🚀 go-pugleaf

 
I'm re-posting this article I received as a response. Permission was
granted to do so, although I claim no responsibility for its contents.
----------------------------------------------------------------------

 
Subject: Re: Looking for a *USEFUL* PROJECT book.
Newsgroups: sci.electronics
Keywords: So far, on
ly JUNK.
Summary: was projects; how about computerized design assistance
References: <553@xroads.UUCP>
 
I too learned much from reviewing, and occasionally building, the simple
examples....learned lots of principles that way. Maybe we are just 10 years
older, and those examples now are too simple.
 
In recent years, after 10 years of "professional design experience" and
[at that time] 4 years of studies of CS and AI, I earned an MS by developing
an automatic circuit synthesis system, Dec 1986. LISP based, for several
good reasons, capable of designing broadband bipolar amplifiers, AC coupled.
 
We hit it with the task of generating an amplifier capable of interfacing
between a phono input and a loudspeaker. It ignored the RIAA compensation
since I had "told it {ACSS} nothing about RIAA", but the solution was a
pretty good solution----given that push-pull buffers were not in the suite
of diagrams/templates making up the knowledge base. The final stage
transistor would have melted since was only a 2N2222, and the collector
resistor would smoked; both those problems are easily corrected since the
knowledge IF/THEN rules to upgrade the components are simple; I just had
not expanded the spec/design/fail/analyze/change_spec behavior to handle
single-point problems as trivial as substituting a much more powerful device
[with the requisite increases in junction capacitance,etc].
 
The system, ACSS, currently performs amp design by satisfying the only overt
constraint between cascaded stages, their Zin and Zout interaction. 
 
For previous design test cases, with R-source nominally that of R-load, 1 or 
more Common-emitter stages were the usual winners. 
For the power amp design we got an unusual selection of
stages. In this case, R-source is 47K, R-load is 8 ohms, and a massive
impedance matching is part of the problem. We also asked for an amp gain of
100.
 
The winning design, after about 2 minutes of VAX11-780 time, had
  input stage           Common-Base-gain&Common-Collector-buffer
 
  middle stage          Common-Emitter-gain
 
  output stage          Common-Base-gain&Common-Collector-buffer
 
 
This was puzzling, but in a day or two, after reviewing the winning stage 
select heuristic, I realized the heart of this problem was not providing gain
but matching to source and load. The several Common-Collector substages had
provided that. AHHH ACSS was smarter than I had thought. (I had learned
several other things about gain stages while "capturing the knowledge".)
 
SO....automatic design EXISTS. For bipolar AC-coupled amps, not worried about
noise floors or linearity,etc. 
 
Currently I am studying for qualifiers, but after March...what would be good
domains to try [emphasis TRY] to add to the system? And in particular, what
would be helpful to the lone home experimenter, trying to learn electronics?
How much explanation is valuable, during the design behavior? and afterwards
for credibility? and if someone different needs to significantly alter the
design without assistance of the original designer [who is on honeymoon, or
at a new job]?
 
Allen Sullivan
sullivan@asuvax.asu.edu         [I think]
sullivan!asuvax!asu!noao!nud    [I think]
 
----> Side question...You wouldn't mind if I asked for a copy of this program
now would you ???
                    -Mike
-- 
==========================================================================
= Mike Carter  !Xroads.wiz | S P Y T R O N I C S   E N G I N E E R I N G =
= N7GYX        Phoenix  AZ | "Imagine your worst nightmares as circuits" =
==========================================================================