The good practices of saving audio samples?

[vinyldinosaur]

I've started recording some sound effects that I intend to use for sampling, to study sound design and synthesis.

What are actually the good practices for saving samples?

I can easily observe that .wav should be the file format.
But what I'm perhaps the most concerned about is the volume level; I must admit that the concept of volume still confuses me in music technology, and I'm looking for a good book on all this "boring but necessary stuff that you don't learn just through serendipitous fun". But now I'd just want to get on with the project anyway, I'm sure I'll come by the ideal comprehensive book eventually.

My questions are:
- Do you normalize samples, and why?
- How loud do you set the samples, and why?
- If you use an algorithm to set the volume, how do you determine which method to use?
- How do you take care that samples (especially ones that belong to a set) are of a similar volume?
- Is there a good way to batch-process samples for volume? (Mac, please)
- In which cases would you save a sample in mono vs. stereo?

Background info that's not a strictly necessary read in order to answer the questions:
Earlier, I was working on a project on my Maschine+ (standalone workstation) and needed to add a layerable sample that I'd recorded in Logic. Made a set of variations, just in case. Not knowing what the good practices are, I took a guess, normalized them, and sent to Maschine+. Fair enough, they sound fine when played back from the computer, but when browsing through samples in Maschine+, they were disturbingly loud compared to the factory stuff as well as the custom downloaded samples that I'd added there. Sure, I put them into the project and adjusted the track volume, but before I start making any more samples, I see that I should understand the process better, so I can establish good habits from the start; I don't want to browse samples with my headphones on, and get jolted by loud surprises! It does actually happen occasionally with factory content as well, but it shouldn't be, and I don't want to become one of those creators!

Thoughts?

 

[Slouching Raymond]

All I can suggest is trial and error, to work out a procedure.
You could 'normalize' your samples, to get them all the same, then work out how much to attenuate to match the factory samples.

I have a (I think Line6) guitar looper, which also has built in drum rhythms.
The only way to select a rhythm is to step through them.
I just needed a simple pattern to play along with, for an older audience.
Trouble was the nice quiet rhythms were all jumbled in with the loud in your face trance garbage.
The audience got a painful earfull of that garbage.
That looper was just a waste of money.
This story goes some way to give a reason to get samples at a similar volume.

 

[ecc83]

Not my field of expertise but my son builds songs from samples which he triggers playing guitar. I think his MO goes something like this...

Track 1 a click
trk 2 a kick
trk 3 a cello
trk 4 a bass
trk 5 a crumbhorn*
Rinse and repeat for the whole of Green Onions by Booker T.

So, "amplitudes" don't come into the equation. Each track will be at the level it was recorded so his next task is to balance all the tracks.

Now I don't know but if'n it were me I would initially set each one to average -20dBFS...Save As and archive that recording. Punch it up again and start mixing?

This is all in Samplitude Pro X 3 but I doubt other DAWs are much different?

*I jest but I think you get my drift?

Dave.

 

[TalismanRich]

Take one of the samples from Maschine+ and look at the peak and LUFS levels. That should tell you where they are. Then look at the peak and LUFS levels of your samples.

As long as a sample's peak is below 0dB, it can be adjusted downward. If I was doing it, I might leave the peak level around -1 or -2 to give headroom.

Normalization is simply finding the absolute largest peak for a sample or track and raising it and all other points in the file.a specified amount such that the highest peak is at 0dB (or whatever level you choose). It doesn't change the signal to noise ratio or distortion characteristics. If your largest peak is -12.4, then you are just raising the fader 12.4dB.