48/50 Linear System Solver

[Dieter]

(02-13-2014 01:30 AM)Tim Wessman Wrote:  Have someone with a complaint about the linear system solver on the 48 series (50g specifically).

The system is exactly determined (three variables in three equations), but its determinant is zero. Which is a quite challenging case for iterative numerical algorithms like a LU-decomposition: due to accumulated roundoff errors the result may be close to, but not exactly zero. Which makes the calculator think a valid solution can be calculated, and so it continues.. with a meaningless result.

I do not have a 48-series calculator here. What does it return if you try to evaluate the determinant? Something very close to (but not equal to) zero?

BTW: that's why I like simple direct methods like Cramer's rule for small linear systems with 2, 3 or maybe even 4 variables. ;-)

Dieter