NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/sys/hp48/6294

Internet Message Format | 1992-12-13 | 3.6 KB

Path: sparky!uunet!spool.mu.edu!uwm.edu!ogicse!news.u.washington.edu!news.uoregon.edu!nntp.uoregon.edu!stevev
From: [e-mail] (Steve VanDevender)
Newsgroups: comp.sys.hp48
Subject: Re: SOLVE questions
Message-ID: <[e-mail]>
Date: 14 Dec 92 00:57:45 GMT
Article-I.D.: miser.STEVEV.92Dec13165745
References: <[e-mail]>
Distribution: usa
Organization: University of Oregon Chemistry Stores
Lines: 65
NNTP-Posting-Host: miser.uoregon.edu
In-reply-to: [e-mail]'s message of Sun, 13 Dec 92 20:19:10 GMT

In article <[e-mail]>
[e-mail] (Bernard Everett Hoefer) writes:

       I do not understand what my HP48SX is doing when I try
   to solve the following equation.

   X*INV(TAN(X))=-4.8749269433

       I get X:  14.4622844612 as the answer, but that is not the
   answer I want.  For some reason, the calculator starts its guesses
   between +1 and -0, then jumps to +1 and +5, then +11 and +32 and
   finally narrows it down to 14.4622844612.
       Why does it ignore 2.64454252521 (the answer I am looking
   for) while it is between 1 and 5?
       Second question:  when I limit the guessing between
   0 and 6.28 (since I am in radians) the calculator jumps right over the
   guessing between 0 and 6.28, and starts looking between 6.28 and 163,
   then 163 to 293!
       I understand that the calculator looks for a sign reversal,
   but it is skipping over two of them when I limit it from 0 to 6.28.
   Can anybody explain why it jumps around so (seemingly) randomly?

You will find comfort and solace on pages 276-282 of The Fine
Manual (if you have the two-volume manual; if you have the more
recent one-volume edition you will have to look up the topic of
"How the HP Solve Application Works" yourself).

What is happening is that the pairs of guesses that you are
providing have function values that are both of the same sign.
When this happens, the solver will effectively draw a line
between the guesses, find where that line intersects the X-axis,
and replace one of the limits with that.  If you give a pair of
guesses, the search is not necessarily limited to the region
between them.

For example, if you give the guesses 0 and 6.28, then you get the
function values 0 and (approx.) -1971.  It then runs a line
through (0, 0) and (6.28, -1971) to obtain a new guess, but by
then it is far enough afield that it won't get a principal value
solution.

One way to make sure you get an solution in the range you want is
to evaluate the function several times in the range of the
desired solution values.  If you find a pair of initial guesses
that yield function values of different sign then you will be
more likely to have the solver converge on a solution between
those guesses.  In general, one guess, or two guesses very close
to each other, will work better than two widely separated
guesses.

Another good way to make sure that you find the solution you want
is to graph the function, then ask for a root in the range of the
desired solution.  You will also get a more accurate picture of
the function or see whether two guesses really have just one root
between them, one or more, or even none.

Using the EXPR= key in the solver will also help tell you whether
a solver-located sign reversal is likely to be a root (left and
right sides match closely) or a disjunction (left and right sides
are of widely different values).
--
Steve VanDevender     [e-mail]
"Bipedalism--an unrecognized disease affecting over 99% of the population.
Symptoms include lack of traffic sense, slow rate of travel, and the
classic, easily recognized behavior known as walking."