A power switch cell (also called a header switch or footer switch) is a specialized standard cell containing a large power-gating transistor that connects or disconnects a power domain from its supply rail — enabling entire blocks of logic to be completely powered down during idle periods to eliminate leakage power.
Why Power Switching?
- At advanced nodes, leakage power can be 30–50% of total power — transistors leak current even when not switching.
- Clock gating saves dynamic power but does nothing for leakage — the transistors remain powered and leaking.
- Power gating (shutting off the supply voltage) is the only way to reduce leakage to near zero.
- Power switches are the physical mechanism that implements power gating.
How Power Switches Work
- Header Switch: A large PMOS transistor between VDD and the local power rail (virtual VDD, or VVDD). When the switch is on, VVDD ≈ VDD. When off, VVDD floats to ground — all logic in the domain loses power.
- Footer Switch: A large NMOS transistor between the local ground (virtual VSS, or VVSS) and VSS. When off, VVSS floats toward VDD.
- Header switches are more common in modern designs — PMOS switches between VDD and virtual VDD.
Power Switch Cell Design
- Large Transistor: The switch transistor must be large enough to carry the entire block's current with minimal voltage drop ($IR$ drop across the switch).
- Low Ron: The switch's on-resistance must be small — typically <50–100 mΩ to keep the voltage drop under 20–50 mV.
- Cell Array: A single switch cell is not large enough for a whole block. Many switch cells are placed in a row/column forming a switch array — all controlled by the same enable signal.
- Daisy Chain Control: Switch cells may be turned on sequentially (daisy chain) rather than simultaneously to limit inrush current during power-up.
Power-Up Sequence
1. Sleep State: Switch off — VVDD = 0V, all logic in dormant state, leakage near zero. 2. Power-Up Signal: Enable signal activates switch cells (may be staggered via daisy chain). 3. Ramp-Up: VVDD ramps from 0 to VDD — supply stabilizes. 4. Isolation Release: Isolation cells release, allowing signals from the powered domain to drive outputs. 5. State Restore: Retention flip-flops restore their saved state. 6. Normal Operation: Block resumes full function.
Design Considerations
- IR Drop: The switch adds resistance in the supply path — must be sized to meet IR drop budget under worst-case current draw.
- Inrush Current: When switching on, the block's decoupling capacitance charges rapidly — creating a current spike. Staggered turn-on mitigates this.
- Always-On Logic: Some cells (retention FFs, isolation cells, control logic) must remain powered — connected to the real (not virtual) VDD.
- Physical Planning: Switch cells must be distributed across the power domain — typically in a ring or grid pattern for uniform IR drop.
Power switch cells are the enabling technology for power gating — they transform leakage power from an unavoidable cost into an engineering choice, providing near-zero leakage for any block that can tolerate being powered down.
Related Topics
Explore 500+ Semiconductor & AI Topics
From EUV lithography to CUDA optimization — search the full knowledge base or chat with our AI assistant.