Delta and Sigma (was: AW: Wavetables....another idea....

[Haible Juergen]

	>The second formant book shows a sigma/delta converter for a delay.
	>With sigma delta you can only process ac, some kind of "reset"
logic
	>has to be applied in order to get proper absolute voltage levels.

My memory is clouded, but I think it was Delta Modulation, not
Sigma-Delta.
Sigma-Delta is good for DC, too. Delta Modulation is not.

As for Delta Modulation, I think Elektor used a special chip
that was original intended for speech bandwith. There was 
a DIY project in ELRAD (and I wouldn't be surprised if that 
would have appeared in ETI, too) called "Delta Delay" where
they build an adaptive Delta Coder / Decoder from standard
components like CA3280 and CMOS shift registers.
I built my own version of this back then, replacing the digital
delay time selection circuit with a HF-VCO, and used the
13700 instead of the 3280. I remember the sound quality
was excellent (*no* noise, and I mean it!) for sounds with
slow transients, but not that good for the bell-like DX-7 sounds
I was fond of in that time. This is strictly according to
adaptive delta modulation theory, of course, where you 
make a 1-bit conversion at a high sampling rate (oversampling),
comparing the momentary input level with a predicted level
that's generated from a certain number of previous bits
(shift register ...), and an exor logic that checks for sucessive
"1" or "0" and ajdusts (OTA ...) the "delta" threshold for the
next samples accordingly. Adjustment takes some time
(some number of samples), so fast transients are distorted.
The great advantages of this technology were:
(1) No "brick wall" band limiting in the audio range required.  
Remember the adverisment for Delta Lab Delays ? What
they didn't tell us was that it was only true for small HF contents ...
(2) No need for exact input level adjustment. The "adaptive"
part was like a built-in compander, too. The most astonishing
thing was how quiet this delay was even at low input levels. A great
plus in comparison with the 8- or 9-bit systems of those days.
(3) As the information was compressed, you needed less memory
for a certain delay.
(4) As the information consited only of one-bit "words", you could
simply hook up a number of 1-bit organized memory chips.

I don't have this delay anymore (another sad story), but I have
a Delta Lab delay now. The sound quality of the ELRAD version
was much better. (The Delta Lab has a simpler coder.)

JH.