* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download 970 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: ANALOG AND DIGITAL SIGNAL...
Survey
Document related concepts
Transcript
970 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 44, NO. 11, NOVEMBER 1997 specifications—N odd, K = 0:2; fp = 0:43; fs = 0:57 and s < 0.01 (040 dB). This filter has (fp + fs )=2 = 0:5: Substituting K = 1=0:2 into (13) and s = 0:01 2 0:2 into (8) gives the estimated value of N as 23.6. The linear programming results of s for filters with N = 23 and N = 25 are 038.0 dB and 041.2 dB, respectively. The selected value of N should be 25. IX. CONCLUSION This paper investigates some of the properties of the linear phase diamond-shaped 2-D FIR low-pass filters. A set of design rules applicable to the design of such filters optimal in the minimax error sense is presented. Over 1000 low-pass diamond-shaped filters were designed using linear programming to form the data base used to formulate the set of design rules. The design rules provide good estimates for the minimum filter support size required to meet a given specification set. From our experience, the maximum relative error in the estimated filter support size is 10%; in most cases, it is less than 5%. ECL Storage Elements: Modeling of Faulty Behavior Sankaran M. Menon, Yashwant K. Malaiya, and Anura P. Jayasumana Abstract— Bipolar emitter coupled logic (ECL) devices can now be fabricated at very high densities and much lower power consumption. Behavior of two different ECL storage element implementations are examined in the presence of physical faults. While fault models for some implementations of CMOS storage elements have been examined, not much attention has been paid to ECL storage elements. The conventional stuck-at fault model termed minimal fault model assumes that an input(output) of a storage element can be stuck-at-1 or 0. The minimal fault model may not model the behavior under certain physical failures in a storage element. The enhanced fault model providing higher coverage of physical failures is presented. Index Terms—Emitter coupled logic, fault modeling, faulty behavior, storage elements. I. INTRODUCTION REFERENCES [1] O. Herrmann, L. R. Rabiner, and D. S. K. Chan, “Practical design rules for optimum finite impulse response low-pass digital filters,” Bell Syst. Tech. J., vol. 52, no. 6, pp. 769–799, July–Aug. 1973. [2] L. R. Rabiner and B. Gold, Theory and Application of Digital Signal Processing. Englewood Cliffs, NJ: Prentice-Hall, 1975. [3] A. V. Oppenheim and R. W. Schafer, Digital Signal Processing. Englewood Cliffs, NJ: Prentice-Hall, 1975, ch. 5. [4] F. Mintzer and B. Liu, “Practical design rules for optimum FIR bandpass digital filters,” IEEE Trans. Acoust., Speech, Signal Processing, vol. ASSP–27, pp. 204–206, Apr. 1979. [5] P. Siohan, “2-D FIR filter design for sampling structure conversion,” IEEE Trans. Circuits Syst. Video Technol., vol. 1, pp. 337–350, Dec. 1991. [6] S.-C. Pei and J.-J. Shyu, “Design of two-dimensional FIR eigenfilters for sampling-structure conversion,” IEEE Trans. Circuits Syst. Video Technol., vol. 3, pp. 158–162, Apr. 1993. [7] Günter Schamel, “Pre- and postfiltering of HDTV signals for sampling rate reduction and display up-conversion,” IEEE Trans. Circuits Syst., vol. CAS–34, pp. 1432–1439, Nov. 1987. [8] R. M. Mersereau and T. C. Speake, “The processing of periodically sampled multidimensional signals,” IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP–31, pp. 188–194, Feb. 1983. [9] E. Dubois, “The sampling and reconstruction of time-varying imagery with applications in video systems,” Proc. IEEE, vol. 73, pp. 502–522, Apr. 1985. [10] R. Manduchi, G. M. Cortelazzo, and G. A. Mian, “Multistage sampling structure conversion of video signals,” IEEE Trans. Circuits Syst. Video Technol., vol. 3, pp. 325–340, Oct. 1993. [11] P. Carrai, G. M. Cortelazzo, and G. A. Mian, “Characteristics of minimax FIR filters for video interpolation/decimation,” IEEE Trans. Circuits Syst. Video Technol., vol. 4, pp. 453–467, Oct. 1994. [12] Y. C. Lim, “Efficient special purpose linear programming for FIR filter design,” IEEE Trans. Acoust., Speech, Signal Processing, vol. ASSP–31, pp. 963–968, Aug. 1983. [13] R. H. Yang and Y. C. Lim, “Grid density for design of one- and twodimensional FIR filters,” Electron. Lett., vol. 27, no. 22, pp. 2053–2055, Oct. 1991. [14] V. Ouvrard and P. Siohan, “Design of two-dimensional video