You are designing an SoC where a 1-bit active-high control pulse originates in a fast DSP clock domain running at 500 MHz (2ns period) and needs to be captured by a slow peripheral domain running at 100 MHz (10ns period). If the pulse in the fast domain is exactly one clock cycle wide (2ns), what is the primary risk of using a standard 2-flop synchronizer in the slow domain, and how should it be mitigated?
A 2ns pulse is significantly shorter than the 10ns clock period of the receiving domain. Depending on the phase relationship between the two asynchronous clocks, the slow clock may never sample the signal while it is high, resulting in complete data loss. A 2-flop synchronizer only resolves metastability; it does not guarantee capture of a signal that is shorter than the destination clock period. The standard solution is to extend the pulse in the transmit domain (often using a toggle flip-flop or an explicit request-acknowledge handshake) so that it remains active long enough to be sampled by the slow clock.