3736 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO.

11, NOVEMBER 2015

A 77-GHz FMCW Radar System Using On-Chip

Waveguide Feeders in 65-nm CMOS
Chenglin Cui, Student Member, IEEE, Seong-Kyun Kim, Member, IEEE, Reem Song, Member, IEEE,
Jae-Hoon Song, Student Member, IEEE, Sangwook Nam, Senior Member, IEEE, and
Byung-Sung Kim, Member, IEEE

Abstract—This paper presents a 77-GHz radar system using The mm-wave frontend of radar system requires high per-
a transmitter (Tx) and receiver (Rx) integrated with on-chip formance, high level of integration, low power, and low manu-
waveguide feeders in 65-nm CMOS. The newly proposed on-chip facturing cost. The main challenge for CMOS technology com-
waveguide feeder shows the insertion loss of about 2 dB and
more than 30% bandwidth. Additionally, both the Tx and Rx are pared to SiGe is the performance. A comparison of CMOS and
integrated with internal 10 frequency multipliers. Therefore, SiGe technologies for car radar applications can be found in [8].
a radar system can be easily implemented without sensitive mil- Advanced CMOS technology is essential to meet the ,
limeter-wave packaging technology by mounting the Tx and Rx requirements for mm-wave design. The noise performance, es-
chips on the waveguide aperture. The interconnections for the pecially the phase noise and noise, is another challenge be-
low-frequency reference and baseband signals can be realized on
the low-cost FR-4 printed circuit board. The radar system shows cause of the homodyne structure of frequency-modulated con-
9-dBm output power and 13-dB down-conversion gain from the tinuous-wave (FMCW) radar system.
waveguide port. In general, the CMOS process shows poorer phase-noise and
Index Terms—Automotive radar, CMOS, frequency multipliers, flicker-noise performance than the SiGe process. However, with
millimeter-wave (mm-wave) packaging, on-chip feeder. advanced design methods, the CMOS voltage-controlled os-
cillator (VCO) showed improved phase-noise performance of
95 dBc/Hz at 1-MHz offset from 90.3 GHz using a Colpitts
I. INTRODUCTION quadrature VCO [9] and 109 dBc/Hz at 10-MHz offset from
77 GHz using high- coplanar-waveguide (CPW) transmis-

R ECENTLY, car radar sensors are increasingly used for

avoiding critical traffic accidents, driving assistance,
and finally, autonomous driving in the near future. Compared
sion-line tank [10]. An alternative solution for good phase noise
is the frequency multiplier based synthesizer [11]–[16]. Phase
noise of multipliers mainly relies on reference phase noise. The
to other radars used for aerospace and military systems, car reported phase-noise characteristics at 1-MHz offset are about
radar sensors should have small size and consume low power. 108 dBc/Hz at 93 GHz using a frequency tripler and an ex-
More importantly, low-cost implementation is essential to ternal source [14], 110 dBc/Hz at 78 GHz using a 10 fre-
be equipped even in compact cars. Recent development of quency multiplier and an external source [15], and 93 dBc/Hz
long-range radar systems at 77 GHz is mainly based on the SiGe at 96 GHz using an integrated tripler and phase-locked loop
technologies with planar packaging using low-loss substrates (PLL) [16].
[1]. Considering the level of integration, power consumption, Besides the challenges from the current CMOS technology,
and manufacturing cost for mass production, the CMOS tech- packaging is a critical issue in mm-wave systems regard-
nology seems to be an alternative solution since the continued less of the technologies. The most popular technique is the
development of CMOS technology successfully demonstrates planar packaging using the wire or flip-chip bonding. The
its millimeter-wave (mm-wave) applications for 60-GHz wire- wire bonding requires careful control of the wire bonds and a
less data link and 77-GHz automotive radar systems [2]–[7]. low-loss substrate to reduce the transition loss. The flip-chip
packaging shows excellent RF performance [1], [17]. However,
Manuscript received March 30, 2015; revised July 04, 2015 and August 17, the cost and the manufacturing complexity are increased. To
2015; accepted August 29, 2015. Date of publication September 25, 2015; date
avoid the high cost, loss, and sensitivity of packaging for silicon
of current version November 03, 2015. This work was supported by the Korea
Government (MSIP) under National Research Foundation of Korea (NRF) mm-wave systems, on-chip antennas have been developed [18],
Grant 2012R1A2A2A06046474. [19]. However, their gain and efficiency are not sufficient for
C. Cui, R. Song, and B.-S. Kim are with the College of Information and Com-
long-range sensing and communications.
munication Engineering, Sungkyunkwan University, Suwon 440-746, Korea
(e-mail: [email protected]). This work focuses on finding the solution to circumvent
S.-K. Kim was with the College of Information and Communication Engi- the sensitive packaging and improve phase noise of CMOS
neering, Sungkyunkwan University, Suwon 440-746, Korea. He is now with
mm-wave systems. This work presents a frequency multiplier
the Electrical and Computer Engineering Department, University of California
at Santa Barbara, Santa Barbara, CA 93106 USA. based 77-GHz CMOS radar transmitter (Tx) and receiver (Rx)
J.-H. Song and S. Nam are with the School of Electrical and Computer Engi- with improved phase noise. In addition, this work proposes a
neering and INMC, Seoul National University, Seoul 151-742, Korea.
novel on-chip waveguide feeder on silicon substrate, which
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. enables the easy integration of radar systems free of mm-wave
Digital Object Identifier 10.1109/TMTT.2015.2477343 packaging issues. The proposed radar system is composed of