filters with spatial constraints,” in Proc. EUSIPCO-92, Aug. 1992. [15] D. W. Marquardt, “An algorithm for least squares estimation of nonlinear parameters,” J. Soc. Ind. Appl. Math., vol. 2, pp. 431–441, 1964. [16] J. G. Reich, C Curve Fitting and Modeling for Scientists and Engineers. New York: McGraw-Hill, 1992. Emitter coupled logic (ECL) using bipolar technology is generally used in applications where switching speed is more important than power dissipation and cost. Conventional bipolar ECL technology represents the state of the art in silicon speed, providing system propagation delays in the subnanosecond range but the price paid for such speeds is very high power dissipation (1.5 mW or more per gate—way too much for VLSI densities) [1]. Recent advances in technology such as BIT1 [1] developed by Bipolar Integrated Technology have made it possible to fabricate bipolar ECL devices that take about 1/20 th the area of conventional ECL devices with speeds comparable to the fastest ECL, consuming only one-tenth the power [1]. With the achievement of low power, high speed, as well as high density, ECL technology is expected to be used widely in various high performance digital circuits [2]. Even more highly integrated bipolar and bipolar/MOS chips are expected in future, further narrowing the gap between low cost workstations and high performance servers [2]. Transistor level shorts and opens model many of the physical failures and defects in IC’s [3]. Analysis of faults in simple logic circuits suggest that transistor level testing provides a higher coverage of faults compared to that at gate level [4]. It is necessary to study the effects of failures at the transistor level and develop accurate fault models at this level [3]. The major fault models at transistor level are stuck-at faults, stuck-shorts and opens of transistor and interconnects, and bridging faults [5]. The issue of fault modeling of 1- and 2-level ECL gates have been addressed in [6] and [7], where augmented fault models were presented which provide higher coverage of physical failures. Effects of bridging faults in ECL circuits was presented in [8] and [9]. Since testing of sequential circuits is difficult, one of the common approaches is to convert the problem into simpler problem of testing Manuscript received October 1, 1994; revised July 3, 1995. This work was supported by a BMDO funded project monitored by ONR. This paper was recommended by Associate Editor M. Soma. S. M. Menon was with the Department of Electrical and Computer Engineering, South Dakota School of Mines and Technology, Rapid City, SD 57701 USA. He is now with Texas Instruments, Dallas, TX 75266 USA. Y. K. Malaiya and A. P. Jayasumana are with the Departments of Computer Science and Electrical Engineering, Colorado State University, Fort Collins, CO 80523 USA. Publisher Item Identifier S 1057-7130(97)07650-7. 1057–7130/97$10.00 1997 IEEE IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 44, NO. 11, NOVEMBER 1997 Fig. 1. ECL storage element-1. combinational circuits. Design for testability approaches are used to provide direct access to inputs and outputs of combinational blocks [4], [10]. Assuming that most faults within a storage element can be modeled as stuck-at-0/1 faults on the outputs, then the same faults need not be explicitly considered. This is because such faults are equivalent to the stuck-at faults in the combinational logic surrounding the storage elements. While much work has been done examining physical failures in CMOS latch cells [11]–[14], not much attention has been paid to ECL storage elements. ECL storage elements are used in designs involving high speed communications and networking. Inadequacy of minimal fault model in representing physical failures for CMOS storage elements is presented in [15], and enhanced fault model is presented providing a higher explicit coverage than the minimal fault model. In this brief, we first examine the behavior of two different ECL storage elements under various physical failures. Minimal enhanced fault models are applied to check for its adequacy. It is observed that the minimal fault model does not provide sufficient coverage. The enhanced fault model provides higher coverage of physical failures. The brief is organized as follows. Sections II and III deal with two different implementations of ECL storage elements and stuckat model, respectively, for ECL storage elements. Section IV deals with analysis