a low-voltage high-performance CMOS four-quadrant analog multiplier is designed and fabricated by 0.8-m N-well double-poly-double-metal CMOS technology. Experimental results have shown that, under a single 1.2-V supply voltage, the circuit has 0.89% linearity error and 1.1% total harmonic distortion under the maximum-scale input 500-mVP-P at both multiplier inputs. The03-dB bandwidth is 2.2 MHz and the dc current is 2.3 mA. By using the proposed multiplier as a mixer-core and connecting a newly designed output buffer, a CMOS RF downconversion mixer is designed and implemented by 0.5-m single-poly-double-metal N-well CMOS technology. The experimental results have shown that, under 3-V supply voltage and 2-dBm LO power, the mixer has01-dB conversion gain, 2.2-GHz input bandwidth, 180-MHz output bandwidth, and 22-dB noise figure. Under the LO frequency 1.9 GHz and the total dc current 21 mA, the third-order input intercept point is +7.5 dBm and the input 1-dB compression point is09 dBm.
Index Terms— Analog multiplier, low voltage, RF mixer,
wire-less communication.
I. INTRODUCTION
I
T is known that the analog multiplier is an important building block in analog signal processing systems. It can be applied to phase comparators, frequency mixers, and neural networks. Generally, the analog multipliers in different applications have different requirements. In the application of phase comparators, the phase delays from both input ports to the output port of the multipliers should be equal. This means that the multipliers should have a symmetric struc-ture. In the application of radio-frequency (RF) mixers, the linearity, frequency response, and the port-to-port isolation of the multipliers are important characteristics. In the application of neural networks, both chip area and power consumption of the multipliers are important issues. So far, many high-performance CMOS analog multipliers have been proposed [1]–[10]. Among them, the proposed multiplier structures in [1]–[3] are asymmetric. The multiplier structure in [4] uses resistive divider and thus has low port-to-port isolation. TheManuscript received April 3, 1997; revised December 1, 1997. This work was supported by the National Science Council (NSC), Taiwan, ROC, under Grant NSC-86-2221-E-009-080.
The authors are with the Integrated Circuits and Systems Lab, Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan 300, ROC.
Publisher Item Identifier S 0018-9200(98)03505-7.
In this paper, a parallel structure for CMOS analog multipli-ers is proposed, which can be designed to satisfy the different requirements in different applications. The proposed multiplier structure is based on the quarter-square identity implemented by using six combiners [11]. It has a simple and symmetric architecture and can be designed to achieve high bandwidth with high port-to-port isolation, or small chip area with small power dissipation for different applications. Moreover, the most noticeable feature of the proposed multiplier structure is its capability of voltage operation. Unlike other low-voltage analog multipliers which use at least two stacked transistors [12]–[15], the proposed multiplier uses only one stacked transistor. Thus it can be operated at lower supply voltage of 1.2 V while sustaining high linearity under high input signal swing. This feature make the proposed multiplier very suitable for the battery-operated portable systems.
Recently, due to the increasing demand on high-performance low-cost wireless communication systems, more and more effort has been devoted to the implementation of CMOS RF mixers [16]–[18]. These mixers have good performance and are suitable for the application of wireless systems. In this paper, as an illustrative application example of the proposed multiplier structure, an RF downconversion mixer is designed by using the proposed multiplier as the mixer-core and a new operational-transconductance-amplifier (OTA) buffer as the output stage. The proposed RF mixer has high input/output bandwidth while sustaining high linearity. Thus it can be applied to either zero-IF or dual-conversion receivers.
In Section II, the operational principle, circuit realization, and experimental results of the proposed analog multiplier are presented. In Section III, the design methodology and the experimental results of the RF downconversion mixer are given. In Section IV, a conclusion is given.
II. MULTIPLIER DESIGN A. Operational Principle
The block diagram of the proposed analog multiplier is shown in Fig. 1. In Fig. 1, the six blocks Com –Com are called the combiners because they combine the input signals to form the output. The output functions of the combiners can
Fig. 1. The proposed structure of the analog multiplier using six combiners.
be expressed as
– (1a)
(1b) where is the output of the combiner Com – and are the inputs to Com , and – and – are constants. Generally, (1) is the second-order functions. If the constants , and in (1) are set to zero, (1) is reduced to the first-order functions similar to those in [4].