0018-9480 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
CUI et al.: 77-GHz FMCW RADAR SYSTEM USING ON-CHIP WAVEGUIDE FEEDERS IN 65-nm CMOS 3737

Fig. 1. Proposed FMCW radar architecture.

separate Tx and Rx chipsets operating at 77 GHz, as shown

in Fig. 1. The Tx consists of a power amplifier (PA), 10
frequency multiplier, and on-chip waveguide feeder. The Rx
includes a low-noise amplifier (LNA), down-conversion mixer,
10 multiplier, and on-chip waveguide feeder. The on-chip
waveguide feeders are directly mounted on the waveguide
Fig. 2. Proposed on-chip waveguide feeder. (a) 3-D view and metal layers.
apertures without additional interconnects. Therefore, the (b) Electrical equivalent circuit of the proposed waveguide feeder. (c) Three
signal of the highest frequency requiring wire interconnects to simulation structures to compare the impedance .
the external world is the reference signal at 7.7 GHz.
To implement the proposed system, the Tx and Rx chipsets
are designed using a 65-nm low-power RF CMOS process with metallization, which are not available in the standard CMOS
1-poly 9-metals. NMOS field-effect transistors (FETs) of the process. Instead of the thru-wafer vias, this work introduces the
process biased at 250- A m current density show and segmented quarter-wave open stub on the chip ground plane in
of 126 and 225 GHz, respectively. the silicon substrate. This composes the microstrip line mode on
This paper is organized as follows. In Section II, details about the waveguide wall and provides the contactless shorting path to
the design of the on-chip waveguide feeder and electromagnetic the waveguide wall from the chip ground. The proposed on-chip
(EM) simulation results are presented. In Section III, designs feeder for the WR-10 waveguide is shown in Fig. 2. The feeder
of the frequency multiplier, and the Rx and Tx chipsets with is directly mounted on the waveguide aperture with a back-
the on-chip feeders are explained. Measurement results of the shorting metal block. Theoretically, the back-short should be
designed chipsets and modules will be shown in Section IV. The placed at above the feeder, where is the guided wave-
radar system implementation and its experimental results will be length in the waveguide. In practice, the height of the back-
presented in Section V followed by a conclusion in Section VI. short location and the dimensions of the waveguide probe are
adjusted to achieve the impedance matching. The metal layers
II. DESIGN OF ON-CHIP WAVEGUIDE FEEDER used in the design are shown in Fig. 2(a). The top aluminum
The proposed on-chip waveguide feeder is composed of an layer is used for the probe and the feeding line because better re-
-plane waveguide probe and two quarter-wave open stubs on turn loss is obtained than the case of using the thick copper layer.
the chip ground plane. These parts are integrated on a silicon Two lowest stacked metal layers form the chip ground plane of
substrate to directly feed the external high-gain antennas, such the active circuit design, as shown in Fig. 2(a). The ground open
as waveguide horn or waveguide slot array antennas without stub uses entire metal layers to reduce the metallic loss. The
additional mm-wave interconnects. electrical equivalent circuit of the proposed waveguide feeder
The on-chip rectangular waveguide feeders have been devel- is given in Fig. 2(b). The system ground reference is formed
oped for compound monolithic microwave integrated circuits on the waveguide wall. The impedance is the grounding
[20], [21]. However, their development required thru-wafer vias impedance through wafer between the chip ground node and
from the chip ground to the waveguide wall and additional back waveguide wall ground. The impedance is the input

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
3738 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 11, NOVEMBER 2015

Fig. 4. Proposed on-chip waveguide feeder. (a) Top view in -plane. (b) Side
view in -plane.

