o Conventional CTS (Buffer/Inverter Tree): The most common approach. The tool starts from the sinks and works backward or starts from the root and works forward, clustering nearby sinks, inserting buffers/inverters to meet skew, latency, and DRC targets, and progressively building a tree structure. The exact topology isn't strictly predefined but emerges based on sink locations and optimization goals. Modern tools use sophisticated algorithms (e.g., clock concurrent optimization - CCOpt) that optimize the clock tree and logic paths concurrently.
o H-Tree:
o Pros:
Theoretically provides perfect zero skew if sinks are perfectly distributed and
loads are balanced. Good for regular structures.
o Cons:
Impractical for real designs where sinks are unevenly distributed, blockages
exist, and loads vary. Consumes significant routing resources and can have high
insertion delay. Pure H-trees are rarely used; however, H-tree topology might
be used for the top levels of the clock tree driving major branches.
o X-Tree:
Similar to H-Tree but uses diagonal connections. Less practical due to
non-Manhattan routing challenges.
o Mesh:
A grid of interconnected clock lines driven by multiple drivers, typically used
for very high-performance designs. Sinks tap off the nearest mesh segment.
Provides very low skew and OCV robustness but consumes high power and routing
resources.
o Multi-Source
CTS (MSCTS): Used when multiple clock sources drive different parts of the
tree; requires careful balancing between sources.
o Types
of Cells Used:
o Clock
Buffers/Inverters: Libraries often contain cells specifically designed for
clock trees ("clock buffers/inverters"). These typically have:
§ Balanced
Rise/Fall Delays: Critical for minimizing duty cycle distortion.
§ Higher
Drive Strength: To drive large fanouts common in clock trees.
§ Lower
Delay Variation: More robust against PVT variations.
o VT
Mix (Threshold Voltage): Clock tree cells are usually chosen from lower
Vt variants (LVT).
§ Reason:
Faster low-Vt cells help minimize insertion delay. They have higher leakage but
the performance benefit on the critical clock network often outweighs the
leakage penalty.
§ Another
reason is it has less delay variation and less OCV impact.
o H-Tree
Pros/Cons: https://vorasaumil.wixsite.com/pdinsight/post/clock-tree-synthesis-part-3-clock-structures-its-implementation-and-analysing-the-results
o Buffers
vs Inverters: http://vlsi-soc.blogspot.com/2014/12/inverter-vs-buffer-based-clock-tree.html
0 comments:
Post a Comment