An Oscillator Design Based on MOS-NDR Inverter Cher-Shiung Tsai1, Chung-Chih Hsiao, Kwang-Jow Gan, Jia-Ming Wu, Ming-Yi Hsieh, Chun-Chieh Liao, Shih-Yu Wang, Feng-Chang Chiang, Chia-Hung Chen, Dong-Shong Liang, Yaw-Hwang Chen Department of Electronic Engineering, Kun Shan University of Technology Tainan, Taiwan 710, R.O.C. e50401@mail.ksut.edu.tw values. Under different voltage, the oscillation Abstract-- First, we create a MOS-NDR (negative frequency will increase as supply voltage increases. It differential resistance) cell and then put two MOS-NDR shows the MOS-NDR oscillator is a voltage control cells in cascode (totem-pole) structure. With MOBILE oscillator (VCO). The MOS-NDR oscillator supplies (monostable-bistable transition logic element) theorem flexibility and easiness in design. we can built a MOS-NDR inverter by placing a NMOS 2. MOS-NDR Device in parallel with the lower part of the cascode structure. The MOS-NDR inverter cascades two Common-Source amplifier and feedback to MOS-NDR inverter. The In the beginning, we present a MOS- NDR circuit will start an oscillation. Such a phenomenon device, which is composed of one PMOS transistor is similar to ring oscillator. The frequency of the and three NMOS transistors, as shown in Fig. 1. This MOS-NDR oscillator will be 116 MHz when voltage is MOS-NDR device can exhibit various negative 2.7 volts. Those MOS-NDR devices are adapted from differential CIC standard 0.35um CMOS process. characteristics by choosing appropriate values for the resistance (NDR) current-voltage transistors parameters. Keywords: monostable-bistable transition logic element (MOBILE); negative differential resistance (NDR), common-source Amplifier, ring oscillator. 1. Introduction There are many kinds of oscillators but we seldom see an oscillator built by MOS-NDR [1-3] devices. In this thesis, we propose a MOS-NDR Fig. 1 Configuration of MOS-NDR device circuit. Inverter [4-5] by MOBILE [6-8] Theorem. Just like ring oscillator [9] theorem, the MOS-NDR inverter In this circuit, transistor mn2 acts like a variable can also implement an oscillator. The MOS-NDR resistor. When supply voltage Vdd equals to zero, mn2 oscillator can reach 100 MHz easily and has low is off and mn3 is on. While supply voltage Vdd noise character. increases, transistor mn2 is on and voltage Vc will The MOS-NDR oscillator can change output decrease to make transistor mn3 turn off. Because frequency by adjusting MOS sizes or resistance transistor mn3 is large size, when supply voltage Vdd increases but total current Idd decreases. It creates a NDR phenomenon. Fig. 2 shows the I-V characteristics of the MOS-NDR device by H-spice simulation. Fig. 3 Configuration of an inverter by MOS-NDR circuit. I Load Driver + Vin Fig. 2 The simulated I-V curve of Fig. 1 Vin=High Vout=Low stable1 3. MOS-NDR Inverter In this work, we demonstrate an inverter gate design that is based on the monostable-bistable transition logic element (MOBILE) theory. A VS Out(L) V MOBILE cell is consisted of two NDR devices in Fig. 4 When IDriver+Vin>ILoad , operating point at cascade structure, as shown in Fig. 3. During a period stable1 and output is in low State when the input signal rises, the voltage at the output node between the two NDR devices goes to one of On the contrary, in Fig. 3 when input (Vin) is in the two stable states (low and high, corresponding to low state then NMOS is off. The driver current is “0” and “1” in binary logic), depending on which smaller than load current, the operating point will be NDR devices is reached the second positive located at point Stable2 and output state is in high differential resistance (PDR) region first. The state as Fig. 5 shows. Such an operation just likes an controlled process obeys a rule: the MOS-NDR inverter. device with the smaller peak current is always reached the second PDR region first. Based on the MOBILE theory and control rule, the measured result I Load Driver Vin=Low Vout=High of inverter logic is implemented. In Fig. 3, when input (Vin) is in high state then stable2 NMOS is on. The summation of driver current and NMOS current is larger than load current, the operating point will be located at point Stable1 and output state is in low state as Fig. 4 shows. Out(H) VS V Fig. 5 When IDriver <ILoad , operating point at stable2 and output is in high state 4. MOS-NDR Oscillator and oscillator circuit. Fig. 7 shows the input (upper) and output (lower) signals in MOS-NDR inverter Fig. 6 shows the oscillator circuit of MOS-NDR circuit of Fig. 3. devices. The 1st stage is NDR inverter and 2nd stage is voltage amplifier which consisted by two common source amplifiers. After MOS-NDR inverter, the output signal could be reduced, it needs voltage amplification. The signal of feedback must be the same as original input signal, it needs two common-source amplifier because common-source amplifier has high voltage gain and phase change of 180 degrees. The theorem of an oscillator is the feedback signal must be the same as the past original input signal but out of phase with 180 degrees. The total change of the MOS-NDR oscillator is 540 Fig. 7 Input signal (upper) and output signal (lower) degrees (= 180 degrees), it satisfies the oscillation reveal inverter operation. phase. The voltage amplifier (2nd stage) also supplies a time delay to decide the oscillation frequency. Such It is an inverter operation, the input is 2 Vp-p and a phenomenon is similar to ring oscillator. All devices output is 0.5Vp-p. The MOS-NDR inverter is the same in Fig.6 are from CIC standard 0.35um CMOS as CMOS inverter except output voltage is not full process except resistor R1 and R2. Resistors R1 and swing. R2 are external devices. We put these elements on Fig. 8 shows the MOS-NDR oscillator output bread board to establish the MOS-NDR oscillator and waveform, output signal is 1.4Vp-p and frequency is measure it. 111.6 MHz as supply voltage is 2.7 Volts. R1 and R2 of the MOS-NDR oscillator (Fig. 6) are small Vs resistance resistors which below 10 Ohms. NDR1 (Load) R2 R1 OUT mn2 Vin mn1 NMOS NDR2 (Driver) Fig. 6 Configuration of the MOS-NDR oscillator circuit. 5. Experimental Results In this thesis, we use Tektronix TDS3034B oscilloscope to measure MOS-NDR inverter circuit Fig.8 Output waveform of MOS-NDR oscillator under 2.7 volts supply voltage. Fig. 9 shows MOS-NDR oscillator frequencies under different supply voltages. The oscillation frequency increases as supply voltage increases. It shows that the MOS-NDR oscillator is a voltage control oscillator (VCO) with good linearity. The start oscillation voltage of the MOS-NDR oscillator is 2.1 Volts. Fig. 10 Configuration of FFT corresponding to Fig. 8 6. Conclusions In this work, we present a MOS-NDR cell which consisted of three NMOS transistors and one PMOS Fig. 9 Oscillator output frequency under different transistor. We use the cell combined with a large size supply voltage. NMOS transistor to build an MOS-NDR inverter by MOBILE We also use Tektronix TDS3034B oscilloscope (monostable-bistable transition element) theorem, then we use the dominant to measure FFT (fast Fourier transform) response of MOS-NDR Fig. 8. In Fig. 10, the MOS-NDR oscillator shows amplifiers to establish MOS-NDR oscillator. FFT diagram with pretty good low-noise logic inverter and two common-source Such an oscillator shows pretty good response in frequency domain and low-noise characteristics. It’s characteristics. In Fig. 10, because the difference between the also a good voltage control oscillator (VCO). The highest peak and 2nd highest peak is about 28 dB, it MOS-NDR oscillator has both easiness and flexibility means that the main frequency signal is twenty-five in design items. (1028/20 =25.12) times stronger than the other frequencies signals. If the output signal 1.5 Vp-p then Acknowledgements the other frequencies signals must be smaller than 0.06 (=1.5/25.12) Vp-p. The authors would like to thank the Chip But the differences between the highest peak Implementation Center (CIC) for their great effort and most other peaks is more than 40 dB in Fig. 10, and assistance in arranging the fabrication of this so most frequencies signals must be smaller than work. This work was supported by the National 0.015 Vp-p (1.5/10 (40/20) = 0.015) . We can see that most signals are low-noise except oscillation frequency signal. Science Council of Republic of China under the contract no. NSC93-2218-E-168-002. References 6. K. Maezawa, H. Matsuzaki , M. Yamamoto, and T. Otsuji, “High-Speed and low-power operation 1. L. O. Chua, J. B. Yu, and Y. Y. Yu, “Negative resistance of a resonant tunneling logic gate MOBILE,” IEEE devices,” Int j. Cir. Appl., vol 11, pp. 161-186, 1983. Electron Device Lett., vol 19, pp. 80-82,1998 2. S. Sen, F. Capasso, A. Y. Cho and D. Sivco, ”Resonant 3. 4. K. Maezawa, T. Akeyoshi, and T. Mizutani, tunneling device with multiple negative differential “Functions and applications of monostable-bistable resistance : digital and signal processing applications transition with reduced circuit complexity,” IEEE Trans. multi-input terminals,” IEEE Trans. Electron Devices, Electron Devices, vol 34 pp. 2185-2191, 1987 vol 41, pp. 148-153, 1994 C. S. Tsai, S. B. Kuo, K. J. Gan, Y. H. Chen, C. C. Hsu, 8. logic elements (MOBILES’s) having Tomoyuki Akeyoshi, Koichi Maezawa, and Takashi K. R. Lee, and D. S. Liang, “Design and Fabrication Mizutani, “Weighted sum threshold logic operation of of MOBILES Prototype MOS-NDR Devices by CMOS (monostable-bistable transition logic Technology,” 2003 EDMS, pp. 235-238. elements) using resonant tunneling transistors,” IEEE K. J. Chen, T. Waho, Electron Device Lett., vol 14, pp.475-477, Dec. 1979 K. Maezawa, and M. Yamamoto, “An exclusive-OR logic circuit based on controlled 9. Adel S. Sedra and Kenneth C. Smith, quenching of series-connected negative differential “Microelectronic Circuits,” 5th edition, pp. 1027, resistance devices,” IEEE Electron Device Lett. vol. 2004. 17, pp. 309-311, 1996 5. 7. Cher-Shiung Tsai and Shin-Bin Kuo, “Investigation of Inverter Design by MOS-NDR Devives and MOBILE Theory,” 2003 International Conference on Informatics, Cybernetic, and Systems, 義守大學, pp. 296-300. Implementation of R-BJT-NDR Devices and Inverter Circuit by BiCMOS Technology Dong-Shong Liang, Kwang-Jow Gan, Chung-Chih Hsiao, Cher-Shiung Tsai, Yaw-Hwang Chen, Shun-Huo Kuo, and Hsin-Da Tesng Department of Electronic Engineering, Kun Shan University of Technology Tainan Hsien, Taiwan, R.O.C. E-mail Address:suln@mail.ksut.edu.tw . ABSTRACT The negative differential resistance (NDR) device studied in this work is composed of the resistors (R) and bipolar junction transistors (BJT) devices. This NDR device can be represented as R-BJT-NDR device. Comparing to the conventional NDR devices such as resonant tunneling diodes (RTD’s), the RTD’s are often implemented by the technique of III-V compound semiconductor. The fabrication cost of these RTD’s is more expensive than that of Si-based technique. However our R-BJT-NDR devices and circuits are composed of the resistors and BJT devices. Therefore, we can fabricate the R-BJT-NDR devices by the standard CMOS or BiCMOS process. Furthermore, the modulation of the current-voltage (I-V) characteristics of this R-BJT-NDR device is easier than that of the RTD device. We can obtain different NDR I-V characteristics by simply adjusting the proper values of resistors. Finally, we demonstrate an inverter circuit design based on the R-BJT-NDR devices and circuits that are implemented by the standard 0.35μm BiCMOS process. Keywords: negative differential resistance (NDR) device, resonant tunneling diodes (RTD), inverter circuit, 0.35μm BiCMOS process 1. INTRODUCTION Functional devices and circuits based on negative differential resistance (NDR) devices have generated substantial research interest owing to their unique NDR characteristics [1]-[3]. The best famous known NDR device is the resonant tunneling diodes (RTD) [4-5]. The fabrications of these NDR devices are often based on the technique of compound semiconductor. Comparing to silicon-based integrated circuit, the cost of compound semiconductor is more expensive. However, we proposed a NDR device that is composed of the resistors (R) and bipolar junction transistors (BTJ) devices. We call this device as R-BJT-NDR device. Because this device is totally composed of the R and BTJ devices, it is suitable for the process of Si-based CMOS or BiCMOS process. In this paper, we investigate the current-voltage (I-V) characteristics by modulating the parameters of the R-BJT-NDR device. The fabrication of the device is utilized by 0.35μm BiCMOS process. Furthermore, we will demonstrate an inverter design based on the R-BJT-NDR device under 3V bias. 2. CIRCUIT ANALYSIS Figure 1 shows this R-BJT-NDR device composed of two n-p-n transistors and five resistors. During suitably arranging the values of the five resistors, we can obtain the I-V curve with NDR characteristics. Figure 2 shows the simulated I-V characteristics by Hspice program. The electrical parameters are R1=4.5kΩ, R2=5.3kΩ, R3=100kΩ, R4=1.2 kΩ, and R5=6kΩ. The I-V characteristics can be divided into four regions. The first segment (I) of the I-V characteristics represents a situation when Q1 and Q2 are cut off, the second segment (II) indicates the case when Q1 is cut off but Q2 is conducted, the third segment (III) refers to the state when both Q1 and Q2 are conducted, and the forth segment (IV) corresponds to the state when Q1 is saturated but Q2 is cut off. values of the resistor R2. Based on the circuit analysis, the peak voltage is approximated to the value of (1+R1/R2)xVr1. The parameter Vr1 is the cut-in voltage of the transistor Q1. Therefore, this R-BJT- NDR device possesses the ability of the adjustable I-V characteristics during suitably designing the resistance. VS R1 R2 R4 R3 Q1 Q2 R5 Fig. 3 The simulated I-V characteristics by modulating the values of R1 resistor. Fig. 1 The NDR device is composed of two transistors and five resistances. Fig. 2 The simulated I-V characteristics by Hspice program. The I-V characteristics of this R-BJTNDR device can be modulated by the relative values of the resistors. Figures 3 and 4 are the simulated I-V characteristics by varying the values of the resistors R1 and R2, respectively. By keeping the other parameters at some fixed values, the peak currents and voltages will be increased by increasing the values of the resistor R1. However, the peak currents and voltages will be decreased by increasing the Fig. 4 The simulated I-V characteristics by modulating the values of R2 resistor. 3. MESURED RESULTS The fabrication of this R-BJT-NDR device is designed based on the standard 0.35μm SiGe process. Figure 5 shows the layout of the R-BJT-NDR device. Figure 6 shows the measured I-V characteristics using the Tektronix 370B Programmable curve tracer. The parameters are designed as R1=5.5kΩ, R2=3.3kΩ, R3=100kΩ, R4=1.2kΩ, and R5=5.5kΩ. The peak current and voltage are about 1.1mA and 1.8V, respectively. The valley 1.6 Current (mA) current and voltage are about 0.6mA and 2.5V, respectively. Figures 7 and 8 show the measured I-V characteristics by varying the values of R1 and R2, respectively. 