casio fx-5500l
| similar in concept to the fx-4000p,
not programmable but with some new dedicated features added. powered by two cr2032 plus a third for memory backup. a rather dull grey styling with a grey plastic wallet style cover. that, although pocket sized, is somewhat more chunky than the modes like the sharp el-545h, for example. the two line dot matrix display is adequate but not especially clear. i preferred the designs with one dot matrix and one segmented line display, which are clearer (like the ti30xiib) despite boldly showing "16x2" on the front, the unit, in fact, only operates to ten significant figures (12 for internal expressions). so that 1/3exe*3 exe gives all 9's. input accepts implied multiplication, standard features include hyperbolics, fractions (but not decimal rationalisation), number bases and 2d stats (no distribution support) and linear regression. the alpha registers a-z are memories. formulae can be stored in the machine. these are expressions involving the alpha registers. on call, they can be evaluated but, unfortunately, they do not operate as a solver (like on the hp17bii) whereby any variable could be solved given the others, making the feature rather weak. this must be the scientific library because there are no built in formulae like the fx-5000f. there are a number of add-on features:
i was interested to see how these extra features weighed up. on most calculators the add-ons are rather lame and somewhat of a disappointment. |
 |
matrices
matrix support is semi-lame, but of some use. you can have mxn matrices in registers A or B of up to 5x5. you can add, subtract, multiply, invert, transpose, scalar product and calculate the determinant. inversion is by gauss elimination with no refinements.
numerical integration
as usual, casio have implemented simpsons rule with no improvements. the unit can choose `n' (the number of steps) or you can supply it. although when chosen, the unit seems only capable of displaying a 5 figure result. results are average (see torture test). results were quite fast, although nowhere near as good as the el-9900.
complex numbers
these were implemented a bit better than i'd expected. several calculators boldly claim to support complex numbers when, in fact, all you can do is basic arithmetic in a very cack-handed way (eg sharp el-545h, hp-20s).
you get to type in generic expressions involving `i' combined with sqrt, 1/x, x^2 and brackets eg. (i+sqrt(3+2i))^-1. for output, the two lines are used to great effect with each able to display the full ten digit answer. one drawback is that the a-z registers are unavailable in complex mode. sadly none of the scientific functions work with complex numbers even x^y and cube root are out.
one of the problems inherent in complex numbers is that you need internally twice the precision for operations as simple as multiplication. for example (1.00001+i)*(1.00001+i) = 2.00001e-5 + i*2.00002, is correctly given by the machine and is just within limits because the 0.00001 is half the precision. going one stage further; (1.000001+i)*(1.000001+i) gives 2e-6+i*2.000002 which is only approximate. the correct 10 digit answer being 2.000001e-6+i*2.000002.
equation features
as well as the ability to store, recall and evaluate formulae (but not solve), there is a special equation mode hardcoded for two other features; solving quadratics and simultaneous linear equations.
the quadratic solver is a simple implementation of the textbook formula with no refinements, although nicely, complex roots are also returned. results were poorer than the el-9900 (see the quadratics article).
the simultaneous equations solver is the same as the matrix support, but saves the effort of entering the coefficient matrix, inverting it and multiplying by the values by doing it all from your (m+1)*m input matrix of coefficients and constants. the sharp el-545h has a similar feature hardcoded for 3*3. naturally this feature uses the matrices memories.