The Logic Off Topic list

GA Moore wrote:

>Sampling involves taking measurements and recording them for
>reproduction.
Depends what you mean by 'recording', I would guess.

>I think the analog modeled softsynths are creating a file which is of the
>same form as one which comes from samples of something,
I am curious. What file are they creating?

>A square wave for example is very simple to create mathematically.
I am curious again. How would you, personally, do this? I had always 
thought it was actually very hard to do.

>These sounds never heard
>the 'light of day' until they are converted to audio.
Well, what I always thought was that if you look carefully at the output 
end of a sampler ... then it really doesn't care two hoots what it's 
original data source was and where it came from. The sampler just computes. 
What it computes is what 'noise' it's going to output given the data stream 
passing through it at that moment. The source of that data is pretty 
inconsequential to the sampler at that point -- although it's obviously 
pretty important to us humans because we tend to gauge the effectiveness of 
the sampler by comparing that output to some original input that we have in 
mind. Can't see that that's the sampler's problem, though, to be honest!!

What I mean is, I'm not exactly clear about this 'light of day' bit.



Strictly speaking, the word 'sampler' comes from the Latin word 'exemplum' 
and simply therefore means 'the example to be followed'. Therefore, when 
you 'store' or 'record' the original data, all you are doing is storing an 
example that is to be followed ... i.e. a string of data that you will 
later use to recreate that 'something' of which you have taken an example. 
The output stage of a sampler does not 'know' that, originally, you got 
that data by taking measurements at a given bit rate. How could it 'know' 
such a thing? All you have is data. Not sound. So ... when it gets to the 
output stage, all the sampler is doing is mathematics and working with its 
data. At least ... that's my impression of things.

It's true that what you are giving the sampler is an example data stream -- 
modelled after a 'real world' exemplum -- that you would really rather like 
it to follow and recreate. In practical terms it's also true, a sampler is 
only really much use to us humans if it follows pretty strictly those 
examples taken from 'the real world'. However, that is because that is what 
we choose to do with it. Once it comes to the output stage, however, in 
every single case the sampler is simply working with the data and 
ultimately spitting out a sound ... but there is no de facto reason WHY -- 
looking strictly at that sampler -- any given sound to be created should in 
fact ever have previously existed 'in the real world'. It's questionable, 
in fact, if it ever did because what the sampler has is data and not a 
sound. All it did was take an example and it's surely a moot point whether 
or not that example is 'the real thing'. I was originally taught that it's 
only in fact a model of the real thing meaning that a sampler is really 
just a modelling device.

It is very easy, in fact, to make a sampler spit out an unearthly sound 
that has never before existed. All one has to do is mess around with data 
-- even if that data is some other sample. This is because all those newly 
designed sounds coming out of some sound designer's mind have never before 
seen the light of day. Peter Gabriel, for example, was famously able to 
transpose some very basic samples up and down a couple of octaves and so 
produce some long and complex timbres ... which had never before seen the 
light of day. Nor had they existed anywhere before ... except in the mind 
of one Peter Gabriel.

Roland for example, when they took to making samplers, used a method of 
differential interpolation to 'recreate' noises. It's hard to see how any 
of that had ever seen the light of day before either.

Far as I know, historically, a sampler is just a common or garden 
synthesizer. I always thought that that's why they always featured in books 
on synthesis techniques. The thing about a sampler -- or so I always 
understood -- is just that it so happens to have an infinite array of 
waveforms available to it simply because at the input stage it is possible 
to throw some data at it that starts off life by being an example of a 
waveform that we humans would immediately associate (when suitably 
reconverted) with some kind of 'natural sound' in 'the real world'. But 
from the point of example-taking onwards a sampler is just another 
synthesizer. The only difference between a sampler and any other kind of 
synthesizer is the source of them waveforms. Samplers do have to deal with 
some pretty complex waveforms -- i.e. mathematical data. But then again ... 
a square wave is pretty complex too! The sampler-as-synthesizer simply does 
not know -- how could it? -- that it is dealing with a waveform that 
originated as an example of something 'in the real world'.

I could well be mistaken, but I was taught when I learned about these 
things (a long time ago, admittedly and age is dimming my memory) that this 
'fact' is actually 100% irrelevant to the mode of operation of a sampler 
when it comes to the output stage. It's simply dealing with a waveform 
represented in numbers and so is just another synthesizer -- meaning that 
they're all equal in their digitalness-ness. They differ only in the source 
of wave forms whether this be physical modelling, granular synthesis or 
anything else.

Come to that there's getting to be very little difference these days 
between sampling and hard disk recording. They both deal with digitizing. 
All we do is use the HD as a real-time playback device. And, just like 
seems to have happened with sampling per se, the 'recording' and 
'manipulation' and 'performance' of data is pretty much all of a muchness, 
really. You can lift get things off your computer -- plug-ins, stimulation 
of MIDI instruments -- and then produce a final audio output that never in 
fact existed 'in the real world'. Your basic song in really just one long 
and complex and manipulable waveform.

But ... then again ... maybe I'm totally mistaken in my understanding of 
samplers and synthesizers ... I am mistaken about a lot of things, I guess, 
although I do like to keep learning.

>Because of that, they may not have many of the ill affects of sampling
>process - which happen during the analog to digital conversion - which
>these sounds never go through.
Can't argue with a 'may'.

>For one thing there should be no
>distortion - at least if the algorhthms are well written.
Can't argue with a 'should'. And ... I guess ... all we really need is a 
well-written algorithm. Lotta highly intelligent and creative people 
working on that one. Maybe they'll do it one day soon. Here's hoping, anyway.

Kool Musick
Keep Musick Kool


_________________________________________________________
Do You Yahoo!?
Get your free @... address at http://mail.yahoo.com