As shown in Fig. 1, the input signals of the multiplier are applied to Com –Com , whereas the output is taken as the difference of the outputs of Com and Com . Assume that the common-mode dc voltages and are imposed upon the input signals and , respectively. Then the inputs to
and and are written as
whereas those to and and as , as shown in Fig. 1. The output of Com –Com – can be obtained by substituting the input voltages of and
– into (1a). After that, and can be derived from (1b). The output signal is given by . Through some calculations, it can be found that if
and , the undesired terms in the expression of can be cancelled, and the resultant is
(2) Thus, the output is a linear multiplication of and with the multiplication constant
. This verifies the multiplication function of the analog multiplier structure in Fig. 1.
The key component in the multiplier of Fig. 1 is the combiner. Any circuit that performs the function in (1) can be used as the combiner. To form the multiplier with the combiners, the constraints are and .
B. Circuit Realization
The second-order transfer function in (1) can be imple-mented by the MOS transistors. To implement the com-biner using the MOS transistors, a parallel circuit structure
Fig. 2. The circuit diagram of the combiner.
is proposed as shown in Fig. 2, where the drain terminals of two MOS transistors and are connected to the resistor . The transfer characteristic of the combiner in Fig. 2 can be modeled by the drain current equation of MOS transistors in the saturation region. Since the nonideal effects of channel-length modulation and mobility degradation can be efficiently suppressed in the analog multiplier structure of Fig. 1, the ideal drain current equation is used to model both combiner and multiplier, and then the nonideal effects will be discussed. Using the ideal square-law current identity of the MOS transistors, the drain current can be expressed as
(3) where is the transconductance param-eter, is the effective surface carrier mobility, is gate oxide capacitance per unit area, is the channel width (length) of the MOS device, is the gate-source voltage, and is the threshold voltage. If and are in the saturation region, the voltage at the drain terminal of the combiner becomes
(4) It can be found that (4) is the same as the required function (1). The complete analog multiplier can be obtained by connect-ing the combiners as shown in Fig. 1. The resultant circuit diagram is shown in Fig. 3. In Com –Com , the value of the resistors – , is , whereas all the MOSFET’s are identical with the same . Substituting , and into (4) and comparing to (1a), one can obtain the corresponding coefficients – as
(5) In Com and Com , the value of the resistors and is , whereas all the MOSFET’s are identical to the same
Fig. 3. The circuit diagram of the proposed analog multiplier.
. Substituting and into (4) and comparing to (1b), one can obtain the corresponding coefficients – as
(6) Thus, the output voltage can be derived by substituting
– in (5) and – in (6) into (2). The result is
(7) The multiplication function is thus realized.
One of the advantages of the new structure in Fig. 3 is that the supply voltage can be very low. The minimum supply voltage of the circuit is determined by the required input signal swing. As in Fig. 3, the input signal range is between and . If the input signal swing is 0.5 V and is 0.7 V, the minimum supply voltage is 1.2 V.
In order to perform the multiplication function, all the transistors in the circuit should be kept in the saturation region under the maximum input signal. Thus, the minimum voltage at the nodes 1– 4, which occurs when both inputs are at ,
should be high enough to keep the transistor on and both the transistors and in the saturation region. Thus the design equation can be written as
(8) The similar restriction is also imposed at the nodes 5 and 6.
The minimum voltage at the nodes 5 or 6 , which occurs
when one of the inputs or is at the maximum voltage
and the other at the minimum, should be high enough to keep the transistor in the saturation region. Thus we have
(9) where
(10) Equations (7)–(10) provide the design guideline for the pro-posed multiplier. For a given , the values of and
can be designed by using (8). After that, the values of and can be determined from (8) and (9).
It is noted that, unlike some analog multipliers whose input range at one input is dependent on that at the other input [19]–[23], both inputs of the proposed multiplier can swing to simultaneously. Thus they have the same input range. Furthermore, the circuit can be designed to be either symmetric or asymmetric by setting the values of and . With the above advantageous features, the proposed multiplier can be applied to various applications.
C. Nonideal Effects
The above derivation of the multiplication function is based on the assumptions of perfect square-law MOSFET char-acteristics and completely matched devices. However, some nonideal effects exist to affect the multiplication function. The nonideal effects include mobility degradation effect and channel-length modulation effect of MOS devices as well as device mismatches due to process variations. For simplicity, the high-order terms are neglected in modeling the nonideal effects on the multiplier structure. Thus mobility degradation effect and channel-length modulation effect can be modeled separately.
TABLE I
THEEXTRA TERMSGENERATED BY THE
ELEMENTMISMATCHES IN THEMULTIPLIER
Taking into account the mobility degradation effect, the current equation of the MOS transistors can be written as
(11) where is the mobility reduction parameter. Using (11) to derive the output voltage of the multiplier and ignoring the high-order terms of , the output voltage becomes
(12) where and are the mobility reduction parameters of the transistors and , respectively, and is the multiplication constant defined in (7).
