One cannot directly connect MCANTX of node-A to MCANRX of node-B and vice versa and expect successful CAN communication. In this case, CAN is unlike other serial interfaces like SCI or SPI. For example, SCI can be made to work with a RS232 transceiver or through a direct connection (SCITX of one node to SCIRX of another node and vice versa). However, CAN bus needs a CAN transceiver for the following reason: In addition to converting the single-ended CAN signal for differential transmission, the transceiver also loops back the CANTX pin to the CANRX pin of a node. This is because a CAN node needs to be able to monitor its own transmission. Why?

• This has to do with the ACK requirement mandated by the CAN protocol. When a node transmits a frame on the CAN bus, it expects an ACK from at least one other node on the network. For the ACK phase, the transmitter puts out a 1 and expects to read back a 0.
• During arbitration, a node with a higher-priority MSGID needs to be able to override a 1 with a 0. Here again, the transmitter needs to be able to read back the transmitted data. When a node puts out a 1 and reads back a 0 during the arbitration phase, it loses arbitration.