Fig. 3. Simulation results of between chip ground node and waveguide

wall ground.

impedance seen by the probe between the probe feeding node

on the chip and the waveguide wall. The ground open stub
guarantees low regardless of the size of the chip ground
plane. To confirm the effectiveness of the ground open stub,
EM simulations for three cases were performed with a port be-
tween the chip ground node and the waveguide wall, as de-
picted in Fig. 2(c). According to the simulation results shown
in Fig. 3, the ground open stub successfully provides a low- Fig. 5. S-parameter simulation results of the on-chip waveguide feeder using
3-D EM simulator.
impedance path to the waveguide wall at 77 GHz, as expected.
The thru-wafer impedance hardly changes even with the chip
ground plane, whereas the solid chip ground plane without the
stub shows much higher inductive impedance, which severely
changes depending on the dimension of the solid chip ground
plane. The nonzero resistive impedance of the ground open stub
is mainly due to the conductivity of the silicon substrate and it
introduces about 0.4-dB transition loss for feeder operation as-
suming an ideal lossless 50- probe.
A U-shaped -plane waveguide probe is devised to reduce
the probe length for saving the die area and achieve an effective
wideband matching. As shown in Fig. 2(a), the ground open stub
is parallel to the bottom section of the U-shaped probe, which
ensures the return current to mostly flow through the open stub Fig. 6. -field distribution in: (a) -plane and (b) -plane.
rather than back-spread to the wide chip ground plane. The top
and side views are shown in Fig. 4(a) and (b) with detailed di-
mensions. The bottom of the silicon substrate of the feeder faces with 35.3-GHz BW, respectively. The wideband matching is
the forward direction of the waveguide. To minimize the sub- achieved without additional matching circuitry. In case of a
strate loss, the substrate is thinned to 50- m thickness, which lossless silicon substrate, the insertion loss of the proposed
still maintains the mechanical ruggedness of the wafer. feeder is reduced to 0.36 dB. Fig. 6 shows the electric field
3-D EM simulation was carried out to estimate the perfor- distribution in the -plane and -plane, which confirms that
mance of the proposed on-chip feeder. As shown in Fig. 4(b), the transmission mode is successfully excited.
one port is at the on-chip feeding point (Port 1) and the other Concerning the assembling tolerance for the position of the
port is placed 5 mm away from the feeder at the waveguide side die attach, EM simulations were performed with various chip
(Port 2). The insertion loss from Port 1 to Port 2 is 1.37 dB at positions within 50- m misalignment in all directions. Simula-
77 GHz, as shown in Fig. 5. The input return losses at Port tion results show that misalignment along the -direction [see
1 and Port 2 are better than 15 dB from 60.3 to 90.4 GHz Fig. 4(a)] increases the insertion loss less than 0.1 dB. For the
with 30.1-GHz bandwidth (BW), and from 62 to 97.3 GHz change in the forward -direction, the insertion loss increases

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
CUI et al.: 77-GHz FMCW RADAR SYSTEM USING ON-CHIP WAVEGUIDE FEEDERS IN 65-nm CMOS 3739

Fig. 8. Block diagram of 10 frequency multiplier.

Fig. 7. Simulation results of radiation patterns of WR10 aperture fed by the

proposed on-chip feeder. The length of the waveguide is 1 mm. Elevation pattern
( -plane), azimuthal pattern ( -plane).

less than 0.2 dB. However, moving backward in the -direction

causes a 0.4-dB increase because it changes the probe length
with respect to the ground wall of the waveguide. Therefore,
the chip should be carefully mounted so that the space in
Fig. 4(a) is not overlapped on the waveguide flange.
The proposed on-chip waveguide feeder can be used with and
without external antennas depending on the required gain for
system implementation. When only the waveguide aperture is
used as an antenna, an ideal mode excitation gives the
gain of 6.5 dBi according to EM simulation. However, due to
the loss of the on-chip feeder, the gain of the aperture antenna
driven by the proposed feeder shows the reduced gain of 5.2 dBi,
as confirmed by the simulation results in Fig. 7. Nonetheless
this result indicates that the proposed feeder on the conducting
package with a rectangular aperture can be used as an antenna as
it is. This configuration can be a very low-cost solution for short-
range radar sensors for side-looking or short-range wireless data
link systems.
Since back-to-back measurements of the proposed feeders are
not plausible due to its short length, we designed the 77-GHz
Tx and Rx with and without on-chip feeders to confirm the
performance of the proposed feeder. The measurement results Fig. 9. Schematic of 10 frequency multiplier. (a) STD circuitry and ring
and the comparison with and without feeders will be given in OSC. (b) Five-push frequency adder. (c) Push–push frequency doubler.
Section IV.