Taking into account the channel-length modulation effect, the current equation of the MOS transistors can be written as (13) where is the channel-length modulation parameter. The out-put voltage considering the channel-length modulation effect can be found in [24].
The linearity of the multiplier has been simulated by HSPICE using a level 6 device model. The results have shown that the maximum-scale total harmonic distortion (THD) without considering the device mismatch effect is very small compared to the THD with considering the mismatch effect. This means the dominant error source of the proposed multiplier is the device mismatch. This is because the multiplication function in (7) is realized by cancelling the undesired terms at the output, which relies on the matching characteristics of the devices. In the multiplier circuit of Fig. 3, the parameters required to be matched are the transconductances and ; the threshold
Fig. 4. The microphotograph of the fabricated analog multiplier.
voltages and ; and the resistances and . Due to the device mismatches, the extra terms other than the multiplication term can be found through theoretical calculation [24]. The extra terms generated by the mismatched parameters are listed in Table I. It can be seen from Table I that the matching characteristics of the parameters
and are more important than other matching parameters. Generally, better matching characteristics can be obtained by using larger dimensions of transistors and resistors. However, the chip area consumption and frequency performance could be degraded at the same time. Thus, tradeoff should be encountered in the circuit design.
In the design of the proposed multiplier structure, both transistor dimensions and resistor values can be chosen to achieve the optimal gain, dynamic range, bandwidth, or noise performance according to the application requirement. In this case, the proposed multiplier structure is designed to be operated at the minimum supply voltage while maintaining high-linearity characteristic and small chip area. The design criterion is to maximize the signal swing while keeping all the active devices in the saturation region. The resultant aspect ratio in micrometers of the transistors are
and , whereas the resistor values are and k with the width 4.2 and 3 m, respectively. By using these parameters, the SPICE Monte Carlo simulations with 0.03-V standard deviation of and 0.02- m standard deviation of dimensions in both the transistors and resistors have been performed. The results show that the THD is about 1%. The THD can be further reduced by using longer channel transistors. The simulations have shown that, with the channel lengths of transistors increased four times and other parameters adjusted to maintain the same signal swing, the THD can be reduced to 0.5% due to the reduced channel-length mismatch errors.
D. Experimental Results
The designed CMOS analog multiplier with V is fabricated by 0.8- m N-well double-poly-double-metal CMOS technology with the nominal threshold voltage
V. The resistors in the circuit are implemented by n poly resistors. The microphotograph of the experimental chip is
Fig. 5. The measured dc transfer characteristics of the fabricated analog multiplier with the input signal voltagesv1andv2between6250 mV.
Fig. 6. The measured maximum-scale linearity error of the fabricated mul-tiplier.
shown in Fig. 4. The active chip area is 114 199 m . In the measurements of the fabricated multiplier, an arbitrary waveform generator with 1-mV resolution is used to generate the required differential input signals.
Fig. 5 shows the measured dc transfer characteristics of the fabricated analog multiplier with the input voltages and between 250 mV and the corresponding maximum output swing 220 mV. The linearity measurement is performed by supplying and with 250 mV dc and 250 mV voltage ramp, respectively, and measuring the voltage dif-ference between and an ideal voltage ramp with equal amplitude. The measured waveforms are shown in Fig. 6 where the top waveform is the measured , the middle waveform is an ideal voltage ramp , and the bottom waveform is the error voltage . As shown in Fig. 6, the maximum-scale linearity error is 3.9 mV/440 mV 0.89%. Fig. 7 shows the measured harmonic distortion
Fig. 7. The measured harmonic distortion of the fabricated analog multiplier under the maximum-scale inputs.
Fig. 8. The measured frequency response of the fabricated multiplier.
of the fabricated multiplier, where 250-mV dc voltage and 20-kHz 500 mV - sinusoidal wave are applied to and , respectively. In Fig. 7, the maximum-scale total harmonic distortion is 1.1%.
The frequency response of the fabricated multiplier is shown in Fig. 8 where the measured 3-dB bandwidth is 2.2 MHz. This bandwidth is limited by the RC low-pass section formed by the multiplier output resistance of 4 k and the package capacitance about 18 pF. If an output buffer is used to reduce the output capacitance of the multiplier, the signal bandwidth can be much higher. This will be verified in the design of RF mixers. Since the multiplier is symmetric, the measurement results remain the same where the input voltages and are interchanged. The measurement results of the fabricated analog multiplier are summarized in Table II.
