How does companding improve the sound ?

[Yuan Han]

I'm looking at this in particular in chorus/delay effects using NE570/1 compander IC.

From what I know it is supposed to give a better signal to noise ratio, but the question is how ?

My Attempt:
Compressing the input signal compresses the noise too, and thus the following signal processing processes the reduced noise signal.
However, when you expand the sound again, wouldn't you expand the noise signal again ?

Han

[Mark Hammer]

Yes, you DO restore whatever was there in the first place to some extent (although remember that in delay-based units, you are also expanding something that has undergone filtering between the compression and  expansion, so the noise is not EXACTLY identical).  The chief reason for using companding, though, is for those situations where there is the risk of *greatly* increasing noise levels by what goes on between the compresor and expandor.  In the case of both tape and BBD,s, there is the risk of a great deal of noise being added in with the signal.

The expansion will restore the initial noise to its original level (minus whatis filtered out) but it will also drastically reduce whatever noise is accumulated "on the way".

The other advantage of companding is that many devices (like tape and BBDs) have a limited dynamic range between the "noise floor" (that point where the noise is as loud as the signal) and the point of clipping.  Push the signal too meekly and the noise is too noticeable.  Push the signal too hard and the noise is replaced by distortion.  Compressing the signal first, allows the user to set the signal level just below the point of clipping, but well above the noise floor.  Think of it as a way of "sneaking" the audio signal through.  Then, once the signal is safely through, expansion restores the normal dynamics, minus both the additional noise and potential clipping.

Because companding DOES require use of envelope followers, which rarely track the signal flawlessly, and perfect restoration of dynamics also needs matched tracking (hard to do unless you precision-match components with a scope), companding will generally introduce some small aberrations in the dynamics, not to mention requiring additional current, cost, space, etc.  Consequently, some folks don't like it.  Inmany instances, though, it is hard to find a better or even equal solution to keeping noise and clipping at a minimum.

[Johnny G]

I have to say i have pondered the same question (tbh i was thinking about it last night with one of my little circuit books in one hand and the toilet paper in the otherbut thats another story lol)

i think it could work by having a higher ratio of expansion to compression so that the actual sound you want is amplified much more that the noise, sort of a more subtle noise gate thats always on i suppose.

i hope thats helpful. in the above mentioned circuit book that i was reading there is a section on the whole companding thing and i might have a it more of an indepth read about it at some point

JG