of physical failures in ECL storage elements. Finally conclusions are given in Section V. II. ECL STORAGE ELEMENTS The ECL logic levels recognized by the storage elements for logic level low are between 01.85 and 01.45 V and for logic level high are between 01.15 and 00.75 V. The intermediate level between 01.45 and 01.15 V is termed “Indeterminate” logic level. Though the input low logic level (VIL ) is between 01.85 and 01.45 V, an input between 05.2 and 01.45 V is valid as logic “0.” Similarly, though the input high logic level (VIH ) is between 01.15 and 00.75 V, an input between 01.15 and 0.00 V is also logic “1.” These voltage levels could occur under some faults in ECL storage elements. Such a situation does not occur in CMOS devices as the output voltage levels reach the extreme voltage levels of 0 and 5 V for logic “0” and “1,” respectively, unlike in ECL devices. An ECL storage element [16] is shown in Fig. 1. Vref1 and Vref2 form the reference voltages for the differential amplifiers. Vref3 along with the transistors Q9 and Q10 and resistors R2 and R3 form 971 Fig. 2. ECL storage element-2. the current sinks. When the clock input is high, the logic level at input D is clocked into the storage element. The clocked input value is maintained by the feedback provided by transistor Q7 : Another storage element [16] is shown in Fig. 2, based on differential amplifier with bilateral drive. In this scheme, true and complementary outputs are directly available. III. STUCK-AT MODEL FOR STORAGE ELEMENTS Most approaches for modeling of faults rely on the assumption that the faults in storage elements can be modeled as stuck-at faults at its inputs and outputs. In many situations, when faults are considered in the combinational logic, it is implicitly assumed that any internal faults in a storage element can either be shown to be equivalent to a stuck-at-output or stuck-at-input, which in turn would appear as a stuck-at-output of the combinational logic surrounding the storage element. This is termed a minimal fault model. Stuck-at-0/1 fault at the input(output) of a storage element is equivalent to the stuck-at-0/1 fault of the combinational logic output feeding (input being fed) by the storage element. Most Design for Testability (DFT) approaches rely on the assumption that the faults in storage elements can be modeled as stuck-at faults at its inputs and outputs. For example, in LSSD implementations, stuck-at faults at the outputs of latches are considered as stuck-at faults at the inputs of the combinational circuit driven by the storage elements. Similarly, stuck-at faults at the inputs of latches are considered as stuck-at faults at the outputs of the combinational circuit driving the storage elements. In addition, the DFT schemes rely on large number of complex storage elements. We examine the effectiveness of the minimal fault model in representing physical failures. The results reveal the need for a more accurate fault model to better represent the physical failures at transistor level. The results inferred from elementary fault model can then be used to obtain fault models for complex storage elements. Improved fault models representing physical failures accurately can help reduce test generation and fault simulation efforts significantly. IV. PHYSICAL FAULTS IN ECL STORAGE ELEMENTS In this section, we evaluate the response of two different ECL storage elements for various faults. Both the storage elements are examined for all possible hard failures. Possible hard failures considered include all possible opens and shorts of resistors, transistor junction opens and shorts. Fig. 3 shows the naming convention used 972 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 44, NO. 11, NOVEMBER 1997 TABLE I BEHAVIOR OF ECL STORAGE ELEMENT-1 UNDER OPENS AND SHORTS (a) (b) Fig. 3. Naming convention used for transistors (a) shorts and (b) opens. for (a) shorts and (b) opens between various junctions of the bipolar transistor. For simulating shorts between various junctions of the bipolar transistors, a low resistance of 1 was connected between the terminals and for opens, a resistance of 10 M was connected between the respective terminals. A good functional fault model which adequately describes the functional behavior of faulty storage elements is sought. The minimal fault model as well as enhanced fault models are examined for their effectiveness in representing the faults. The enhanced fault model proposed in [15], was shown to provide a higher explicit fault coverage for CMOS storage elements. Data feedthrough faults, which are remarkably different from stuck-at-0/1 faults were proposed. Such faults cause the storage element to become either Data or Data-feed-through, which could lead to timing problem or coupling between combinational blocks separated by the storage elements [17]. In the two different ECL storage elements examined in this brief, Data and Data feed-through faults have been observed and are defined [15] as follows. Definition 1: A faulty storage element is said to have a feedthrough fault if it becomes Data feed-through or Data feed-through. 