15.01.2013 Views

U. Glaeser

U. Glaeser

U. Glaeser

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

In some literature, the authors like to group P leakage and P static together and call it as the static component<br />

of power.<br />

The switching component of the power occurs when energy is drawn from the power supply to charge<br />

parasitic capacitors made up of gate, diffusion, and interconnect capacitance. For properly designed<br />

circuits, the switching component will contribute more than 90% of the power consumption, making<br />

it the primary target for power reduction [12]. A system level approach, which involves optimizing<br />

algorithms, architectures, logic design, circuit design, and physical design, can be used to minimize<br />

power. The physical capacitance can be minimized through choice of substrate, layout optimization,<br />

device sizing, and choice of logic styles. The choice of supply voltage has the greatest impact on the<br />

power-delay product, which is the amount of energy required to perform a given function. From the<br />

expression for the switching component of power (P switching = C LV 2 f ), it is clear that if the supply voltage<br />

is reduced, the power delay-product will improve quadratically. Unfortunately, a reduction in supply<br />

voltage is associated with a reduction in circuit speed. However, if the goal is to increase the MIPS/Watt<br />

in general purpose computing for a fixed level, then various architectural schemes can be used for<br />

voltage reduction.<br />

The short-circuit power dissipation, P short-circuit, is due to short-circuit current, I sc. Finite rise and fall<br />

time of the input waveforms result in a direct current path between supply voltage V DD and ground, which<br />

exists for a short period of time during switching. Such a path never exists in dynamic circuits, as precharge<br />

and evaluate transistors should never be ON simultaneously, as this would lead to incorrect evaluation.<br />

Short-circuit currents are, therefore, a problem encountered only in static designs. Through proper choices<br />

of transistor sizes, the short-circuit component of power dissipation can be kept to less than 10%.<br />

Leakage power, P leakage, is due to the leakage current, I leakage. Two types of leakage currents seen through<br />

in CMOS VLSI circuits: reverse biased diode leakage current at the transistor drain, and the subthreshold<br />

leakage current through the channel of an “OFF” device. The magnitude of these leakage currents is set<br />

predominantly by the processing technology. The sub-threshold leakage occurs due to carrier diffusion<br />

between the source and the drain when the gate-source voltage, V gs, has exceeded the weak inversion<br />

point, but still below the threshold voltage V t. In this regime, the MOSFET behaves almost like a bipolar<br />

transistor, and the subthreshold current is exponentially dependent on V gs. At present P leakage is a small<br />

percentage of total power dissipation, but as the transistor size becomes smaller and smaller and the<br />

number of transistors that can be integrated into a single silicon die increases, this component of power<br />

dissipation is expected to become more significant.<br />

Static power, P static, is due to constant static current, I static, from V DD to ground when the circuit is not<br />

switching. As we have seen earlier, complementary CMOS combines pull-up and pull-down networks<br />

and only one of them is ON at any given time. Therefore, in true complementary CMOS design, there<br />

is no static power dissipation. There are times when deviations from the CMOS design style are necessary.<br />

For example in special circuits such as ROMs or register files, it may be useful to use pseudo NMOS<br />

logic circuit due to its area efficiency. In such a circuit under certain output conditions there is a constant<br />

static current flow, I static, from V DD to ground, which dissipates power.<br />

The power reduction techniques at the circuit level are quite limited when compared with the other<br />

techniques at higher abstraction levels. At the circuit level, percentage power reduction in the teens is<br />

considered good [11]; however, low-power circuit techniques can have major impact because some<br />

circuits are repeated several times to complete the design. For example, adders are one of the most<br />

often used arithmetic circuits in digital systems. Adders are used to perform subtraction, multiplication,<br />

and division operations. Reducing power consumption in adders will result in reduced power consumption<br />

of many digital systems. Various different types of adders have different speeds, areas, power<br />

dissipations, and configurations available for the VLSI circuit designer. Adders or subsystems consisting<br />

of adder circuits are often in the critical path of microcomputers and digital signal processing circuit;<br />

thus a lot of effort has been spent on optimizing them. As shown in Fig. 2.10, the straight forward<br />

realization of a CMOS full adder will require 28 transistors. This adder is not optimized for power<br />

dissipation. Recently, there has been tremendous research effort in the design and characterization of<br />

low-power adders. The 14-transistor (14T) full adder proposed by Abu Shama et al. [13], dual value<br />

© 2002 by CRC Press LLC

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!