Dubstepforum.com

Why are people normalising tracks?

mthrfnk wrote:Why are people normalising tracks?

Misinformation.
They think they need to.
They haven't separated mixing and mastering in their mind.

No one is normalising individual tracks. You normalise the final product so that you don't have to turn up your volume to hear it. I don't see what the harm is.

jonahmann wrote:No one is normalising individual tracks. You normalise the final product so that you don't have to turn up your volume to hear it. I don't see what the harm is.

Raising volume is a mastering job. Do that with compressors, limiters, gainers, channel faders etc.

You're putting your signal through another set of quality diminishing calculations. There should literally be no need to normalise (your 2buss, normalise the fuck out of samples, I don't care).

We're talking about mastering.

jonahmann wrote:No one is normalising individual tracks. You normalise the final product so that you don't have to turn up your volume to hear it. I don't see what the harm is.

Normalising the final product isn't the same as mastering. If you want to boost the volume, use a limiter.

Normalising raises the noise level in your track and will only make it as loud as the loudest peak in your waveform.

Normalization is not a destructive process. It simply scans the file for the loudest maximum peak and linearly adjust gain to bring that peak up to 0 (or -0.1 or whatever threshold is selected), so the balance between quietest and loudest parts remains the same. Also there seems to be lots of confusion regarding peaks and perceived volume (usually measured in RMS). Your track can never go over 0dBFS, so if you are "mastering" your own material any samples above 0 will just be clipped off. But the RMS and dynamic range you should be aiming for is a much more complicated and relevant question. I usually aim for the very loudest sections (of loud material) to be around -7.5 dBFS+3 RMS, occasionally breaking -7 if the bass gets louder in certain parts. Going above that you start to have serious tradeoffs between punch and volume. A very useful and free tool for analysis is voxengo's Span plugin. There is tons of usefull info on the subject at turnmeup.org

mthrfnk wrote:

jonahmann wrote:No one is normalising individual tracks. You normalise the final product so that you don't have to turn up your volume to hear it. I don't see what the harm is.

Normalising the final product isn't the same as mastering. If you want to boost the volume, use a limiter.

Normalising raises the noise level in your track and will only make it as loud as the loudest peak in your waveform.

That's what he wants.

In other words, you didn't read the thread title. I didn't say normalising was the same as mastering, but yes it is the end of the whole process. Increasing percieved loudness is not mastering either, and it increases noise. Hypercompression is not something I agree with.

Anyway, the guys question has been answered.

jonahmann wrote:

mthrfnk wrote:

jonahmann wrote:No one is normalising individual tracks. You normalise the final product so that you don't have to turn up your volume to hear it. I don't see what the harm is.

Normalising the final product isn't the same as mastering. If you want to boost the volume, use a limiter.

Normalising raises the noise level in your track and will only make it as loud as the loudest peak in your waveform.

That's what he wants.

In other words, you didn't read the thread title.

Wtf... the title is what dB should the final mastered project be at. Why the hell would you normalise a mastered track?

jonahmann wrote:

mthrfnk wrote:

jonahmann wrote:No one is normalising individual tracks. You normalise the final product so that you don't have to turn up your volume to hear it. I don't see what the harm is.

Normalising the final product isn't the same as mastering. If you want to boost the volume, use a limiter.

Normalising raises the noise level in your track and will only make it as loud as the loudest peak in your waveform.

That's what he wants.

In other words, you didn't read the thread title. I didn't say normalising was the same as mastering, but yes it is the end of the whole process. Increasing percieved loudness is not mastering either, and it increases noise. Hypercompression is not something I agree with.

Anyway, the guys question has been answered.

Two things. One, normalizing is the worst way of increasing volume to zero. All modern limiters have something called upward expansion that tries to keep the noisefloor down in different ways (varying from denoising to negative compression to gating) and does a hell of a better job than normalization, which is a basic scientific term than cannot be improved on and does nothing to stop the raising of the noise floor and degradation of your sound.

Also the majority of the time you normalize, you create inter-sample peaks because you're forcing the level up to 0.0 without screening.it for them. Using a limiter, even if you don't intend to slam your track, is very important. Master to -0.3dB if you haven't got an inter-sample limiter and -0.1 if you do (for compatibility with certain old devices that see 0.0 PCM as a command signal apparently)

All this info comes from my college principal. He has two gold disks in his office and mastered both of the tracks he went gold with so I tend to listen to him when he teaches.

doublepost

All right, but if you limit to -6, then normalise, you won't hear the difference, especially at high bit depths.. unless he is using analogue.

