Re: Synthesis (was Re:Amelio's ouster)

[nobody@REPLAY.COM (Anonymous)]

"Geoff Kaiser" <superkais@macconnect.com> wrote:
>(Mitchell Spector) wrote:
>
>>    ....The Sound Canvas does not have
>>a digitized recording of _each_ and _every_ note on the keyboard
>scale!
>
>Never said they did....they map several samples over the audible range of
>notes available, just like a sample keyboard. A typical piano sample may
>have 10-20 notes sampled, and played back at higher/lower rates for the
>notes in between them.

Like NoiseTracker GS (all versions), SoundSmith, and MOD players.
True.  However, the "synthesis" in wavetable synthesis refers not
to these sample players, but rather to something like synthLAB.
synthLAB takes a small sample and modifies the wave.  Specifically,
you can have one sample and, using the CPU, map 15 or more different
synthesized instruments from the same wave.  This is done by shaping
attack, decay, sustain, and so forth.  As you know, from the specs
of the Mirage, the Ensoniq 5503 chip that the GS uses has 32
oscillators.

synthLAB uses 4 of these together to produce a single instrument.
Two oscillators are paired to form a generator.  Two generators can
interact with each other (feeding gen one into gen two or vice versa)
or with itself (one generator plays straight through while the other
feeds back).  When you are done synthesizing your instruments from
one wave, each can sound very different than the others, although
they all originated from the same wave.

This is different from "sampler" programs such as NoiseTracker GS.
NoiseTracker GS allows you to have "patches" of samples limited only
by system RAM.  If you have 8 megs, you can have close to 8 megs
of samples for the instruments.  However, except for minor modifications
(such as fade in, fade out, pitch blend, etc.), they only map the
samples over a range of frequencies.  Here, only a single generator
(two oscillators) per voice are needed.  More flexible (near 8 megs
of sample space and up to 16 different simultaneous voices), but I
agree that it is not synthesis in the sense that synthLAB is.


>>The Ensoniq 5503 in the IIgs worked the same way, which I wonder, do
>>you not consider that a synthesizer either? Just what IS it then?
>>And what was the Mirage keyboard that used the same chip back in
>1985?
>
>I'm not too familiar with the IIgs's built-in card, but the Mirage was a
>sampler, not a synthesizer. You had to load each and every sound on floppies
>to hear them. The Mirage, like most any sampler, doesn't create sound, it
>only re-creates sound.

Then by your definition, the IIgs is a sampler as well.  The IIgs
sound hardware works the same as the Mirage (same Ensoniq 5503 DOC
chip).

However, being a computer, it can use the CPU to do wavetable
synthesis (generating different sounding instruments from a single
wave).  You can, of course, use it _as a straight sampler_ in which
case, using NoiseTracker GS, you have up to near 8 megs of space to
hold your samples (rather than only 128K as in the Mirage).


>>    I suppose if you want to be really picky, the AdLib, SoundBlaster,
>>Commodore VIC20, C64, TI99/4a, Mockingboard, Phasor, Atari ST, etc all
>>had true "synthesizers". They generated sound from simple waves (i.e.
>>squarewave, sinewave, triangle) or complex waveforms (FM synth).
>
>Yes, that would be correct. I'm not saying that a true synthesized waveform
>is the most desirable. Sampled ones usually are. I was only saying that
>samplers are not synthesizers.

But Geoff, in wavetable synthesis, you are taking a single, sampled
wave and modifying it (in real time) based on parameters that you
input into multiple instruments.  That's not just sampling.  It is
also synthesis.  Hence the term "wavetable synthesis."

Again, the GS can also act as a very good sampler (better than the
Amiga, even, because you can use 16 as opposed to 4 voices...the
GS can stretch to 32 voices with Diversi-Tune while the Amiga can
stretch to 8 with OctoMed, but those are exceptions), but it can
also do synthesis.

Diversi-Tune uses totally synthethic (FM synthesis type) waves, btw.


>>    This is new, now if a Macintosh doesn't have something, it not only
>>is an unnecessary addition, but you can try and mess with terms to the
>>point the thing doesn't exist at all in other computers. My point was:
>>The Macintosh has no synthesizer (call it what you wish) while PC's can
>>and often do have one, through add-on boards.
>
>My god, you are missing my point! Again! A wavetable card is NOT a true
>synthesizer! Neither is Quicktime's built-in Roland sounds! I'm not arguing
>whether having a card is better or not (the advantages of either are
>obvious). I'm just saying that in the end it's the same thing. It is
>COMPLETELY INCORRECT to say a generalized statement like "the Mac has no
>synth, PCs have one". They each have the same resulting type of sound, one
>does it through hardware, the other through software. But neither are
>actually creating waveforms from "mathematical equations", as you say.

But the "wavetable synthesis" devices _are_ doing synthesis in that
they are modifying waves on the fly.

However, I think Mitchell's point is that the Mac has no music hardware
chip (call it a sampler if you wish) and depends on the CPU to do all
the work (and he can hear the difference).

>
>>... Now the
>>chip has to be in a keyboard to be consider a synthesizer? I'll bet
>>the Mirage would be a true-synthesizer, but the IIgs not, right?
>
>See above...the Mirage was a sampler.  IIgs I haven't heard....

The IIgs _can_ work as a sampler, but you can also use it to do
synthesis (see above).

As a sampler, it is pretty good.  Take a look at it.  On a resonably
fast hard drive, you can run oversampling to 4X on the wave, resulting
in a playback rate of >100kHz.  Sampling programs that can record
go up to 56kHz.  The playback is very clean, but I've only heard
crappy old SoundBlasters (the original models) and the whine of a Mac
Plus to compare (listening to each computer through headphones, of
course).

The Ensoniq 5503 in the GS has 32 independent oscillators (sound
generators).  It has a dedicated 64K RAM buffer for holding samples
without CPU intervention (but samples can also be held in main RAM).
All _synthesis_ programs currently available using the MIDIsynth tools
are limited to 64K of samples.  However, _as a sampler_, the GS
can access up to near 8 megs of samples with the right program
(NoiseTracker GS) and enough memory installed because sampler music
programs can store samples in main RAM.

To hear MIDI files as they were meant to be heard, I sometimes pass
the MIDI feed into SoundSmith, run it through NoiseTracker GS into
16 voices, and then add each instrument patch into NoiseTracker GS'
RAM bank (I have 8 megs, so there's basically no chance of running
out of sample space).  The results are great!  Because GS synth
programs don't use main RAM to store sounds but instead are limited
to the buffer (while GS sampling programs are not), I find it is a
better sampler than a synthesizer.

It is a nice MIDI machine too.  Check out the MasterTracks and Music
Writer demos.  The "Pro" versions support some 16 to 32 staves and
work much like the Mac versions.  They use the same MIDI interface
as the Mac.  You can also use the GS as a MIDI synth on a MIDI setup
(driven by your Mac) by having it in "input" mode in synthLAB or
Diversi-Tune.

-Scott G.