Normalize with RMS or peak?

[Todd24]

How do you render and normalize?

I have been using the peak setting, but my CDs are always a bit quiter than store bought CDs.

[AD80]

I dont. I try to get the original material to its loudest level before I render. Normalizing can add a lot of extra noise depending on the material.

Store bought CD's are ran through like a million diffrent things before they get burned on CD. The mastering process EQ's and compresses which adds a lot of volume. Then after thant usually they'll use a maxmizing limiter to squash the hell out of it which adds even more volume.

Try something like Voxengo Elephant, TL's Maximizer, PSP Vintage Warmer, Waves L2-L3, to add a lot of volume to your final tracks.

[Graeme]

Normalizing can add a lot of extra noise depending on the material.

It most certainly shouldn't.

All normalising does is to ascertain the peak level of the track and then amplify the whole track, such that the peak hits a point you determine (usually at or just below 0dBFS). This should not add any noise to the track at all.

[haydxn]

yes, but any noise floor is also increased with normalising. for example, if you have a very quiet sound to begin with (recorded from a microphone) then if you normalise it to bring it up loud enough to go with other sounds, you're also bringing up the noise floor. thus it's always best to have things as loud as possible beforehand. this includes the output from vsts, because if they're rendered as a final wav file with low numbers, normalising will be simply scale the numbers up. if they're very low then there is inherently less dynamic range; scaling the level up will cause the innacuracies to be greater than if they were rendered at a higher level in the first place. therefore normalisation of a rendered file will be noisier than if it were rendered at a louder level. it's not likely to be noticable unless the volume was very low, but nonetheless it's still something to be aware of.

it's only really a concern when dealing with recorded material - recording fairly quietly then relying on normalising/boosting the level will definitely be worse than simply recording at the maximum possible input level.

[IIRs]

it's only really a concern when dealing with recorded material - recording fairly quietly then relying on normalising/boosting the level will definitely be worse than simply recording at the maximum possible input level.

This is true mainly of 16-bit recordings, and is a powerful argument for higher resolutions.. the only time normalising might make sense is when dealing with the final stereo mix before it is dropped to 16-bit (assuming it will be), and only then if you intend to burn it straight to CD with no further processing.. otherwise, don't normalise: it's pointless at best.

[multree]

well sometimes normalizing in T will bring the noise floor down....

how does it do that you might ask.... well when the peaks are beyond zero dB then normalising (using peak mode - RMS would make the f**ker clip) will lower the volume and hence the noise floor

[multree]

but then again I don't really use normalize for that but rather the find peak based on the maximum level of the whole edit.... (and lower it by 0.3dB

[Graeme]

yes, but any noise floor is also increased with normalising.

Yes, of course it is, but the S/N ratio remains the same.

it's only really a concern when dealing with recorded material - recording fairly quietly then relying on normalising/boosting the level will definitely be worse than simply recording at the maximum possible input level.

It is true that recording at a low level does reduce the effective bit rate of the recording (broadly, every 6dB below peak equates to the loss of 1 bit, so if you record at a peak of -18dB, you've only got a 13 bit recording). In turn, this means the s/N ratio is worse than if you recorded to peak at a higher level.

Even so, normalising does not change this ratio - same argument applies. The S/N ratio will remain constant, normalising or no. Of course if the original level was low, then it may be necessary to increase the signal level somewhere further down the chain - and this may well introduce extra noise - but the S/N ratio of the original recording remains the same.

[IIRs]

yes, but any noise floor is also increased with normalising.

Yes, of course it is, but the S/N ratio remains the same.

Except for the extra rounding error.. which could be significant with a 16-bit file.

[Graeme]

Except for the extra rounding error.. which could be significant with a 16-bit file.

I think I would replace 'significant' with 'miniscule' .

[haydxn]

i'm not disputing the S/N ratio thing! fact is that if you normalise something quiet, you'll very likely hear more noise. simple as that

[Spitfire31]

"Normalize with RMS or peak?"

You should do neither. Normalizing sets the level of your whole song according to what may be a single freak peak of a few samples - not the best way to attain a balanced level between songs... Normalizing is for exporting to a sampler, nothing else.

To raise the apparent loudness of your finished song, use your ears (so you don't squash the dynamics too much) and a good _limiter_ on the master outs, before going from 24 to 16 bits.

//Joey

[semiquaver]

on a mix its best to first "find normalized level" then render (no nead to normalize there). if its just a single track I would use "by peak", so that it will be as loud as it can be without distortion. Where by RMS comes in handy is when you are doing several different tracks and want the levels to be comparable. Or if you don't mind clipping a few transients and "by peak" yields a quiet mix (in this case your mix is probably shy on compression.)

for clips or tracks I think the best least destructive thing if you're going to bring them back to T is to render them 32 bits floats then there could be no need to normalise (floats have the same level of detail whether the signal is loud or soft - samples are encoded with a scaling factor -- pretty neat).

It would be good to know though whether t's normalize in export and render is done in 32 or 64 bit floats (ie before conversion to 16 or 24) or after. In the former case the cost is about nil - in the latter case there would be a little penalty.

[djsubject]

peak

but a good ere for mastering allways helps


Subject

[Graeme]

i'm not disputing the S/N ratio thing! fact is that if you normalise something quiet, you'll very likely hear more noise. simple as that

If you play both the original and the normalised files at the same level - I doubt if you'd hear any difference at all.

But people don't do this - they play the normalised file at the same monitor level as they played the original - ergo it's louder and yes, you hear more noise (because that's louder as well).

Anyway, I think we've gone as far with this as we need to - I'll go back to whinging about the way Mackie have treated users outside the USA . I mean, just how long am I expected to put up with the hiss?