III. RF MIXER A. Mixer Characteristics
In the RF mixers, the important design parameters are noise figure (NF), conversion gain (CG), third-order
input-intercept-TABLE II
THEMEASUREDCHARACTERISTICS OF THEFABRICATEDANALOGMULTIPLIER
point (IIP ), input bandwidth, output bandwidth, port-to-port isolation, and local oscillator (LO) power. These parameters should be designed to meet the requirements of various stan-dards for different wireless communication systems.
In this paper, the proposed mixer is intended to be used as the RF downconversion mixer in the wireless receiver. The RF downconversion mixer is often placed after a low-noise amplifier (LNA). The LNA provides sufficient power gain to mask the noise contribution of the subsequent stages. Thus the noise figure contributed by the mixer can be ignored if its value is lower than the total gain of the previous stages. For a receiver with 20-dB gain contributed by LNA, the NF of the mixer should be lower than 20 dB.
Since the LNA has provided sufficient gain, the CG of the mixer should not be high to overdrive the subsequent stages. Higher gain also implies higher signal swing in the circuit, which could degrade the linearity and the dynamic range. Nevertheless, very low gain far below 0 dB is also unacceptable because the noise contributed by the stages after the mixer becomes higher. Thus the value of CG around 0 dB is acceptable.
In modern wireless systems, the receivers could be subject to an environment with large adjacent-channel interfering signals. Due to the nonlinearity of the receiver, those interfering signals produce co-channel interference which degrades the signal-to-noise-ratio of received signals. Thus, IIP of the receiver, which indicates the ability of the receiver to reject the interfering signals, becomes a very important feature of the RF receiver. In most cases, the signal power handled by the RF mixer is higher than those by the other stages in the receiver. Thus, IIP of RF mixers is a critical parameter in the receiver design. In order to sustain a high receiver linearity, IIP of the mixer should be as high as possible.
To cover the interested signal frequency ranges according to the standard, both the input bandwidth in the RF port and the output bandwidth in the IF port of the mixer should be high enough. For the application of modern wireless systems, the required input bandwidth is 900 MHz or 1.9 GHz. But the output bandwidth requirement depends on the architecture of the receiver. In a dual-conversion receiver, the IF is about 70–100 MHz for the 900-MHz system or 90–240 MHz for the 1.9-GHz system [25]–[28]. In a zero-IF receiver, the output bandwidth of a few megahertz is enough.
The port-to-port isolation and LO power are also important issues in the mixer design. The LO power in the order of a few
dBm is often required by the mixer to obtain high linearity and high dynamic range. Such a high LO power causes the LO energy leaks through the RF port and radiates from the antenna if the port isolation of the mixer is not high enough. The design criteria of the mixer is to keep the LO power as low as possible and increase the port isolation.
B. Buffer Design
When the mixer is not integrated with the IF circuits, the IF output of the mixer is required to drive a low-impedance load such as an external filter or the instrument impedance. However, the circuit in Fig. 3 is not suitable for driving the low-impedance load. Thus an output buffer is required. In order to keep the IF signals undistorted, the buffer should have high driving capability with high linearity and high bandwidth. The conventional buffer amplifier has sufficient driving capability, but its bandwidth is not high enough [29]–[31]. Therefore, a new high-performance OTA buffer is designed.
Fig. 9 shows the circuit diagram of the proposed OTA buffer which consists of an input stage, a predriver stage, and an output stage. The input stage performs the subtraction of the input signals and provides a little gain to compensate the loss of the predriver stage. The polysilicon resistors and are chosen as load elements because they have higher frequency response and higher signal swing than the active loads. The predriver stage performs level-shifting and single-to-differential conversion of the signals from the input stage. The output stage is a push–pull stage driven by two level-shifted signals from the opposite side. This cross-coupled scheme provides additional common-mode rejection for the circuit.
In order to operate the circuit under the best condition, the level-shifted signals driving the output stage should have equal amplitude, which requires the transistors – to have matched transconductances. Moreover, the dimensions of the transistors – should be matched to those of – . If these matching requirements are met, the currents flow through both PMOS and NMOS transistors of the output stage are matched and the second-harmonic distortion of the output current can be dramatically reduced. This phenomenon is illustrated in Fig. 10 where the currents flow through both PMOS and NMOS transistors of the output stage and the combined output current on a 50- load are drawn as a function of the input voltage. As seen in Fig. 10, the combined output current has a larger linearity range than those obtained by driving a single transistor. Thus high linearity can be obtained with low output transistor bias currents. In addition to the high-linearity characteristic, the buffer also benefits from the low-impedance nodes 1– 6 in the circuit. Thus,
high-frequency response can be obtained. The buffer is intended for open-loop operation and no compensation is employed.
