The "Error Rate Calculation" block does not support variable-sized signals, as confirmed by this documentation link:
One workaround you could try: Find the maximum output size for your "Rx" block. Then, if the output of Rx is less than this maximum size, add zeros to the output signal until it is the same size as the maximum size.
For example, say the output of the Rx block is [1 1 1] for one time step, and then [1 1 1 1 1] for another time step. The maximum size of this signal is 1x5, so the first time step's output should be "padded" with zeros, so that it becomes [1 1 1 0 0]. You could do this in a MATLAB function block.
