[sdiy] Using analog for physical modeling

[Tom Wiltshire]

On 20 Sep 2011, at 21:08, Chris Muir wrote:

> 
> On Sep 18, 2011, at 12:55 PM, aankrom wrote:
> 
>> Quantization & its error aside, BBD's are truly analog. An actual analog voltage is passed through the chain of capacitors at each stage. At no time is the information transformed into 1' & 0's. That's a good enough definition of analog for me.
> 
> 
> Well, BBDs are a sampled system, albeit one where the samples are stored in an analog fashion. They have all the same issues as other sampled systems, with the added problems of drooping analog samples. This is veering into a philosophical "what is _true_ analog" discussion, I guess.

It needn't get too philosophical or complicated. There's two different dimensions that can be discrete or continuous - time and level. A full digital system uses both discrete steps in time (sampling), and discrete steps in level (quantization). BBDs use discrete steps in time, but continuous levels. Full analogue systems are continuous for both time and level.

For me, the significant thing about the proposed BBD-based KS synthesis system is not the continuous levels, but rather the variable sample rate. One of the big problems with typical digital systems is the fixed sample rate, which means that you finish up needing to interpolate between the available data to get sub-sample increments (not that it hasn't been extensively studied and dealt with - it has). Everyone who's worked with fixed sample rates will have come up against this at some point. A genuinely variable sample rate such as is provided by a VCO BBD clock avoids this issue completely.

Tom