1.2 0.8 0.4 0 5.5K 6.5K 7.5K 0 1 2 3 4 5 Voltage (V) Fig. 7 The measured I-V characteristics of the R-BJT-NDR device by varying the R1 values. 1.6 Fig. 5 The layout of the R-BJT-NDR device. Current (mA) Current (mA) 1.6 1.2 0.8 0.4 0 1.2 0.8 0.4 0 0 1 2 3 4 5 Voltage (V) Fig. 6 The measured I-V characteristics of the R-BJT-NDR device. 5.3K 4.3K 3.3K 0 1 2 3 4 5 Voltage (V) Fig. 8 The measured I-V characteristics of the R-BJT-NDR device by varying the R2 values. Based on the measured results, this R-BJT-NDR device could exhibit the NDR characteristics by suitably controlling the parameters. The tendency of the measured I-V characteristics is satisfactory by comparing to the simulated results. 4. INVERTER DESIGN Figure 9 shows the inverter circuit with a NDR device in series with a NMOS device. The input voltage is the gate voltage of the NMOS device. Fig. 10 The load-line analysis of the inverter circuit. Fig. 9 The inverter circuit design based on this R-BJT-NDR device and a NMOS. The operating principle of the static inverter can be explained by the load-line method as shown in Fig. 10. When two devices are connected in series, the upper R-BJT-NDR device is treated as a load device to the pull-down NMOS device. The operation point (Q) at any bias (VS) can be obtained by the stable intersection point of the two I-V characteristics. The voltage value of the operation point Q is the output voltage of the circuit. In Fig. 9, the load-line curve is the I-V characteristics of the R-BJT-NDR device. The driver curve is the I-V characteristics of the NMOS. When the input voltage is at low level (0V), the gate voltage of the NMOS is lower than the VT, so the NMOS will be located at OFF state. The operating point will be located at the intersection point (Q) of two I-V characteristics of the load and driver devices. The output voltage will be approximately the same value as the voltage VS. The output state is at the high level. However when the input voltage is at high level (3V), the NMOS will turn on. The operating point will be located at the intersection point (Q) of two I-V characteristics. As seen, the output state is at the low level. The key point on designing this inverter circuit is the I-V characteristics of the NMOS device at the high state should be higher than the peak current of the R-BJT-NDR device. The result of the operation of the inverter circuit is shown in Fig. 11. The supply voltage VS is chosen at 3V. As seen, when the input voltage has reached the low state (0V), the output voltage is at the high state (3V). When the input voltage has reached the high state (3V), the output voltage is at the low state (0.2V). Therefore, this circuit can be operated as an inverter. Fig. 11 The results of the inverter circuit. 5. CONCLUSIONS We have demonstrated the design and fabrication of the R-BJT-NDR device and inverter circuit based on the standard 0.35μm BiCMOS process. The modulation of the I-V characteristics of this R-BJT-NDR device is easier than that of the RTD device. We can obtain different NDR I-V characteristics by simply adjusting the proper values of resistors. This R-BJT-NDR device is more convenient to integrate with other active MOS and BJT devices and passive R, L and C devices. These phenomena might provide the practical application in some NDR-based circuits. It means that the R-BJT-NDR device might have the potential to achieve the design of system-on-a-chip (SOC). ACKNOWLEDGMENTS The authors would like to thank the Chip Implementation Center (CIC) for their great effort and assistance in arranging the fabrication of this chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2215-E-168-002. REFERENCES [1] S. Sen, F. Capasso, A. Y. Cho, and D. Sivco,“Resonant tunneling device with multiple negative differential resistance: digital and signal processing applications with reduced circuit complexity,"IEEE Trans. Electron Devices, vol. 34, pp. 2185-2191, 1987. [2] K. J. Chen, K. Maezawa, and M. Yamamoto, “InP-based high-performance monostable-bistable transition logic elements (MOBILE's) using integrated multiple-input resonant-tunneling devices”, IEEE Electron Device Lett., vol. 17, pp. 127-129, 1996. [3] Z. X. Yan and M. J. Deen, ”A new resonant-tunnel diode-based multivalued memory circuit using a MESFET depletion load,” IEEE J. Solid-State Circuits, vol. 27, pp. 1198-1202, 1992. [4] K. Maezawa, H. Matsuzaki, M. Yamamoto, and T. Otsuji, “High-speed and low-power operation of a resonant tunneling logic gate MOBILE,” IEEE Electron Device Lett., vol. 19, pp. 80–82, Mar. 1998. [5] J. P. A. van der Wagt, H. Tang, T. P. E. Broekaert, A. C. Seabaugh, and Y. C. Kao, ”Multibit resonant tunneling diode SRAM cell based on slew-rate addressing,” IEEE Trans. Electron Devices, vol. 46, pp. 55-62, 1999. High-Frequency Voltage-Controlled Oscillator Design by MOS-MDR Devices and Circuits Kwang-Jow Gan, Dong-Shong Liang*, Shih-Yu Wang, Chung-Chih Hsiao, Cher-Shiung Tsai, Yaw-Hwang Chen, and Feng-Chang Chiang Department of Electronic Engineering, Kun Shan University of Technology This paper describes the design of a novel voltage-controlled oscillator (VCO) based on the new-type MOS-NDR devices and circuits. The MOS-NDR device is composed of three N-type metal-oxide-semiconductor field-effect-transistor (MOS) devices that can exhibit the negative differential resistance (NDR) characteristics in its current-voltage (I-V) curve by suitably arranging the parameters. The length of three MOS devices is fixed at 0.35μm. The width is designed as 1μm, 10μm, and 40μm, respectively. The I-V curve of the MOS-NDR device can exhibit a Λ-type characteristic. When we connect a NMOS device with a MOS-NDR device in parallel, the total current is the sum of the currents through the NMOS and MOS-NDR devices. Therefore we can modulate the total current by the gate voltage of the NMOS. Firstly, we design a low-power inverter by combining two series-connecting MOS-NDR devices according to the monostable-bistable transition logic element (MOBILE) theory. Secondly, we design an oscillator by cascading this MOS-NDR inverter with the other two common-source amplifiers, as shown in Fig. 1. The signal will be feedback from the output of the second common-source amplifier to the input of the MOS-NDR inverter. Under suitable design, we can obtain an oscillator with its frequency proportional to the magnitude of input bias, Vds1 or Vds2. This new VCO has a wide range of operation frequency from 0.65MHz to 1.55GHz. It consumes 7mW in its central frequency of 1.2GHz using a 2.4V power supply. The fabrication of this VCO is based on the standard 0.35μm CMOS process and occupied an area of 128 x 45 μm2. Vds1 Vds2 MOS-NDR1 OUT MOS-NDR2 CS INV Fig. 1 The circuit configuration of a novel voltage-controlled oscillator (VCO). Logic Circuit Design Based on MOS-NDR Devices and Circuits Fabricated by CMOS Process Kwang-Jow Gan, Dong-Shong Liang*, Chung-Chih Hsiao, Shih-Yu Wang, Feng-Chang Chiang, Cher-Shiung Tsai, Yaw-Hwang Chen, Shun-Huo Kuo, and Chi-Pin Chen Department of Electronic Engineering, Kun Shan University of Technology, Taiwan, R.O.C. Abstract--We propose a new MOS-NDR device that is composed of the metal-oxide-semiconductor field-effecttransistor (MOS) devices. This device could exhibit the negative differential resistance (NDR) characteristics in the current-voltage characteristics by suitably modulating the MOS parameters. We design a logic circuit which can operate the inverter, NOR, and NAND gates. The devices and circuits are fabricated by the standard 0.35μm CMOS process. is cutoff. The third situation: mn1 is saturation, mn2 is saturation, mn3 is saturation, and mp4 is cutoff. Finally, the fourth situation: mn1 is saturation, mn2 is linear, mn3 is cutoff, and mp4 is linear until saturation. Vdd mn1 1. INTRODUCTION Functional devices and circuits based on negative differential resistance (NDR) devices have generated substantial research interest owing to their unique NDR characteristic [1]-[2]. Taking advantage of the NDR feature, circuit complexity can be greatly reduced and novel circuit applications have also obtained. Some applications make use of the monostablebistable transition logic element (MOBILE) as a highly functional logic gate [3]-[4]. A MOBILE consists of two NDR devices connected in series and is driven by a suitable bias voltage. The voltage at the output node between the two NDR devices could hold on one of the two possible stable states (low and high, corresponding to “0” and “1”), depending on the relative difference of peak current (IP) between two devices. We propose a MOS-NDR device that is composed of the metal-oxide-semiconductor field-effect-transistor (MOS) devices. Then we design a logic circuit that can operate the inverter, NOR, and NAND function in the same circuit. It is different from the CMOS logic family constructed by NMOS and PMOS devices. Finally, we fabricate the MOS-NDR devices and logic circuits by standard 0.35μm CMOS process. 2. DEVICE STRUCTURE AND OPERATION Figure 1 shows a MOS-NDR device, which is composed of three NMOS devices and one PMOS device. This circuit is derived from a Λ-type topology described in [5]-[6]. This MOS-NDR device can exhibit various NDR current-voltage (I-V) characteristics by choosing appropriate MOS parameters. Figure 2 shows the simulation results by HSPICE program. If the Vgg voltage is fixed at some value, the operation of this MOS-NDR device can be divided into four situations by gradually increasing the bias Vdd. The first situation: mn1 is saturation, mn2 is cutoff, mn3 is linear, and mp4 is cutoff. The second situation: mn1 is saturation, mn2 is saturation, mn3 is from linear to saturation, and mp4 mn3 mp4 Vgg mn2 Fig. 1 The circuit configuration for a MOS-NDR device. Fig2 Simulation result for the MOS-NDR device by HSPICE program. The value of Vgg voltage will affect the I-V curve, especially in its peak current. Figure 3 shows the I-V characteristics, measured by Tektronix 370B, with the Vgg voltage varied from 1.5V to 3.3V, gradually. The width parameters are designed as mn1=5μm, mn2=100μm, mn3=10μm, and mp4=100μm. The length parameters are all fixed at 0.35μm. The I-V characteristic could be divided into three segments as: the first PDR segment, the NDR segment, and the second PDR segment in sequence. 3.0 2.0 Vgg=3.3V 1st PDR 2nd PDR NDR Vgg=3V 1.0 Vgg=2.5V Current (mA) Current (mA) 6.0 T1 varied from 0V to 3.3V 3.3V 4.0 2.0 3V 2.5V 2V 1.5V 1V 0V Vgg=2V Vgg=1.5V 0.0 0 0.4 0.0 0.8 1.2 1.6 0 0.4 0.8 1.2 1.6 Voltage (V) Voltage (V) Fig. 3 The measured I-V characteristics for a MOS-NDR devices with different Vgg voltages. Fig. 5 Measured I-V characteristics with different T1 voltages. 3. LOGIC CIRCUIT DESIGN If the IP of the driver is smaller than IP of the load (with T2 or T3 OFF), the circuit switches to the stable point Q corresponding to a high output voltage, as shown in Fig. 6. The operation point Q is located at the second PDR region of the driver’s I-V curve. On the other hand, a bigger peak current (with T2 or T3 ON) of the driver will result in the stable point Q with low output voltage. At this time, the operation point Q is located at the second PDR region of the load’s I-V curve. Therefore, an inverter operation can be obtained at the output node between the two NDR devices. Similarly, if we design the logic function according to the MOBILE theory and truth table, we can obtain the NOR and NAND gate logic operation. Our logic circuit design is based on the MOBILE theory. A MOBILE circuit consists of two NDR devices connected in series and is driven by a bias voltage VS. The logic circuit configuration is shown in Fig. 4. T1 is used as the controlled gate. T2 and T3 are the input gates with square wave signal. The width parameters are all 5μm. The upper NDR1 device is treated as a load device to the pull-down NDR2 driver device. When the bias voltage is smaller than twice the peak voltage (2VP), there is one stable point (monostable) in the series circuit. However when the bias voltage is larger than two peak voltages but smaller than two valley voltages (2VV), there is two possible stable points (bistable) that respect the low and high states (corresponding to “0” and “1”), respectively. A small difference between the peak currents of the series-connected NDR devices determines the state of the circuit. I Driver + T2 or T3 Driver VS T1 Q NDR1 Load Q Load Vout T2 NDR2 Driver T3 Out(L) Out(H) VS V Fig. 6 The MOBILE operation is dependent on the relative difference in the peak current of load and driver devices. Fig. 4 Configuration of the logic circuit. The circuit shown at upper left corner of Fig. 4 is a MOS-NDR device with parallel connection of a T1 NMOS. The total current ITotal is the sum of the currents through the MOS-NDR and NMOS devices: ITotal=INDR+IMOS. Since IMOS is modulated by the gate voltage (T1), so is ITotal. Figure 5 shows the measure I-V characteristics of a MOS-NDR device with T1 varying from 0 to 3.3V, gradually. As seen, we can modulate the peak current of the MOS-NDR device through T1 gate. 4. EXPERIMENT RESULTS The MOS-NDR devices and logic circuits are fabricated by the standard 0.35μm CMOS process. The length parameters for all MOS are fixed at 0.35μm. The width parameters of the two MOS-NDR devices are designed as mn1=5μm, mn2=100μm, mn3=10μm, and mp4=100μm. The width parameters for T1, T2, and T3 are fixed at 5μm. For the inverter operation, T2 gate is chosen as the input square wave with signal varied from 0 to 2V. The supply voltage VS is fixed at 1.5V that is bigger than 2VP but is small than 2VV voltage. The operation of T1 control gate is cutoff. The Vgg voltages are 3.3V and 2.8V for NDR1 and NDR2, respectively. The inverter operation is shown in Fig. 7. As seen, when the input voltage has reached the low state (0V), the output voltage will be at the high state. On the other hand, when the input voltage has reached the high state (2V), the output voltage must be at the low state. The parameters are designed and shown in Table I. The operation results for this logic circuit are shown in Fig. 7, respectively. As seen, we can obtain the inverter, NOR, and NAND gate operation at the same circuit only by modulating the relative parameters. Table I. The parameters condition for the logic circuit. Inverter NOR NAND VS 1.5V 1.5V 1.5V T1 OFF 1.2V 3.3V T2 INPUT INPUT INPUT T3 OFF INPUT INPUT 3.3V 3.3V 3.3V 2.8V 2.5V 2.5V NDR1 Vgg NDR2 Vgg 5. CONCLUSIONS Fig. 7 The measured results for the logic circuit. For the NOR gate operation, T2 and T3 gates both are the input square wave with signal varied from 0 to 2V. The period of T2 is two times of the period of T3 signal. The controlling gate T1 is fixed at 1.2V. According to the truth table, when the input T2 and T3 gates are located at low level, the peak current of ITotal of NDR2 device must be smaller than the peak current of ITotal of NDR1 device. Then the stable operating point will be located at the relative “high” level at the output node. If one (or both) of the input T2 and T3 is (are) located at high voltage level, the peak current of ITotal of NDR2 device must be bigger than the peak current of ITotal of NDR1 devices. The stable operating point will be located at