Access Official Technical PDF

Share Summary:LinkedInEmail

Power gating techniques can reduce leakage in CMOS circuits, but their impact on timing, area and circuit behavior varies with the logic implementation. For industrial automation engineers evaluating low-power digital hardware, a comparison of 4-bit serial-in, serial-out (SISO) shift registers illustrates an important design principle: no single leakage-control method is best across every circuit architecture and performance objective.

The reported simulations compare several power-gating approaches alongside multiplexer- and gate-diffusion-input-based (GDI) register designs. The results show meaningful differences in average power and propagation delay, making it important to assess both measures rather than selecting a technique by its power result alone.

Why leakage control matters in CMOS circuits

CMOS power consumption includes dynamic power associated with switching and leakage current that persists when transistors are not actively changing state. Leakage includes subthreshold current and reverse-biased junction leakage. As transistor dimensions scale, controlling this unwanted current becomes an important part of low-power VLSI design.

Power gating addresses leakage by changing how a logic block connects to its supply during inactive periods. Sleep transistors, commonly arranged as header or footer devices, can isolate a block from the power rail. Other methods combine supply control with transistor stacking or alter the arrangement of control devices. These choices affect not only leakage, but also wake-up behavior, circuit area and the delay of the operating circuit.

Power gating techniques: how the approaches differ

Sleep and stack-based methods

The sleep approach adds header and footer control transistors to a logic block. Their control signals determine whether the block operates normally or enters a leakage-reduction state. The approach is straightforward, but its effectiveness and operating impact depend on the implementation.

Stack-based methods use series transistors to reduce subthreshold current. The sleepy stack approach combines stacked devices with sleep control, seeking to use the leakage benefits of stacking while retaining an active operating mode. This can require more transistors and area than a basic sleep implementation.

Dual-sleep uses additional leakage-control devices in the header and footer paths. The extra control structure can increase circuit area, so its potential power benefit needs to be considered alongside implementation cost and delay.

Zig-zag control

The zig-zag approach alternates true and complementary control signals between header and footer devices in successive blocks. This arrangement is intended to reduce wake-up time while controlling leakage. Its results, like those of other methods, depend on the circuit and the way the gating scheme is applied.

Why the SISO register implementation matters

A 4-bit SISO shift register is built by cascading four D flip-flops, with each stage passing stored data to the next. The study compares register implementations based on multiplexers, GDI and self-resetting GDI. These circuit structures do not behave identically, so a power-gating method’s result on one implementation should not automatically be assumed for another.

In a multiplexer-based D flip-flop, two 2:1 multiplexers form the storage element, with complementary clock signals controlling the two stages. GDI is an alternative logic implementation intended to reduce power in digital circuits. The self-resetting GDI design adds resetting and level-restoration logic, which can change both power consumption and timing characteristics.

What the simulation results show

The circuits were simulated in Cadence Virtuoso using 180 nm technology and a 1.8 V supply. The SISO-register tests used a maximum frequency of 25 MHz, and average power and propagation delay were measured through transient analysis. The reported values are specific to these simulation conditions and circuit implementations.

  • Multiplexer-based SISO: The zig-zag configuration reported 233.0 µW, compared with 253.0 µW for sleep, 279.0 µW for sleepy stack and 236.5 µW for dual sleep. Its reported delay was 4.967 ns; the sleepy-stack and dual-sleep results were 4.88 ns and 4.92 ns, respectively. The sleep result had a substantially longer reported delay of 79.45 ns.
  • GDI-based SISO: Zig-zag reported the lowest power among the listed configurations at 459.4 µW. Sleep, sleepy stack and dual sleep reported 555.9 µW, 652.6 µW and 653.3 µW, respectively. Delay results also varied considerably, so the power ranking alone does not establish the best timing choice.
  • Self-resetting GDI SISO: Sleep reported the lowest power in this group at 6.235 mW, followed by zig-zag at 6.651 mW, sleepy stack at 6.841 mW and dual sleep at 7.268 mW. The reported delays differed substantially across methods.

These comparisons show why the preferred option can change with the register design. Zig-zag has the lowest reported power in the GDI-based group, while sleep is lowest for the self-resetting GDI implementation. In the multiplexer-based case, zig-zag also reports the lowest power, but the sleep configuration’s much longer delay highlights a potential trade-off.

Applying the findings to design decisions

For engineers assessing low-power sequential logic, the practical takeaway is to evaluate power gating at the level of the actual circuit, not as an isolated technique. A useful comparison should include:

  • Average power under representative input and clock conditions.
  • Propagation delay across relevant switching transitions.
  • Area and implementation overhead associated with added control transistors.
  • Behavior when the circuit enters and exits its low-power state.

The study reports reductions in power for gated designs relative to conventional implementations, but its results are tied to the simulated 4-bit register structures and stated operating conditions. They should therefore guide design evaluation rather than be treated as universal performance guarantees. Verification in the target process, at the intended supply and timing conditions, remains essential.

Conclusion

Power gating techniques offer several ways to manage leakage in CMOS sequential circuits, from basic sleep control to stack-based and zig-zag arrangements. The 4-bit SISO comparisons demonstrate that the lowest-power option depends on whether the register uses multiplexer, GDI or self-resetting GDI logic, and that power savings must be weighed against delay and implementation overhead. For automation hardware designers, the practical approach is to benchmark candidate methods on the intended circuit and operating conditions before selecting a power-management strategy.

Access Official Technical PDF

About The Author