1) A faulty storage cell is said to be Data-feed-through when its behavior becomes combinational such that R(s; ti ) = f (y ) for each ti 2 T ; where y is the data part of ti : The above definition is modified to include data feed-through faults, which is defined as shown in Definition 2. Definition 2: A faulty storage element is said to have a feedthrough fault if it becomes data-feed-through. 1) A faulty storage cell is said to be data-feed-through when its behavior becomes combinational such that R(s; ti ) = f (y ) for each ti 2 T ; where y is the data part of ti : where T = ft1 ; 1 1 1 ; tn g are the set of all possible input combinations. For an elementary synchronous storage element with input D and a control signal C LK; and n = 4. R(s; ti ) is the response of the cell to the input vector ti applied to the cell when the cell is at state s: To avoid glitches and hazards in clocked storage elements, clock is applied when the data is stable. The latch is race-free if the data is stable when clock is active [11]. Hence, data (D) is allowed to change only when clock (C LK ) is low. Certain faults cause change in behavior of the synchronous storage element in latch phase, while functioning properly in the transparent phase. This causes the cell to be unable-to-latch 1(0) and is defined [15] below. Definition 3: Let the state of a faulty synchronous elementary storage element during the transition from the transparent phase to the latch phase be Q = 1(0). If the state of the cell becomes 0(1) during the latch phase irrespective of data input, then the cell is said to be unable-to-latch 1(0). The notation U L 0 1(U L 0 0) is used to describe this. The ECL storage element-1 was analyzed for all input vectors under shorts and opens between the junctions of all transistors. The outputs were verified by performing SPICE [18] simulations and the results are tabulated in Table I. While some faults do not cause any appreciable change in the logic level exhibiting fault-free level, some faults manifest as stuck-at-1 or stuck-at-0. Six faults cause the cell op = = = Open, sh Short, (e, b, c emitter, base, collector), @ Abnormal Current. = to be transparent with Q output being the same as D and therefore the cell exhibits Data-feed-through. For one of the faults, the output becomes a logical function of the clock signal and is termed as complex behavior. Behavior exhibiting unable to latch-1 and unable to latch-0 are observed for 2 and 9 faults, respectively. Out of the 68 faults considered, 46 faults (67.6%), are covered by the minimal fault model, while the enhanced fault model covers 17 more resulting in 63 faults (92.6%). Three faults manifesting as indeterminate output and 2 faults exhibiting complex behavior cannot be modeled either by minimal fault model or by the enhanced fault model. Results obtained for ECL storage element-2 Q output is summarized in Table II. Twenty faults do not cause any appreciable change in the logic level at Q output exhibiting fault-free output. Some faults manifest as stuck-at-1 or stuck-at-0. Five faults cause the cell to exhibit data feed-through. For four of the faults, the output becomes a logical function of the clock signal and is termed as complex behavior. Behavior exhibiting unable to latch-1 and unable to latch0 are observed for five and 11 faults, respectively. Out of the 90 faults considered, 53 faults (58.8%), are covered by the minimal fault model, while the enhanced fault model covers 21 more resulting in 74 faults (82.2%). Twelve faults manifesting as indeterminate output and three faults exhibiting complex behavior cannot be modeled either by minimal fault model or by the enhanced fault model. Results obtained for ECL storage element-2 Q output is summarized in Table III. Twenty faults do not cause any appreciable change in the logic level at Q output exhibiting fault-free output. Some faults manifest as stuck-at-1 or stuck-at-0. Seven faults cause the cell to exhibit Data feed-through. For two of the faults, the output becomes a logical function of the clock signal and is termed as complex behavior. Behavior exhibiting unable to latch-1 and unable to latch-0 are observed for 5 and 10 faults, respectively. Out of the 90 faults considered, 53 faults (58.8%), are covered by the minimal fault model, while the enhanced fault model covers 22 more resulting in 75 faults (83.3%). 