamics at all.

The only way to know what that area is, is to look at the waveform. But you should be using your ears anyways - again, go extreme with limiter settings to hear what it does to your sound, then back off gradually noting again the difference in sound. Eventually, you should be able to see some gain reduction on the limiter's meters, yet not hear the dynamics being squashed.

Note that this is all very context-sensitive -- it depends completely on the signal involved.
Click to expand...


Bruce,

thanks for the education!!

Al
 
For anybody who prefers to follow math (like myself), here is what is going on with normalizing:

Say we are dealing with 16bit samples. The amplitude value can be anywhere from 0 to 65,536. In this case, 65,536 represents the loudest possible signal before clipping, or 0 db. Anything less that that is a negative decibal number, all the way down to a value of 0 for negative infinity db (silent).

For simplicity's sake, we'll look at wave data that has 4 samples instead of millions of samples.

537
7438
20,743
60,023

Those are the sample values. So, if we are normalizing to 0 db, the loudest peak must end up at 65,536 and everything else must be scaled up accordingly. Here's how to do that:

Find the largest value of all your samples. It doesn't matter if you have 4 samples like we have or millions of samples. You can only take the single largest value. In our case it is 60,023. Divide every other number by that amount. After the division, we are left with these values for our samples:

0.0089466
0.12392
0.34558
1

The result is that every single sample in the entire audio clip is between the values of 0 and 1. From here it is easy to bring the clip to any maximum peak you can think of. In this case we want to normalize to 0db (a value of 65,536), so we simply multiply every number by 65,536.

586
8121
22648
65536

And there we have our final normalized numbers. Those numbers will result in a wave that sounds the same but louder. You will also notice if you follow along with Windows calculator that many many digits had to be thrown away two times due to rounding. Hence, the rounding errors.


Here's why you should rarely normalize. Moving the fader will do the exact same thing mathematically. If you move the fader the exact right amount, you can end up with the same 4 values we ended up with durring normalization (586, 8121, etc.) However, the value of the clip on the disk remains at the original values (537, 7438 etc.) and the math is applied with real time processing. With normalization, those orriginal values are gone and you are only left with the end result, errors and all included. If you decide that normalization made you too loud, you have to move the fader back down, only now you are doing your calculations off of the normalized numbers that already have errors in them.

If you moved the fader up in the first place (making the clip exactly as loud as normalization would have made it) and then decided that the fader had to come back down a bit, you are still doing your calculations off of the orriginal values.



In short, try moving your faders before you normalize.
 
For anybody who prefers to follow math (like myself), here is what is going on with normalizing:

Say we are dealing with 16bit samples. The amplitude value can be anywhere from 0 to 65,536. In this case, 65,536 represents the loudest possible signal before clipping, or 0 db. Anything less that that is a negative decibal number, all the way down to a value of 0 for negative infinity db (silent).

For simplicity's sake, we'll look at wave data that has 4 samples instead of millions of samples.

537
7438
20,743
60,023

Those are the sample values. So, if we are normalizing to 0 db, the loudest peak must end up at 65,536 and everything else must be scaled up accordingly. Here's how to do that:

Find the largest value of all your samples. It doesn't matter if you have 4 samples like we have or millions of samples. You can only take the single largest value. In our case it is 60,023. Divide every other number by that amount. After the division, we are left with these values for our samples:

0.0089466
0.12392
0.34558
1

The result is that every single sample in the entire audio clip is between the values of 0 and 1. From here it is easy to bring the clip to any maximum peak you can think of. In this case we want to normalize to 0db (a value of 65,536), so we simply multiply every number by 65,536.

586
8121
22648
65536

And there we have our final normalized numbers. Those numbers will result in a wave that sounds the same but louder. You will also notice if you follow along with Windows calculator that many many digits had to be thrown away two times due to rounding. Hence, the rounding errors.


Here's why you should rarely normalize. Moving the fader will do the exact same thing mathematically. If you move the fader the exact right amount, you can end up with the same 4 values we ended up with durring normalization (586, 8121, etc.) However, the value of the clip on the disk remains at the original values (537, 7438 etc.) and the math is applied with real time processing. With normalization, those orriginal values are gone and you are only left with the end result, errors and all included. If you decide that normalization made you too loud, you have to move the fader back down, only now you are doing your calculations off of the normalized numbers that already have errors in them.

If you moved the fader up in the first place (making the clip exactly as loud as normalization would have made it) and then decided that the fader had to come back down a bit, you are still doing your calculations off of the orriginal values.



