The problem of instrumentation network design and upgrade consists of determining the optimal set of instruments to obtain sufficiently accurate and reliable estimates of variables. An existing formulation of the cost optimal sensor network design and upgrade problem (Bagajewicz, 1997) is in the form of a mathematical programming problem, but it has the disadvantage that the constraints are not stated explicitly in analytical form, forcing therefore a special tree search solution procedure. An alternative MILP formulation is presented, which is useful for small to medium size problems, where all constraints are explicit.