[Elecraft] Re: Homebrew Autotuner

[David A.Belsley]

Joseph:
   Genetic Algorithms would be an extremely inefficient method for  
solving such a problem where the objective value is so easily measured  
and the direct effects of the arguments of the objective function  
(capacitance and inductance) are so easily determined.  Wayne's  
solution is relatively direct and clearly effective.  Kevin Rock's  
proposal, using some method of steepest decent is also of interest.   
However, anyone who has used a transmatch with tunable input and output  
will know there arise situations where simple iteration between the two  
tuning capacitors does not simply walk down a SWR valley.  Occasionally  
one must significantly off-tune one cap to get a lower reading with the  
other.  This, then, is a situation where first derivatives are  
inadequate and second derivatives are best brought into play.  Steepest  
decent is a "first-derivative only" algorithm.  Some form of a DFP or,  
better yet, Newton's method -- ones that also use the Hessian  
information -- would be of interest in this context.  However, all  
these kinds of optimization routines are assuming continuous  
variability of the arguments, C and L.  With a very finite number of  
values of C and L to consider, a tree network with branch-and-bound  
really seems the way to go.  Switches in values that increase SWR are  
pretty effective means for pruning branches off the tree.

best wishes,

david belsley, w1euy



On Sep 14, 2005, at 9:57 AM, Sanger, Joseph wrote:

> Stephen Kercel's suggestion to use a Genetic Algoirthm is brilliant ...
> this sounds like a perfect application for these astounding and amazing
> techniques.  I have had two passions in my life (at least, PG-rated)  
> ...
> ham radio and genetic algorithms ... own every book ever written on  
> GA's
> as well (have even read most of them!) ... (not a lot of real world
> experience with GA's of course ... I am more of a theoretician!).
>
>
>
> The beauty of the GA's, loosely modeled on the way real DNA and natural
> selection work, is that they often finds good empirical solutions to
> difficult problems like tuner settings optimization, without the need
> for the articulation of the exact methodology used.  In other words,  
> all
> of the factors that Wayne wrestled with become moot ... the GA will  
> find
> good solutions to the problem without all the intellectual knowledge of
> antennas and radio physics.
>
>
>
> I  would encourage any of you who want to have some real fun to search
> google and read about GA's.
>
>
>
>
>
> Joseph Sanger, M.D.
>
> WB2SSB
>
>
>
>
> ----------------------------------------------------------------------- 
> --------
> This email message, including any attachments, is for the sole use of  
> the intended recipient(s) and may contain information that is  
> proprietary, confidential, and exempt from disclosure under applicable  
> law. Any unauthorized review, use, disclosure, or distribution is  
> prohibited. If you have received this email in error please notify the  
> sender by return email and delete the original message. Please note,  
> the recipient should check this email and any attachments for the  
> presence of viruses. The organization accepts no liability for any  
> damage caused by any virus transmitted by this email.
>
> ======================================================================= 
> =======
> _______________________________________________
> Elecraft mailing list
> Post to: Elecraft at mailman.qth.net
> You must be a subscriber to post to the list.
> Subscriber Info (Addr. Change, sub, unsub etc.):
>  http://mailman.qth.net/mailman/listinfo/elecraft
>
> Help: http://mailman.qth.net/subscribers.htm
> Elecraft web page: http://www.elecraft.com