The supply voltage of the buffer is determined by the required signal swing at the nodes 3– 6. The maximum signal
swing at the nodes 3–6 can be expressed as
Fig. 9. The circuit diagram of the proposed OTA buffer.
where is the minimum saturation drain-to-source volt-age of the transistor . In this design, given
V, V, and V, is 0.7 V. The simulated gain and bandwidth of the buffer with 50- loads are 1 dB and 180 MHz, respectively. The dc current is 6 mA.
C. Mixer Design
The multiplier in Fig. 3 is used as a mixer-core which is directly connected to the OTA buffer to form a complete mixer. An intuitive operational principle of the mixer-core is given as follows. Assume the port in Fig. 3 is supplied with a large enough LO signal to drive the transistors ON and OFF. When LO is at high voltage, the transistors and are ON, whereas the transistors and are OFF. Thus the nodes 1 and 3 are at low voltage that disable the
transistors and . In this case, the RF signal can be transmitted to the output through the transistors
and and the output is out-phase. When LO is at low voltage, the transistors and are ON, whereas the transistors and are OFF. Thus the nodes 2 and 4 are at low voltage that disable the transistors and .
In this case, the RF signal can be transmitted to the output through the transistors and and the output is in-phase. Thus the RF signal is switched by the LO signal.
For the mixer-core of an RF downconversion mixer, the element values in Section II-C are not the optimum design.
Fig. 10. The simulated currents flow through both PMOS and NMOS transistors of the output stage and the combined output current on a 50- load as a function of input voltage.
Redesigning of the element values is required to meet the requirements given in Section III-A.
The voltage conversion gain of the mixer can be derived from (7) and rewritten in terms of small-signal
pa-rameters. The result is
(15) where and are the transconductances of the transistors and , respectively, is the LO amplitude expressed in , and is the voltage gain of the buffer.
The input bandwidth of the mixer is determined by the time constant at the nodes 1–4 in Fig. 3, which is
given by
(16) where is the total capacitance on the nodes, which can be expressed as
(17) where and are the drain terminal capacitances of transistors and , respectively, is the gate terminal capacitance of transistor , and is the parasitic capacitance of resistor .
The noise factor of the mixer can be derived based on the following assumptions. 1) Under the large LO signal, the average transconductances of the transistors are equal to the quiescent transconductances. 2) The image-band noise power is equal to the RF-band noise power. Both of them are transferred to the IF frequency band completely. 3) Only thermal noise is considered. 4) The noise contributed by the buffer stage is ignored. The derived is given as (18), shown at the bottom of the page, where is the transconductance of the transistor , and is the source resistance 50 . In the numerator of (18), the four terms represent the noise contributed by the transistors and , the transistor , the resistor , and the resistor . Hand calculation of (18) shows that the most noisy sources are the transistors and
. Their contribution is about 80% of the total noise. The tradeoffs among the element values in the mixer-core design can be observed from (15)–(18). As in (15) and (16), the resistance affects both and . Thus, it is an important parameter in the mixer design. Since is a dominant parameter, the design of the element values can be started from the determination of to achieve . In fact, the large parasitic capacitance associated with the resistor leaves very few choices of for the gigahertz range operation. The value of a few hundred ohms is reasonable. While is set, the parasitic capacitance can be determined from the dimension of . At the same time, the upper limit of is also set by the required
from (16). Substituting the obtained and into (17), the maximum terminal capacitances of the transistors can be determined, which implies the channel widths of the transistors are limited.
Since is designed and , , and are given, the unknown parameters , , , and in (15)
TABLE III
THEELEMENTVALUES OF THEPROPOSEDMIXER
can be determined. Since the channel widths of the transistors are set, and can be obtained by giving proper channel length and dc bias. Then the value of and can be determined from (15). With all the device parameters determined, the noise factor can be calculated from (18). If the resultant is not low enough, and should be increased. This can be achieved by increasing the bias currents or decreasing the channel lengths of and . However, increasing the bias current increases the power consumption, whereas decreasing the channel length increases the mismatch errors that degrade the linearity. Thus, the tradeoffs among noise factor, power consumption, and linearity performance should be made. If the result is still not satisfactory after using the above two design methods, the design process should be resumed with another value of .
