Could someone be so kind as to explain to me the difference between compression / mulitband compression vs. doing equilization, limiting, and normalizing?

edit: The application is FLStudio, if that helps any.

Some background: multiband compressors are usually used by mastering engineer's where the mastering is done on a frequency by frequency basis. They are a littled more complicated to use. Hard limiting the peak would be compressing it to a certain decibal level so that it has a constant peak. Some Nu Metal records are hard limited to 6 dB if you analyze the waveform. That's the easiest example I can think of. Compressors are most commonly used in AM/FM broadcasts to make the music sound more dynamic and "punchier" too.

Compression is a process of limiting the dynamic range. Basically it is saying, here is the loudest part of the audio (the highest peaks). Make the quieter parts closer to this but leave the louder parts alone. It can also be made to work from the other end, de-amplifying the louder without effecting the lower levels. Multi-band compression is doing this independently for limited frequency ranges, i.e. treat the audio from 250Hz to 700Hz independently from the audio at 701Hz to 1200Hz, independently from ... .

Equalization is amplifying, or de-amplifying, specific frequency ranges. It does not act like multi-band compression in that it acts equally for all audio within the specified frequency band, raising or lowering the level equally across all amplitudes.

Limiting is picking a maximum amplitude and saying 'nothing shall go above this." Anything above that level gets squash to that amplitude. Anything below that level is not effected.

Normalization is amplification, or de-amplification, to some predetermined level. Everything is amplified equally, there is no change of relationship between quieter and louder parts. It is like using a volume control while listening except that it is changing the data in the file so that, at a given volume control setting, the audio will be either louder or softer than it was before normalization.

this is totally nitpicky, but actually what you're describing here is an expander, not a compressor. a compressor squishes down the loudest parts and then compensates for it with make-up gain.

I thought an expander did the opposite, IE make the quiet parts quieter, thus expanding the difference between soft and loud.

Compression reduces the dynamic range, expansion increases the dynamic range. Compression may indeed be more frequently done by reducing the loudest parts rather than raising the quieter parts, my second expressed option. I confess ignorance on most general practices; I use it infrequently myself. Reducing the range from the top may have an advantage vis a vis the noise floor, but it was still compression, not expansion that I described. The make-up gain is an option that can be applied or not applied as desired.

No. AndyH-ha was right. In both cases the louder parts and quieter parts get "closer" to each other. This is compression. Expanding is moving the louder parts farther away from the quieter parts ("distance" measured as volume difference).

By your definition an expander is equivalent to a compressor followed by a positive gain change which is not true.

Sebastian

Does any of these processes allow the chopping off (not squashing) of signals above a certain level or below a certain level?

What do you mean by "chopping off"? Do you want the resultant waveform to be the same as if clipping occurred at whatever signal level you select?

I don't know what might be available, but it seems unlikely anyone would much expend effort to produce a transform that deliberately does what everyone tries to avoid. However, you could always accomplish that result by amplifying beyond 0dB (e.g. normalize to 110%). If you are working in integer, the clipping would be immediate. If working in floating point, there would be no damage to the waveform unless and until you converted the file from floating point to integer.

This doesn't work on the low signal level end of course, but it's hard for me to imagine just how that might work in the data or why anyone would want it

Ok so on the high end that is clipping. I was actually going to ask about clipping, but you've already answered.

I vaguely recall from many years ago seeing Adobe Audition having a function that could apply multi-band equalization within a range of levels, without affecting content above or below that range. Was wondering if there is a proper name for this? Or I may have misinterpreted the function.

For an example on the lower end, consider material recorded live which has low level content (such as audience chatter) which is too random to be filtered out by simple noise reduction.

Equalization is by its nature 'multi-band.' To EQ something means to amplify or attenuate some limited frequency components. This may be done only in one range or 'band' or (to different amounts) in many, but the idea is that different frequencies are treated differently. If you involve the entire audio range you are doing simple amplification or normalization.

I'm still not sure but it sounds like you just want to make lower level sounds less audible. This is expansion, increasing the dynamic range. You move the loudest and softest further apart and the softer stuff becomes less audible. This can be done gradually over the entire range or it can be set to cut in at some particular level (e.g. at -33dB, effecting nothing louder) and can have whatever slope you want. There may be several different points (levels) where the slope changes.