A remark on the convolution with the box spline