13 faults manifesting as indeterminate output and 2 faults exhibiting complex behavior cannot be modeled either by minimal fault model or by the enhanced fault model. Two faults (R2sh and R3sh ) in ECL storage element-1 and four faults (R3sh ; R4sh ; R5sh and R6sh ) in ECL storage element-2 cause an IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 44, NO. 11, NOVEMBER 1997 TABLE II Q OUTPUT OF ECL STORAGE ELEMENT-1 UNDER OPENS AND SHORTS 973 times. A current monitor to monitor the power supply current (IEE ) can be used to detect the enhanced current drawn by the device and work is underway to develop such a scheme. On careful observation of Tables II and III indicate that for 14 of the physical failures (15%) marked 3; the input test vectors cause both the true and complementary outputs to exhibit similar outputs (00 or 11), i.e., exhibits loss of complementarity. Under fault-free conditions, the Q and Q outputs exhibit fault-free dissimilar outputs (01 or 10). A simple testable design using an exclusive-OR or an exclusive-NOR gate can be used to detect the loss of complementarity occurring at the outputs of ECL storage elements. Use of an exclusive-OR or exclusiveNOR to detect loss of complementarity may be a severe penalty to pay in terms of area overhead. However, this approach is effective and a scheme to implement this logic with less number of transistors is underway. V. CONCLUSIONS op = Open, sh = Short, (e, b, c, = emitter, base, collector), * = Loss of Complementarity, @ = Abnormal Current. Physical failures in two different ECL storage elements have been examined. High fault coverage may not be obtained using the minimal fault model for ECL storage elements. The enhanced fault model provides higher explicit coverage of physical failures compared to the minimal fault model. The enhanced fault model includes faults that cause the cell become Data or Data feed-through as well as faults that cause the cell unable to latch High or Low signals. When modest fault coverage is needed, the minimal fault model may be sufficient. However, in situations where higher fault coverage is sought, the enhanced fault model may be able to provide higher coverage of physical failures. Power supply current monitoring can be used for detection of certain types of faults in ECL storage elements. TABLE III Q OUTPUT OF ECL STORAGE ELEMENT-1 UNDER OPENS AND SHORTS op = Open, sh = Short, (e, b, c = emitter, base, collector), 3 = Loss of Complementarity, @ = Abnormal Current. increase current drawn by the device. Normal current drawn by the device is 6 mA but under the above mentioned faults, the current drawn by the device increases to 1.3A, an increase of about 200 REFERENCES [1] G. Wilson, “Creating low-power bipolar ECL at VLSI densities,” VLSI Syst. Design, May 1986, pp. 84–86. [2] E. W. Brown, A. Agrawal, T. Creary, M. F. Klein, D. Murata, and J. Petolino, “Implementing sparc in ECL,” IEEE Micro, pp. 10–22, Feb. 1990. [3] J. A. Abraham and W. K. Fuchs, “Fault and error models for VLSI,” Proc. IEEE, vol. 74, pp. 639–653, May 1986. [4] Y. K. Malaiya, B. Gupta, Anura P. Jayasumana, R. Rajsuman, S. Menon, and S. Yang, “Functional fault modeling for elementary static storage elements,” in IEEE Comput. Soc. 12th Annual IEEE Workshop Design for Testabil., Apr. 18–21, 1989. [5] C. C. Beh, K. H. Arya, C. E. Radke, and K. E. Torku, “Do stuck faults models reflect manufacturing defects?” in Proc. IEEE Test Conf., Nov. 1982, pp. 35–42. [6] S. M. Menon, A. P. Jayasumana, and Y. K. Malaiya, “Fault modeling of ECL devices,” Electron. Lett., pp. 1105–1108, July 1990. [7] S. M. Menon, Y. K. Malaiya, and A. P. Jayasumana, “Fault modeling and testable design of 2-level complex ECL gates,” in Proc. 4th CSI/IEEE Int. Symp. VLSI Design, Jan. 1991, pp. 23–28. [8] S. M. Menon, Y. K. Malaiya, and A. P. Jayasumana, “On Bridging Faults in ECL Circuits,” in Proc. 5th CSI/IEEE