batch normalize sample collection

[taifun]

Hi guys, out of practicality and eardrum-preservation considerations, I am thinking about normalizing my huge library of one-shot drum hits. I've just had it with continually having to adjust the output volume of my sampler when browsing through different samples. Before going ahead with this, I wanted to see if anyone has any advice about this. I was thinking of normalizing everything to -3dB, which I think would be a good middle ground in terms of s/n ratio. This entails that for samples that are originally close to full scale I will lose some bit resolution and for other (very quiet) samples I will be lifting the noisefloor up. The -3dB serves a couple of purposes though:

1) I never need samples to be at full level in my mixes, so this will trim them down a bit beforehand.
2) for the samples that were originally close to full scale, -3dB creates a bit of headroom for further processing.

My main concern is with bringing up the noise of the samples that were originally rather low in level. I think the latter might be worse than losing 3db of bit resolution for loud samples. This sort of makes me want to normalize even lower, which would be even closer to most real life mixing levels I will want the samples at in the end. Any thoughts/pointers to good threads welcome!

[thecontrolcentre]

I wouldn't do it. As you already said, you'll end with a load of "noisy" samples where the noisefloor has been pushed right up. Would make as much sense to separate the loud samples from the quiet one's imho.

Are you sure you want your hihats to be as loud as your snare's and kicks?

[whyterabbyt]

I wouldn't do it. As you already said, you'll end with a load of "noisy" samples where the noisefloor has been pushed right up. Would make as much sense to separate the loud samples from the quiet one's imho.

Are you sure you want your hihats to be as loud as your snare's and kicks?

I concur. Normalise the copies in use, not the originals.

[taifun]

I get your point about noisy samples, but the thing is that the noise is there anyway, even when they are left alone. The quiet samples would have to be turned up anyway to the targeted level if I would use them as they are now.

If I would normalize the quiet ones, there is the good chance that I make 2 volume adjustments to those samples: first they get turned up a whole lot, only to lower the level again later, the end sum of which is loss of signal to noise ratio. On the other hand, if all samples are at the same level, it might be easy to spot the samples with poor s/n ratio and just toss them from the collection. This could help me clean out my collection which is way too big anyway.

[thecontrolcentre]

I get your point about noisy samples, but the thing is that the noise is there anyway, even when they are left alone.

So there's little point spending time normalising them, is there? It's better to edit samples in the context of how they're being used. I always make copies of the samples I'm using in a track (collect all and save).

my collection which is way too big anyway.

So is mine.