the relative “low” level at this case. Therefore, we design the Vgg voltages with 3.3V and 2.5V for NDR1 and NDR2, respectively. As for the operation of a NAND gate, when the input T2 and T3 gates both are located at high level, the peak current of ITotal of NDR2 device must be bigger than the peak current of ITotal of NDR1 device. The stable operating point will be located at the relative “low” level. On the other hand, the peak current of ITotal of NDR2 device must be smaller than the peak current of ITotal of NDR1 device for the other cases. The stable operating point will be located at the relative “high” level. Therefore, the control gate T1 is designed fixed at 3.3V. The Vgg voltages are fixed at 3.3V and 2.5V for NDR1 and NDR2, respectively. We have demonstrated the logic circuit design based on the MOS-NDR devices and circuits. This circuit design is operated according to the principle of MOBILE theory. Our MOS-NDR devices and circuits are fabricated by the technique of Si-based CMOS technique. The fabrication cost will be cheaper than that of resonant tunneling diodes (RTD) implemented by the technique of compound semiconductor. Furthermore, our NDR devices are composed of the MOS devices, so these NDR devices are easy to combine with other devices and circuit to achieve the system-on-a-chip (SoC). If the integrated circuit is fabricated with transistor dimensions less than 0.1μm technique, the applications based on the MOS-NDR devices and circuits are still useful. Therefore, the MOS-NDR devices and circuit have high potential in the SoC applications. ACKNOWLEDGMENTS The authors would like to thank the Chip Implementation Center (CIC) of Taiwan for their great effort and assistance in arranging the fabrication of this chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2218- E-168-002. REFERENCES [1] S. Sen, F. Capasso, A. Y. Cho, and D. Sivco, “Resonant tunneling device with multiple negative differential resistance: digital and signal processing applications with reduced circuit complexity,"IEEE Trans. Electron Devices, vol. 34, pp. 2185-2191, 1987. [2] T. H. Kuo, H. C. Lin, R. C. Potter, and D. Shupe,“A novel A/D converter using resonant tunneling [3] [4] [5] [6] diodes,"IEEE J. Solid-State Circuits, vol. 26, pp. 145-149, 1991. K. Maezawa, H. Matsuzaki, M. Yamamoto, and T. Otsuji, “High-speed and low-power operation of a resonant tunneling logic gate MOBILE,” IEEE Electron Device Lett., vol. 19, pp. 80-82, 1998. K. J. Chen, K. Maezawa, and M. Yamamoto, “InP-based high-performance monostable-bistable transition logic elements (MOBILE's) using integrated multiple-input resonant-tunneling devices”, IEEE Electron Device Lett., vol. 17, pp. 127-129, 1996. C. Y. Wu and K. N. Lai, “Integrated Λ -type differential negative resistance MOSFET device,” IEEE J. Solid-State Circuits, vol. 14, pp. 1094-1101, 1979. A. F. Gonzalez and P. Mazumder, “Multiple-valued signed-digit adder using negative differentialresistance devices”, IEEE Trans. Comput, vol. 47, pp. 947-959, 1998. Novel Voltage-Controlled Oscillator Design by MOS-NDR Devices and Circuits Dong-Shong Liang*, Kwang-Jow Gan, Chung-Chih Hsiao, Cher-Shiung Tsai, Yaw-Hwang Chen, Shih-Yu Wang, Shun-Huo Kuo, Feng-Chang Chiang, and Long-Xian Su Department of Electronic Engineering, Kun Shan University of Technology, Taiwan, R.O.C. In recent years, several new applications based on resonant tunneling diode (RTD) have been reported [1]-[4]. The negative-differential-resistance (NDR) current-voltage (I-V) characteristics of the RTD devices have several advantages, and they may have high potential as functional devices due to their unique folding I-V characteristics. However these RTD devices are fabricated by the compound semiconductor and process. Therefore this kind of NDR device is difficult to combine with other devices and circuits to achieve the system-on-a-chip (SoC). Therefore, we proposed a new MOS-NDR device that is fully composed of metal-oxide-semiconductor field-effect-transistor (MOS) devices. During suitably arranging the MOS parameters, we can obtain the NDR characteristic in its I-V curve. We call this NDR device as MOS-NDR device. Because this NDR device is consisted of the MOS devices, yet it is much more convenient to combine other devices and circuits to achieve the SoC by standard CMOS process. We will demonstrate a novel voltage-controlled oscillator (VCO) designed based on the MOS-NDR devices and circuit. The VCO is constructed by three cascading low-power MOS-NDR inverters. The basic inverter is composed by two series-connected MOS-NDR devices according to the monostable-bistable transition logic element (MOBILE). The signal will be feedback from the output of the third MOS-NDR inverter to the input of the first MOS-NDR inverter. Under suitable design, we can obtain an oscillator with its frequency proportional to the magnitude of input bias. We design and implement this VCO by the 0.35μm CMOS process. II. MOS-NDR Device and Inverter DESIGN Figure 1 shows a MOS-NDR device which is composed of three NMOS transistors. This circuit can show the Λ-type NDR I-V characteristic by suitably arranging the MOS parameters. Figure 2 shows the I-V curve, measured by Tektronix-370B curve trace, with the width parameters of the three MOS devices as 1μm, 10μm, and 40μm, respectively. The MOS length is fixed at 0.35μm. The VGG is fixed at 3.3V. Fig. 1 The circuit configuration of a Λ-type MOS-NDR device. 2 W1=1u,W2=10u,W3=40u VGG=3.3V 1.6 Current(mA) Abstract--This paper describes the design of a voltage-controlled oscillator (VCO) based on the negative differential resistance (NDR) devices. The NDR devices used in the work is fully composed by the metal-oxidesemiconductor field-effect-transistor (MOS) devices. This MOS-NDR device can exhibit the NDR characteristic in its current-voltage curve by suitably arranging MOS parameters. The VCO is constructed by three low-power MOS-NDR inverter. This novel VCO has a range of operation frequency from 151MHz to 268MHz. It consumes 24.5mW in its central frequency of 260MHz using a 2V power supply. This VCO is fabricated by 0.35μm CMOS process and occupied an area of 120 x 86 μm2. I. Introduction (2) (1) (3) (4) 1.2 0.8 0.4 0 0 0.5 1 1.5 2 2.5 Voltage(V) Fig. 2 The measured I-V characteristic for a MOS-NDR device.. If we connect a NMOS with a MOS-NDR device in parallel, we can modulate the peak current of MOS-NDR device’s I-V curve through the gate voltage of NMOS. The circuit is shown in Fig. 3. The total current ITotal is the sum of the currents through the MOS-NDR and NMOS devices: ITotal=INDR+IMOS. Since IMOS is modulated by the gate voltage (VG), so is ITotal. Our inverter circuit design is based on two series-connected MOS-NDR devices as shown in Fig. 4. This circuit is called as monostable-bistable transition logic element (MOBILE) [5]-[6]. The input node is located at the VG gate. The output node is located between the two MOS-NDR devices. When the bias voltage is smaller than twice the peak voltage (2VP), there is one stable point (monostable) in the series circuit. However when the bias voltage is larger than two peak voltages but smaller than two valley voltages (2VV), there will be two possible stable points (bistable). The intersection point located at the NDR region for two I-V curve will be regarded as unstable point. The two stable points that respect the low and high states (corresponding to “0” and “1”), respectively, are shown in Fig. 5. A small difference between the peak currents of the two devices determines the state of the circuit. If the peak current of the driver device is smaller than that of the load device, the operation point will be located at the high state. On the other hand, if the peak current of the driver device is bigger than that of the load device through the VG voltage, the operation point will be located at the low state. Fig. 5 The bistable states when VS is bigger than2VP but smaller than 2VV. By suitably determining the parameters of devices and circuits, we can obtain the inverter result as shown in Fig. 6. Figure 7 illustrates the relationship between the input bias VS and the power dissipation for a MOS-NDR inverter. The power dissipation of this inverter is 5.75mW using a 2V power supply. Fig. 3 The peak current of MOS-NDR device can be controlled by the VG voltage. Fig. 6 The MOS-NDR inverter result. Fig. 4 The MOS-NDR inverter designed by MOBILE theory. Power Dissipation (mW) 10 8 6 4 2 0 1.6 2 Input Bias (V) 2.4 Fig. 7 The power dissipation for a MOS-NDR inverter. III. VCO DESIGN The circuit configuration of the novel VCO is shown in Fig. 8. The oscillator circuit is composed of three cascading MOS-NDR inverters. The signal will be feedback from the output of the third MOS-NDR inverter to the input of the first MOS-NDR inverter. This VCO is designed and fabricated by the standard 0.35μm CMOS process and occupied an area of 120 x 86 μm2. Figure 9 shows the layout of VCO. Fig. 10 The simulated waveform of the oscillator. Fig. 11 The simulated waveform of the oscillator. Fig. 8 The circuit configuration of the VCO. Frequency (MHz) 280 240 200 160 120 1.6 2 Input Bias (V) 2.4 Fig. 12 The relationship between the input bias and the oscillation frequency. Fig. 9 The layout of VCO. Under suitable parameters design, we can obtain an oscillator with its frequency proportional to the magnitude of input bias VS. Figure 10 shows the simulated waveform of the oscillator under 2V bias voltage by HSPICE program. Figure 11 shows its spectrum. Figure 12 illustrates the relationship between the input bias and the oscillation frequency. This VCO has a range of operation frequency from 151MHz to 268MHz. The relationship between the power dissipation and input bias is shown in Fig. 13. It consumes 24.5mW in its central frequency of 260MHz using a 2V power supply. V. Conclusions Power Dissipation (mW) IV. SIMULATION RESULTS 40 30 20 10 0 1.6 2 Input Bias (V) 2.4 Fig. 13 The power dissipation for a VCO. We have designed a voltage-controlled oscillator (VCO) based on the Λ-type MOS-NDR devices and circuits. The VCO is composed of three cascading low-power MOS-NDR inverters. For chip area die consideration, we use fully MOS devices instead of LC-tank structure. The oscillation frequency is from 151MHz to 268MHz. This VCO is designed by the standard 0.35μm CMOS process. If we fabricated this VCO by 0.18μm CMOS process, the oscillation frequency will be increased above GHz. ACKNOWLEDGMENTS The authors would like to thank the Chip Implementation Center (CIC) of Taiwan for their great effort and assistance in arranging the fabrication of this chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2218- E-168-002. REFERENCES [7] S. Sen, F. Capasso, A. Y. Cho, and D. Sivco, “Resonant tunneling device with multiple negative differential resistance: digital and signal processing applications with reduced circuit complexity,"IEEE Trans. Electron Devices, vol. 34, pp. 2185-2191, 1987. [8] T. H. Kuo, H. C. Lin, R. C. Potter, and D. Shupe,“A novel A/D converter using resonant tunneling diodes,"IEEE J. Solid-State Circuits, vol. 26, pp. 145-149, 1991. [9] L.O. Chua, “Simplicial RTD-Based Cellular Nonlinear Networks“, IEEE Trans. on Cir. and Sys.-I: Fundamental Theo. and Appl., vol.50, no.4, April 2003. [10] Y. Kawano, Y. Ohno, S. Kishimoto, K. Maezawa, T. Mizutani, and K. Sano, “88 GHz dynamic 2:1 frequency divider using resonant tunnelling chaos circuit”, Electron. Lett., vol 39, no. 21, pp. 1546-48, 2003. [11] K. J. Chen, K. Maezawa, and M. Yamamoto, “InP-based high-performance monostable-bistable transition logic elements (MOBILE's) using integrated multiple-input resonant-tunneling devices”, IEEE Electron Device Lett., vol. 17, pp. 127-129, 1996. [12] K. Maezawa, H. Matsuzaki, M. Yamamoto, and T. Otsuji, “High-speed and low-power operation of a resonant tunneling logic gate MOBILE,” IEEE Electron Device Lett., vol. 19, pp. 80-82, 1998. Design and Fabrication of Voltage-Controlled Oscillator by Novel MOS-NDR Device Kwang-Jow Gan, Chung-Chih Hsiao, Shih-Yu Wang, Feng-Chang Chiang, Cher-Shiung Tsai, Yaw-Hwang Chen, Dong-Shong Liang, Chun-Da Tu, Wei-Lun Sun, Chia-Hung Chen, Chun-Ming Wen, Chun-Chieh Liao, Jia-Ming Wu, and Ming-Yi Hsieh Department of Electronic Engineering, Kun Shan University, Tainan County, 710, Taiwan Abstract — This paper describes the design of a voltage-controlled oscillator (VCO) using the novel MOS-NDR devices. The VCO circuit is constructed by cascading a MOS-NDR inverter with the other two common-source amplifiers. This novel VCO has a range of operation frequency from 119MHz to 188MHz under the bias voltage from 1.8V to 2.8V. The fabrication of this VCO is based on the standard 0.35μm CMOS process and occupied an area of 0.03mm2. Index Terms — voltage-controlled oscillator, MOS- NDR inverter, 0.35μm CMOS process. I. INTRODUCTION Negative differential resistance (NDR) devices have attracted considerable attention in recent years because of their potential as functional devices for circuit applications [1]-[3]. Taking full advantages of NDR, we can reduce circuit complexity to enhance circuit performance in terms of operation speed, chip area, and power consumption. However these NDR devices such as resonant tunneling diodes (RTD) are fabricated by the III-V compound semiconductor. Therefore this kind of NDR device is difficult to combine with other devices and circuits to achieve the system-on-a-chip (SoC). In this paper, we proposed a MOS-NDR device that is made of metal-oxide-semiconductor fieldeffect-transistor (MOS) devices. During suitably controlling the MOS parameters, we can obtain the NDR characteristic in its I-V curve. Because this NDR device is consisted of the MOS devices, it is much more convenient to combine other devices and circuits to achieve the SoC by standard Si-based CMOS or BiCMOS process. The voltage-controlled oscillator (VCO) is one of the important components in the phase-locked loop. We will demonstrate a VCO circuit based on the MOS-NDR devices and circuits. This oscillator is composed of an NDR-based inverter, two common source amplifiers and a CMOS inverter. This novel VCO has a range of operation frequency from 119MHz to 188MHz under the bias voltage from 1.8V to 2.8V. The fabrication of this VCO is based on the standard 0.35μm CMOS process and occupied an area of 0.03mm2. II. MOS-NDR INVERTER Our logic circuit design is based on the Monostable-Bistable transition Logic Element (MOBILE) theory [4]-[5]. A MOBILE circuit consists of two MOS-NDR devices connected in series and is driven by a bias voltage VS, as shown in Fig. 1. The MOS-NDR device used in this work that is made of three NMOS and an external bias VGG. It can present a λ-type I-V curve with suitable control the width of each MOS device [6]. The value of Vgg voltage will affect the peak current of the I-V curve of the MOS-NDR device. Fig. 1 Inverter circuit configuration for a MOBILE MOS-NDR device. Figure 2 shows the I-V curve with the VGG voltage varying from 1.5V to 3.3V by a step of 0.3V. The width parameters of the three MOS devices are designed as 1μm, 10μm, and 40μm, respectively. The MOS lengths are all fixed at 0.35μm. The MOS-NDR2 device is connected with a NMOS in parallel. The total current ITotal is the sum of the currents through the NMOS and MOS-NDR2 That is ITotal= IMOS+IMOS-NDR2. Since IMOS is modulated by the Vin voltage, so is ITotal. Therefore, we can modulate the peak current of this combined circuit through the magnitude of the Vin., as shown in Fig. 3. The Vin voltages are varied from 1V to 2.2V with a step 0.3V. 2 W1=1u,W2=10u,W3=40u VGG=1.5V~3.3V Step=0.3V Current(mA) 1.6 1.2 0.8 VGG=3.3V 0.4 VGG=1.5V 0 0 0.5 1 1.5 2 If the peak current of the driver is smaller than the peak current of the load, the circuit switches to the stable point Q corresponding to a high output voltage, as shown in Fig. 4. As shown, the operation point Q is located at the positive differential resistance (PDR) region of the load’s I-V curve. In order to obtain a small peak current in MOS-NDR2 device, we can adjust the value of the Vgg voltage to be smaller than that of MOS-NDR1 device. On the other hand, a bigger peak current in the driver will result in the stable point Q with low output voltage. At this time, the operation point Q is located at the PDR region of the driver’s I-V curve. Thus the state of the circuit is determined by the magnitude of the peak current of the MOS-NDR2 driver device that can be controlled by the Vin voltage. Therefore, an inverter operation can be obtained at the output node between the two MOS-NDR devices. By suitably determining the parameters of devices and circuits, we can obtain one of the inverter logic results as shown in Fig. 5. 