In short, try moving your faders before you normalize. :)
 
For anybody who prefers to follow math (like myself), here is what is going on with normalizing:

Say we are dealing with 16bit samples. The amplitude value can be anywhere from 0 to 65,536. In this case, 65,536 represents the loudest possible signal before clipping, or 0 db. Anything less that that is a negative decibal number, all the way down to a value of 0 for negative infinity db (silent).

For simplicity's sake, we'll look at wave data that has 4 samples instead of millions of samples.

537
7438
20,743
60,023

Those are the sample values. So, if we are normalizing to 0 db, the loudest peak must end up at 65,536 and everything else must be scaled up accordingly. Here's how to do that:

Find the largest value of all your samples. It doesn't matter if you have 4 samples like we have or millions of samples. You can only take the single largest value. In our case it is 60,023. Divide every other number by that amount. After the division, we are left with these values for our samples:

0.0089466
0.12392
0.34558
1

The result is that every single sample in the entire audio clip is between the values of 0 and 1. From here it is easy to bring the clip to any maximum peak you can think of. In this case we want to normalize to 0db (a value of 65,536), so we simply multiply every number by 65,536.

586
8121
22648
65536

And there we have our final normalized numbers. Those numbers will result in a wave that sounds the same but louder. You will also notice if you follow along with Windows calculator that many many digits had to be thrown away two times due to rounding. Hence, the rounding errors.


Here's why you should rarely normalize. Moving the fader will do the exact same thing mathematically. If you move the fader the exact right amount, you can end up with the same 4 values we ended up with durring normalization (586, 8121, etc.) However, the value of the clip on the disk remains at the original values (537, 7438 etc.) and the math is applied with real time processing. With normalization, those orriginal values are gone and you are only left with the end result, errors and all included. If you decide that normalization made you too loud, you have to move the fader back down, only now you are doing your calculations off of the normalized numbers that already have errors in them.

If you moved the fader up in the first place (making the clip exactly as loud as normalization would have made it) and then decided that the fader had to come back down a bit, you are still doing your calculations off of the orriginal values.



In short, try moving your faders before you normalize. :)
 
For anybody who prefers to follow math (like myself), here is what is going on with normalizing:

Say we are dealing with 16bit samples. The amplitude value can be anywhere from 0 to 65,536. In this case, 65,536 represents the loudest possible signal before clipping, or 0 db. Anything less that that is a negative decibal number, all the way down to a value of 0 for negative infinity db (silent).

For simplicity's sake, we'll look at wave data that has 4 samples instead of millions of samples.

537
7438
20,743
60,023

Those are the sample values. So, if we are normalizing to 0 db, the loudest peak must end up at 65,536 and everything else must be scaled up accordingly. Here's how to do that:

Find the largest value of all your samples. It doesn't matter if you have 4 samples like we have or millions of samples. You can only take the single largest value. In our case it is 60,023. Divide every other number by that amount. After the division, we are left with these values for our samples:

0.0089466
0.12392
0.34558
1

The result is that every single sample in the entire audio clip is between the values of 0 and 1. From here it is easy to bring the clip to any maximum peak you can think of. In this case we want to normalize to 0db (a value of 65,536), so we simply multiply every number by 65,536.

586
8121
22648
65536

And there we have our final normalized numbers. Those numbers will result in a wave that sounds the same but louder. You will also notice if you follow along with Windows calculator that many many digits had to be thrown away two times due to rounding. Hence, the rounding errors.


Here's why you should rarely normalize. Moving the fader will do the exact same thing mathematically. If you move the fader the exact right amount, you can end up with the same 4 values we ended up with durring normalization (586, 8121, etc.) However, the value of the clip on the disk remains at the original values (537, 7438 etc.) and the math is applied with real time processing. With normalization, those orriginal values are gone and you are only left with the end result, errors and all included. If you decide that normalization made you too loud, you have to move the fader back down, only now you are doing your calculations off of the normalized numbers that already have errors in them.

If you moved the fader up in the first place (making the clip exactly as loud as normalization would have made it) and then decided that the fader had to come back down a bit, you are still doing your calculations off of the orriginal values.



In short, try moving your faders before you normalize. :)
 
In fact that's what Ozone2 does when you use its 'normalize' function. It calculates how high to move the input signal fader and moves it instead of doing a destructive dsp operation like the commonly referred to 'normalize' in many audio editors.
kylen
 
kylen said:
If you check out 'The Mastering Engineers Handbook' (Bobby Owsinski) there's info about how the pro's re-amp and when. I'm sure the word normalize is in there too - many times !
Click to expand...
If you have the book -- check out page 15....
 
Yes, I have that one - I was just reading one of the interviews last night ! Bob O. , What's happening on p15 ? A nice surprise ? Ha Ha ! :cool:
kylen
 
You must log in or register to reply here.
Back
Top