A STUDY ON CMOS NEGATIVE RESISTANCE CIRCUITS Vishal Patel, R. Raut Concordia University - ECE Department EV 5.139, 1515 St.

Catherine West Montreal, Quebec H4G 2W1 Email: rabinr@ece.concordia.ca

ABSTRACT An in-depth study of CMOS transconductor designed negative resistance circuits is presented. Important large signal and small signal characteristics including noise, stability and bandwidth are investigated. A strategy of designing large bandwidth active resistors is proposed with supporting analysis. Key stability issues that have not previously been reported are discussed. Finally, applications which include the design of a low phase noise current mode oscillator and a high bandwidth inverting amplifier are demonstrated with simulation results. Figure 1 Grounded negative resistor Index Terms CMOS Transconductor, negative resistance circuits, stability, oscillator, amplifier 1. INTRODUCTION Active resistors have become increasingly popular in VLSI and integrated circuit technology. A popular approach used is the conversion of a transconductance device to a driving-point impedance, via connection of the input and output terminals. This design technique results in active resistors whose values are controlled by the DC power supplies; a feature that permits the design of analog negative resistance circuits. Negative resistors have the potential of being a key building block for larger electronic systems in VLSI technology. The tunnel diode was the first ever negative resistance device discovered in the 1950s, after which numerous alternate and improved circuits have been reported [1]-[3]. Extensive research in this field is a result of applications in various areas of CMOS and VLSI technology [3]-[6], with the most common being the correction of non-ideal behavior of an electronic system caused by parasitic impedances. No work on detailed characterization of active negative resistance circuits in CMOS technology seems to have appeared in the current literature. This paper attempts to fill this void by reporting results on several negative resistance circuits that have been designed in 0.5 micron CMOS technology. The large and small signal characteristics such as linearity and stability are investigated. Applications, which include the design of a low phase noise oscillator and bandwidth extension of an inverting amplifier, are analyzed. 2. CHARACTERTICS OF THE NEGATIVE RESISTORS A total of six negative resistors were investigated (three grounded and three floating), with all having comparable characteristics. The 978-1-4244-1643-1/08/$25.00 2008 IEEE results for two of these resistors are presented in this paper. 2.1. Large Signal Characteristics Grounded negative resistor Figure 1 presents the schematic of the grounded negative resistor [6]. The resistor is formed using a transconductance amplifier that is designed with two cross coupled differential pairs having identical MOS devices. Equation (23) in [6] expresses that the resistance (Zin) is negative as long as the control voltage Vr is less than Vc. Table 1 summarizes some of the important large signal characteristics of the grounded negative resistor that was designed in a 0.5 micron CMOS process with transistors of W = 2.5 m, L = 1m and a power supply rail of 1.5 V. Table 1 Large signal characteristics of grounded negative resistor Range of resistance (0.6 < Vr < 1.4) Linearity (function of Vr) Noise: R = -20 K Power consumption: R = -20 K -8 K to -250 K 5th order polynomial 3.6e-15 sq. V/Hz 3.73e-4 Watts

Of the three grounded negative resistors studied, this one has the widest resistance range, with the resistor of [7] having the least power consumption and noise, and [3] having the best linearity. Floating Negative Resistance Figure 2 presents the schematic of the floating negative resistor that has been designed with a wide band transconductance amplifier [8]. The resistance is made negative as long as control voltage Vc1 is less than Vc2. Table 2 presents some of the

001283

C gs

C gb

g p1v2

Rds

C db

C db

Rds

g p 3v1

C gb

C gs

C gs

C gb

g p 2 v2 Rds

C db

C db

Rds g p 4 v1

C gb

C gs

C gd
v2

v1

v2

C gd

v2

C gd

v1

C gs

C gb

g n1v2

Rds

C db

C gs

C gb

g n2 v 2 Rds

C db

Figure 2 Floating negative resistor important large signal characteristics that have been investigated. This resistor has the best linearity, with [6] having a very wide resistance range but the most noise, and [4] having the least power consumption. Table 2 Large signal characteristics of floating negative resistor Range of resistance (0.9 < Vc1 < 1.4) Linearity (function of Vc1) Noise: R = -10 K Power consumption: R = -10 K 2.2. Small Signal Characteristics The AC equivalent model of the grounded negative resistor is developed and presented in Figure 3. This model includes the transistors parasitic capacitances that influence the response at higher frequencies. The nodal admittance matrix is derived and given as
sB + G1 1 sB21 sB v1 i1 g p4v1 gn2v2 12 = g v g v g v sB2 + G2 v2 p3 1 p1 2 n1 2

