From: Larry Sayre 
To: atm{at}shore.net
Reply-To: Larry Sayre 


I ran PLOP several times, each time allowing 'angles to vary' on a design
for a 27 pont cell.   My first run simply used "Automatic Cell
Design".  Subsequent runs were done with nothing "fixed",
everything set to "optomize", and I ran through a series of finer
and finer step sizes, causing PLOP to run through many more iterations to
come to a final solution each time.  My last run went through several
thousand trial iterations before coming to an end.  Each time I allowed
PLOP to run more iterations (finer step sizes) before coming to a solution,
the RMS error went down a bit more vs. the previous run(s), but the Peak to
Valley error went up a bit more.  By my last run, the Peak to Valley error
was up appreciably vs. my initial "Auto Cell Design" run, but
this run yielded the best RMS error result.  Is this as it should be? 
Should I be concentrating on lowering the RMS error and ignoring the rise
in P/V error?  Overall the RMS error did not come down as much each time as
the P/V error went up.

Lawrence Sayre

--
My philosophy, in essence, is the concept of man as a moral being, with his
own happiness as the moral purpose of his life, with productive achievement
as his noblest activity, and reason as his only absolute.

Ayn Rand (in the appendix to 'Atlas Shrugged')

--- BBBS/NT v4.00 MP