Normalize with RMS or peak?

[AD80]

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).

So what would be the point of normalizing it if you're playing it at the same level you had it before? Ofcourse people dont do that becuase it doesnt make sense. Either way you look at it, the noise floor will be increased along with the material so its better to amplify at the source.

[Todd24]

For kicks I tried using RMS at -10db. All five tracks fit together well and were as loud or louder than the CD iused for comparison.

Of course they all cliped most of the way through as well. Doh!

[IIRs]

if you are exporting mixes to be mastering together as an album, RMS normalising might make sense.. but -24dB or lower would be a more sensible figure than -10! Doing this would provide mixes that were all at a consistent level to start with, so then you would just need to adjust for musical differences.

[Todd24]

For the sake of getting a demo out, at consistant levels, i am using Final mix and exporting at 16/44. I thought this would be the easy way to go. It seems like I would do better to export a 32/96 file and master with some thing else.

[semiquaver]

Not a bad plan - export at 32, don't need to worry about the levels really. If you like final mix, you could then bring all your final mixes into a new T edit - each w/final mix till they all sound good and the relative and absolute levels are good. Then export again without normalizing.

[Graeme]

So what would be the point of normalizing it if you're playing it at the same level you had it before? Ofcourse people dont do that becuase it doesnt make sense. Either way you look at it, the noise floor will be increased along with the material so its better to amplify at the source.

The 'point' of normalising - well, there could be any number of reasons. Maybe there isn't enough make-up gain avaialable, so it's not possible to get the track lound enough in the mix. Perhaps you want to use a software compressor preset which would rely on the track hitting 0dB. Some mixers have a fixed return level, so a quiet track would mean the fader pushed to the top and still not loud enough - whereas a signal around the 0dB mark would mean the fader sitting at just about the 'right' level, etc., etc.

Although I'm the first person to say you should record close to peak level (to maintain the bit depth, if nothing else) there may be a good reason for a recording to be lower than it ought to be, at which point normalising comes to the rescue. Also, it may be that you might want to trade bit depth against a noisy pre-amp.

There are all sort of reasons why one might want to normalise a track - but the point I was originally trying to make was that it shouldn't make any discernable difference to the actual sound of that track - and it certainly shouldn't introduce any additional noise.

[IIRs]

The 'point' of normalising - well, there could be any number of reasons. Maybe there isn't enough make-up gain avaialable, so it's not possible to get the track lound enough in the mix.

There is always enough make-up gain available in Tracktion: each clip has its own gain & pan parameters, and you can create as many Vol/Pan filters as you need..

Perhaps you want to use a software compressor preset which would rely on the track hitting 0dB.

Why would a compressor preset need there to be a single 0dBFS peak somewhere in the file? That makes no sense.. though, come to think of it, compressor presets don't make much sense to start with!

Some mixers have a fixed return level, so a quiet track would mean the fader pushed to the top and still not loud enough - whereas a signal around the 0dB mark would mean the fader sitting at just about the 'right' level, etc., etc.

In what way does this relate to a 32/64-bit DAW?

- but the point I was originally trying to make was that it shouldn't make any discernable difference to the actual sound of that track - and it certainly shouldn't introduce any additional noise.

If the file happens to be 32-bit, then you are right, there are no quality implications to a destructive normalise.. you just can't change your mind. If it's any lower than that, then normalising will introduce extra distortion due to rounding errors, that would not occur if you changed the gain non-destructively in Tracktion. That's a mathematical fact. If you decide its insignificant, then you go ahead and do it if you want.. but if everyone thought it was insignificant, no-one would have bothered to invent dither..

[mandolarian]

Oh, come on Plati, you know dither wasn't invented. It was a shrewd marketing ploy by those analog-only companies who had all this leftover hiss and no market for it. Just don't get me started on the fluoride conspiracy...

[Warmonger]

Normalizing adds noise due to the nature of number representations on a computer. If you want to raise the gain by a certain amount, you do a simple multiply. But, the result of that multiplication is almost never going to be exact, so there will be a little bit of noise added to the signal. The amount of noise at a given sample is (if memory from my numerical computing class serves) <= .5 of the LSB. At 32-bit resolution, 1/2 bit of extra noise isn't much to worry about. At 16-bits, it'll be a lot more significant, but still nothing to lose sleep over, IMO.

[Graeme]

If the file happens to be 32-bit, then you are right, there are no quality implications to a destructive normalise.. you just can't change your mind. If it's any lower than that, then normalising will introduce extra distortion due to rounding errors, that would not occur if you changed the gain non-destructively in Tracktion. That's a mathematical fact. If you decide its insignificant, then you go ahead and do it if you want.. but if everyone thought it was insignificant, no-one would have bothered to invent dither..

Although I appreciate this is the Tracktion forum, I was trying to make my comments as broadly applicable as possible, not Tracktion specific. I can certainly 'change my mind' with a lot of the software I use. You can in Traction as well - surely, that's what the undo function is for?

Dither is something else entirely and has no place in this particular discussion. The difference between a 16 bit file (after normalising) are, indeed, insignificant - and yes, I do it all the time (well, to be honest, not all the time, but it wouldn't concern me if it were necessary for some reason).

[IIRs]

My comments were not Tracktion specific, with the possible exception of 'there is always enough gain', and they apply to any 32-bit mixing envronment.

Dither is entirely relevant: it is a technique to minimise the quality loss when truncating data.. which you do every time you normalise.

[Todd24]

Good stuff guys. Thanks.

I did it with my ears and I like the results MUCH better. I have moved to the no normalizing, side of this debate....Or is everyone already there?