2.5 Voltage(V) Fig. 2 The measured results for the MOS-NDR device by modulating the VGG voltage. Fig. 4 MOBILE operation of an inverter circuit. Fig. 3 Modulation of peak current by Vin voltage. The upper MOS-NDR1 device is treated as a load device to the pull-down MOS-NDR2 driver device. When the bias voltage is smaller than twice the peak voltage (2VP), there is one stable point (monostable) in the series circuit. However when the bias voltage is bigger than 2VP, but is smaller than 2VV, there is two possible stable points (bistable) that respect the low and high states (corresponding to “0” and “1”), respectively. A small difference between the peak currents of the series-connected NDR devices determines the state of the circuit. The selection of the stable operating point could be explained by means of the load-line method, as shown in Fig. 4. Fig. 5 Measured result for an inverter circuit. III. VCO Design The CMOS LC-tank oscillators have shown an excellent phase noise performance with low power consumption. However, the large chip area of the inductor L has become critical drawback in LC VCOs. On the other hand, the ring VCOs have the advantages such as the ease of integration with CMOS technique, and the smaller chip area. The oscillation frequency of the ring oscillator is inversely proportional to the number of delay stages N. Employing fewer delay stages can also reduce the power dissipation, and chip area. The oscillator circuit is composed of a NDR inverter, two common source amplifier and a CMOS inverter, as shown in Fig. 6. Fig. 8 The 156MHz oscillation frequency under 2.3V bias voltage. 250 Fig. 6 The circuit configuration of a novel VCO. The VCO signal will be feedback from the output of the second common source amplifier to the input of the MOS-NDR inverter. The output of every circuit is affected by the previous one, and then the last circuit output goes back to affect the first MOS-NDR inverter. There has been transmission delay between every circuit, so that it makes circuit produce oscillation. The novel VCO is fabricated using 0.35μm CMOS technology as shown in Fig. 7. The layout of this VCO is occupied an area of 0.03 mm2. Frequency (MHz) 200 150 100 50 0 1.6 2 2.4 2.8 Voltage (V) Fig. 9 Oscillation frequencies of the VCO. IV. CONCLUSIONS Fig. 7 Layout and PAD of the VCO chip. Figure 8 shows the output waveform with 156MHz oscillation frequency under 2.3V bias voltage. Under suitable design, we can obtain an oscillator with its frequency proportional to the magnitude of input bias. Figure 9 shows the relationship between the input bias and the oscillation frequency. This novel VCO has a range of operation frequency from 119MHz to 188MHz under the bias voltage from 1.8V to 2.8V. We have designed a voltage-controlled-oscillator (VCO) based on the Λ-type NDR-based devices and circuits. The NDR devices are all made of the MOS devices. Therefore, we can fabricate the VCO circuit based on the standard 0.35μm CMOS process. In comparison with the LC-tank VCO structure, this novel MOS-NDR VCO will occupy less chip area. This novel VCO has a range of operation frequency from 119MHz to 188MHz under the bias voltage from 1.8V to 2.8V. Furthermore, even if the integrated circuit is fabricated with transistor dimensions less than 0.35μm technique, the applications based on the MOS-NDR devices and circuits are still useful. Therefore, it can provide some useful applications in the NDR-based circuit and system design. ACKNOWLEDGMENTS The authors would like to thank the Chip Implementation Center (CIC) for their great effort and assistance in arranging the fabrication of this chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC94-2815-C-168-005-E. REFERENCES [1] [2] [3] [4] [5] [6] F. Capasso, S. Sen, A. C. Gossard, A. L. Hutchinson, and J. H. English, “Quantum-well resonant tunneling bipolar transistor operating at room temperature,” IEEE Electron Device Lett., vol. EDL-7, pp. 573-576, 1986. H. Matsuzaki, T. Itoh and M. Yamamoto: “A Novel High-Speed Flip-Flop Circuit Using RTDs and HEMTs”, Proc. IEEE 9th Great Lakes Sympo. on VLSI, Michigan, 1999, pp.154-157. H. L. Chan, S. Mohan, P. Mazumder, and G. I. Haddad, "Compact Multiple Valued Multiplexers Using Negative Differential Resistance Devices," IEEE J Solid-state Circuits, vol. 31, pp. 1151-1156, 1996. K. Maezawa and T. Mizutani, “A new resonant tunneling logic gate employing monostable-bistable transition,” Japan J. Appl.Phys., vol. 32, pt. 2, no. 1A/B, p. L42, Jan. 15, 1993. K. J. Chen, K. Maezawa, and M. Yamamoto, “InP-based high-performance monostable-bistable transition logic elements (MOBILE's) using integrated multiple-input resonant-tunneling devices”, IEEE Electron Device Lett., vol. 17, pp. 127-129, 1996. C. Y. Wu and K. N. Lai, “Integrated λ-type differential negative resistance MOSFET device,” IEEE J. Solid-State Circuits, vol. 14, pp. 1094-1101, 1979. Investigation of MOS-NDR Voltage Controlled Ring Oscillator Fabricated by CMOS Process Kwang-Jow Gan, Dong-Shong Liang*, Chung-Chih Hsiao, Cher-Shiung Tsai, and Yaw-Hwang Chen Abstract- A voltage-controlled ring oscillator (VCO) based on novel MOS-NDR circuit is described. This MOS-NDR circuit is made of metal-oxidesemiconductor field-effect-transistor (MOS) devices that can exhibit the negative differential resistance (NDR) current-voltage characteristic by suitably arranging the MOS parameters. The VCO is constructed by three low-power MOS-NDR inverters. This novel VCO has a range of operation frequency from 38MHz to 162MHz. It consumes 24mW in its central frequency of 118MHz using a 2V power supply. This VCO is fabricated by 0.35μm CMOS process and occupy an area of 0.015 mm2. Our logic circuit design is based on the MOBILE theory. A MOBILE circuit consists of two MOS-NDR devices connected in series and is driven by a bias voltage VS, as shown in Fig. 1. The MOS-NDR device used in this work that is made of three NMOS and an external bias VGG. It can present a λ-type I-V curve with suitable control the width of each MOS device [6]. The value of Vgg voltage will affect the peak current of the I-V curve of the MOS-NDR device. I. Introduction Fig. 1 Inverter circuit configuration for a MOBILE MOS-NDR device. 2 W1=1u,W2=10u,W3=40u VGG=1.5V~3.3V Step=0.3V 1.6 Current(mA) In recent years, several new applications based on resonant tunneling diode (RTD) have been reported [1]-[3]. The negative-differential-resistance (NDR) current-voltage (I-V) characteristics of the RTD devices have several advantages, and they may have high potential as functional devices due to their unique folding I-V characteristics. However these RTD devices are fabricated by the III-V compound material and process. Therefore this kind of NDR device is difficult to combine with other devices and circuits to achieve the system-on-a-chip (SoC). Therefore, we proposed a new NDR device that is totally composed of metal-oxide-semiconductor field-effect-transistor (MOS) devices. We can obtain the NDR characteristic in its I-V curve by suitably arranging the MOS parameters. We call the NDR device to be as MOS-NDR device in this work. Because this NDR device is consisted of the MOS devices, yet it is much more convenient to combine other devices and circuits to achieve the SoC by standard CMOS process. Our novel voltage-controlled ring oscillator (VCO) is constructed by three low-power MOS-NDR inverters. The MOS-NDR inverter logic gate can be obtained by connecting two MOS-NDR devices in series according to the Monostable-Bistable transition Logic Element (MOBILE) theory [4]-[5]. The signal will be feedback from the output of the third MOS-NDR inverter to the input of the first MOS-NDR inverter. Under suitable design, we can obtain an oscillator with its frequency proportional to the magnitude of input bias. We demonstrate this VCO based on the standard 0.35μm CMOS process. 1.2 0.8 VGG=3.3V 0.4 VGG=1.5V 0 0 0.5 1 1.5 2 2.5 Voltage(V) Fig. 2 The measured results for the MOS-NDR device by Kwang-Jow Gan, Dong-Shong Liang*, Chung-Chih Hsiao, modulating the VGG voltage. Cher-Shiung Tsai, and Yaw-Hwang Chen are with the Department of Electronic Engineering, the Kun Shan University, Taiwan, Figure 2 shows the I-V curve with the VGG voltage E-mail: suln@mail.ksut.edu.tw varying from 1.5V to 3.3V by a step of 0.3V. The width II. MOS-NDR DEVICE AND INVERTER parameters of the three MOS devices are designed as 1μm, 10μm, and 40μm, respectively. The MOS lengths are all fixed at 0.35μm. The MOS-NDR2 device is connected with a NMOS in parallel. The total current Itotal is the sum of the currents through the NMOS and MOS-NDR2 That is Itotal= IMOS+IMOS-NDR2. Since IMOS is modulated by the Vin voltage, so is ITotal. Therefore, we can modulate the peak current of this combined circuit through the magnitude of the Vin., as shown in Fig. 3. The Vin voltages are varied from 1V to 2.2V with a step 0.3V. Fig. 4 MOBILE operation of an inverter circuit. Fig. 3 Modulation of peak current by Vin voltage. The upper MOS-NDR1 device is treated as a load device to the pull-down MOS-NDR2 driver device. When the bias voltage is smaller than twice the peak voltage (2VP), there is one stable point (monostable) in the series circuit. However when the bias voltage is bigger than 2VP, but is smaller than 2VV, there is two possible stable points (bistable) that respect the low and high states (corresponding to “0” and “1”), respectively. A small difference between the peak currents of the series-connected NDR devices determines the state of the circuit. The selection of the stable operating point could be explained by means of the load-line method, as shown in Fig. 4. If the peak current of the driver is smaller than the peak current of the load, the circuit switches to the stable point Q corresponding to a high output voltage, as shown in Fig. 4. As shown, the operation point Q is located at the positive differential resistance (PDR) region of the load’s I-V curve. In order to obtain a small peak current in MOS-NDR2 device, we can adjust the value of the Vgg voltage to be smaller than that of MOS-NDR1 device. On the other hand, a bigger peak current in the driver will result in the stable point Q with low output voltage. At this time, the operation point Q is located at the PDR region of the driver’s I-V curve. Thus the state of the circuit is determined by the magnitude of the peak current of the MOS-NDR2 driver device that can be controlled by the Vin voltage. Therefore, an inverter operation can be obtained at the output node between the two MOS-NDR devices. By suitably determining the parameters of devices and circuits, we can obtain one of the inverter logic results as shown in Fig. 5. Fig. 5 Measured result for an inverter circuit. III. VCO DESIGN The CMOS LC-tank oscillators have shown an excellent phase noise performance with low power consumption. However, the large chip area of the inductor L has become critical drawback in LC VCOs. On the other hand, the ring VCOs have the advantages such as the ease of integration with CMOS technique, and the smaller chip area. The oscillation frequency of the ring oscillator is inversely proportional to the number of delay stages N. Employing fewer delay stages can also reduce the power dissipation, and chip area. The circuit configuration of the novel VCO is shown in Fig. 6. The oscillator circuit is composed of three low-power MOS-NDR inverters. Every MOS-NDR inverter is regarded as a delay stage. With the 2V supply voltage, the current consumption is about 12mA. The signal will be feedback from the output of the third MOS-NDR inverter to the input of the first MOS-NDR inverter. The output of every circuit is affected by the previous one, and then the last circuit output goes back to affect the first MOS-NDR inverter. There has been transmission delay between every circuit, so that it makes circuit produce oscillation. The novel VCO is fabricated using 0.35μm CMOS technology as shown in Fig. 7. The layout of this VCO is occupied an area of 0.015 mm2. voltage from 1.6V from 2.4V step 0.2V, the oscillator frequency is increased from 38GHz to 162MHz. IV. CONCLUSIONS We have designed a voltage-controlled-oscillator (VCO) based on the Λ-type NDR-based devices and circuits. The VCO is composed of three cascading low-power MOS-NDR inverters. The NDR devices are all made of the MOS devices. Therefore, we can fabricate the VCO circuit based on the standard 0.35μm CMOS process. In comparison with the LC-tank VCO structure, this novel MOS-NDR VCO will occupy less chip area. This novel VCO has a range of operation frequency from 38MHz to 162MHz. It consumes 24mW in its central frequency of 118MHz using a 2V power supply. Fig. 6 Circuit configuration of the ring VCO. Furthermore, even if the integrated circuit is fabricated with transistor dimensions less than 0.35μm technique, the applications based on the MOS-NDR devices and circuits are still useful. We can predict the oscillation frequency will be capable for GHz operation by the Hspice simulated results from the 0.18μm CMOS process. Therefore, it can provide some useful applications in the NDR-based circuit and system design. ACKNOWLEDGMENTS The authors would like to thank the Chip Implementation Center (CIC) for their great effort and assistance in arranging the fabrication of this chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC94-2215-E-168-001. REFERENCES F. Capasso, S. Sen, A. C. Gossard, A. L. Hutchinson, and J. H. English, “Quantum-well resonant tunneling bipolar transistor operating at room temperature,” IEEE Electron Device Lett., vol. EDL-7, pp. 573-576, 1986. Fig. 7 Layout and PAD of the VCO chip. 200 H. Matsuzaki, T. Itoh and M. Yamamoto: “A Novel High-Speed Flip-Flop Circuit Using RTDs and HEMTs”, Proc. IEEE 9th Great Lakes Sympo. on VLSI, Michigan, 1999, pp.154-157. Frequency (MHz) 160 120 H. L. Chan, S. Mohan, P. Mazumder, and G. I. Haddad, "Compact Multiple Valued Multiplexers Using Negative Differential Resistance Devices," IEEE J Solid-state Circuits, vol. 31, pp. 1151-1156, 1996. 