This paper deals with the output synchronization problem for linear heterogeneous multi-agent systems. It is shown that in cycle-free communication networks, synchronization can be ensured by static state feedback laws under mild assumptions. Moreover, we present an observer-based strategy for the case that the agents have only access to relative output differences. While in general additional controller states must be communicated to synchronize the agents by observer-based methods, we prove that in cycle-free networks, only output differences are needed to achieve an observer-based synchronization.