RESEARCH ARTICLE

Design and Optimization of A Direct-Conversion Double-Balanced

Mixer for RF Receiver Front-End
Frederick Ray I. Gomez
New Product Introduction Department, Back-End Manufacturing & Technology, STMicroelectronics, Inc.
9 Mountain Drive, Light Industry & Science Park II, Brgy. La Mesa, Calamba City, Laguna, Philippines 4027

Abstract: Differential implementation is becoming highly favoured in RFIC (radio frequency integrated circuit) design,
notably its high immunity to common-mode noises, acceptable rejection of parasitic coupling, and increased dynamic
range. One specific RF front-end building block that is usually designed as a differential circuit is the mixer. This
technical paper presents a study of a differential mixer, notably the double-balanced mixer implemented on a
direct-conversion architecture in a standard 90nm CMOS (complementary metal-oxide semiconductor) process.
Operating frequency is set at 5GHz, which is a typical frequency for RF (radio frequency) receiver. Impedance
matching was essential to fully optimize the mixer design. The direct-conversion double-balance mixer design
eventually achieved conversion gain of 11.463dB and noise figure of 16.529dB, comparable to mixer designs from past
research and studies.
Keywords: Double-Balanced Mixer; Direct Conversion; Conversion Gain; Noise Figure; RF Front-End

1. Introduction
The front-end of a RF (radio frequency) wireless receiver is of particular interest to many RFIC (radio frequency
integrated circuit) designers and researchers as it attests to be the most critical part in many communication systems and
wireless applications like the Bluetooth, WiFi (wireless fidelity), and WiMAX (worldwide interoperability for
microwave access). The block diagram of a typical receiver is shown in Figure 1.

Figure 1; Block diagram of a typical RF receiver.

Mixer is among the front-end building blocks of an RF receiver. It is also referred to as a converter because it
converts RF signals into a lower intermediate frequency (IF) by mixing with an offset local oscillator (LO). Depending
on the RF receiver requirements, mixers must undergo a careful design process since complex tradeoffs exists among
different performance parameters.

2. Review of Related Literature

RF receivers can be categorized as superheterodyne (high-IF), low-IF, and homodyne or direct-conversion
orzero-IF based on the resulting IF signal they operate. For the direct-conversion receiver, IF is designed to be centered

Copyright © 2018 Frederick Ray I. Gomez

doi: 10.18063/peee.v1i3.800
This is an open-access article distributed under the terms of the Creative Commons Attribution Unported License
(http://creativecommons.org/licenses/by-nc/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original
work is properly cited.

Progress of Electrical and Electronic Engineering Volume 1 Issue 3 | 2018 | 1

provided by Whioce Journals

View metadata, citation and similar papers at core.ac.uk brought to you by CORE
at frequency zero. This means that the LO frequency is equal to the input RF frequency. With zero-IF, image signal is
avoided and the analog filtering problem can be easily handled. Moreover, the desired signal is translated directly to
the baseband, allowing analog-to-digital converter (ADC) and digital signal processing (DSP) circuits to perform
modulation and other supplementary functions[1,2]. This eliminates the need for highly complex filters since channel
selection only requires a low-pass filter (LPF) as shown in Figure 2. Smaller and cheaper receivers with low power
consumption could be realized for various wireless applications such as Bluetooth, WiFi, and WiMAX. Many of the
implemented receivers in WiMAX[3-5] use the zero-IF architecture since the LPFs make sure that the closely-spaced
carrier signals do not cause interference with each other.

Figure 2; Direct-conversion or zero-IF receiver.

Designing a mixer must take into account the trade-offs among the performance parameters. As earlier mentioned,
a careful study of these important parameters must be done in order to design a fully-functional and optimized mixer. A
mixer’s efficiency on frequency conversion from RF to IF is characterized by conversion gain (CG) or loss, expressed
in Eq. (1) and (2). CG is the ratio of the desired IF output to the value of the RF input, and may be expressed in voltage
or power. In cases when CG is less than unity or 0dB, it is fittingly termed as a conversion loss.

