I was a grader for an electromagnetics class once. The professor would assign a simulation project that typically would involve numericaly finding (for example) the charge distribution on plates of a capacitor or the impedance and speed of propagation in a transmission line. This was useful because most available estimates of these quantities used simplifying assumptions (for example, infinite plate width on a capacitor) that simplified analytical expressions but added inaccuracies. The students would compare their numeric calculations to what they'd expect from these standard expressions.

One year, the assignment was to find the characteristic impedance on a particular transmission line. Picture a rectangular tube with a small rectangular tube inside it (somewhat like the cross-section of a rectangular coaxial cable). To find the per-unit-length capacitance of the transmission line, it was necessary to find the voltage gradient right at the skin of the two conductors. That is, if you put 1V of potential on the center conductor and 0V on the outside conductor, there would be a voltage profile between the two that would look like a downward sloping hill. If you could describe how that slope looks very close to the conductors, you could do some useful things. So the simulation would take a cross section of this transmission line and cut it up into lots of discrete little squares. Electromagnetics gave expressions relating each square to the squares around it, and this set up the difference equations that made the simulation go. Boundary conditions were set on the two conductors (put 1V on one conductor and 0V on the others), the potential ("voltage") on each square was set to 0V, and then new voltage values could be generated iteratively (via the electromagnetics difference equations) which gradually raised each 0V cell to its proper level.

Most people would use MATLAB for this simulation. However, because it was a class for younger students (sophomores primarily) many students were not familiar with MATLAB at the time. Because of that, one student used Excel to do the processing. Each cell held the formula for the difference equations expressing the relationship between each discrete voltage unit and the adjacent discrete voltage unit. As he ran his simulation, Excel would update each cell continuously until there was very little change from the previous value to the next value. The simulation would stop at this point.

Now, what was neat about solving the simulation this way is that the spreadsheet produced a really interesting graphic display of the voltage profile as it evolved. It was possible to do this in MATLAB after all the processing was done, but it would typically be done on a graph. In this case, you actually watched the numbers move in space. That is, you could watch the voltage "tent" raise by watching the numbers toward the center conductor get closer to 1 and the numbers farther away get closer to some value in between 0 and 1.

The professor said that this was the first time anyone had ever used Excel (instead of MATLAB, for example) to do the simulation.