Clock Tree Synthesis (CTS) is the automated physical design process of building a balanced, optimized clock distribution network that delivers the clock signal from its source to every sequential element (flip-flop, register, latch) in the design — with minimal skew, controlled insertion delay, acceptable transition times, and low power consumption.
Why CTS Is Critical
- A modern SoC can have millions of flip-flops — all needing a clean, well-timed clock.
- The clock is the highest switching-activity net on the chip — it toggles every cycle at every flip-flop, so it dominates dynamic power.
- Clock quality (skew, jitter, transition time) directly determines the maximum operating frequency and timing margin of the design.
CTS Objectives
- Skew Minimization: All flip-flops should see the clock edge at approximately the same time. Target skew depends on the clock period — typically <5% of the period.
- Insertion Delay Control: Total delay from clock source to leaf flip-flops should be reasonable and consistent.
- Transition Time: Clock edges should be sharp (fast rise/fall) — slow edges increase short-circuit power and degrade timing margins.
- Power Optimization: Minimize the number and size of clock buffers — clock tree power can be 30–50% of total dynamic power.
- DRV Fixing: Ensure all clock nets meet design rule constraints (max capacitance, max transition, max fanout).
CTS Methodology
- Clustering: Group nearby flip-flops into clusters that share a common clock buffer.
- Buffer/Inverter Insertion: Insert a tree of buffers (or inverters for balanced rise/fall) to drive the clock from the source to all clusters.
- Balancing: Adjust buffer sizes, wire lengths, and topology to equalize delay to all sinks.
- NDR (Non-Default Rules): Route clock wires with wider width and spacing for better signal quality and reduced coupling.
- Shielding: Add grounded guard wires adjacent to clock routes for noise isolation.
- Multi-Source CTS: For large designs, use multiple clock roots (from a clock mesh or multiple PLLs) to reduce tree depth.
Clock Tree Topologies
- Balanced Tree (H-Tree): Symmetric branching where each branch has equal length — inherently low skew.
- Mesh: A grid of interconnected clock wires — low skew through averaging, but higher power.
- Spine: A central spine with branches — used for structured layouts.
- Hybrid: Combination of tree and mesh — mesh at the top level for global balance, trees at the local level for efficiency.
CTS in the Design Flow
- CTS runs after placement and before or during routing — flip-flop locations must be known.
- Pre-CTS Timing: Timing is estimated with ideal (zero-skew) clocks.
- Post-CTS Timing: Real clock tree delays and skew are included — timing may change significantly.
- Post-CTS Optimization: Additional optimization (gate sizing, buffer insertion, useful skew) to fix timing violations introduced by real clock delays.
Clock tree synthesis is arguably the most impactful single step in physical design — the quality of the clock tree directly determines chip frequency, power, and timing closure difficulty.