Audacity - horrible clipping!

[Speed Demon]

I like to use a plugin to adjust the dynamic range of a track. Especially on vocals, drums and percussion instruments.
It's great for maintaining and brightening those soft passages that can easily get lost in the mix.

[Hilary]

I've had the same problem before. If you normalize all you're tracks and then leave them all on 0dB, there will be clipping, because the result will be more than 0dB. So just adjust the volume down on all tracks.

I can't remember ever having this problem on the BR's.

I am not a fan of NORMALIZING as it really takes away the dynamics of your recordings...IMHO.  I want those subtle lows and highs of an instrument - normalizing, to me, is like making everything scream at you.

Record with the instruments not so hot coming in.  You can always increase gain in the software, but getting rid of too hot an instrument in the mix... will also get rid of subtle sounds you may want to keep.

Mike

This is interesting as I always normalise my tracks (won"t be from now on) and the headspace tips. As far as the original clipping, I'd just highlight all the tracks (ctrl a) and reduce the sound until it didn't clip (effects amplification) unless I'm missing something, which is often the case!

[AndyR]

Normalising of itself is not a problem, Hilary, it's very useful. All it does is analyse the audio file you're normalising, looking for the highest peak. When it's found that, it calculates the distance between that peak and the ceiling, then it amplifies EVERYTHING in the file by that exact amount and prints the result. You don't lose any dynamics at all with normalising on its own. Feel free to keep using it!

The way you lose the dynamics is in what you do with it next... eg compressing it so you can turn it up further. Compressing or limiting will lower peaks above the threshold you set and leave the rest untouched. So when you amplify it some more (maybe by normalising again), the quieter bits get louder in comparison to the loud bits - that's what compressors/limiters are for, they reduce the dynamics to a level you are happy with. If you do it right, you make your performances sound more musical than they were, if you do it too much, you remove the expression you put into the performance. Sometimes that's what you want, sometimes it isn't - it's all up to what you want and what you can hear yourself... there is no rule of thumb on this, really, even though we all need one! Certainly beware of using anyone else's rule of thumb - they have different listening preferences and are trying to achieve something else with their own recordings. The way I look at it is - I'll try someone else's idea, and if I like it, I hang on to it... but I'm continually trying to understand better and questioning what I do (whether it's my bright idea or someone else's).

I always normalise my finished mixes to 0db (100%) before mastering them - that way I always know exactly where I'm at (at "max") before the signal goes into the mastering tools. I understand some professional mastering engineers prefer 80-85% because that gives them head-room - but that's just the way they've learnt to use their tools. The main thing is knowing exactly where you are relation to the ceiling, that way you can use your usual tricks knowing you won't clip. For me, during the amateur mastering I do, I prefer being "at the ceiling" and making sure I never amplify (turn anything up - volumes or eq) from there before the signal has been lowered in some way by something else in the mastering tool chain. That just seems to work for me.

For example (I'm using the BR1600's mastering tools here), the first thing in my mastering chain is an EQ. Usually I don't need this (if the EQ is wrong, perhaps the mix is wrong!), but sometimes I do. There's an input volume on the EQ, if all I want to do is cut some EQ, eg cut some bass or mids, then I leave that volume at 0, and everything is cool. But if I want to boost something, I reduce that input volume. Otherwise, because the audio is already normalised to 100%, the peak(s) are already at the ceiling, ANY amplification (and boosting EQ is amplification) will take it through the ceiling and introduce clipping, whether it's audible or not.

After mastering, I always normalise the final file, I want it as loud as possible (remember, normalising does NOT remove dynamics itself, it just turns EVERYTHING up to the ceiling you set). I then check the normalised version against the non-normalised. If it's not louder, I bin it. If it is, and it's still clean, that's what I publish. For a while, I tried to avoid this last normalising stage by using the mastering tool to amplify (partly because I'd heard that "normalising is wrong") ... but I found that my judgement is not good enough to avoid pushing it too hot. I've since realised normalising is never wrong - it's just that it can lead you to do some other stuff later that might be "wrong". So now I always master to safe levels and then normalise to amplify the results cleanly to 100%.

BEFORE MIXING, I very rarely use normalising on individual tracks anymore. The way I look at it is this - if I do normalise, it makes it louder, and then I have to turn the fader down further when I'm mixing!!

I'll normalise an individual track for one of two reasons:

1. The recorded audio is too quiet and I can't turn it up enough (and I have to turn the rest of the tracks down to get a balance). But turning anything up, either with the fader or by normalising it, also turns up the noise that came with the performance... so, if the audio is too quiet, I'll often re-record it at a better level - less noise introduced then.

2. If I want to use a compressor for effect on the track during the mix. And I mean for an audible effect, not just to protect against peaks during mixing. I've just used this yesterday - I have this spikey guitar part that's on the upbeats and has to stick out. It wasn't working, it sounded not as good as I wanted... then I realised I was playing it a bit inconsistently, and I can't do it any better. I don't want to lose any dynamics, but these spikes are the only thing on the track, so I'm in less danger. So I want to use heavy compression (limiting) to even them out. I could have spent AGES fiddling with the threshold on the compressor to hear where to set it to cut about 6db off the top of the performance (then the ones that stick out will be nearer to the rest so it sounds like I knew what I was doing when playing)... but I don't know where the peaks actually are! SO - normalise the track to 100%, I know exactly where the peaks are now ... all I had to do was set the compressor to -6db, and hey-presto, exactly what I wanted... I bounced the thing to another track that I could then turn down and use in the mix - sounds loads better now... I removed a bunch of dynamics that I didn't want, that were making the whole mix sound naff (of course, the real answer was to get a session musician who plays that style of music better than I do! - or delay the process so that I could practice the part for a few weeks!)


Sorry, went on a bit (I'm meant to be cleaning the oven )... but here's the conclusion:

Think of normalising as a volume control that enables you to automatically turn the track up to max without clipping. It affects everything in the track equally, just like the physical volume control does... BUT, always remember, once you've normalised it, you CANNOT turn it up or boost its eq anymore without introducing clipping (unless you compress it first).


Now, having said that, my experience of audacity's normalising is not great (or maybe the BR's is too forgiving?). If I take an un-normalised track and normalise it to 0db (100%) in audacity, and do NOTHING else (levels set to unity etc) ... it ALWAYS clips. That strikes me as plain wrong. You've asked it to set the peaks to 0db, and it's put the bluddy things over 0db!!! I don't use audacity for stuff like this!!