80 40 0 1.6 1.8 2 2.2 2.4 Voltage (V) Fig. 8 The relationship between the oscillation frequency and the bias VS. Figure 8 shows the measured results for the ring VCO circuit. If we changed the value of VS voltage, we can find that the oscillation frequency is proportional to the magnitude of input bias. If we modulate the VS K. Maezawa and T. Mizutani, “A new resonant tunneling logic gate employing monostable-bistable transition,” Japan J. Appl.Phys., vol. 32, pt. 2, no. 1A/B, p. L42, Jan. 15, 1993. K. J. Chen, K. Maezawa, and M. Yamamoto, “InP-based high-performance monostable-bistable transition logic elements (MOBILE's) using integrated multiple-input resonant-tunneling devices”, IEEE Electron Device Lett., vol. 17, pp. 127-129, 1996. C. Y. Wu and K. N. Lai, “Integrated λ-type differential negative resistance MOSFET device,” IEEE J. Solid-State Circuits, vol. 14, pp. 1094-1101, 1979. A NAND Gate Design Based on MOS-NDR Devices and Circuits Implemented by CMOS Technology Kwang-Jow Gan, Chi-Pin Chen, Long-Xian Su, To-Kai Liang, Cher-Shiung Tsai, Yaw-Hwang Chen, Chung-Chih Hsiao, Shih-Yu Wang, and Feng-Chang Chiang Department of Electronic Engineering, Kun Shan University of Technology Abstract In this paper we propose a MOS-NDR device that is A NAND logic circuit design based on the composed of the metal-oxide-semiconductor field- MOS-NDR devices is demonstrated. The MOS-NDR effect-transistor (MOS) devices. Then we design a logic device is consisted of the metal-oxide-semiconductor circuit that can operate the NAND function. We fabricate field-effect-transistor (MOS) devices. This device could the circuit by standard CMOS process. exhibit the negative differential resistance (NDR) II. The MOS-NDR Devices characteristics in the current-voltage curve by suitably Figure 1 shows a MOS-NDR device, which is arranging the parameters of the MOS devices. As a result, composed of three NMOS devices and one PMOS device. we design a NAND logic operation based on MOS-NDR This circuit is derived from a Λ-type topology described devices and circuits which is implemented by the in [5]-[6]. This MOS-NDR device can exhibit various standard CMOS process. NDR current-voltage (I-V) characteristics by choosing Keywords: NAND logic circuit, negative differential appropriate parameters for transistors. resistance (NDR) characteristics, .35μm CMOS process I. Introduction Functional devices and circuits based on negative differential resistance (NDR) devices have generated substantial research interest owing to their unique NDR characteristic [1]-[2]. Taking advantage of the NDR feature, circuit complexity can be greatly reduced and novel circuit applications have also obtained. Some applications make use of the monostablebistable transition logic element (MOBILE) as a highly functional logic gate [3]-[4]. A MOBILE consists of two NDR devices connected in series and is driven by a suitable bias voltage. The voltage at the output node between the two NDR devices could hold on one of the two possible stable states (low and high, corresponding to “0” and “1”), depending on the relative difference of peak current between two devices. Fig. 1 The circuit configuration for a MOS-NDR device. The value of Vgg voltage will affect the I-V curve, especially in its peak current. Figure 2 shows the I-V characteristics, measured by Tektronix 370B, with the Vgg voltage varied from 1.5V to 3.3V, gradually. The width parameters of the gate of the MOS devices are described as mn1=5μm, mn2=100μm, mn3=10μm, and mp4=100μm. The length parameters of four MOS are all MOS-NDR device with parallel connection of a T1 fixed at 0.35μm. The I-V characteristics could be divided NMOS. The total current ITotal is the sum of the currents into three segments as: the first PDR segment, the NDR through segment, and the second PDR segment in sequence. ITotal=INDR+IMOS. Since IMOS is modulated by the gate 3.0 the MOS-NDR and NMOS devices: voltage (T1), so is ITotal. Figure 4 shows the measure I-V Current (mA) characteristics of a MOS-NDR device with T1 varying from 0 to 3.3V, gradually. As seen, we can modulate the 2.0 peak current of the MOS-NDR device through T1 gate. Vgg=3.3V 1st PDR 2nd PDR NDR Vgg=3V 1.0 Vgg=2.5V Vgg=2V Vgg=1.5V 0.0 0 0.4 0.8 1.2 1.6 Voltage (V) Fig. 2 The measured I-V characteristics for a MOS-NDR devices with different Vgg voltages. III. Logic Circuit Design Our logic circuit design is based on the MOBILE Fig. 3 The logic circuit configuration. theory. A MOBILE circuit consists of two NDR devices connected in series and is driven by a bias voltage VS. 6.0 control gate. T2 and T3 are the input square wave signal. The width parameters are all 5μm. The upper NDR1 device is treated as a load device to the pull-down NDR2 driver device. When the bias voltage is smaller than twice the peak voltage (2VP), there is one stable point Current (mA) Figure 3 shows the logic circuit configuration. T1 is the T1 varied from 0V to 3.3V 3.3V 4.0 2.0 3V 2.5V 2V 1.5V 1V 0V (monostable) in the series circuit. However when the bias voltage is larger than two peak voltages but smaller than two valley voltages (2VV), there is two possible stable 0.0 0 points (bistable) that respect the low and high states (corresponding to “0” and “1”), respectively. A small difference between the peak currents of the series-connected NDR devices determines the state of the circuit. The circuit shown at upper left corner of Fig. 4 is a 0.4 0.8 1.2 1.6 Voltage (V) Fig. 4 The measured I-V characteristics with different T1 voltages. If the IP of the driver is smaller than IP of the load (with T2 or T3 OFF), the circuit switches to the stable point Q corresponding to a high output voltage, as shown operation point is dependent on the relative difference of in Fig. 5. The operation point Q is located at the second peak current between two devices. Figure 6 shows the PDR region of the driver’s I-V curve. On the other hand, measured result for the NAND logic circuit. a bigger peak current (with T2 or T3 ON) of the driver will result in the stable point Q with low output voltage. At this time, the operation point Q is located at the second PDR region of the load’s I-V curve. Therefore, an inverter operation can be obtained at the output node between the two NDR devices. Similarly, if we design the logic function according to the MOBILE theory and truth table, we can obtain the NAND gate logic operation. Fig. 6 The measured results for the logic circuit. V. Conclusions We have demonstrated the NAND logic circuit design based on the MOS-NDR devices and circuits. This circuit design is operated according to the principle of MOBILE theory. Our NDR devices and circuits can be fabricated by the technique of Si-based CMOS technique. The fabrication cost will be cheaper than that of resonant tunneling diodes (RTD) implemented by the technique of compound semiconductor. Fig. 5 The MOBILE operation is dependent on the Acknowledgement relative difference of peak current. IV. Experiment Results The authors would like to thank the Chip Implementation Center (CIC) for their great effort and assistance in arranging the fabrication of this chip. This The MOS-NDR devices and logic circuits are fabricated based on the standard 0.35μm CMOS process. The length parameters for all MOS are fixed at 0.35μm. work was supported by the National Science Council of Republic of are the input square wave. The period of T2 is two times of the period of T3 signal. The control T1 gates is 3.3V. The Vgg voltages are 3.3V and 2.5V for NDR1 and NDR2, respectively. We design the logic circuit based on the truth table and MOBILE theory. The selection of the 2. T. H. Kuo, H. C. Lin, R. C. Potter, and D. Shupe, under the contract no. NSC93-2218-E-168-002. The supply voltage VS is fixed at 1.5V. For the NAND gate operation, T2 and T3 gates both China References 1. S. Sen, F. Capasso, A. Y. Cho, and D. Sivco, “ Resonant tunneling device with multiple negative differential resistance: digital and signal processing applications with reduced circuit complexity,"IEEE Trans. Electron Devices, vol. 34, pp. 2185-2191, 1987. “A novel A/D converter using resonant tunneling 3. 4. diodes,"IEEE J. Solid-State Circuits, vol. 26, pp. devices”, IEEE Electron Device Lett., vol. 17, pp. 145-149, 1991. 127-129, 1996. K. Maezawa, H. Matsuzaki, M. Yamamoto, and T. 5. C. Y. Wu and K. N. Lai, “Integrated Λ -type Otsuji, “High-speed and low-power operation of a differential negative resistance MOSFET device,” resonant tunneling logic gate MOBILE,” IEEE IEEE J. Solid-State Circuits, vol. 14, pp. Electron Device Lett., vol. 19, pp. 80-82, 1998. 1094-1101, 1979. K. J. Chen, K. Maezawa, and M. Yamamoto, 6. A. F. Gonzalez and P. Mazumder, “Multiple-valued “InP-based high-performance monostable-bistable signed-digit transition differential-resistance integrated logic elements multiple-input (MOBILE's) using resonant-tunneling adder using devices”, Comput, vol. 47, pp. 947-959, 1998. negative IEEE Trans. Oscillator Design by MOS-NDR Devices and Circuits To-Kai Liang, Kwang-Jow Gan, Cher-Shiung Tsai, Shih-Yu Wang, Chung-Chih Hsiao, Feng-Chang Chiang, Yaw-Hwang Chen, Chi-Pin Chen, and Long-Xian Su Department of Electronic Engineering, Kun Shan University of Technology Abstract amplifier and a CMOS inverter. The measured results The oscillator design based on the negative differential resistance (NDR) devices and circuits is show the oscillator frequency is about 100MHz. II. The MOS-NDR Devices demonstrated. The NDR device is composed of Figure 1 shows a prototype MOS-NDR device, metal-oxide-semiconductor field-effect-transistor (MOS) which is composed of three NMOS transistors and one devices. An inverter logic gate can be obtained by PMOS transistor. This circuit is derived from a Λ-type connecting two MOS-NDR devices in series according to topology described in [4]. The gate of mn3 is connected the element to the output of the inverter formed by mn1 and mn2. It (MOBILE) theory. As a result, we will demonstrate an can exhibit various NDR current-voltage characteristics oscillator circuit with frequency 100MHz. by choosing appropriate values for transistors. monostable-bistable transition logic Vdd Keywords: MOS-NDR device, MOBILE, oscillator circuit I. Introduction Functional devices and circuits based on negative differential resistance (NDR) devices have generated substantial research interest owing to their unique NDR mn1 mn3 characteristics [1]-[3]. The best famous known NDR devices are Esaki diodes and resonant tunneling diodes (RTD). The fabrications of these NDR devices are based on the technique of compound semiconductor. Comparing to Silicon-based integrated circuit, the cost of compound semiconductor is more expensive. However, we proposed a new NDR device that is composed of metal-oxide-semiconductor field-effect-transistor (MOS) devices. We call this device to be as MOS-NDR device. Because this device is totally composed of MOS devices, it is suitable for the process of Si-based CMOS process. First, we demonstrate an inverter circuit based on the MOS-NDR device. Then we design an oscillator circuit composed of a NDR inverter, a common source mn2 Fig. 1 The circuit configuration for a MOS-NDR device. The operation of this MOS-NDR device can be divided into four situations by gradually increasing the bias Vdd. The first situation represents a condition when mn1 is saturated, mn2 is cutoff, mn3 is linear, and mp4 is cutoff. The second situation indicates the case when mn1 is saturated, mn2 is saturation, mn3 is from linear to saturated, and mp4 is cutoff. The third situation refers to the state when mn1 is saturated, mn2 is saturated, mn3 is saturated, and mp4 is cutoff. Finally, the fourth situation corresponds to the state when mn1 is saturated, mn2 is linear, mn3 is cutoff, and mp4 is saturated. Therefore, we Since IMOS is modulated by the gate voltage (VG), so is can obtain various NDR I-V characteristics by choosing ITotal. appropriate values for MOS parameters. III. Inverter Circuit Design Figure 2 shows the measured I-V characteristics by modulating the width of mn1. The other width Our inverter circuit design is based on the MOBILE parameters of the gate of the MOS devices are described theory. A MOBILE circuit consists of two NDR devices as mn2=100μm, mn3=10μm, and mp4=100μm. The connected in series and is driven by a bias voltage VS. length parameters of four MOS are all fixed at 0.35μm. Figure 4 shows the configuration of a MOBILE circuit. The Vgg voltage is fixed at 3.3V. The VG gate is used as the input signal gate. The upper 2.5 NDR1 device is treated as a load device to the pull-down modulation of mn1 10μm 2 Current (mA) NDR2 driver device. When the bias voltage is smaller than twice the peak voltage (2VP), there is one stable point (monostable) in the series circuit. However when 5μm 1.5 the bias voltage is bigger than 2VP, but is smaller than 1μm 1 2VV, there is two possible stable points (bistable) that respect the low and high states (corresponding to “0” and 0.5 0 0 “1”), respectively. A small difference between the peak currents of the series-connected NDR devices determines 0.4 0.8 1.2 1.6 2 Voltage (V) the state of the circuit. VS Fig. 2 The measured I-V characteristics for a MOS-NDR devices by modulating the width of mn1. NDR1 (Load) VG Out NDR2 (Driver) Fig. 4 The circuit configuration of an inverter. Fig. 3 The peak current of MOS-NDR device could be controlled by the VG voltage. If the peak current of the driver is smaller than the peak current of the load, the circuit switches to the stable Figure 3 shows the circuit with the parallel point Q corresponding to a high output voltage, as shown connection of a NMOS and a MOS-NDR device. The in Fig. 5. The operation point Q is located at the second total current ITotal is the sum of the currents through the PDR region of the driver’s I-V curve. In order to obtain a MOS-NDR and NMOS devices: ITotal=INDR+IMOS. small peak current in NDR2 device, we can adjust the IV. Oscillator Design value of the Vgg voltage to be smaller than that of NDR1 device. On the other hand, a bigger peak current in the driver will result in the stable point Q with low output voltage. At this time, the operation point Q is located at the second PDR region of the load’s I-V curve. Thus the The oscillator circuit is composed of a NDR inverter, a common source amplifier and a CMOS inverter, as shown in Figure 7. state of the circuit is determined by the magnitude of the peak current of the NDR2 driver device that can be controlled by the input VG gate. Therefore, an inverter operation can be obtained at the output node between the two NDR devices. Figure 6 shows the simulated result for an inverter logic circuit. The VS is fixed at 1.5V. Fig. 7 The circuit configuration of an oscillator based on the MOS-NDR devices. When the output of NDR inverter rises from VOL to VOH, it makes impact to the next CMOS inverter. The output of voltage will rise from VOH to VOL. The output of INV1 will connect to the common source amplifier. The Fig. 5 The MOBILE operation of an inverter circuit. common source amplifier can provide the 180∘phase reverse. The voltage output will rise form VOL to VOH. As shown in Fig. 7, every circuit output is affected by the previous one, and then the last circuit output goes back to affect the first circuit-NDR inverter. There has been transmission delay between every circuit, so that it makes circuit produce oscillation. We add a CMOS inverter (INV2) in back of the output of the common source amplifier. We design and modulate the kN/kP ratio to a smaller value which makes the output of the common source amplifier obtain a perfect wave. The INV2 can be used as a load as well. Figure 8 shows the measured results of an oscillator. The INV1 and INV2 with number SN74HC04 and Fig. 6 The simulated results of an inverter logic based on MM74HC04, respectively, are used in the measurement. the MOBILE theory. The oscillator frequency is about 100MHz. diodes,"IEEE J. Solid-State Circuits, vol. 26, pp. 145-149, 1991. 