Beruflich Dokumente
Kultur Dokumente
9, SEPTEMBER 2009
The layouts of the presented CMOS PAs were designed floor plan of the chip layout must be designed based on active
by modeling the interconnection wires in simulation and then and passive device sizing and positioning. Once the devices are
optimizing their widths for minimum parasitic effects and placed on the layout, the length of the interconnections can be
hence optimum measured circuit performance. The intercon- found, and the metal layers can be chosen. The interconnection
nection modeling and layout design considerations resulted models are then added to the circuit schematic in the simulator,
in improvements from 8% to 35% in the drain efficiency of and the effects of the layout parasitics can then be carefully
the class-E PA. The design was further improved by using a accounted for. Generally, minimizing the resistive losses is the
standard CMOS process with a 2-m-thick top-metal layer, major concern in PA designs. However, in the case of tuned
which was double the thickness of the first design. This thicker amplifiers, such as class-E and class-F, the parasitic inductances
metal resulted in an increase in inductor quality factor and, and capacitances can significantly affect the tuning if not taken
hence, an increase in drain efficiency from 35% to 55%. Due to into account during the simulations. Once the parasitic compo-
the careful layout design and interconnection optimization, the nents are taken into account, the circuit component values can
implemented circuits had a good match between the measured be fine tuned, and the performance can be enhanced. Finally, a
and simulated performance characteristics. layout-versus-schematic step is necessary to ensure that there
This paper is organized as follows: Section II contains a are no errors in the layout.
description of the interconnection modeling and circuit simu- The work in this paper focuses on using a simple three-
lation procedure. In Section III, the PA layouts and layout de- element T-section equivalent circuit model of the interconnec-
sign considerations following the parasitic-aware circuit design tion. Fig. 3(a) shows a screen capture of the InterConnect
approach are presented and discussed. The LNA design and application interface that was designed to calculate the in-
measurement results that were used to verify the parasitic-aware terconnection equivalent circuit component values. The figure
design approach are also presented in this section. Finally, the also shows the equivalent circuit model used to represent the
conclusions are stated in Section IV. interconnections. The InterConnect program takes layout in-
formation regarding the interconnections width (w), length
II. INTERCONNECTION MODEL AND DESIGN PROCEDURE (l), and metal layer used, and produces the value of the ca-
pacitance, resistance, and inductance of the interconnection
On-chip interconnections add parasitics to the RF circuit, parasitics based on the following equations, where Lmetal is in
which can have various effects, and depending on the RF nanohenries [7]:
circuit being designed, these effects need to be traded off.
The tradeoff between the parasitic capacitance, inductance, and Rmetal = R l/w (1)
resistance of interconnections can be optimized by modeling Cmetal = C w l (2)
every interconnection with a lumped element circuit model, Lmetal = 2l [ln (2l/(w + t)) 0. 50049 + (w + t)/3l] (3)
which is generally not available from the foundries.
where Rmetal , Cmetal , and Lmetal are the interconnection
equivalent resistance, capacitance, and inductance, respectively.
A. Interconnection Model
For the selected metal layer, t, R , and C are the interconnec-
Fig. 2 shows a flowchart of the RF design procedure that was tions thickness, sheet resistance, and capacitance of a single
used in this paper [12][17]. After obtaining a schematic of the metal square, respectively, which are usually provided in the
circuit design that achieves the desired design specifications, a foundry documentation supplied to the designers.
1884 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 56, NO. 9, SEPTEMBER 2009
B. Simulation Setup
The PA circuit that was designed to operate at 2.4 GHz for
low-voltage biomedical applications consisted of two stages:
a class-E output stage and a class-F driver stage. Fig. 4(a)
shows the schematic of the designed two-stage PA with all the
components fully integrated. Class-E PAs have an ideal drain
efficiency of 100% when the active device acts as a switch with
zero ON-resistance [24]. For optimum class-E operation, the
following three criteria must be met [24], [25]:
1) the voltage across the switch in the OFF-mode should not
rise until after the transistor is OFF;
2) the voltage across the switch should go back to zero
immediately before turn-ON;
3) the slope of the switch voltage should be zero at turn-on
to achieve soft switching.
Many design equations exist to aid in achieving optimum
Fig. 3. (a) Screen capture of the InterConnect application interface showing operation by accurately tuning the output network [25], [26].
the three-component equivalent circuit model used. (b) Equivalent circuit used
in [14] to model and meander-line resistor. (c) Cross-sectional example of the
The theory of the class-F driver stage PA is based on peaking
upper four metal layers of a six-metal-layer CMOS process showing the vias. the odd harmonics and attenuating the even ones to obtain a
square-shaped waveform. This results in less overlap between
Using the equivalent circuit shown in Fig. 3(a) is sufficient
the voltage and current signals, which leads to less power
for applications at frequencies below 10 GHz, after which, the
dissipation in the active device. In theory, class-F can achieve
lumped element model must be refined [18]. At higher frequen-
very high efficiencies; however, the networks used for harmonic
cies, i.e., above 10 GHz, the resistance of the metal intercon-
peaking must be accurately tuned.
nections begins to rise due to skin and proximity effects, which
require the use of frequency-dependent elements to model the Class-F has proven to be a good driver for class-E PAs, since
interconnections resistance and inductance [19]. In addition, class-E is more efficient when driven with a square wave [27],
since every interconnection is characterized separately, issues [28]. The detailed operation of the circuit shown in Fig. 4(a)
such as interconnection overlap capacitance, crosstalk, and is described in [16] and [17]. From the brief explanation given
antenna effects are not accounted for. However, these effects of how this PA topology operates, it can be realized that the
can be avoided by careful placement of the components during successful operation of this circuit depends greatly on the how
the layout. The substrate resistance, which is neglected in this well the components are tuned. This, in addition to the effect of
model, can be justified by careful layout and ensuring that resistive losses on the PAs drain efficiency, makes this circuit
there are abundant substrate ground contacts near all the circuit extremely sensitive to layout parasitics.