Eq. (1)

Eq. (2)
VIF and VRF are the root mean square (RMS) voltages of the IF and RF signals, respectively, while PIF and PRF are
the equivalent power of the IF and RF signals, respectively. Conversion gain is preferred over conversion loss because
amplification along with frequency translation. Nevertheless, it should be noted that conversion gain directly affects the
noise figure and linearity of the overall receiver. Hence, design tradeoffs concerning these parameters are inevitable.
Noise figure (NF) is another important parameter of the mixer. It is a measure of the amount of
signal-to-noise-ratio (SNR) degradation introduced by the mixer as seen at the output. Eq. (3) shows the
relation between the SNR at the input port and the SNR at the output port of the mixer.

Eq. (3)
Noise figures of mixers tend to be higher than amplifiers (i.e. low-noise amplifiers, power amplifiers) because of
the contribution of noise from other frequencies (apart from input RF signal) that can mix down to the IF. This
considerable noise in mixers is the main reason why low-noise amplifiers (LNA) are used in the front-end, before the
mixer[6].
A popular solution for the mixer is based on the double-balanced topology, with the schematic shown in Figure 3.
Double-balanced mixer is also commonly known as Gilbert cell mixer. It operates with differential LO and RF inputs.
In this topology, LO products are prevented from getting to the output by combining two single-balanced mixers. As
illustrated in Figure 3, the two single-balanced mixers are connected in anti-parallel as far as the LO is concerned, but in
parallel for the RF signal. Thus, the LO terms sum to zero in the output, whereas the converted RF signal is doubled in
the output[6]. This is most desirable for high port-to-port isolation and spurious output rejection applications.

2 | Frederick Ray I. Gomez Progress of Electrical and Electronic Engineering

 Figure 3; Double-balanced mixer topology.

3. Double-Balanced Mixer Design

Double-balanced mixer topology can provide high conversion gain, very low noise figure, and high degree of
LO-IF isolation. The main disadvantage of this topology is its physical implementation. A balun transformer is required
to convert the single-ended input to a differential RF input signal of the mixer. Transformers with very low insertion
loss are difficult to realize in monolithic integration, hence this forces the use of an off-chip transformer which occupies
more board space and cost[7]. An alternative solution is to use active baluns in place of their passive counterparts, as
they can produce gain, occupy smaller chip area and can operate at RF and higher frequencies[8].
The goal of the paper is then to design and optimize a direct-conversion double-balanced mixer implemented in a
standard 90nm CMOS (complementary metal-oxide semiconductor) process, operating at frequency of 5GHz which is a
typical frequency for an RF receiver. The target specifications of the figures of merit are based on the performance
comparison in terms of conversion gain and noise figure of past researches on direct-conversion active mixer topologies
given in Table 1.
Conversion Gain
Design RF (GHz) Noise Figure (dB) Topology
(dB)
[4] 3.4~3.85 10 10 Single-balanced
[5] 2~11 21.5~22.8 21.5~25.8 Double-balanced
[9] 2 19.5 10.2 BiCMOS Double-balanced
[10] 20~40 16 -- BiCMOS Single-balanced
[11] 5.2 9.3 10.5 Double-balanced
Table 1. Performance comparison of mixer designs.
It is imperative to determine the proper biasing and sizing of all the transistors such that RF transistors (M1-M2)
will operate in the saturation region while the LO transistors (M3-M6) operate near the boundary of the saturation and
linear regions. The mixer design used nsvt (NMOS standard Vt) which is the typical model for the transistor.
One way to increase the performance of the mixer in terms of conversion gain and noise figure is to apply
impedance matching in the circuit. Z11 and Z22 can be obtained using sp-analysis which is swept from f1 = 700MHz to f2
= 6GHz. Actual inductor and capacitor values at f = 5GHz can be computed from the L-network reactances. .
Inductors and capacitors are key passive components that are crucial for impedance matching, and are specifically
designed such that they would satisfy the conversion gain of the mixer. Impedance matching is necessary in RF circuit
design to provide maximum possible power transfer between the source or the generator and the load[12]. The schematic
diagram of the whole circuit is shown in Figure 4.
To supply differential LO input to the mixer, a port PORT2 with a matching resistor (set to 50Ω) is used which is
then fed into an ideal passive balun to convert the single-ended signal into differential. For the differential RF input of
the mixer, same setup as the LO is used with PORT1. To use the differential output for measurements, matching the IF
output port PORT3 to the output impedance of the mixer is necessary. PORT1 is set to DC source type with pacmag