3. K. J. Chen, K. Maezawa, and M. Yamamoto, “InP-based high-performance monostable-bistable transition integrated logic elements multiple-input (MOBILE's) using resonant-tunneling devices”, IEEE Electron Device Lett., vol. 17, pp. 127-129, 1996. 4. C. Y. Wu and K. N. Lai, “Integrated Λ -type differential negative resistance MOSFET device,” IEEE J. Solid-State Circuits, vol. 14, pp. 1094-1101, Fig. 8 The measured results of an oscillator. V. Conclusions We have designed and demonstrated an oscillator circuit based on the MOS-NDR devices and MOBILE theory. The circuit configuration is similar to the ring oscillator. We first obtain an inverter output from the MOS-NDR MOBILE circuit. Then we connect the output of the MOS-NDR inverter with a common source amplifier and a CMOS inverter. The result shows that the oscillator frequency is about 100MHz. Acknowledgement The authors would like to thank the Chip Implementation Center (CIC) for their great effort and assistance in arranging the fabrication of this MOS-NDR chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2218-E-168-002. References 1. S. Sen, F. Capasso, A. Y. Cho, and D. Sivco, “Resonant tunneling device with multiple negative differential resistance: digital and signal processing applications with reduced circuit complexity, " IEEE Trans. Electron Devices, vol. 34, pp. 2185-2191, 1987. 2. T. H. Kuo, H. C. Lin, R. C. Potter, and D. Shupe, “A novel A/D converter using resonant tunneling 1979. Frequency Multiplier Using Multiple-Peak MOS-NDR Devices and Circuits To-Kai Liang, Kwang-Jow Gan, Cher-Shiung Tsai, Yaw-Hwang Chen, Chi-Pin Chen, Long-Xian Su, Chung-Chih Hsiao, Feng-Chang Chiang, and Shih-Yu Wang Department of Electronic Engineering, Kun Shan University of Technology Abstract circuit, the fabrication cost of compound semiconductor The design of a frequency multiplier based on the multiple-peak negative differential resistance (NDR) devices and circuits is demonstrated. The NDR device used in this work is composed of metal-oxide-semiconductor field-effect-transistor (MOS) devices. The multiple-peak NDR circuit is obtained by connecting two MOS-NDR devices in series. This circuit can exhibit three stable operating points. These phenomena can be used in a circuit that multiplied the input signal frequency by a factor of three. We design and fabricate the MOS-NDR device and frequency multiplier circuit by the standard 0.35μm CMOS process. Keywords: multiple-peak NDR circuit, MOBILE, oscillator circuit The negative differential resistance (NDR) devices have attracted considerable attention in recent years for potential as functional devices device that is composed of metal-oxide-semiconductor field-effect-transistor (MOS) devices. We call this device as MOS-NDR device. Because this device is totally composed of MOS devices, it is suitable for the process of Si-based CMOS process. In this paper, we demonstrate a two-peak NDR device by combining two discrete MOS-NDR devices in series and obtain two NDR regions with characteristics suitable for applications such as frequency multipliers. II. The Multiple-Peak MOS-NDR Devices Figure 1 shows a MOS-NDR device, which is composed of three NMOS transistors and one PMOS transistor. The gate of mn3 is connected to the output of the inverter formed by mn1 and mn2. This circuit is I. Introduction their is more expensive. However, we proposed a new NDR for circuit derived from a Λ-type topology described in [3]. It can exhibit various NDR I-V characteristics by choosing appropriate values for transistors. Vdd application. There are many applications in signal processing for a device that exhibits more than one NDR region. To obtain the multiple-peak current-voltage (I-V) NDR characteristics, we can connect two or more NDR devices in series or parallel. The best famous known NDR device is the resonant mn1 mn3 tunneling diode (RTD) [1]-[2]. The fabrications of the RTD devices are based on the technique of compound semiconductor. Comparing to Silicon-based integrated mn2 Fig. 1 The circuit configuration for a MOS-NDR device. I-V characteristics. I IT IP IV Therefore, the forward I-V characteristics follow the 1 RT 1 RN breakpoints A, B, C, D and E in order, as shown in Figs. 3(a)-3(f). The combined two-peak I-V characteristics are 1 RP shown in Fig. 4. VV VP VT V Fig. 2 The piecewise-linear PWL approximation of the I-V characteristics of a MOS-NDR device. We use the piecewise-linear (PWL) approximation method to describe the I-V characteristics of a MOS-NDR device. According to the PWL approximation method, we could model the I-V curve of a one-peak NDR device by three segments as shown in Fig. 3 The load-line analysis for two MOS-NDR devices Fig. 2. The electrical parameters are defined as follows: in series. RP is the positive differential resistance, RN is the negative differential resistance, and RT is the resistance of the right-most region. VP is peak voltage, and VV is the valley voltage. IP is the peak current, and IV is the valley current. VT and IT are the corresponding voltage and current at one point on RT segment. We have supposed that R N > R P ≅ R T . When two identical NDR devices are connected in series, there are two peaks and valleys in the combined Fig. 4 The combined I-V characteristics for two I-V characteristics. Their combined I-V characteristics series-connected MOS-NDR devices. can be obtained by the load-line technique as shown in Fig. 3. The upper NDR device is treated as a load device to the pull-down NDR device (driver device). Therefore, the load line is the I-V characteristics of the upper NDR device and represented by the dashed lines in Fig. 3. The I-V characteristics of the driver device are shown by the solid lines. As seen in Fig. 3(a), the load line will move towards the right along the voltage axis as the bias voltage is increased gradually from zero voltage to high voltage. We can obtain the series current for different bias voltage by the stable intersection point of the two We fabricate the MOS-NDR devices by the standard 0.35μm CMOS process. Figure 5 shows the layout of a MOS-NDR device. The width parameters of the gate of the MOS devices are described as mn1=10μm, mn2=100μm, mn3=10μm, and mp4=100μm. The length parameters of four MOS are all fixed at 0.35μm.The bias Vgg is fixed at 3.3V for both two devices. The experimental I-V curve, measured by HP-4155A, for two series-connected MOS-NDR devices connected in series is shown in Fig. 6. As shown, we can obtain two peaks and valleys in the combined I-V characteristics. The measured results points. It can be used for signal processing applications are similar to our load-line analysis as shown in Fig. 4. such as frequency multiplying and waveform scrambling Since absolutely identical NDR devices do not exist in and practice, the peak and valley currents between two multiplication is illustrated in Fig. 7. descrambling. An example of frequency devices are slightly different. Therefore, there are small differences between the measured and analytic results. Fig. 7 Circuit configuration of a frequency multiplier. When a sawtooth waveform was applied to the circuit in Fig. 8(a), the output was also a sawtooth with a Fig. 5 The layout of a MOS-NDR device. frequency three times higher than the input frequency, as 2.5 shown in Fig. 8(b). 2 Load Line (Rout) 1.6 Voltage (V) Current (mA) 2 Q1 1.5 Q2 1 1.2 0.8 0.4 Q3 0 -0.4 0.5 0 0.4 0.8 0.4 0.8 Time (ms) (a) 0 0 1 2 3 0.8 4 Voltage (V) Voltage (V) Fig. 6 The measured I-V results for two MOS-NDR 0.6 0.4 0.2 devices connected in series. 0 -0.4 III. Frequency Multiplier Design 0 Time (ms) (b) The two-peak I-V characteristics can form three Fig. 8 (a) Sawtooth waveform that was applied to the stable states in their positive differential resistance circuit shown in Fig. 7. (b)Output waveform across the segments. Figure 6 shows a 2KΩ load line which resistor. intersects the I-V characteristics to form three stable IV. Conclusions In this work, we have demonstrated the two-peak NDR circuit with two MOS-NDR devices in series. The two-peak I-V characteristics can form three stable states in their positive differential resistance segments. As a chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2218-E-168-002. References 7. “ Resonant tunneling device with multiple result, we demonstrated a frequency multiplier using two negative differential resistance: digital and signal MOS-NDR devices and circuits that can multiply the processing input signal frequency by a factor of three. technique of III-V compound semiconductor. The 8. frequency multiplier circuit by the standard CMOS process. Acknowledgement The authors would like to thank the Chip Implementation Center (CIC) for their great effort and assistance in arranging the fabrication of this MOS-NDR reduced circuit T. H. Kuo, H. C. Lin, R. C. Potter, and D. Shupe, “A novel A/D converter using resonant tunneling diodes,"IEEE J. Solid-State Circuits, vol. 26, pp. than that of Si-based technique. However our NDR Therefore, we can fabricate the MOS-NDR device and with 34, pp. 2185-2191, 1987. fabrication cost of these RTD devices is more expensive devices and circuits are composed of MOS devices. applications complexity,"IEEE Trans. Electron Devices, vol. The conventional NDR devices such as resonant tunneling diodes (RTD) are implemented by the S. Sen, F. Capasso, A. Y. Cho, and D. Sivco, 145-149, 1991. 9. C. Y. Wu and K. N. Lai, “Integrated Λ -type differential negative resistance MOSFET device,” IEEE J. Solid-State Circuits, vol. 14, pp. 1094-1101, 1979. Design and Simulation of Voltage Controlled Oscillator with High Frequency by Negative Differential Resistance Devices and Integrated Circuits To-Kai Liang, Shih-Yu Wang, Kwang-Jow Gan, Cher-Shiung Tsai, Chung-Chih Hsiao, Feng-Chang Chiang, Yaw-Hwang Chen, Chi-Pin Chen, and Long-Xian Su Department of Electronic Engineering, Kun Shan University of Technology Abstract Then we design an oscillator circuit based on the The voltage controlled oscillator (VCO) design 0.35μm CMOS technique. This oscillator is composed of based on the negative differential resistance (NDR) an NDR-based inverter, two common source amplifiers devices and integrated circuits is demonstrated. The and a CMOS inverter. We find that the oscillator oscillator frequency is proportional to the magnitude of frequency is proportional to the magnitude of input bias. input Therefore, it is suitable for the voltage controlled bias. The NDR device is composed of metal-oxide-semiconductor field-effect-transistor (MOS) devices. Therefore, the oscillator can be fabricated by standard CMOS process. We discuss the simulated results under 0.35μm CMOS process. Keywords: voltage controlled oscillator, oscillator (VCO) design. II. The Λ-type NDR Device Figure 1 shows a Λ-type MOS-NDR device, which is composed of three NMOS transistors. This circuit can negative show the a Λ-type I-V characteristic [4]. differential resistance, CMOS process I. Introduction In recent years, several new memory and logic circuits based on negative differential resistance (NDR) devices have been reported [1]-[3]. The current-voltage (I-V) characteristics of the NDR devices have several advantages, and they may have high potential as functional devices due to their unique folding I-V characteristics. These novel devices also might overcome the limits imposed by the complexity of conventional transistor. First, we design an inverter based on the negative differential resistance (NDR) device. We present a NDR device that is composed of metal-oxide-semiconductor field-effect-transistor (MOS) devices. We call this device to be as MOS-NDR device. Because this device is totally composed of MOS devices, it is suitable for the CMOS process. Fig. 1 TheΛ-type NDR device circuit. Figure 2 shows the I-V curve, measured by Tektronix-370B curve trace, with the width parameters of the gate of the MOS devices : WQ1=1μm ; WQ2 =10μm;WQ3 =40μm. Each MOS length is fixed at 0.35μm. The VGG is fixed at 3.3V. to 1.9V. 2 W1=1u,W2=10u,W3=40u VGG=3.3V Current(mA) 1.6 (2) (1) (3) (4) 1.2 0.8 0.4 0 0 0.5 1 1.5 2 2.5 Voltage(V) Fig. 4 The peak current of Λ-type MOS-NDR device Fig. 2 The measured I-V characteristics of a Λ-type could be controlled by the VG voltage. MOS-NDR device. The I-V characteristics of this NDR device could be controlled by the VGG voltage. Figure 3 shows the I-V curve with the VGG voltage varying from 1.5V to 3.3V by a step of 0.3V. 2 W1=1u,W2=10u,W3=40u VGG=1.5V~3.3V Step=0.3V Current(mA) 1.6 1.2 0.8 VGG=3.3V 0.4 VGG=1.5V Fig. 5 The measured I-V characteristics of the Λ-type 0 0 0.5 1 1.5 2 MOS-NDR device with VG varying from 0V to 1.9V. 2.5 Voltage(V) Our inverter circuit design is based on two series-connected MOS-NDR devices as shown in Fig. 6. Fig. 3 The measured I-V characteristics of a NDR device This circuit is called as monostable-bistable transition by varying the VGG voltages. logic element (MOBILE) [5]. The input node is located III. The Inverter based on Λ-type NDR Device at the VG gate. The output node is located between the two MOS-NDR devices. When the bias VS is bigger than Figure 4 shows the circuit with the parallel two peak voltage (2VP), but is smaller than two valley connection of a NMOS and a MOS-NDR device. The voltage (2VV), there is two possible stable points total current ITotal is the sum of the currents through the (bistable) that respect the low and high states MOS-NDR and NMOS devices: ITotal=INDR+IMOS. (corresponding to “0” and “1”), respectively. A small Since IMOS is modulated by the gate voltage (VG), so is difference ITotal. Figure 5 shows the measured I-V characteristics of series-connected NDR devices determines the state of the the Λ-type MOS-NDR device with VG varying from 0V circuit. By suitably determining the parameters of between the peak currents of the devices and circuits, we can obtain the inverter logic gate 360∘reverse for two cascade common source amplifiers. as shown in Fig. 7. When it works to amplify the output of voltage, it makes the next circuit work. The output of every circuit is affected