Figure 3 AC model of grounded negative resistor frequency than the others, this pole will have the most influence on the frequency response, and is labeled as the dominant pole. In the presence of a dominant pole the bandwidth can be estimated by [9]
BW 1 1 + 1 + .. 1 2 1 + 1 + .. 1 p2 z2 2 2 2 2 p2 pn z2 zn 1 1

..(3)

-8 K to -11 K 2nd order polynomial 2.5e-16 sq. V/Hz 9.6e-4 Watts

The grounded negative resistor of Figure 1 and the negative resistor of [3] were both found to have a dominant pole. This dominant pole can be used to approximate a single-pole model of the resistor. The resistor whose frequency response is presented in Figure 4 has a dominant pole located at 2.5E8 rad/s, with the bandwidth measured to be 2.7E8 rad/s. In an attempt to optimize the bandwidth, the effect of transforming a transconductance device into an active resistor was explored. An experiment was carried out on a CMOS inverter, which could be employed as a simple transconductance amplifier. The nodal admittance matrix of the CMOS inverter is derived as
sB 1 + G 1 sB o 1 + g p 1 + g n1 sB o + G o sB 1 o v1 v o i1 = i o

..(1)

..(4)

The details regarding the matrix elements are described in Table 3. The expression for the resistance is evaluated as
R= v1 i1 = s B2 + G2 + g p1 + gn1 B1(g p1 + gn1 + G2 ) + B2 (G1 + g p4 ) (G1 + g p4 )(g p1 + gn1 + G2 ) + s 2 (B1B2 B12B21) + s B21(g p2 + gn2 ) B12g p3 g p3 (g p2 + gn2 )

..(2)

The resistance value at low frequency and the phase characteristics can be identified from (2). Figure 4 presents the simulation result relating the magnitude and phase of the resistor as a function of frequency. All of the circuits that were investigated have a resistance expression that are of the form given in (2), and a similar frequency response to that of Figure 4. Bandwidth Investigation The bandwidth of the resistors is dependent on the distribution of the poles and zeros. If one of the poles is at a much lower Table 3 Admittance matrix elements
B 1 = C gsp 4 + C gbp 4 + C dbp 4 + C gdp 2 + C dbp 2 + C dbn 2 + C gdn 2 + C gsp 3 + C gdp 3 + C gbp 3 ; B 2 = C dbp 1 + C gsp 1 + C gbp 1 + C gsn 1 + C gbn 1 + C dbn 1 + C gdp 3 + C dbp 3 + C gsp 2 + C gbp 2 + C gdp 2 + C gsn 2 + C gbn 2 + C gdn 2 B 12 = B 21 = C gdp
3

The transconductance of the inverter is determined as element Y21 of the nodal admittance matrix. Shorting the input and output terminals results in all the elements of the matrix being added together, causing the real and imaginary parts of the complex expression to become larger. This translates to smaller pole and zero values; hence a lower bandwidth. This experiment was repeated for the grounded negative resistor with the simulation results presented in Figure 5. The poles and zeros of the transconductor are found to be located at higher frequencies than those of the resistor, with Figure 5 validating the fact that the bandwidth of the resistor is smaller. We have come to recognize that we encounter a decrease in the bandwidth when a transconductance device is transformed to a driving-point impedance. We can conclude that the reduction in the bandwidth of the active resistor is a result of the parasitic capacitances at the input and output terminals of the transconductor adding up. Therefore, for the design of a large bandwidth active resistor, it is necessary to use a high bandwidth transconductance amplifier. Stability The expressions of the negative resistances as a function of frequency are of the form

C gdp 2 C gdn 2 ; G 1 = g dp 2 + G 2 = g dp 1 + g dp 3 + g dn 1

g dp 4 + g dn 2 ;

001284

12 10 Magnitude (abs) 8

x 10

Bode Diagram 100 80 60

Bode Diagram

Resistor

6 40 4 2 0 -45 Magnitude (dB) 20 0 -20 -40 -60

Phase (deg)

-90

-135 -80 -180 10

2

Transconductor
10
2

-100 10
4

10

10

10

10

10

10

10

10

10

10

12

Frequency (rad/sec)

Frequency (rad/sec)

Figure 4 Simulation of grounded negative resistor

R= d n s n + d n 1s n 1.... + d1s d 0 a n s n + a n 1s n 1.... + a1s + a0

Figure 5 Bandwidth results of grounded negative resistor results of the LC tank formed with a 3.3 pF capacitor, and a negative resistor of value -105 K ohms (approximately 100 times larger in magnitude than that of the parasitic resistor). The grounded negative resistor of Figure 1 was used for this design. It is seen that the negative resistance has no significant effect on the resonant frequency; however it reduces the bandwidth of the tank. The reduction in the bandwidth is proven by the analytical expressions given by
BW 1 = R R 1 BW 2 = L L CN

...(5)