Progress of Electrical and Electronic Engineering Volume 1 Issue 3 | 2018 | 3

(periodic ac magnitude) set to 1. PORT3, which is the IF port, is set also to DC source type. The only large signal is
from PORT2 which is a sine wave with flo=5GHz and PLO = 0dBm.
For the impedance matching, the input matching network is applied before the differential RF input of the mixer
instead of placing it before the balun. This will result to an adjustment on the value of L1, which will decrease, since
the balun circuit has self-inductance. The adjusted value of L1 can be determined using the sp-analysis swept from f1
= 700MHz to f2 = 6GHz. Moreover, inductors with small inductances are more realizable in actual designs than their
larger counterparts. The final values of the L-matching network are summarized in Table 2.
Software design tools ASITIC[13,14] and SpiralCalc[15,16] were used for the design of the inductors. Both software
tools are available for academic and non-commercial purposes. Table 3 shows the design parameters obtained for the
design of the spiral inductors using ASITIC while Table 4 shows that of using SpiralCalc.

Figure 4; Schematic design of double-balanced mixer.

L and C Value

L1 12.7nH
C1 1.116087pF
L2 1.328515nH
C2 267.127fF

Table 2. Final values of L-network elements.

4 | Frederick Ray I. Gomez Progress of Electrical and Electronic Engineering

 Inductors
Parameters
L1 L2
Desired L 12.7nH 1.328515nH
No. of sides 4 8
Length, D 300µm 190µm
Metal width, W 10.886µm 10.901775µm
Spacing, S 1 1
No. of turns, N 4.25 2.5
Metal layer 7 7
Inductance, L 12.711nH 1.329nH
Q-factor, Q 2.322 5.694
Table 3. Inductor design using ASITIC.

Inductors
Parameters
L1 L2
Desired L 12.7nH 1.328515nH
No. of sides 4 8
Length, D 300µm 190µm
Metal width, W 11.2µm 10µm
Spacing, S 1 1
No. of turns, N 6 2
Inductance, L:
Modified Wheeler 12.885µm 1.326µm
Current Sheet 12.739µm 1.326µm
Monomial Fit 12.624µm 1.394µm
Table 4. Inductor design using SpiralCalc.
In ASITIC, the spiral inductors are designed such that desired inductances are achieved and the Q-factors are
optimized with eddy-current option enabled to include the effects of substrate induced eddy current losses. L1 have
smaller Q-factor than L2 because of its high inductance value. For the inductor design using SpiralCalc, same parameter
values from the ASITIC parameters are used except for the metal width and the number of turns of the spiral inductor.
These parameters are tweaked such that the desired inductances are achieved for the inductors.
The n2port from the analogLib library is used as a model block for all the ASITIC inductors. Touchstone format of
S-parameter file is used as file input of the n2port component since the actual S-parameters using ASITIC are given in
touchstone format. The figures of merit such as conversion gain and noise figure are determined using SpectreRF in the
Analog Design Environment.
A mixer’s frequency converting action is characterized by conversion gain or loss. Voltage conversion gain is the
ratio of the RMS voltages of the IF and RF signals, earlier given in Eq. (1) and (2). The variations of conversion gain
with the power of LO signal (PLO) can be measured using swept PSS (Periodic Steady-State) analysis with PAC
(Periodic AC) analysis. The PAC analysis will then compute the voltage conversion gain in dB20 of the whole circuit
with PORT3 as the output port (with output harmonic of 0, which is 5GHz) and PORT1 as the input port (with input
harmonic of -1, which is 0GHz). Setting the input port to RF+ port, which is located after the balun circuit, will
compute the voltage conversion gain of the mixer only. Simulation plots of the conversion gain swept from PLO =
-10dBm to PLO = 30dBm are shown in Figure 5-7.

