Re: Neural/Digital convergence: dynamic memory and bucket brigades

From: James Rogers (
Date: Thu Oct 09 2003 - 22:04:00 MDT

On 10/9/03 7:12 PM, "" <> wrote:
> I can see what you are talking about. We can say that a buffer is 'any method
> of delaying data without losing it'; simply sending the signal down some
> neural path which does nothing but return it will do the job.

It is a little more complicated than that, since you need a little more than
a delay. Just because it can be used as a delay (an essential property)
doesn't mean it has to be used for that literal purpose. You actually have
a fair amount of control via side-chains. It isn't just a simple cascading
FIFO buffer nor is the cascade necessarily automatic or continuous depending
on the nature of the side-chain; you can actually execute more complicated
conditional patterns within it while the sequential values are in-flight.
It is a slightly strange way to do things, but quite functional.

The only reason this even occurred to me is that a few years ago I was
working on modeling an archaic audio circuit in software that relied on
analog bucket brigade type circuits and logic. I thought the whole concept
was fascinating and peculiar, but nobody has used those in ages since
digital has been vastly cheaper and better for decades. It was a very
short-lived kind of circuit design, and if it wasn't for that project a few
years ago I probably wouldn't have been familiar with it.

While those types of circuits are way cool and have somewhat unique
properties for a physical circuit, there is a reason they never saw much
use. The signal-to-noise ratio was relatively poor (at best maybe 8-10 bits
usable dynamic range typically), something digital could easily trump and
did so cheaply shortly after these types of circuits became commercially
viable. Since these types of circuits can be emulated pretty simply in
software on digital circuits, little reason existed to continue making the
real thing for most purposes.


-James Rogers

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:43 MDT