CMOS:Dynamic CMOS.

Dynamic CMOS

Dynamic CMOS [4] is a CMOS logic gate that works in a manner very different from the CMOS logic gates discussed so far, which are usually called static CMOS. Using clock pulse, a parasitic capacitance is precharged, and then it is evaluated whether the parasitic capacitance is discharged or not, depending on the values of input variables. Dynamic CMOS has been used often for high speed because parasitic capacitance can be made small due to unique connection configuration of MOSFETs inside a logic gate, although good layout is not easy. Power consumption is not necessarily small. (In static CMOS, a current flows from the power supply to the ground during transition period. But in dynamic CMOS, there is no such current. But this does not mean that dynamic CMOS consumes less power because when input variables do not change their values, static CMOS does not consume power at all, whereas dynamic CMOS may consume power by repeating precharging. Dynamic CMOS and static CMOS have completely different power consumption mechanisms.)

Domino CMOS

Domino CMOS, illustrated in Figure 39.9, consists of pairs of a CMOS logic gate and an inverter CMOS logic gate [4]. The first CMOS logic gate in each pair (such as logic gates 1, 3, and 5) has the pMOS subcircuit, consisting of a single pMOSFET with clock, and the nMOS subcircuit, consisting of many nMOSFETs with logic inputs and a single nMOSFET with a clock. The first CMOS logic gate is followed by an inverter CMOS logic gate (such as logic gates 2, 4, and 6). When a clock pulse is absent at all terminals labeled c, all parasitic capacitances (shown by dotted lines) are charged to value 1 (i.e., a high voltage) because all pMOSFETs are conductive. This process is called precharging. Thus, the outputs of all inverters become value 0. Suppose that x = v = 1 (i.e., a high voltage) and y = z = u = 0 (i.e., a low voltage). When a clock pulse appears, that is, c = 1, all pMOSFETs become non-conductive but the nMOS subcircuit in each of logic gates 1 and 5 becomes conductive, discharging parasitic capacitance. Then the outputs of logic gates 1, 2, 3, 4, 5, and 6 become 0, 1, 1, 0, 0, and 1, respectively. Notice that the output of logic gate 3 remains precharged because its nMOSFET for u remains non-conductive. Domino CMOS has the following advantages:

• It has a small area because the pMOS subcircuit in each logic gate consists of a single pMOSFET.

• It is faster (about twice) than the static CMOS discussed so far because parasitic capacitances are reduced by using a single pMOS in each logic gate and the first logic gate is buffered by an inverter. Also, an inverter, such as logic gate 2, has smaller parasitic capacitance at its output because it connects to only nMOSFET in logic gate 3, for example, compared to static CMOS where it connects to both pMOSFET and nMOSFET in each of next static CMOS logic gates, to which the output of this static CMOS is connected. This also makes domino CMOS faster.

• It is free of glitches (i.e., transition is smooth) because at the output of each logic gate, a high voltage remains or decreases, but no voltage increases from low to high [7].

Domino CMOS has the following disadvantage:

• Only positive functions with respect to input variables can be realized. (If both xi and xi for each xi is available as network inputs, the network can realize any function. But if only one of them, say xi , is available, functions that are dependent xi on cannot be realized by a domino CMOS logic network.)

So we have to have domino CMOS networks in double-rail input logic (e.g., Ref. 2), or to add inverters, whenever necessary. Thus, although the number of MOSFETs in domino CMOS networks in single-rail input logic, such as Figure 39.9, is almost half of static CMOS networks, the number of MOSFETs in such domino CMOS networks to realize any logic functions may become comparable to the number of MOSFETs in static CMOS networks in single-rail input logic.

Dynamic CVSL

Static CVSL, which is previously described, can be easily converted into dynamic CVSL which is faster, as illustrated in Figure 39.10. The parasitic capacitance of two output terminals are precharged through each of the two pMOSFETs during the absence of a clock pulse. Dynamic CVSL works in a similar manner to domino CMOS. Notice that two pMOSFETs are not cross-connected like static CVSL and we have essentially two independent logic gates. Dynamic CSVL with two outputs, f and f , is in double-rail logic, so unlike

domino CMOS, not only positive functions but also any logic functions can be realized. It is fast because the pMOS subcircuit of static CMOS is replaced by one pMOSFET and consequently parasitic capacitance is small and also because the output of a dynamic CSVL is connected only to the nMOS subcircuits, instead of to both the nMOS and pMOS of a next static CMOS logic gate. It is also free of glitches.

Problems of Dynamic CMOS

Dynamic CMOS, such as domino CMOS and differential CMOS logic, is increasingly important for circuits that require high speed, such as arithmetic/logic units [5], although design and layout of appropriate distribution of voltages and currents are far trickier than static CMOS. Dynamic CMOS with a single-phase-clock has advantage of simple clock distribution lines [3].