Progress of Electrical and Electronic Engineering Volume 1 Issue 3 | 2018 | 5

 Figure 5; Conversion gain (in dB) vs. PLO (w/o matching).

Figure 6; Conversion gain (in dB) vs. PLO (ideal L).

Figure 7; Conversion gain (in dB) vs. PLO (ASITIC L).

For the noise figure, Pnoise (Periodic Noise) analysis with PSS analysis is used. In addition, PSP (Periodic
S-Parameters) analysis with PSS analysis can also be used to determine the noise figure of the circuit. Noise figure
from a sweep range of -10dBm to 30dBm can now be determined and plotted using these analyses. The resulting plots
are shown in Figure 8-9.

6 | Frederick Ray I. Gomez Progress of Electrical and Electronic Engineering

 Figure 8; Noise figure (in dB) vs. PLO of (Pnoise analysis).

Figure 9; Noise figure (in dB) vs. PLO of (PSP analysis).

Table 5 summarizes the simulation results for conversion gain and noise figure at PLO = 0dBm.
Figures of Merit Value
CG (mixer+balun) [w/o matching] -17.431dB
CG (mixer) [w/o matching] -11.410dB
CG (mixer+balun) [ideal L] 5.443dB
CG (mixer) [ideal L] 11.463dB
CG (mixer+balun) [ASITIC L] -10.966dB
CG (mixer) [ASITIC L] -4.946dB

Pnoise NF [w/o matching] 125.355dB

Pnoise NF [ideal L] 16.529dB
Pnoise NF [ASITIC L] 33.713dB
PSP NF [w/o matching] 8.127dB
PSP NF [ideal L] 7.136dB
PSP NF [ASITIC L] 21.437dB
Table 5. Simulation results (at PLO = 0).

4. Discussion of Results
Based on the conversion gain simulation results as shown in Figure 5-7 and in Table 5, input and output
impedance matching contribute to better performance. Furthermore, using ideal inductors for impedance matching
produced better performance as compared to using non-ideal ASITIC inductors through the n2port. It can be observed
from the simulation plots that the conversion gain of the mixer only is higher than the conversion gain of the whole
circuit consisting of the mixer and the balun. This is because the balun in the circuit, which is a passive balun, has
insertion loss and thus incapable of producing gain and degrading the overall gain of the cascaded network.

Progress of Electrical and Electronic Engineering Volume 1 Issue 3 | 2018 | 7

 Input and output impedance matching still contribute to better noise figure performance of the circuit, as indicated
in Figure 8-9 and Table 5. It can also be observed that using ideal inductors for impedance matching produced lower
noise figure as compared to using non-ideal ASITIC inductors through the n2port model. The model block n2port
introduces noise to the system, thus, adding to the total noise figure of the circuit

5. Conclusions and Recommendations

A design of direct-conversion double-balanced mixer was implemented and optimized on this paper.
Proper biasing and sizing of all the transistors were necessary to ensure the required mode of operation for all the
transistors. Conversion gain and noise figure were determined to measure the performance of the mixer design.
These performance parameters can be optimized by introducing impedance matching in the circuit. The effect of the
passive balun in the mixer design was also noted, resulting to the decrease in the conversion gain of the overall circuit.
Ultimately, the direct-conversion double-balance mixer design achieved conversion gain of 11.463dB and noise figure
of 16.529dB (using Pnoise analysis) at 5GHz, comparable to other mixer designs from past researches.
For future studies, an active balun can be used instead of passive balun. Active baluns are capable of producing
gain and if cascaded in a double-balanced mixer to supply the differential RF and LO inputs, the overall performance of
the mixer can be improved. Although active baluns are unidirectional converters, they are also used for their
large bandwidth, which is beyond what non-ideal passive baluns can provide[8].