(OSC), five-push adder, and push–push doubler. The schematic

III. DESIGN OF FREQUENCY MULTIPLIER, Rx, AND Tx is shown in Fig. 9.
The input reference is converted to the differential signal
A. Frequency Multiplier Design
through the inverter-based STD circuitry, as shown in
As we explained in Section I, the frequency multiplier is Fig. 9(a).The free-running operating frequency of the ring OSC
adopted instead of PLLs for local oscillator (LO) signal gen- is set to 7.7 GHz. The differential reference signal is injected
eration because of its better phase noise and smaller size. The into one of the ring OSC cells shown in Fig. 9(a) and the
reproduced block diagram of the 10 frequency multiplier is injection-locked five differential signals with phase offset
shown in Fig. 8, which was published in [15]. The 10 fre- at 7.7 GHz are obtained. The 7.7-GHz voltage signals with rich
quency multiplier is composed of a single-to-differential (STD) harmonic components from the ring OSC are combined in the
converter, differential five-phase injection-locked ring oscillator current domain by the following adder stage. The adder is a

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
3740 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 11, NOVEMBER 2015

Fig. 10. Frequency multiplier based Tx and Rx with on-chip feeders. (a) Rx with an on-chip feeder. (b) Tx with an on-chip feeder.

common-source (CS) differential amplifier sharing the output Fig. 9(c). The second harmonic signal generated at the common
load, as shown in Fig. 9(b). Assuming the linear operation of node is injected into the injection-locked OSC to obtain the dif-
the adder, the harmonic currents are combined at the common ferential output signal. Dummy cells are used to improve the
output of the adder. In principle, due to the phase relation of output balance.
the fundamental and the harmonic currents, only the harmonic There is a tradeoff between the lock BW and the output swing
currents of fifth and higher orders are constructively added and depending on the quality factor of the LC tank for both multi-
other lower harmonic currents are canceled out [12]. plier stages. With the high quality factor, the high output swing
For an ideal ring OSC output, the combined harmonic com- is obtained at the cost of lock BW.
ponents at the output of the adder generate the magnitude as
large as that of the fundamental signal. However, in practice, B. Rx Design
they are attenuated due to the low-pass nature of active devices The LNA and the mixer in Fig. 10(a) are based on the work
for both of the ring OSCs and adder amplifiers. Additionally, demonstrated in [23]. The LO signal for the mixer is driven
asymmetry caused by the injection cell in the ring OSC results by the 10 multiplier integrated in the Rx. The on-chip feeder is
in uneven skew of the multi-phase output of the ring OSC, which directly attached to the input of the LNA. The LNA consists of
causes the unwanted spurious signals. Therefore, although the three-stage CS amplifiers. Two lowest metal layers are stacked
large number of ring OSC stages is preferred for higher multipli- as ground plane and 3- m-thick top copper and 1.3- m-thick
cation ratio, the number of the stage should be carefully selected aluminum layers are stacked as the signal line. The signal line
considering the magnitude and spurious signal generation. The is 4.6 m wide and 3.8 m above the ground plane. The trans-
strength of the fifth harmonic component is inherently enforced mission line has 32- characteristic impedance and 2-dB/mm
by adequately biasing the adder amplifier. In theory, the fifth insertion loss. The transmission line and the inductor
harmonic current components of class-A and class-B amplifiers are used for the input matching. The inductors and are
are zero, but the class-AB operation produces a nonzero fifth used for inter-stage matching and , , and
harmonic current [22]. Therefore, the class-AB-biased adder work as loads. On-wafer measurement results of the LNA alone
amplifier helps to enhance the fifth-harmonic component. The are given in Fig. 11. The peak gain is 10 dB and the noise figure
fifth harmonic is additionally boosted by the cross-coupled pair (NF) is 7.95 dB at 77 GHz, consuming 22.8 mW. The mea-
and the resonant load. The resonant load also rejects the un- sured gain is far less than the simulated one. Considering the
wanted harmonics and spurs due to the phase and amplitude correspondence between the resonant frequencies, it is consid-
mismatch. Therefore, only the desired 38.5-GHz signal can be ered that the uncounted degeneration inductance of the CS stage
coupled through the transformer to the push–push doubler. The seems to be the main cause of gain reduction.
push–push doubler is implemented by tying the drain nodes of The high-gain mixer with low LO power using a split self-
the differential CS pair and loading this node with a second har- driven switching cell [24] is used as a down-conversion mixer.
monic resonant tank composed of the transformer, as shown in The use of the split cross-coupled switching cell can reduce the

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
CUI et al.: 77-GHz FMCW RADAR SYSTEM USING ON-CHIP WAVEGUIDE FEEDERS IN 65-nm CMOS 3741

Fig. 12. Die micrographs. (a) Rx chip. (b) Tx chip.

