here we have a rather nice 10 digit, mixed segment and dot
matrix display, scientific calculator from sharp, 1993. somewhat more modern
than the older, but nice, el512 styling and
before the later, and more tacky plastic dal style.
as a scientific model, its pretty good with all the usual stuff you
would expect including hyperbolics, fractions, number bases, 2d stats with
regression, numerical integration and blind formula programmable.
the operational style is an interesting mix somewhere between algebraic
entry and formula entry. essentially, you calculate using a formula style,
but operations can be performed directly on a result (like with aos) by
entering an additional = to force evaluation. entering a formula
based expression is a bit odd because the machine can only display the
last part of the entry in the small dot matrix portion on the left of the
display eg. press sin cos tan 1 = to compute sin(cos(tan(1))).
the right arrow key can be pressed to delete digits of number entry,
but do not delete operations of a pending formula when not programming.
12 digits are used internally and, when not in number bases, the
letters AF and M (the main memory) can be used to store
values and indeed can be used as variables in the formula aer mode.
algebraic expression reserve (aer) mode
this is, in fact, the programmable mode of the machine. there are two
areas 1: and 2: that can store a total of 40 steps. the
formulae can involve variables AF and M (these are the only
alpha characters available, despite an alpha key). statements can
be appended with comma. for example, A+1=>A,3A^2+7A+9 is valid.
a limited set of branch conditions are allowed (now quite normal for blind
program formulae). these are; x=m,x!=m,x<m, x>m, x<=m, x>=m
the condition always causes a branch to the start of the
program. program editing is possible with the arrow keys, del
and ins.
numerical integration
like a lot of modern formula programmables, here is numerical
integration implemented by unmodified simpson's rule. as usual, the
integration limits are evaluated and must exist. furthermore, the number
of subdivisions must be supplied, directly and not as a power of 2. there
is no error tracking and no indication of how much of the numerical result
is correct. nevertheless, this is useful albeit a bit slow (see torture
test).