by the previous one, and then the last circuit output goes back to affect the first NDR-based inverter. There has been transmission delay between every circuit, so that it makes circuit produce oscillation. The final stage of the inverter is composed of a CMOS inverter. We design the kN/kP ratio to a smaller value that makes the output of common source amplifiers Fig. 6 The inverter circuit design based on the MOBILE. to obtain a perfect wave. All of the circuit components are made of NMOS devices, so we can design the oscillator according to the standard 0.35μm CMOS process. Figure 9 shows the 0.35μm CMOS layout of the circuit. Fig. 7 The measured result for an inverter circuit. IV. Oscillator Design The oscillator circuit is composed of a NDR inverter, two common source amplifier and a CMOS inverter, as shown in Figure 8. Fig. 9 The layout of the oscillator. Fig. 8 The circuit configuration of an oscillator. When the output of NDR-based inverter rises from VOL to VOH, it makes impact to the common source amplifier. Because it is an invert amplifier, it provides a V. Simulation Results Figure 10 shows the simulated results for the output of every stage. If we changed the value of Vds2 voltage, we can find that the oscillation frequency is proportional to the magnitude of input bias. If we modulate the Vds2 voltage from 1.8V from 3V step 0.2V, the oscillator frequency is increased from 0.65GHz to 1.55GHz. Therefore, this circuit can be regarded as the voltage-controlled-oscillator (VCO). It can provide some useful applications in the NDR-based circuit and system design. Further theory analysis and experiment results are under progressing. It will be discussed elsewhere in the future. circuits according to the standard 0.35μm CMOS process. The oscillator circuit configuration is similar to the ring oscillator. From the simulated analysis, we find that the oscillation frequency is proportional to the magnitude of bias. The oscillator frequency is about 1.55 GHz under 3V. Acknowledgement This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2218-E-168-002. References 10. S. Sen, F. Capasso, A. Y. Cho, and D. Sivco, “ Resonant tunneling device with multiple negative differential resistance: digital and signal processing applications with reduced circuit complexity,"IEEE Trans. Electron Devices, vol. 34, pp. 2185-2191, 1987. 11. T. H. Kuo, H. C. Lin, R. C. Potter, and D. Shupe, “A novel A/D converter using resonant tunneling diodes,"IEEE J. Solid-State Circuits, vol. 26, pp. 145-149, 1991. Fig. 10 The simulated results for the output of every 12. stage shown in Fig. 8. signed-digit adder differential-resistance 1.6 using devices”, negative IEEE Trans. Comput, vol. 47, pp. 947-959, 1998. 1.4 Frequency (GHz) A. F. Gonzalez and P. Mazumder, “Multiple-valued 13. 1.2 C. Y. Wu and K. N. Lai, “Integrated Λ -type differential negative resistance MOSFET device,” IEEE J. Solid-State Circuits, vol. 14, pp. 1 1094-1101, 1979. 0.8 0.6 1.6 14. K. J. Chen, K. Maezawa, and M. Yamamoto, “InP-based high-performance monostable-bistable 2 2.4 2.8 3.2 Voltage (V) Fig. 11 The relationship between the frequency and the bias Vds2. VI. Conclusions We have designed a voltage-controlled-oscillator (VCO) based on the Λ-type NDR-based devices and transition integrated logic elements multiple-input (MOBILE's) using resonant-tunneling devices”, IEEE Electron Device Lett., vol. 17, pp. 127-129, 1996. Four-Valued Memory Circuit Designed by Multiple-Peak MOS-NDR Devices and Circuits Dong-Shong Liang*, Kwang-Jow Gan, Long-Xian Su, Chi-Pin Chen, Chung-Chih Hsiao, Cher-Shiung Tsai, Yaw-Hwang Chen, Shih-Yu Wang, Shun-Huo Kuo, and Feng-Chang Chiang Department of Electronic Engineering, Kun Shan University of Technology, Taiwan, R.O.C. Abstract--This paper describes the design of a four-valued memory cell based on a three-peak MOS-NDR circuit. We connect three MOS-NDR devices in parallel that can create a three-peak current-voltage curve by suitably arranging the parameters. Due to its folding I-V characteristics, multiple -peak NDR device is a very promising device for multiple -valued logic application. This memory cell structure can be easily extended to implement more states in a memory circuit. all fixed at 0.35μm. The VGG is 3.3V. This circuit also can show the Λ-type NDR I-V characteristic as shown in Fig. 3, if we take off the mp4 device from Fig. 1. VDD mn1 I. Introduction The multiple-peak negative differential resistance (NDR) devices offer much promise for multi-valued memory circuits [1]-[4]. Taking full advantages of the NDR device, we can reduce circuit complexity to enhance circuit performance in terms of operation speed, chip area, and power consumption. Previously published memory circuits usually consisted of resonant tunneling diode (RTD) in series. However these RTD devices are fabricated by the III-V compound semiconductor and process. Therefore the RTD devices are difficult to combine with other devices and circuits to achieve the system-on-a-chip (SoC). However the MOS-NDR device used in this work is fully composed of metal-oxide-semiconductor field-effecttransistor (MOS) devices. During suitably arranging the MOS parameters, we can obtain the NDR characteristic in its I-V curve. Therefore, we can implement the MOS-NDR devices and related circuits by Si-based CMOS or BiCMOS process. So it is much more convenient to achieve the SoC. Numerous multiple-peak applications using two or more NDR devices in series or vertical integration create the multiple-peak I-V characteristics. These devices could be treated as independent devices connected in series. However we use three MOS-NDR devices connected in parallel to create the three-peak I-V characteristics. The three-peak MOS-NDR device can have four stable states at its positive differential resistance (PDR) segments during suitable load design. Using this three-peak MOS-NDR device with a constant current source load, a four-state multi-valued memory circuit is demonstrated by the 0.35μm CMOS process. II. the multiple-peak MOS-NDR Device The prototype MOS-NDR device as shown in Fig. 1 is composed of three NMOS transistors and one PMOS transistor. The gate of mn3 is connected to the output of the inverter formed by mn1 and mn2. It can exhibit various NDR current-voltage characteristics by choosing appropriate values for transistors. Figure 2 shows the simulation results by HSPICE program. The width parameters are designed as mn1=5μm, mn2=100μm, mn3=10μm, and mp4=100μm. The length parameters are 3 mp4 VGG mn2 Fig. 1 Circuit configuration of a MOS-NDR device. Fig. 2 Simulated result for the MOS-NDR device by HSPICE program. Numerous multiple-peak applications often use two or more NDR devices in series or vertical integration to create the multiple-peak I-V characteristics. But it is difficult to simulate the series circuit of NDR devices [5]. Therefore, we design the multiple-peak NDR characteristics by connecting the MOS-NDR devices in parallel, as shown in Fig. 4. The MOS-NDR1 and MOS-NDR2 are designed by the N-type devices. The MOS-NDR3 is designed by the Λ-type device. The mn4, mn8 and mn9 devices are used to delay the turn-on voltage of MOS-NDR2 and MOS-NDR3, respectively. The simulated result is shown in Fig. 5. The Vgg voltages are all fixed at 3.3V. R V1 V2 current source V3 V4 Fig. 3 Simulated result for the Λ-type I-V characteristic. Fig. 6 The I-V characteristics for multiple-valued memory circuit using either a resistor load or a constant current source. Fig. 4 Circuit configuration of the multiple-peak MOSNDR device. If a constant current source is used as the load device, the current could be adjusted through appropriate device parameters to a value approximately halfway between the peak and valley current of the multiple-peak NDR device. By comparing to the resistor load configuration, the memory circuit using the constant current source as the load device has two advantages, the better noise margins for the four stable states and the low power dissipation [6]. Fig. 5 Simulated result for the multiple-peak MOS-NDR device. III. MULTIPLE –VALUED MEMORY CIRCUIT DESIGN Due to the folding I-V characteristics, multiple-peak NDR device is a very promising device for multi-valued logic application. The obvious method to bias the multiple-peak NDR device into the multiple stable states is to use a resistor or a constant current source as a load, as shown in Fig. 6, respectively. For the three-peak MOS-NDR device, there are at most four stable states from V1 to V4. When the series resistor exceeds the magnitude of the NDR of the multiple-peak device, the series resistor can be considered as a load resistance which intersects the PDR regions of the multiple-peak device at multiple stable states. When one locates the stable points, one always locates the points of intersection on the PDR region because all the intersection points on the NDR region are always unstable. Fig. 7 The four-valued MOS-NDR memory circuit. Figure 7 shows the four-state MOS-NDR memory circuit. The voltage to be stored is provided at the cell input and loaded by momentarily enabling the write clock. The input voltage then controls the voltage at the multiple-peak MOS-NDR node. When the write clock is disabled, the voltage across the MOS-NDR device adjusts to the nearest stable operating point, thereby storing the input value at one of the four discrete levels. A saw-tooth wave is applied to the input of the memory circuit with amplitude of 3.5V. A square wave is then applied to the write gate input which alternately turns the MOS on and off. The output of this memory circuit gives the four stable memory values, as shown in Fig. 8. We can understand the attractive feature for this memory circuit for the constant-current-source load is that the noise margin would remain the same with a further increase in the number of current peaks of the multiple-peak MOS-NDR device. The design of this memory circuit is demonstrated by the standard 0.35μm CMOS process. We design the constant current source by the MOS devices. The authors would like to thank the Chip Implementation Center (CIC) of Taiwan for their great effort and assistance in arranging the fabrication of this chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2218- E-168-002. References Fig. 8 Simulated results for the MOS-NDR memory circuit. V. CONCLUSIONS In this work, we have demonstrated the three-peak MOS-NDR circuit with three MOS-NDR devices connected in parallel. The series combination of this device with a constant-current-source load is used to demonstrate the operation of a four stable state memory cell. Because all of the devices used in this circuit are fully composed of MOS devices, we can fabricate this memory circuit by the standard Si-based CMOS or BiCMOS process. Furthermore, this MOS-NDR device and circuit will be easy to integrate with other device and circuit to achieve the goal of SoC. Acknowledgments [13] S. Sen, F. Capasso, A. Y. Cho, and D. Sivco, “Resonant tunneling device with multiple negative differential resistance: digital and signal processing applications with reduced circuit complexity,"IEEE Trans. Electron Devices, vol. 34, pp. 2185-2191, 1987. [14] J. P. A. van der Wagt, H. Tang, T. P. E. Broekaert, A. C. Seabaugh, and Y. C. Kao, ”Multibit resonant tunneling diode SRAM cell based on slew-rate addressing,” IEEE Trans. Electron Devices, vol. 46, pp. 55-62, 1999. [15] S. J. Wei and H. C. Lin, “Multivalued SRAM cell using resonant tunneling diodes,” IEEE J. Solid-St. Circuits, vol. 27, pp. 212-216, 1992. [16] A. C. Seabaugh, Y. C. Kao, and H. T. Yuan, ”Nine-state resonant tunneling diode memory,” IEEE Electron Device Lett., vol. 13, pp. 479-481, 1992. [17] K. J. Gan“Hysteresis phenomena for the series circuit of two identical negative differential resistance devices,"Japanese Journal of Applied Physics, Vol. 40, No. 4A, pp. 2159-2164, 2001. [18] Z. X. Yan and M. J. Deen, ”A new resonant-tunnel diode-based multivalued memory circuit using a MESFET depletion load,” IEEE J. Solid-State Circuits, vol. 27, pp. 1198-1202, 1992. An oscillator design based on MOS differential amplifier by simulation Cher-Shiung Tsai, Ming-Yi Hsieh, Jia-Ming Wu, Chun-Chieh Liao, Jeng-Lung Wu, Kwang-Jow Gan, Yaw-Hwang Chen, Dong-Shong Liang, Chia-Hung Chen and Chung-Chih Hsiao Department of Electronic Engineering, Kun Shan University, Tainan, Taiwan 710, R.O.C. Abstract — In this thesis, we present an oscillator mainly composed by a MOS differential amplifier. The simulations use H-spice to verify the differential amplifier oscillator under CIC 0.35μm Si-Ge process parameters. We have used discrete devices on bread board to prove such circuit is an oscillator circuit successfully [1-2]. Simulations show such an oscillator can work stably from 0.8 volts to 3.3 volts supply voltage. When supply voltage is close to 3.3 volts, the output frequency will be more than 1.4 GHz. The differential amplifier oscillator can start oscillating at low voltage when supply voltage is only 0.8 volts and output frequency is about 32 MHz. We use FFT (Fast Fourier Transform) diagram to analyze the oscillator and shows the oscillator is with low noise characteristic. Finally, those simulation results reveal that the oscillator is not only an excellent voltage controlled oscillator (VCO) but also low power consumption. Keyword — Differential amplifier, VCO, FFT. I. INTRODUCTION We use the high input resistance, high output resistance and high voltage gain characteristics of MOS differential amplifier [3-5] to create an oscillator. Such oscillator is based on differential amplifier have two outputs, one output is high voltage state and the other will be in low voltage state. We connect one CMOS inverter and two CMOS inverters after two different outputs. It is an asymmetric structure and most output waveform tends to be square waveform. The oscillator frequencies are decided by PMOS NMOS resistors, MOS transistors and CMOS inverters. In this thesis, we present a different type oscillator and use simulation results to prove such oscillator is useful, easiness and flexibility in design. saturation and the other is off. Owing to we can’t fabricate two identical transistors M1 and M2, so most conditions are M1 saturation and M2 off, or M1 off and M2 saturation. In Fig.1 we assume M1 saturation and M2 off, so voltage OP1 is in low state and voltage OP2 is in high state. In the meanwhile, voltage G1 is in high state and voltage G2 is low state. After CMOS inverter (INV1) time delay, voltage G2 becomes high state to turn on transistor M2, so voltage OP2 changes into low state. After CMOS inverters (INV2, INV3) time delay, voltage G1 changes into low state to turn off transistor M1. Base on same analysis, M2 will be off and M1 will be saturation in the next run. After a fixed period, M1 and M2 will toggle their states. Such on/off continuous switching phenomena will cause oscillation. The nodes or transistors status in Fig.1 shows in Table.1. Vcc PM1 CMOS INV1 H/L PM1 CMOS INV2 OP1 L/H OUTPUT OP2 H/L G2 G1 M1 CMOS INV3 M2 *L/H H/L H/L NM1 * : Change State Fig.1 The MOS differential amplifier oscillator. II. CIRCUIT THEOREM The oscillator is composed of MOS differential amplifier by adding three CMOS inverters as shown in Fig.1. A formal differential amplifier needs a constant current source but we use resistor NM1 to replace the constant current source for simplicity. According to differential amplifier operation, transistor M1 and M2 can’t be both off or both in triode state in the same time. Transistors M1, M2 will both in saturation state or one is M1 M2 OP1 OP2 G1 G2 State1 Sat Off L H H L State2 Off Sat H L L H Table1. Status of transistors and nodes of Fig.1. III. SIMULATION RESULTS In this thesis, we use H-Spice and CIC 0.35um Si-Ge process parameter to run simulations. The discrete devices are MOS transistors M1 (M2), resistors PM1, NM1 and CMOS inverters. The channel length of all MOS devices in simulations is fixed and minimum value (l=0.35μm). The output waveform under 0.8 volts supply voltage and oscillation frequency is 32 MHz as shown in Fig.2. M1 (or M2) channel width is 15μm. Resistor PM1 and NM1 are PMOS and NMOS respectively, their gates are connected to ground or VDD to become resistors. In Fig.1, the channel width of PM1 and NM1 are 10μm, 40μm respectively. other signals. If the main signal is 1.5 volts then the most others signals shall be smaller than 0.06 volts. Most conditions are noise signals will become larger as output frequency increases in the MOS differential amplifier oscillator. Fig.4 Typical FFT diagram of the MOS differential amplifier oscillator. Fig.5 shows resistor NM1 (width=40μm) and transistor Q1 (Q2) unchanged but PM1 channel width is changed from 4μm to 10μm. Fig.5 reveals larger channel width of PM1 will have higher output frequency because larger channel width has smaller resistance. PM1 1600 Fig.2 Output waveform under 0.8 volts. 