Page 462 - DSP Integrated Circuits
P. 462

10.5 Clocking of Synchronous Systems                                 447

            Another disadvantage is the voltage variations that may occur at the output
        of a block due to evaluation time in the block. This results in peaks in the output
        voltage of a p-block, when it is low for two consecutive clock periods, on a negative
        clock transition. For an n-block the opposite situation occurs with voltage drops on
        positive clock transitions when the output is high. Neither of these cases cause
        problems in single-phase logic, but connections to other kinds of logic have to be
        made with caution.
            An advantage is that
        only one clock wire needs to
        be distributed over the logic
        circuitry, resulting in fewer
        clock drivers and no clock
        skew. However, the require-
        ment on sharp clock edges is
        strict.
            A dynamic shift regis-
        ter using a single-phase
        clock is shown in Figure
        10.18. When 0 = 0, the first
        inverter is in its precharge
        mode    and   the second
        inverter is off. When 0 = 1,
        the first inverter is in its  Figure 10.18 CMOS shift register; using a single-
        evaluate mode and the sec-              phase clock
        ond inverter is turned on.
        The third inverter is in its
        precharge mode and the fourth inverter is off. When 0=0 the next time, the third
        inverter is in its evaluate mode and the fourth inverter is on. At the end of the
        clock period, the output becomes equal to the input.


        10.5.3 Two-Phase Clock
        The logic circuit shown in Figure
        10.14 may still malfunction if the
        clock signals, 0 and &, are skewed. If
        the clock skew is large enough, both
        0 and €> are high for a short time,
        and the signal may propagate
        through two adjacent logic blocks. To
        prevent this happening, a nonover-
        lapped two-phase clock, 4>i and 02,
        can be used. The nonoverlapping
        parts of the clock interval allow for
                                  1
        some skew of 0\ and 02- I*  Figure
        10.19, the two-phase clock, 0± and  Figure 10.19 Two-phase non-overlapping
        3>2, is shown.                                 clock, <PI(£) and &z(t)
            The switches needed in a
        dynamic latch can be implemented by
        using either pass-transistors or transmission gates. A dynamic shift register can
        easily be implemented by alternately cascading transmission gates (TG) and
   457   458   459   460   461   462   463   464   465   466   467