A new expression of a non-conservative robust stability condition is derived for sampled-data systems taking into consideration its convenience in numerical computation. Our expression is described using a global maximum of a certain differentiable function, while the conventional expression includes a global maximum of a structured singular value, which is expensive to compute. When the sampling period is long, the advantages of our expression become clear. A possible problem is that the maximized function may be multimodal. However, the conventional expression has the same problem and even a local maximum is shown to give useful information. Our expression is applicable to a fairly large class of systems including the multi-input multi-output additive and multiplicative perturbation models. Using an example, we show how our expression is computed and make a comparison with the conventional expression.