1200 Fig.3 shows the oscillator with same condition under 3.3 volts supply voltage and its output frequency is 1.4 GHz. 10u MHZ 4u 800 400 0 1 1.5 2 2.5 3 3.5 V Fig.5 Oscillator output frequencies under different width of PM1 and supply voltages. Fig.3 Output waveform under 3.3 volts. Fig.4 is the typical fast Fourier transform (FFT) diagram of the MOS differential amplifier oscillator. Fig.4 shows the oscillator with low noise characteristics. The highest signal is more than the most other signals about 28 db in Fig.4. It means the main oscillation signal is twenty-five (1028/20=25.1) times stronger than the most Fig.6 shows resistor PM1 (width=10μm) and transistor M1 (M2) unchanged but NM1 is changed from 20μm to 40μm. Fig.8 also reveals larger channel width of NM1 will have higher output frequency because larger channel width has resistance. M1 (M2) in Fig.4 or Fig.5 is the same as Fig.2 which the channel width of M1 (M2) is 15μm. Fig.7 shows resistors PM1 (width=10μm) and NM1 (width=40μm) unchanged but transistor M1 (or M2) channel width is changed. The channel widths of transistor M1 (or M2) are 7μm and 15μm respectively. VCO NM1 1600 1600 1200 1200 40u 20u MHZ MHZ 800 800 400 400 0 0.5 0 1 1.5 2 2.5 3 3.5 V Fig.6 Oscillator output frequencies under different width of NM1 and supply voltages. 1 1.5 2 2.5 3 3.5 V Fig.8 Voltage controlled oscillator (VCO) characteristics of Fig.1. Pdd 25 The effect of M1 (or M2) is the same as those effects of resistors PM1 and NM1 on output frequency response. The larger channel width transistor has higher output frequency because of higher gm value. 20 M1(M2) 15 1600 mW 10 1200 15u 5 7u MHZ 800 0 0.5 1 1.5 2 2.5 3 3.5 V Fig.9 Oscillator power dissipation under different output voltages. 400 0 0.8 1.2 1.6 2 2.4 2.8 3.2 V Fig.7 Oscillator output frequencies under different M1 (M2) and supply voltages. From Fig.5 to Fig.7, the oscillation frequency increases as supply voltage increases. It shows the MOS differential amplifier oscillator is also a voltage controlled oscillator (VCO). The MOS differential amplifier oscillator shows excellent VCO linearity from 0.8 volts to 3.3 volts supply voltage as shown in Fig.8. We will take use of the nice VCO linearity in our next new phase lock loop (PLL) project. Fig.9 shows the power dissipation of MOS differential oscillator under different output frequencies. It’s only 10μW under 0.8 volts supply voltage and output frequency is 32 MHz. While supply voltage is 3.3 volts, its output frequency is 1.4 GHz but power dissipation will be only 22.5mW. The power dissipation is proportional to the square of output voltages in Fig.9. It is consistent with CMOS devices power dissipation PDD = fCV2DD [6]. Fig.10 shows the layout of the MOS differential amplifier oscillator with 0.35μm Si-Ge process. We have applied CIC process successfully and will receive those chips in this September. The applied CIC chip includes three different sizes to get 500MHz, 700MHz and 900MHz output frequencies under 2.0 volts supply voltage respectively. Fig.10 Layout of MOS differential amplifier oscillator. IV. CONCLUSIONS ACKNOWLEDGEMENTS The MOS differential amplifier oscillator generates square wave not the same as traditional oscillators, such as quarts oscillator or ring oscillator can generate sinusoidal waves. It is the same as general oscillators that the noises of MOS differential amplifier oscillator are proportional to output frequencies. The MOS differential amplifier oscillator still has three advantages of good linearity voltage controlled (VCO) characteristic, low noise and low power dissipation characteristics. In our simulations, increase those channel widths of PM1, NM1 or M1 (M2) values can increase output frequency. As we know, VCO is the leading role of phase lock loop (PLL) circuit. Our new project is phase lock loop circuits, the MOS differential amplifier oscillator will be the first priority in our PLL VCO block. If we implement those MOS differential amplifier oscillators into IC chips in this September, then we will measure output frequencies and power dissipations under different supply voltages. We will also get FFT (Fast Fourier Transform) diagram to analyze the noise characteristics. When we have those measurement data, then we will compare those differences between simulations and experiments. Finally, we will find out the limitations of the H-spice on frequency domain, power dissipation and noise characters. The authors would like to thank the National Science Council of Republic of China for their generous and kind support. This work was supported by the National Science Council of Republic of China under the contract no. NSC93-2218-E-168-002. REFERENCES [1] [2] [3] [4] [5] Cher-Shiung Tsai, Chun-Chieh Liao, Jia- Ming Wu, Ming-Yi Hsieh, Kwang-Jow Gan, Dong-Shong Liang, Yaw-Hwang Chen, Chia-Hung Chen, and To-Kai Liang, “An Oscillator Design Based on BJT Differential Amplifier”, 第三屆微電子技術發展與應用研討會, 國 立高雄海洋科技大學, 94/5/20, 2005. Cher-Shiung Tsai, Jia-Ming Wu, Ming-Yi Hsieh, Chun-Chieh Liao, Tien-Hung Chang, Kwang-Jow Gan, Dong-Shong Liang, Yaw-Hwang Chen, Chia-Hung Chen, Chun-Ming Wen, “A MOS DIFFERENT- IAL AMPLIFIER OSCILLATOR”, 16th VLSI Design/CAD Symposium, 花蓮美崙大飯店, 94/8/9~12, 2005. Richard C. Jaeger and Travis N. Blalock, “Microelectronic Circuit Design,” 2nd edition, pp. 1087-1108, 2003. Adel S. Sedra and Kenneth C. Smith, “Microelectronic Circuits,” 5th edition, pp. 687-719, 2004. Randall L. Geiger, Phillip E. Allen and Noel R. Strader, “VLSI Design Techniques for Analog and Digital Circuits,” pp. 431-454, 1990. Adel S. Sedra and Kenneth C. Smith, “Microelectronic Circuits,” 5 Frequency Divider Design Using Negative Differential Resistance Device Kwang-Jow Gan, Chun-Da Tu, Chi-Pin Chen, Yaw-Hwang Chen, Cher-Shiung Tsai, Dong-Shong Liang, Wei-Lun Sun, Chia-Hung Chen, Chun-Ming Wen, Chung-Chih Hsiao, Shih-Yu Wang, and Feng-Chang Chiang Department of Electronic Engineering, Kun Shan University, Tainan County, 710, Taiwan Abstract — This paper demonstrate a chaos generator using a R-BJT-NDR device. We utilize two bipolar junction transistors and four resistors to construct the R-BJT-NDR device which can show the negative differential resistance (NDR) characteristic in its current-voltage curve by suitably modulating the resistances. We will demonstrate a frequency divider using the chaos generator. This circuit is consisted of a R-BJT-NDR device, an inductor, and a capacitor. We investigate the effects of the input frequency and the bias voltage on the operation. Index Terms — chaos generator, R-BJT-NDR device, frequency divider. I. INTRODUCTION The negative differential resistance (NDR) devices are attractive for high-frequency applications and high-speed logic applications [1]-[3]. One of the most important features of the NDR devices is their strong nonlinearity. A circuit consisting of such nonlinear characteristic often shows the chaos phenomenon. In this paper, we will design a chaos generator circuit using an R-BJT-NDR device and demonstrate its application to a dynamic frequency divider. The frequency divider is simply made of a R-BJT-NDR device, an inductor, and a capacitor. The R-BJT-NDR device is composed of two bipolar junction transistors and four resistors. During suitably arranging the resistances, we can obtain the NDR characteristic in its current-voltage (I-V) curve. In comparison with the traditional NDR devices such as resonant tunneling diode (RTD), the fabrication of R-BJT-NDR device will be much cheaper and easier than RTD. The applications using nonlinear circuits using the chaotic behavior have been studied intensively in the field of information processing and communication systems [4]. In this work, we investigate the operating margins of this circuit with respect to input signal frequency and the bias voltage. II. R-BJT-NDR DEVICE We proposed a NDR device that is composed of the bipolar junction transistors (BTJ) and resistors (R) devices. We call this device as R-BJT-NDR device. Because this device is fully composed of the BJT and R, it is suitable for the process of Si-based CMOS or BiCMOS technique. This R-BJT-NDR device is composed of two n-p-n transistors and five resistors, as shown in Fig. 1. During suitably arranging the values of the five resistors, we can obtain the I-V curve with NDR characteristics. Figure 2 shows the simulated I-V characteristics by Hspice program. The electrical parameters are R1=4.5kΩ, R2=5.3kΩ, R3=100kΩ, R4=1.2kΩ, and R5=6kΩ. VS R1 R2 R4 R3 Q1 Q2 R5 Fig. 1 NDR device is composed of two transistors and five resistances. Fig. 4 Simulated waveforms of Vout in a chaotic state. Fig. 2 Simulated I-V characteristics by Hspice program. III. CHAOS CIRCUIT Figure 3 shows the configuration of a frequency divider. This frequency divider consists of a R-BJT-NDR device, an inductor, and a capacitor. This is a kind of van der Pol oscillator having an input terminal. The differential equations determining the system’s behavior are expressed as di L (t ) Vin (t ) − Vout (t ) = dt L (1) dVout (t ) iL (t ) − i NDR (Vout ) = dt Cout (2) We can also observe the long-period behavior in some regions. This is called the bifurcation phenomenon. In the bifurcation region, the system’s output period is the integer-multiple of the input period. It should be noted that the operation frequency range depends on the LC characteristic frequency, determined as (1/2π LC ). Therefore, a frequency range close to the cutoff frequency of the R-BJT-NDR device is expected if the appropriate values are chosen for L and C. Figure 5 shows the circuit’s output period as a function of the input frequency. Figures 5(a)-(b) illustrate the waveforms obtained from the simulation for 2T and 3T under input frequency 1.5MHz and 1.8MHz, respectively. The waveforms show the results for 1/2 and 1/3 frequency dividing states, respectively. (a) When an external oscillating signal, Vin=Vo+ Asin(2πft), is applied, this circuit will generate the chaos phenomenon in such nonlinear system. Where Vo, 2A, and f are input bias, amplitude and frequency, respectively. This circuit can output various types of signal patterns including chaos, as shown in Fig. 4. (b) Fig. 5 (a) a 1/2 frequency dividing state, (b) a 1/3 frequency dividing state. Fig. 3 Configuration of the chaos generator circuit. The frequency dividing operation is explained by the relationship between the charging time of the capacitance Cout, and the input period. At low input frequency, charges are supplied to the output capacitance during a cycle of the input. That will be sufficient to switch the R-BJT-NDR device. However, when the input frequency increases, the charges supplied to the capacitance during a cycle decrease due to the shorter period and the increased impedance of the inductor. Then, the amount of charges supplied to the capacitance is not sufficient to switch the R-BJT-NDR device. Therefore, two or more cycles are necessary to switch the MOS-BJT-NDR device. This causes the frequency dividing operations of 1/2, 1/3, and so on. The phenomena are called period-adding bifurcation. Figure 6 shows the circuit’s output period as a function of the input frequency. The output period increases discontinuously with increasing frequency. There will be a chaos phenomenon existed at the transition. Fig. 6 The output period will be a function of input signal. The magnitude of the input bias will affect the results of frequency divider. The values of L, C, and f are fixed at 1mH, 10pF, and 1.6MHz, respectively. Figure 7 shows the dividing ratio as a function of the bias voltage. As seen, stable 1/2, 1/3, and 1/4 dividing operations can be obtained in the simulation results. The results demonstrate the dividing ratio can be selected by changing the input bias. Figure 8 illustrates the bias regions for the frequency divider. Black areas denote the chaos regions. Fig. 8 The dividing ratio as a function of the input bias. IV. CONCLUSIONS We have proposed a NDR-based chaos circuit and its application to a novel frequency divider. This chaos circuit is based on the strong nonlinearity of an R-BJT-NDR device. We have studied the operation of the frequency divider circuit with respect to the input signal frequency and input bias. The frequency dividing operation is based on the bifurcation phenomenon which appears in the chaotic system. The dividing ration can be selected by changing the input frequency and input bias. ACKNOWLEDGMENTS (a) The authors would like to thank the Chip Implementation Center (CIC) for their great effort and assistance in arranging the fabrication of this chip. This work was supported by the National Science Council of Republic of China under the contract no. NSC94-2215-E-168-001. References [7] (b) (c) Fig. 7 (a) a 1/2 frequency dividing state, (b) a 1/3 frequency dividing state, (b) a 1/4 frequency dividing state. K. Maezawa, H. Matsuzaki, M. Yamamoto, and T. Otsuji, “High-speed and low-power operation of a resonant tunneling logic gate MOBILE,” IEEE Electron Device Lett., vol. 19, pp. 80-82, 1998. [8] A. F. Gonzalez and P. Mazumder, “Multiple-valued signed-digit adder using negative differential-resistance devices”, IEEE Trans. Comput, vol. 47, pp. 947-959, 1998. [9] H. L. Chan, S. Mohan, P. Mazumder, and G. I. Haddad, "Compact Multiple Valued Multiplexers Using Negative Differential Resistance Devices," IEEE J Solid-state Circuits, vol. 31, pp. 1151-1156, 1996. [10] M. J. Ogorzalek, Chaos & Complexity in Nonlinear Electron Circuit, World Scientific, 1997.