In this design, the nominal supply voltage is 3 V. This value of supply voltage is required for the buffer operation. If the mixer is to be on-chip connected to other circuits, the buffer is not required and the supply voltage can be lower. The element values of the mixer-core and the buffer are listed in Table III. The input terminals of the mixer-core and are served as LO port and RF port, respectively. Both the dc bias of RF and LO ports are 1.5 V. The width of the resistor is 4 m, which results in 0.06 pF . The total capacitance is 0.15 pF; thus, the input bandwidth is 2.1 GHz. The simulated is 0.89 ( dB) while the LO power is 2 dBm (0.283 V ). The noise figure calculated using (18) is 23.8 dB. The dc current of the mixer-core is 15 mA. This value is designed after the tradeoff with both noise figure and linearity. The dc current could be decreased if the required input bandwidth is not so high. With the bandwidth requirement decreased to 1 GHz, the designed value of is increased. This means the transistor dimensions can be increased; thus, the same transconductances can be obtained with lower dc bias currents. The simulations have shown that, with the device dimensions and resistor values of the mixer-core changed to , , and dc bias voltage V, the mixer-core can reach the same performance as the previous design while decreasing the dc current to 9 mA.
Fig. 11. The measured input bandwidth of the fabricated mixer.
D. Experimental Results
The experimental chip of the proposed mixer is fabricated by 0.5- m single-poly-double-metal N-well CMOS technol-ogy. The active chip area is 590 220 m . In order to perform the high-frequency measurements, the experimental chip is mounted on the PCB directly, which effectively reduces the I/O parasitics. In the experimental board, both RF and LO ports are terminated with matching resistors; thus, the port reflections can be lower than 10 dB in the measured fre-quency band. The differential signals in the measurements are generated through three passive single/differential converters connected to the RF, LO, and IF ports. The bandwidth of the converters is 2 GHz. The accuracy of the measured signal power level is about 0.5 dB.
The measured power conversion gain versus RF signal frequency with IF fixed to 20 MHz is shown in Fig. 11. In the measurement, the high frequency is limited by the bandwidth of the single/differential converters. However, the extrapolated input bandwidth shown in Fig. 11 is 2.2 GHz which is close to the calculated value. The measured output bandwidth of the fabricated mixer with LO fixed to 1.9 GHz is shown in Fig. 12, where the measured 3-dB bandwidth is 180 MHz, which is consistent with the simulated bandwidth of the output buffer. The intrinsic output bandwidth of the mixer-core is much higher than this value. The simulations have shown that the internal bandwidth at the output nodes of the mixer-core is up to 620 MHz. With proper changes of the component values, this value can be designed to be higher than 1 GHz. Thus an upconversion mixer can also be implemented by the proposed multiplier structure.
Since the most popular frequency bands of the modern wireless systems are 900 MHz and 1.9 GHz, the LO frequen-cies are set to the above two frequenfrequen-cies in the following measurements. The measured NF versus IF frequency of the fabricated mixer is shown in Fig. 13, where the minimum IF
Fig. 12. The measured output bandwidth of the fabricated mixer.
Fig. 13. The measured NF versus IF frequency of the fabricated mixer.
frequency in the measurement is limited to 10 MHz by the instrument. As seen in Fig. 13, the measured NF is about 22 1 dB in the entire IF frequency band. This value is slightly lower than the calculated value (23.8 dB) because the average transconductances of the transistors under the large LO signal swing are higher than the quiescent transconductances used in the calculation.
The measured third-order input intercept point IIP and the measured input 1 dB compression point IP of the fabricated mixer with MHz and 1.9 GHz are shown in Fig. 14(a) and (b), respectively. The two-tone frequencies in the measurement with MHz (1.9 GHz) are 930 MHz (1.93 GHz) and 940 MHz (1.94 GHz). As seen in Fig. 14(a), the measured IIP and IP are 7.5 dBm and 10 dBm, respectively. In Fig. 14(b), IIP and IP are 7.5
(a)
(b)
Fig. 14. The measured, IIP3 and IP1dB of the fabricated mixer under (a) 900-MHz LO and (b) 1.9-GHz LO.
dBm and 9 dBm, respectively. It has been found that the value of IIP measured at one of the differential output ports of the mixer is the same as that measured at the differential output ports of the mixer. This means that the intermodulation distortion of the mixer is mainly contributed by the mixer-core and the output buffer only contributes negligible distortion. The high-linearity characteristic of the buffer is thus proved.