Good call on using a limiter with ISP detection, though.

Why would you limit to -6dB? If you mean adding 6B of gain and limiting at 0dBFS then likely not, but it may sound slightly different after conversion on soundcloud/youtube etc or under different speakers. The difference would be minimal, but lots of minimal differences is what mastering is all about.

If you mean limiting to a ceiling of -6dBFS and normalising after then that is a flat out stupid idea. -6dbFS is half signal strength so if you do that you are effectively reducing the bitrate for your music to half its value before noiseshaping then resampling it at a higher value. Sure, most summing inside DAW's these days is done floating point so is at ridiculously high bitrate ,but that is no reason to cut the sample accuracy of your DAW in half. If you need the track to be quieter then use your volume knob. Never reduce the volume inside the DAW if you intend to produce a master. Master to at the lowest -1dbFS.

Mixes are a different story. Make sure you bounce out your mix at the highest samplerate and bit res possible ready to import in again for mastering (except for Logic users, don't bounce to 32bit, bounce to 24).

jonahmann wrote:All right, but if you limit to -6, then normalise, you won't hear the difference, especially at high bit depths.. unless he is using analogue.

knobgoblin wrote:Normalization is not a destructive process. It simply scans the file for the loudest maximum peak and linearly adjust gain to bring that peak up to 0 (or -0.1 or whatever threshold is selected), so the balance between quietest and loudest parts remains the same. Also there seems to be lots of confusion regarding peaks and perceived volume (usually measured in RMS). Your track can never go over 0dBFS, so if you are "mastering" your own material any samples above 0 will just be clipped off. But the RMS and dynamic range you should be aiming for is a much more complicated and relevant question. I usually aim for the very loudest sections (of loud material) to be around -7.5 dBFS+3 RMS, occasionally breaking -7 if the bass gets louder in certain parts. Going above that you start to have serious tradeoffs between punch and volume. A very useful and free tool for analysis is voxengo's Span plugin. There is tons of usefull info on the subject at turnmeup.org

Normalisation is a destructive process, the calculations degrade the signal quality (obviously not a great amount), which is why it is not adviseable to normalise when there should be literally no need to. If your track is peaking at -7.5dbFS post-mastering then you should probably have a word with your mastering engineer.

Will confirm what dididub said. Normalization IS destructive. It's worse than destructive because even if it wasn't destructive and we had completely analogue infinite floating point normalization it still raises the noise floor already in your record.

This thread has gone on far longer than needed.

mthrfnk wrote:This thread has gone on far longer than needed.

Bad advice and misinformation needs to be cleared up or refuted (for new producers' sake)

dididub wrote:
Bad advice and misinformation needs to be cleared up or refuted (for new producers' sake)

I couldn't agree more.

I didnt say anywhere in my post that your track should be peaking at -7.5dBFS post mastering. Post mastering I assume the use of a brickwall limiter, fixing the peaks around 0. I was talking about the loudness of the track, as defined in RMS using a dBFS+3 calibration. The whole reason I posted in the first place was because the OP didnt seem to grasp the fundamental difference between peak and RMS, which your post also seems to imply.

Also, normalization is not a destructive process. A destructive process I would define as a function where there is no corresponding function to reverse the process. That would be anything to do with dynamics, where you fundamentally alter the relationship between the peaks and the noise floor. Normalization is simply a gain adjustment, linearly across an entire file. A normalized track and a non normalized version of the same track played at the same volume will null when one is flipped out of phase. You are not altering the S/N ratio at all. Inter-sample peaks, I believe, have more to do with the reconstruction process than with the file itself (which is why most masters are not actually limited to 0dBFS, but around -0.1dBFS or slightly lower as to not clip crappy D/A converters)

It isn't advisable to normalize a track in preparation for mastering however. Not because it "raises the noise" in a destructive way, but because if there is no headroom in a digital file, any processing done to it will most likely result in clipping before it even reaches the limiter. Also, when sending audio out to analog mastering equipment, you need to attenuate the signal as most gear is expecting audio around -18dB. So, the reason not to normalize is to provide the headroom for processing that a mastering engineer will do, and if you do normalize, they will just turn it down anyway to work on it.

Here is a good article on the subject:
http://www.hometracked.com/2008/04/20/1 ... alization/

this turned into a shitstorm. it becomes shit for me as an uneducated music producer when i dont know who to listen to.. so many conflicting opinions here.. :s