Real gates are built from transistors, and transistors take time to switch. So when an input changes, the output doesn't follow instantly. The propagation delay tpd is the worst case: the longest you could ever wait before the output is guaranteed to show its new, correct value.
Think of it as a promise from the gate's designer: "after tpd, you can trust me." Before that, the output might still be old, or halfway through changing.
Why it matters:
- It sets how fast a circuit can run. A clock can't tick again until every output it depends on has settled.
- Delays add along a chain of gates, so a long chain is slow. The slowest chain is the critical path.
- It has a partner, the contamination delay
tcd, which is the shortest time before the output can start to move. Alwaystcd≤tpd.
Gate delays are tiny, so they're quoted in nanoseconds (1 ns = 10⁻⁹ s) or picoseconds (1 ps = 10⁻¹² s). 1 ns = 1000 ps. In this course wires count as zero delay, and each gate type is given its own tpd. For a flip-flop, the equivalent number is the clock to q delay.