The measured LO-to-RF isolation versus LO frequency is shown in Fig. 15, where the isolation degrades from 71 to 27 dB as the frequency increases from 100 MHz to 1.9 GHz. This means that the coupling between RF and LO ports is capacitive. The capacitive coupling is mainly caused by the asymmetry and crossover of the interconnection lines in the circuit layout. The measured characteristics of the fabricated downconversion mixer are summarized in Table IV.
Fig. 15. The measured LO-to-RF isolation of the fabricated downconversion mixer.
TABLE IV
THEMEASUREDCHARACTERISTICS OF THEMIXER
IV. CONCLUSION
A new CMOS analog multiplier based on the square-law characteristics of MOS devices has been designed and analyzed. The multiplier can be operated at very low sup-ply voltage while sustaining a high linearity characteristic. Moreover, it has the advantageous characteristics of symmetric structure, high-frequency response, and small chip area. Thus, the proposed CMOS analog multiplier is very feasible in various applications. By using the proposed analog multiplier structure, an RF downconversion mixer has been successfully designed and fabricated. The performance of the experimental chip has been verified through the measurement. It has also been shown from the experimental results that the fabricated RF mixer can meet the requirements of 900 MHz and 1.9 GHz wireless communication systems. Further research will be con-ducted to integrate the designed RF mixer with an RF bandpass amplifier [32] to form a high-integration CMOS receiver.
ACKNOWLEDGMENT
The authors would like to thank the reviewers for their valuable comments and suggestions. They also wish to thank
State Circuits, vol. 26, pp. 1293–1301, Sept. 1991.
[6] S. L. Wong, N. Kalyanasundaram, and C. A. T. Salama, “Wide dynamic range four-quadrant CMOS analog multiplier using linearized transcon-ductance stages,” IEEE J. Solid-State Circuits, vol. 21, pp. 1120–1122, Dec. 1986.
[7] J. N. Babanezhad and G. C. Temes, “A 20-V four-quadrant CMOS analog multiplier,” IEEE J. Solid-State Circuits, vol. 20, pp. 1158–1168, Dec. 1985.
[8] S.-I. Liu and C.-C. Chang, “CMOS analog divider and four-quadrant multiplier using pool circuits,” IEEE J. Solid-State Circuits, vol. 30, pp. 1025–1029, Sept. 1995.
[9] D. C. Soo and R. G. Meyer, “A four-quadrant NMOS analog multiplier,”
IEEE J. Solid-State Circuits, vol. 17, pp. 1174–1178, Dec. 1982.
[10] J. S. Pena-Finol and J. A. Connelly, “A MOS four-quadrant analog multiplier using the quarter-square technique,” IEEE J. Solid-State
Circuits, vol. 22, pp. 1064–1073, Dec. 1987.
[11] S.-Y. Hsiao and C.-Y. Wu, “A 1.2 V CMOS four-quadrant analog multiplier,” in Proc. ISCAS, June 1997, pp. 241–244.
[12] A. L. Coban, P. E. Allen, and X. Shi, “Low-voltage analog IC design in CMOS technology,” IEEE Trans. Circuits Syst.-I, vol. 42, pp. 955–958, Nov. 1995.
[13] S.-I. Liu, “Low voltage CMOS four-quadrant multiplier,” Electron. Lett., vol. 30, no. 25, pp. 2125–2126, Dec. 1994.
[14] G. Colli and F. Montecchi, “Low voltage low power CMOS four-quadrant analog multiplier for neural network applications,” in Proc.
ISCAS, May 1996, pp. 496–499.
[15] S.-I. Liu and C.-C. Chang, “Low-voltage CMOS four-quadrant multi-plier,” Electron. Lett., vol. 33, no. 3, pp. 207–208, Jan. 1997. [16] P. Y. Chan, A. Rofougaran, K. A. Ahmed, and A. A. Abidi, “A highly
linear 1-GHz CMOS downconversion mixer,” in Proc. ESSCIRC, Sept. 1993, pp. 210–213.
[17] J. Crols and M. S. J. Steyaert, “A 1.5 GHz highly linear CMOS downconversion mixer,” IEEE J. Solid-State Circuits, vol. 30, pp. 736–742, July 1995.
[18] A. Rofougaran, J. Y.-C. Chang, M. Rofougaran, and A. A. Abidi, “A 1 GHz CMOS RF front-end IC for a direct-conversion wireless receiver,”
IEEE J. Solid-State Circuits, vol. 31, pp. 880–889, July 1996.