According to the Hurwitz criteria, for a system to be stable it is mandatory that the denominator be a Hurwitz polynomial. A polynomial is said to be Hurwitzian if all the coefficients are positive, which translate to roots with real negative parts. The negative coefficient in the denominator is unwanted as it introduces a positive real pole leading to instability; however it is what constrains the resistance to be negative at DC. Potential instability in the system can also be predicted from the positive slope phase response of the resistor [10], shown in the simulation result of Figure 4. The amount of lead angle can be related to the damping ratio of the negative resistance element and the attendant percent overshoot in response to a step signal input. The overshoot will provide an indication of the potential instability that a bigger system may encounter when using the negative resistance as a sub- system. This knowledge can be used either (i) to ascertain stability of the system, or (ii) to achieve an oscillatory system. 3. APPLICATIONS In this section we present applications of the CMOS negative resistors with supporting analysis and simulation results. 3.1. Current Mode LC-tank Oscillator A current mode oscillator design in which an LC tank is driven by an active transconductance device has been reported [11]. The quality factor of an oscillator is defined as the ratio between the stored energy to the energy dissipated per one radian of the oscillation. The objective is to minimize the energy dissipated, therefore achieve a high quality factor oscillator design. For this particular design the quality factor of the oscillator is highly dependant on the quality factor of the LC tank circuit being used. The inductor that forms the LC tank generally has some parasitic resistance associated with it which degrades the quality factor of the tank. Adding a negative resistance in parallel with the LC tank circuit can cancel the effect of the inductors parasitic resistance, and enhance the characteristics of the system in which the tank is employed. A 365 H active inductor designed using two VCTs [3] in a 0.5 micron CMOS process was determined to have a parasitic resistance of 1092 . Table 4 summarizes the simulation

..(6)

where R is the parasitic resistor of the inductor and N is the negative resistor. Table 4 LC tank simulation results Resonant Freq 4.4 MHz 4.4 MHz Bandwidth 650 KHz 152 KHz Quality Factor 7 29

Without neg R With neg R

The phase noise of an oscillator, described as the short term random frequency fluctuations of the output signal, is inversely proportional to the quality factor of the resonator. An ideal oscillator would produce a pure sine wave, with all of the signals power being at a single frequency. The increasing demand for high resolution wireless communication has stressed the importance of low phase noise oscillators. The quality factor of the resonator, which in this case is the LC tank circuit, is one of the significant sources that generate phase noise in the output of an oscillator. This application proves how a negative resistance circuit can be used to improve the quality factor of a resonator; hence permit the design of a low phase noise current mode oscillator. The stability of the LC tank network which accommodates the unstable negative resistor is analyzed. The expression of the negative resistance given in (5) can be written on pole-zero form as
R
Neg

R DC p 1 p 2 (s + z 1 z 1 (s + p 1 )(s p 2

) ..(7) )

with the expression for the impedance of the LCR circuit given by
Z
LCR

Z z
2

(x

DC 2

(s

+ z

)
0

..(8)
)

+ x1 s + x

When the two impedances are combined in parallel the final expression simplifies to

001285

Z =

(s

(s

+ p

)(s

+ z2
2

)(s

+ z1

+ x1 s + x 0

..(9)

Bode Diagram 20 With neg R Without neg R 10

The denominator of (9) is a Hurwitz polynomial, therefore represents a stable system. It has been determined that this desirable outcome will be retained as long as the resistance RDC is larger than Z DC . When the impedances of (7) and (8) are added in a parallel configuration, the resulting impedance of (9) has a denominator with coefficients that are a summation of terms involving products of RDC and terms involving products of Z DC . Because RDC is greater, the denominator becomes Hurwitzian and the system has all poles with negative real parts. As long as this is ascertained, the negative resistance will not introduce instability into the final system. Only the quality factor will be changed in the desired direction to improve the selectivity. 3.2. Inverting Feedback Amplifier [12] Applying feedback is a common technique used for stabilizing and improving an amplifiers operating characteristics. A universal configuration often used is the inverting feedback amplifier. The closed loop transfer function of an inverting amplifier using an opamp with finite open loop gain A, input admittance of Yi and bandwidth 0 is given by
Rf Vo A0 = Vi Rg s Rf (Yf +Yg +Yi ) + A0 + Rf (Yf +Yg +Yi )0

0 Magnitude (dB)

-10

-20

-30

-40 5 10

10

10

10

10

10

10

Frequency (rad/sec)

Figure 6 Inverting feedback amplifier simulation result

Vo A (s + z 3 )(s + z 4 ) = (s + p 4 )(s + p 5 )(s + p 6 )( s + p 7 ) Vi

..(12)

