Re: yes!

john larkin <jlarkin_highland_tech> wrote:

On Mon, 19 Aug 2024 14:44:39 -0000 (UTC), Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:
 

john larkin <jlarkin_highland_tech> wrote:

On Mon, 19 Aug 2024 07:19:40 GMT, Jan Panteltje <alien@comet.invalid>
wrote:
 

On a sunny day (Mon, 19 Aug 2024 15:27:38 +1000) it happened Bill Sloman
<bill.sloman@ieee.org> wrote in <v9ul4a$2ogi5$2@dont-email.me>:
 

On 19/08/2024 1:14 am, john larkin wrote:

On Sun, 18 Aug 2024 15:39:46 +1000, Bill Sloman <bill.sloman@ieee.org>
wrote:
 

On 18/08/2024 11:16 am, john larkin wrote:

On Sat, 17 Aug 2024 17:54:38 +0100, Martin Brown
<'''newspam'''@nonad.co.uk> wrote:

<snip off-topic maunderings>

How about programming a computer to generate random character
substitutions in, say, a Python program, and test various resulting
versions to see if they improve, or better yet, perform some wonderful
new unexpected function.
 
That would be neo-darwinian programming, random mutation and
selection.
 
Actually, that scheme has been tried for circuit design. It didn't
work well.
 
Random mutation and selection does work to design LC filters, up to
3rd order or so. At higher orders, it diverges to nonsense.

 
If you parameterize using the LC values, I believe that. It’s very
difficult to tune a high-order filter unless you start out pretty close.
 
Parameterizing f_0 and Q for each section works much much better.
 

 
Active filters are easier, where the sections don't interact.
 

Yeah, you just pick the poles and zeros.  Of course the GBW has to be
several times f_0*Q.

Cheers

Phil Hobbs



--
Dr Philip C D Hobbs  Principal Consultant  ElectroOptical Innovations LLC /
Hobbs ElectroOptics  Optics, Electro-optics, Photonics, Analog Electronics