[19] S.-I. Liu and Y.-S. Hwang, “CMOS squarer and four-quadrant multi-plier,” IEEE Trans. Circuits Syst.-I, vol. 42, pp. 119–122, Feb. 1995. [20] Z. Hong and H. Melchior, “Analogue four-quadrant CMOS multiplier
with resistors,” Electron. Lett., vol. 21, no. 12, pp. 531–532, June 1985. [21] H.-J. Song and C.-K. Kim, “An MOS four-quadrant analog multiplier using simple two-input squaring circuits with source followers,” IEEE
J. Solid-State Circuits, vol. 25, pp. 841–848, June 1990.
[22] S. Sakurai and M. Ismail, “High frequency wide range CMOS analogue multiplier,” Electron. Lett., vol. 28, no. 24, pp. 2228–2229, Nov. 1992. [23] Y. H. Kim and S. B. Park, “Four-quadrant CMOS analogue multiplier,”
Electron. Lett., vol. 28, no. 7, pp. 649–650, Mar. 1992.
[24] S.-Y. Hsiao, “The design and analysis of new CMOS RF receiver front-end circuits,” Ph.D. dissertation, Nat. Chiao Tung Univ., Taiwan. [25] A. A. Abidi, “Direct-conversion radio transceivers for digital commu-nications,” IEEE J. Solid-State Circuits, vol. 30, pp. 1399–1410, Dec. 1995.
[26] P. R. Gray and R. G. Meyer, “Future directions in silicon IC’s for RF personal communications,” in Proc. CICC, 1995, pp. 6.1.1–6.1.8. [27] I. A. Koullias, J. H. Havens, I. G. Post, and P. E. Bronner, “A 900 MHz
transceiver chip set for dual-mode cellular radio mobile terminals,” in
ISSCC Dig. Tech. Papers, 1993, pp. 140–141.
J. Solid-State Circuits, vol. 31, pp. 1939–1944, Dec. 1996.
[37] J. C. Rudell, J.-J. Ou, T. B. Cho, G. Chien, F. Brianti, J. A. Weldon, and P. R. Gray, “A 1.9 GHz wide-band IF double conversion CMOS integrated receiver for cordless telephone applications,” in ISSCC Dig.
Tech. Papers, 1997, pp. 304–305.
[38] K. Bult and H. Wallinga, “A class of analog CMOS circuits based on the square-law characteristic of an MOS transistors in saturation,” IEEE
J. Solid-State Circuits, vol. 22, pp. 357–365, June 1987.
[39] H. R. Mehrvarz and C. Y. Kwok, “A novel multi-input floating-gate MOS four-quadrant analog multiplier,” IEEE J. Solid-State Circuits, vol. 31, pp. 1123–1131, Aug. 1996.
[40] K. Kimura, “Some circuit design techniques for low-voltage analog functional elements using squaring circuits,” IEEE Trans. Circuits
Syst.-I, vol. 43, pp. 559–576, July 1996.
Shuo-Yuan Hsiao (S’93) was born in Taichung, Taiwan, China, in 1969. He received the B.S. degree from the Department of Electronic Engineering, Fu-Jen Catholic University, Taiwan, in 1991, and the Ph.D. degree from the Department of Electronics Engineering and the Institute of Electronics, Na-tional Chiao-Tung University, Taiwan, in 1997.
His research interests include analog integrated circuits design and RF integrated circuits design.
Chung-Yu Wu (M’76–SM’96–F’98) was born in Chiayi, Taiwan, China, in 1950. He received the M.S. and Ph.D. degrees from the Department of Electronics Engineering, National Chiao-Tung Uni-versity, Taiwan, in 1976 and 1980, respectively.
From 1980 to 1984 he was an Associate Professor at the National Chiao-Tung University. From 1984 to 1986, he was a Visiting Associate Professor at the Department of Electrical Engineering, Portland State University, Oregon. Since 1987, he has been a Professor at the National Chiao-Tung University. From 1991 to 1995, he served as Director of the Division of Engineering and Applied Science in the National Science Council. Currently, he is the Centennial Honorary Chair Professor at the National Chiao-Tung University. He has published more than 77 journal papers and 107 conference papers. He also has 18 patents including nine U.S. patents. His current research interests include low-voltage low-power mixed-mode integrated circuit design, hardware implementation of visual and auditory neural systems, and RF integrated circuit design.
Dr. Wu is a member of Eta Kappa Nu and Phi Tau Phi. He was awarded the Outstanding Research Award by the National Science Council in 1989 and 1995, and was named Outstanding Engineering Professor by the Chinese Engineer Association in 1996.