components. By following the parasitic-aware design procedure explained
A more advanced model, such as the one shown in Fig. 3(b) in Section II-A, the schematic representation shown in Fig. 4(b)
[20], [21], can be used for frequencies above 10 GHz, which can be obtained, where there is an interconnection model repre-
takes into account the substrate resistance (Repi ). This model senting every on-chip interconnection wire. It is noted that this
EL-DESOUKI et al.: IMPACT OF ON-CHIP INTERCONNECTIONS ON CMOS RF INTEGRATED CIRCUITS 1885
Fig. 5. Photomicrograph of (a) the first fabricated PA design and (b) the
improved design. These designs include an output filter, which is why there
is an additional inductor when compared to (c), which is the PA fabricated with
a thick top-metal option [16].
Fig. 10. Photomicrograph of the 5-GHz LNA showing the variability of the
interconnection widths [23].
[27] T. Sowlati, C. A. T. Salama, J. Sitch, G. Rabjohn, and D. Smith, Low M. Jamal Deen (F02) was born in Georgetown,
voltage, high efficiency class E GaAs power amplifiers for mobile com- Guyana, South America. He received the Ph.D. de-
munications, in Proc. Tech. Dig. Gallium Arsenide Integr. Circuit Symp., gree in electrical engineering and applied physics
Oct. 1994, pp. 171174. from Case Western Reserve University, Cleveland,
[28] C. C. Ho, C. W. Kuo, C. C. Hsiao, and Y. J. Chan, A fully integrated OH, in 1985. His Ph.D. dissertation was on the
class-E CMOS amplifier with a class-F driver stage, in Proc. IEEE RFIC design and modeling of a new CARS spectrometer
Symp. Dig., Jun. 2003, pp. 211214. for dynamic temperature measurements and combus-
tion optimization in rocket and jet engines, and was
sponsored by NASA, Cleveland.
He is currently a Professor of electrical and
computer engineering with McMaster University,
Hamilton, ON, Canada, and holds of the Senior Canada Research Chair in infor-
mation technology. His research interests are microelectronics/ nanoelectronics,
optoelectronics, nanotechnology, and their emerging applications.
Dr. Deen was a Fulbright Scholar (under the Latin American scholarship
program) from 1980 to 1982, an American Vacuum Society Scholar from 1983
to 1984, and a Natural Science and Engineering Research Council of Canada
Senior Industrial Fellow in 1993. He was awarded the 2002 Thomas D. Callinan
Award from the Electrochemical Society, the Distinguished Researcher Award,
Munir M. El-Desouki (S03) was born in Toronto, Province of Ontario, in July 2001, a Humboldt Research Award in 2006, an IBM
Canada, in 1980. He received the B.Sc. degree Faculty award in 2006, and the Eadie Medal from the Royal Society of Canada
in electrical engineering in 2002 from King Fahd in 2008, and has won seven best paper awards. He is a Distinguished Lecturer
University of Petroleum and Minerals (KFUPM), of the IEEE Electron Devices Society. His research record includes about
Dhahran, Saudi Arabia, and the M.A.Sc. degree in 400 peer-reviewed articles (83 are invited), 14 invited book chapters, and
electrical engineering and the M.Eng. degree in en- 7 awarded patents. He is currently an Editor of the IEEE TRANSACTIONS
gineering entrepreneurship and innovation in 2006 ON E LECTRON D EVICES , Executive Editor of Fluctuations and Noise Letters,
and 2007, respectively, from McMaster University, and a member of the Editorial Board of The Journal of Nanoscience and
Hamilton, ON, Canada, where he is currently work- Nanotechnology, the Microelectronics Journal, and the International Journal
ing toward the Ph.D. degree in electrical engineering. of High Speed Electronics and Systems. He has been elected a Fellow of
During the summer of 2001, he was a Field The Royal Society of Canada (FRSC)The Academies of Arts, Humanities
Engineer with General Electric Medical Systems. Since 2002, he has been and Sciences of Canada, a Fellow of the Canadian Academy of Engineering
with the Computer and Electronics Institute, King Abdul-Aziz City for Science (FCAE), a Foreign Fellow of the Indian National Academy of Engineering
and Technology (KACST), Riyadh, Saudi Arabia, where he is currently an (FINAE-Foreign), a Fellow of The American Physical Society (FAPS), a Fellow
Electronics Researcher attending McMaster University with a scholarship from of The Electrochemical Society (FECS), a Fellow of The American Association
KACST to complete his graduate studies. In April 2006, he was awarded the for the Advancement of Science (FAAAS), a Fellow of The Engineering
Ontario Graduate Scholarship award, and in March 2006, he was awarded Institute of Canada (FEIC), and an Honorary Member of the World Innovation
the Natural Science and Engineering Research Council of Canada three-year Foundationthe foundations highest honor.
doctorate award. He was also awarded one patent on a project completed
during his B.Sc. degree at KFUPM. His current research interests include high-
sensitivity and ultrahigh-speed CMOS imagers and camera-on-a-chip design
for biomedical applications. His previous research interests include high-
efficiency circuits and building blocks for low-power microwave transceivers,
mixed-signal IC design, biotelemetry, and wireless sensor networks.
Mr. El-Desouki was a member of the Saudi Engineering Committee from
1998 to 1999 and the Saudi Osteoporosis Club from 1999 to 2001. He has also
been a member of the Canadian Blood Services since 2005.