Clock domain crossing

From Wikipedia, the free encyclopedia
Jump to: navigation, search

In digital electronic design a clock domain crossing (CDC), or simply clock crossing, is the traversal of a signal in a synchronous digital circuit from one clock domain into another. If a signal does not assert long enough and is not registered, it may appear asynchronous on the incoming clock boundary.[1]

A synchronous system is composed of a single electronic oscillator that generates a clock signal, and its clock domain—the memory elements directly clocked by that signal from that oscillator, and the combinational logic attached to the outputs of those memory elements.

Because of speed-of-light delays, timing skew, etc., the size of a clock domain in such a synchronous system is inversely proportional to the frequency of the clock. In early computers, typically all the digital logic ran in a single clock domain. Because it is difficult to carry digital signals above 66 MHz on standard PCB traces (and the clock signal is the highest frequency in a synchronous digital system), CPUs that run faster than that speed invariably are single-chip CPUs with a PLL or other on-chip oscillator, so all the really high-frequency signals stay inside the chip and are never carried by the PCB traces. At first each CPU chip ran in its own single clock domain, and the rest of the digital logic of the computer ran in another slower clock domain. A few modern CPUs have such a high speed clock that designers are forced to create several different clock domains on a single CPU chip.

Different clock domains have clocks which have a different frequency, a different phase (due to either differing clock latency or a different clock source), or both. Either way the relationship between the clock edges in the two domains can not be relied upon.

Synchronizing a single bit signal to a clock domain with a higher frequency can be accomplished by registering the signal through a flip-flop that is clocked by the source domain, thus holding the signal long enough to be detected by the higher frequency clocked destination domain.

To avoid issues with meta-stability in the destination domain, a minimum of 2 stages of re-synchronization flip-flops are included in the destination domain.

Synchronizing a single bit signal traversing into clock domain with a slower frequency is more cumbersome. This typically requires a register in each clock domain with a form of feedback from the destination domain to the source domain, indicating that the signal was detected.[2]

In some cases, clock gating can result in two clock domains where the "slower" domain changes from one second to the next.

See also[edit]

References[edit]

  1. ^ Parker, Roy H., Caution: Clock Crossing A prescription for uncontaminated data across clock domains, Chip Design Magazine, Issue 5, Article 32, July 2004.
  2. ^ Stein, Mike, Crossing the abyss: asynchronous signals in a synchronous world, Paradigm Works, EDN Magazine, 24 July 2003.

External links[edit]