It is recognized that the denominator of (12) is a Hurwitz polynomial, which indicates a stable system. It has been determined that when the two admittances Y = Yi + Yg + Y f and Y N are added, the resulting active impedance will be Hurwitz stable if R DC > Ri || R f || R g . As long as this inequality is satisfied the inverting amplifier system will be stable, as is portrayed in (12). The two applications presented in this paper justify the importance of negative resistors in CMOS technology. An interesting conclusion drawn from the final results is that when solely used, an active transconductor designed negative resistor is potentially unstable, however may not cause a larger electronic system to be unstable. 4. CONCLUSIONS In this paper we have presented results of investigations on the characteristics of CMOS negative resistance circuits. Critical bandwidth and stability issues regarding the negative resistors were brought forward. Finally, applications which include the design of a low phase noise current mode oscillator are reported. ACKNOWLEDGMENT The research was supported by a research grant awarded to Dr. Raut by the Natural Science and Engineering Research Council (NSERC) of Canada, and by an internal grant (SRT) provided by Concordia University. REFERENCES [1] T. Oura, T. Yoneyama, S. Tantry, H. Asai, A CMOS Floating Resistor Circuit Having Both Positive and Negative Resistance Values, IEICE Trans. Fundamentals, Vol. E85-A, No. 2, Feb 2002 [2] Niksa Tadic, A Floating, Negative-Resistance VoltageControlled Resistor, IEEE Instrumentation and Measurement Technology Conference, Budapest, Hungary, May 2001 [3] R. Raut, A CMOS Building Block for Analogue VLSI Systems, Int. J. Electronics, Vol. 80, No. 1, pages 77-98, 1996

..(10)

The one-pole model of an op-amp was used to derive (10). The opamp was designed is a 0.5 micron CMOS process with the following parameters: A = 1160 V/V, Z1= 1 = 5e11 and 0 = 226
Yi

Hz. The output impedance of the op-amp was ignored. Equation (10) represents the transfer function of a single pole system. Adding a grounded negative resistor at the inverting input terminal of the amplifier will cancel the effect of the op-amp input admittance and extend the bandwidth. This effect can be seen analytically from (10). Adding a negative resistor, such that Yi = Yi Y N , results in the denominator having smaller coefficients. This leads to the pole having a larger magnitude, therefore an increase in the bandwidth. Figure 6 shows the simulation result of the inverting amplifier with R f = 10 K ohms, R g = 1 K and
RN = 1300 . An increase in the bandwidth from 2 MHz to 15 MHz was achieved. The negative resistance used was designed with a CMOS current conveyor [7], and is a three pole system of the form (as an admittance)
z 1 z 2 (s + p 1 )(s + p 2 R DC p 1 p 2 p 3 (s + z 1

)(s )(s

p3 + z2

) )

..(11)

Adding the admittance of (11) to the admittance Y = Y f +Y g + Yi and substituting the result into (10) results in the final feedback system given by

001286

[4] F. Yang, P. Loumeau, P. Senn, Novel Output Stage for DC Gain Enhancement of OPAMP and OTA, Electronics Letters, Vol. 29, No. 11, May 1993 [5] Dallas Semiconductor, Application note 815, Negative Resistance Load Canceller Allows Voltage Reference to Drive Heavy Loads, Sept 2001, http://www.maximic.com/appnotes.cfm/appnote_number/815 [6] S. Szczepanski, J. Jakusz, R. Schaumann, A Linear Fully Balanced CMOS OTA for VHF Filtering Applications, IEEE Tans. On Circuits and Systems-||: Analog and Digital Signal Processing, Vol. 44, No. 3, Mar 1997 [7] Analog IC Design: The Current-Mode Approach, edited by C. Toumazou, F.J Lidgey, D.G. Haigh (Peter Peregrinus Ltd, ISBN 0863412157), Ch. 3 [8] R. Raut, Wideband CMOS Transconductor for Analog VLSI Systems, IEEE Trans. on Circuits and Systems 2: Analog and Digital Signal Processing, Vol. 43, No. 11, Nov 1996 [9] A.S. Sedra, K.C. Smith, Microelectronic Circuits, Oxford University Press: Fifth Edition 2004, ISBN 9780195142518 [10] J. Alvin Connelly, David R. Hertling, Using Spice to Determine Stability, School of Electrical Engineering and Computer Engineering, Georgia Institute of Technology, Atlanta GA USA, IEEE 1999 [11] R. Raut, Current Mode Oscillators Using a Voltage to Current Transducer in CMOS Technology, Int. J. Electronics, Vol. 76, No. 2, pg 257-264, 1994 [12] R. Wu, F.J. Lidgey, K. Hayatleh, Design of Amplifiers with High Gain Accuracy and High Linearity, Proceedings of IEEE International Midwest Symposium on Circuits and Systems, Montreal, Canada, pp. 269-272, Aug 2007

001287

Intentional Blank Page

001288