Acknowledgment
The author would like to thank Prof. Maria Theresa De Leon, Ph.D. and Prof. John Richard E. Hizon, Ph.D. of
Microelectronics and Microprocessors Laboratory at University of the Philippines for the technical support during the
course of the study, and to the STMicroelectronics Calamba NPI Team and the Management Team.
References
1. B. Razavi, “RF microelectronics,” Upper Saddle River, NJ, USA: Prentice Hall Press, 1998.
2. W. Namgoong and T. Meng, “Direct-conversion RF receiver design,” IEEE Transactions on Communications, vol.
49, no. 3, March 2001.
3. Y. Zhou, C.P. Yoong, L.S. Weng, Y.J. Khoi, M.C.Y. Wah, K.A C. Moy, and D.W.T. Fatt, “A 5 GHz dual-mode
WiMAX/WLAN direct-conversion receiver,” in Proc. IEEE International Symposium on Circuits and Systems,
May 2006.
4. J.G. Atallah, S. Rodriguez, L.R. Zeng, and M. Ismail, “A direct conversion WiMAX RF receiver front-end in
CMOS technology,” in Proc. International Symposium on Signals, Circuits and Systems, vol. 1, July 2007.
5. J. Y. Lyu and Z.M. Lin, “A 2~11 GHz direct-conversion mixer for WiMAX applications,” TENCON 2007 – IEEE
Region 10 Conference, pp. 1-4, October 2007.
6. T. Lee, “The design of CMOS radio-frequency integrated circuits,” Cambridge: Cambridge University Press,
1998.
7. M. Voltti, T. Koivisto, and E. Tiiliharju, “Comparison of active and passive mixers,” 18th European Conference on
Circuit Theory and Design, pp. 890-893. August 2007.
8. F.R. Gomez, M.T. de Leon, and C.R. Roque, “Active balun circuits for WiMAX receiver front-end,” TENCON
2010 – IEEE Region 10 Conference, pp. 1156-1161, November 2010.
9. T. Tikka, J. Ryynanen, M. Hotti, and K. Halonen, “Design of a high linearity mixer for
direct-conversion base-station receiver,” in Proc. IEEE International Symposium on Circuits and Systems, May
2006.
10. K.W. Hamed, A.P. Freundorfer, and Y.M.M. Antar, “A monolithic double-balanced direct conversion mixer with an
integrated wideband passive balun,” IEEE Journal of Solid-State Circuits, vol. 40, no. 3, pp. 622-629, March 2005.
11. J. Park, C.H. Lee, B.S. Kim, and J. Laskar, “Design and analysis of low flicker-noise CMOS mixers for
direct-conversion receivers,” IEEE Transactions on Microwave Theory and Techniques, vol. 54, no. 12, December
2006.
12. F.R. Gomez, “Design of impedance matching networks for RF applications,” Asian Journal of Engineering and
Technology, vol. 6, no. 4, September 2018.
13. A.M. Niknejad and R.G. Meyer, “Analysis and optimization of monolithic inductors and transformers for RF ICs,”
in Proc. IEEE Custom Integrated Circuits Conference, Santa Clara, CA, pp. 375-378, May 1997.
14. R.G. Meyer and A.M. Niknejad, “ASITIC for Windows NT/2000,” Research in RFIC Design,

8 | Frederick Ray I. Gomez Progress of Electrical and Electronic Engineering

 http://rfic.eecs.berkeley.edu/~niknejad/Asitic/ grackle/cygwin_info.html.
15. Stanford Microwave Integrated Circuits Laboratory. Integrated Spiral Inductor Calculator,
http://www-smirc.stanford.edu/ spiralCalc.html.
16. M. del Mar Hershenson, S.S. Mohan, S.P. Boyd and T.H. Lee, “Optimization of inductor circuits via geometric
programming,” in Proc. Design Automation Conference, New Orleans, LA, pp. 994-998, June 1999.

Progress of Electrical and Electronic Engineering Volume 1 Issue 3 | 2018 | 9