The Controller Area Network (CAN) is a serial bus with high speed, high reliability, and low cost for distributed real time control applications. The CAN is a desirable, cheap solution for networks in industrial environments, but there is a limit on the maximum length of a single CAN bus. A solution is to divide the CAN network into segments and connect them using bridges. Bridges are high performance devices that are used to interconnect LANs at the logical link control (LLC) or medium access control (MAC) level in the protocol hierarchy. Unlike many serial communication protocols the CAN message contains no information relating to the destination and source addresses. Instead the message contains an identifier which indicates the type of information contained in the message. Because of this feature of the CAN message it is not possible to use traditional address based bridges to connect CAN segments. The aim of this study is not only to design and implement a bridge to connect CAN segments based on the CAN protocol features but also to do a performance analysis of the designed bridge. The designed bridge is a transparent bridge, and uses a learning process to build-up a forwarding database.