The trade-off between flatness, quantization error, and robustness on the one side and penumbra width on the other side can be described analytically for equally weighted spots. Treatment planning systems often perform a least-squares optimization of the individual spot weights which results in smaller lateral penumbras and smaller quantization errors than for uniform discrete scanning. However, the benefit of this weight optimization decreases with increasing lambda (in the regime lambda > sigma(b)). The spot spacing, which is obtained from the scenario that the optimization objective is met by uniform discrete scanning, poses a sharp upper limit for the spot spacing lambda in weight optimization methods.