TABLE OF CONTENTS This section is about how to determine whether a MIMO model is good. Good models have been thoroughly validated to reflect precisely the aspect of a MIMO one is interested in. Thus, before adopting a MIMO model, one has to have a clear idea about:
Which aspect of MIMO shall be modelled? Multiplexing, beamforming or diversity gain?
At which level should the model function? Propagation, channel, link, system?
Has the model been validated?
If yes, by which procedure? Theoretically, or by a single measurement campaign?
Does the metric chosen for validation correlate with the desired MIMO system performance?
Model validation is so important because how can one otherwise be sure that one is modelling what one wants to model? The result of careful model validation should be twofold:
Where are the limitations of the model used?
Are some models better suited to predict a certain aspect of MIMO system performance than others?
Surprisingly, much more effort has gone into the development of new models than into their validation.
One goal of this section is the comparison of measures of quality, in short metrics: which is the most suitable to predict a certain specified MIMO properties best? Single-number metrics are desirable for model validation because they are easy to compare. It should be kept in mind, though, that any model is a reduction of reality, and single-number metrics are even more so. Obviously, the more complex a metric is, the more information...
