?
Good paper.
As I read the linked article I was questioning how the many potentially unknown parameters were handled. One is phasing which was cleverly addressed by setting the measured tower phasing according to the model. Good approach but it requires measurement and adjustment and not simply crunching the numbers.
Modeling is useful in predicting performance and trying "what if" experiments. I do this often for circuit designs with good effect but the user must know the limitations.
I recently posted measured data taken from a short wire antenna and the reason this was not just modeled is a very common situation which arises in many engineering problems. Modeling depends on accurately knowing the system parameters often which must be measured. In my reported antenna experiment the actual antenna Z was needed. Numbers could have been assumed from published works or from NEC simulations but the results would be based on assumed values which probably would have been wrong for this specific antenna giving wrong results.
An example I recall from my college days was an analysis of a gun turret servo control system which required knowing the moment of inertia of rotating parts, gear slop, and the friction in the bearings and gears. The professor, an ex Navy engineer, explained how these had to be measured for each particular turret in order to perform the servo system analysis. There was no one size fits all set of such data.
A simpler example is the gain of a FET common source amplifier is = gmRd which is very easy to calculate but the gm varies greatly from one device to another so the model requires measuring gm for each device. Might just as well build it and measure the gain or use some Monte Carlo analysis. A counter example is that the gain of a BJT amplifier with an emitter resistor is = Rc/Re. Since both resistances are known to within a few percent the modeling gives very accurate results applicable to large numbers of such amplifiers. Modeling of filters where the Rs, Ls, and Cs are accurately known also gives accurate results.
Modeling is very useful but often it has to be used in conjunction with measurement and experimentation. The two techniques are synergistic but modeling is not a stand alone technique.
Neil
Just to note that prior to the development and FCC acceptance of "Method of Moments" software such as NEC, the pattern and gain characteristics of antenna systems including those of directional MW broadcast antenna arrays were calculated by the considerable and manual effort of consulting engineers, based on Maxwell's equations (Google for details about Maxwell's equations).
NEC software removed a lot of the manual effort in this, but does require its proper and educated use.
A valid NEC model of a given antenna system and its propagation environment closely predicts the real-world, accurately-measured performance of that system after its physical installation.
Rich's remarks about how things were done before modern software solutions rings true. Most electronics hobbyists know how to calculate voltages and currents for simple networks by combining resistors in series or parallel and working with the simplified networks with fewer components in the model. What many may not know is that there are two techniques, mesh and nodal analysis, which work for solving all passive network problems and that the simplification techniques commonly used are shortcuts in order to avoid the tedious arithmetic of mesh and nodal analysis.
For example, a network with five nodes will have a set of five simultaneous equations with five unknowns. Writing the equations is easy...solving them is tedious. Modern calculators and computers can solve these virtually instantaneously but the results are useless unless the equations are properly formed. Thus Rich's comment "... does require its proper and educated use..." is certainly true.
Neil