Int. Symp. VLSI Design, Jan. 1992, pp. 55–60. [9] S. M. Menon, A. P. Jayasumana, and Y. K. Malaiya, “Modeling and analysis of bridging faults in emitter coupled logic (ECL) circuits,” IEE Proc., Part E: Comput. Digital Tech., vol. 140, no. 4, July 1993, pp. 220–226. [10] V. D. Agrawal, S. K. Jain, and D. M. Singer, “Automation in design for testability,” in IEEE 1984 Custom Integrated Circuits Conf., 1984, pp. 159–163. [11] M. K. Reddy and S. M. Reddy, “Detecting FET stuck-open faults in CMOS latches and flipflops,” IEEE Design Test Comput., pp. 17–26, Oct. 1986. [12] D. L. Liu and E. J. McCluskey, “A CMOS cell library design for testability,” VLSI Systems Design, May 4, 1987, pp. 58–65. [13] R. Anglada and R. Rubio, “Functional fault models for sequential circuits,” Res. Rep. DEE-3, Elect. Eng. Dept, Polytech. Univ. Catalunya, Barcelona, Spain, 1987. 974 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS—II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 44, NO. 11, NOVEMBER 1997 [14] K. J. Lee and M. A. Breuer, “A universal test sequence for CMOS scan registers,” in IEEE 1990 Custom Integrated Circuits Conf., 1990, pp. 28.5.1–28.5.4. [15] W. K. Al-Assadi, Y. K. Malaiya, and A. P. Jayasuamana, “Use of storage elements as primitives for modeling faults in synchronous sequential circuits,” in 6th Int. Conf. VLSI Design, Jan. 1993, pp. 118–123. [16] C. Morandi, L. Niccolai, F. Fantini, and S. Gaviraghi, “ECL fault modeling,” in IEE Proc. Part E, Computers and digital techniques, Nov. 1988, pp. 312–317. [17] W. K. Al-Assadi, Y. K. Malaiya, and A. P. Jayasuamana, “Detection of feed-through faults in CMOS storage elements,” in 4th NASA Symp. VLSI Design, Oct. 1992, pp. 7.2.1–7.2.5. [18] L. W. Nagel, “SPICE2 : A computer program to simulate semiconductor circuits,” Electron. Res. Lab., Univ. California, Berkeley, Memo. ERLM520, May 1975. Fig. 1. Elimination of redundant branches. Synthesis of High-Speed Pass-Transistor Logic Vojin G. Oklobdzija and Benoit Duchêne Abstract—This brief presents new pass-transistor logic which contains fewer transistors and has better performance than Hitachi’s double passtransistor logic (DPL). The new logic is characterized by excellent speed and low power. Index Terms—CMOS, pass-transistor logic. I. INTRODUCTION New CMOS logic using pass-transistor circuits has been proposed recently with the objective of improving speed and power consumption [2]–[4]. The double pass-transistor logic (DPL), developed by Hitachi demonstrated a 1.5-nS 32-bit ALU and 4.4-nS 54-bit multiplier in 0.25-m technology [2], [4]. However, DPL has not yet been fully adopted because of its high transistor count. The logic proposed here minimizes the number of transistors used in DPL and preserves the speed. The simulations and tests were performed using 1-m CMOS. Fig. 2. Signal rearrangement. II. NEW LOGIC GATE DVL (dual value logic) gate was obtained by eliminating the redundant branches and rearrangement of signals in DPL. Signal rearrangement results in NAND gate configuration which is faster than DPL (60 pS versus 75 pS), where the AND half is faster. These simplifications, illustrated in Figs. 1–3, preserve the advantages of DPL gates. Finally, we chose a faster half from Fig. 1 and from Fig. 2. The resulting DVL gate contains total of eight transistors compared to DPL consisting of four transistors of each type. There is a total of nine inputs in DVL versus 12 in DPL resulting in a smaller capacitive load of DVL gate. In DVL, three inputs are connected to the transistor source and six to the gate (three to p-type and three to n-type). In Manuscript received June 10, 1995; revised November 12, 1995. This paper was recommended by Associate Editor S. Kiaei. V. G. Oklobdzija is with the Advanced Computer System Engineering Laboratory, Electrical and Computer Engineering Department, University of California at Davis, Davis, CA 95616 USA. B. Duchêne is with Ecole Superieure d’Ingenieurs en Electrotechnique et Electronique, 93162 Noisy le Grand CEDEX, France. Publisher Item Identifier S 1057-7130(97)07654-4. Fig. 3. Resulting DVL gate. TABLE I COMPARISON BETWEEN DVL AND CMOS DPL 4, inputs are connected to the source and eight to the gate (four to p-type and four to n-type transistors). A similar method can be used to build the NOR/OR gates. Comparison with CMOS: A comparison between DVL and conventional CMOS for given function F2 = BC + ABC is shown in Fig. 4 and in Table I. 1057–7130/97$10.00 1997 IEEE