Clock Routing, Root, and Distribution - 2022.1 English

UltraFast Design Methodology Guide for Xilinx FPGAs and SoCs (UG949)

Document ID
UG949
Release Date
2022-06-08
Version
2022.1 English

To properly understand the clocking capacity of an UltraScale device and the clocking utilization of a design, it is important to know how the clock routes use the dedicated routing resources:

  • From the clock buffer to the clock root, the clock signal goes through one or several segments of vertical and horizontal routing. Each segment must use the same track ID (between 0 and 23).
  • At the clock root, the clock signal transitions from the routing track to the distribution track with the same track ID. To reduce skew, the clock root is usually in the clock region located in the center of the clock window. The clock window is the rectangular area that includes all the clock regions where the clock net loads are placed. For skew optimization reasons, the Vivado IDE might move the clock root to off center.
  • From the clock root to the CLB columns where the loads are located, the clock signal travels on the vertical distribution (both up and down the device as needed) and then onto the horizontal distribution (both to the left and right as needed).
  • The CLB columns are split into two halves, which are located above and below the horizontal distribution resources. Each half of the CLB column contains several leaf clock routing resources that can be reached by any of the horizontal distribution tracks.

In some cases, a clock buffer can directly drive onto the clock distribution track. This usually happens when the clock root is located in the same clock region as the clock buffer or when the clock buffer only drives non-clock pins (for example, high fanout nets).

Because clock routing resources are segmented, only the routing and distribution segments used to traverse a clock region or to reach a load in a clock region are consumed.

The following figure shows how a clock buffer located in clock region X2Y1 reaches its loads placed inside the clock window, which is formed by a rectangle of clock regions from X1Y3 to X5Y5.

Figure 1. UltraScale Device Clock Routing from Driver to Loads

In the following figure, a routed device view shows an example of a global clock that spans most of the device. The clock buffer driving the network is marked in blue in clock region X2Y0 and drives onto the horizontal routing in that clock region. The net then transitions from the horizontal routing onto the vertical routing in clock region X2Y0 reaching the clock root in clock region X2Y5. All clock routing is marked in blue. The clock root is marked in red in the clock region X2Y5. From the clock root in X2Y5, the net transitions onto the vertical distribution and then the horizontal distribution to the clock leaf pins. The distribution layer and the leaf clock routing resources in the CLB columns are marked in red.

Figure 2. Routed Device View of a Routed Clock Network