Bus

A point-to-point link joins two pins. A bus is a set of nets that more than two parts connect to, so every part on it sees every transition and the parts take turns talking. That sharing is the point of a bus. Because of it, a standard has a lot to say about a bus. The parts have to agree on levels, speed, addressing, and who is allowed to drive when.

The word does two jobs on this site, and each is checked somewhere different.

On a schematic it is drawing shorthand. DATA[7:0] is one line standing for eight signals. Downstream, every member has to come out of the read as its own net. A diff, a highlight, or any rule over those signals is unreliable until it does. bus-not-modeled is that check, and it reports on the reader rather than on the design.

As an interface it is a standard. CAN, I2C and SPI each name their signals and demand things of them. One wants a termination resistor at each end of the run. An open-drain bus wants a pull-up on every line. And a transceiver sits between the processor’s logic pins and whatever voltages the standard actually uses. A two-wire bus is usually a differential pair, for reasons of noise rather than of protocol.

None of that is visible in a netlist as a requirement. A board with CAN wired wrongly is a perfectly valid netlist, so an interface profile declares the shape and the engine compiles rules from the declaration. profile-signal-missing reports a role that no net carries, and profile-signal-dangling reports one wired at a single end. profile-termination checks the resistor across the pair. I2C is common enough to be checked without a profile at all, by i2c-pull-up.

The course teaches it in chapter 10, which argues that a bus is a contract, and in chapter 1, which reads a terminator off the two nets it bridges.