Fig. 13. Rx and Tx modules.

Fig. 11. On-wafer measurement results of the standalone LNA. (a) S-parame- voltage swing with low power consumption and reduce capaci-
ters. (b) NF. tive loading of the frequency multiplier. The second stage em-
ploys the cross-coupled neutralization capacitance between the
gate and the drain to improve the stability as well as the gain
required LO power and enhance the gain. The output current [25]. Transformers are used for the impedance matching and
from the LNA is directly applied to the mixer switching core. the coupling for both stages. The output of the PA is directly
The input parasitic capacitance of the mixer is resonated by the connected to the on-chip feeder. The same on-chip feeder used
load of the LNA. Therefore, the chip size and the layout com- in the Rx is attached to the output of the PA without addi-
plexity are significantly reduced because no additional inductor tional tuning. As shown in Section II, the input impedance of
is required to neutralize the parasitic capacitance at the tail of the the feeder was matched to 50 for wide BW. The transformer
switching stage. The die micrograph of the Rx with the feeder is and the TLIN are used to transform 50- load impedance to
shown in Fig. 12(a). The Rx chip occupies 0.91 mm 1.46 mm an optimum impedance.
area including the on-chip feeder. The Rxs with and without the The performance of the PA alone can be referred to the pre-
feeders were fabricated for comparison. The measurement re- vious work [23]. The PA has a gain of 8.5 dB and output power
sults will be given in Section III-C. of 9.6 dBm at saturation. The peak power-added efficiency
(PAE) is 6.45% and dc power consumption is 112 mW. The
C. Tx Design die micrograph of the Tx chip with the on-chip feeder is shown
The frequency multiplier based Tx with the on-chip feeder is in Fig. 12(b). The Tx chip occupies 0.91 mm 1.36 mm area
shown in Fig. 10(b). The 77-GHz output signal of the 10 fre- including the on-chip feeder. The frequency multiplier based
quency multiplier is coupled to the PA by a transformer. The Tx with the on-chip feeder and the one without the feeder
PA is designed using two-stage CS amplifiers [23]. The first were assembled for comparison and the results will be given in
stage includes a cross-coupled pair structure to obtain a high- Section IV.

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
3742 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 11, NOVEMBER 2015

Fig. 16. Fabricated Rx module with an on-chip feeder mounted on the aperture.

Fig. 14. Fabricated Tx module with an on-chip feeder mounted on the aperture.

Fig. 17. Rx module measurement results. (a) Conversion gain with and without
feeder. (b) Input return loss. (c) Measured conversion gain.

configuration in Fig. 1. The horn antennas can be easily mounted

at the apertures of the modules. The assembled modules with
24-dBi horn antennas are shown in Fig. 13.

A. Tx Module
The Tx module using the developed chip is shown in Fig. 14.
The aluminum metal housing is used to build WR10 waveguide
and attach the chip. The chip is directly mounted on the wave-
guide aperture to feed the 77-GHz signal. The other side of the
aperture is a standard UG-387/U waveguide flange to connect
the horn antenna. The back-short of the waveguide is realized
using the metal cover. The reference signal and the dc bias are
distributed using the low cost FR-4 printed circuit board (PCB)
and low-frequency wire bonding. The PCB is installed inside
the metal housing. The reference signal is applied to the module
through an SMA connecter. The dimension of the module is
4 cm 4.5 cm 1.4 cm.
Fig. 15. Tx module measurement results. (a) Output power. (b) Phase noise. The signal generator model HP 8340B is used to provide
a 7.7-GHz reference signal of 0-dBm power to test the Tx
IV. FMCW RADAR SYSTEM IMPLEMENTATION module. The Tx chip without the on-chip feeder is measured
The Tx and the Rx modules were implemented using the using on-wafer probing. The measurement results of the Tx
chips designed in the previous sections. An FMCW radar system module with and without the on-chip feeder are shown in
was simply built using these two modules, commercial horn an- Fig. 15. The module achieves 8.7–9.1-dBm output power within
tennas, and an external signal generation module following the a 76.8–77.4-GHz lock range. The output power is 1.5 dB lower

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
CUI et al.: 77-GHz FMCW RADAR SYSTEM USING ON-CHIP WAVEGUIDE FEEDERS IN 65-nm CMOS 3743

TABLE I
COMPARISON OFTRX PERFORMANCE

than that of the on-wafer measurement of the chip without

the feeder. The output power of the Tx seems to be saturation
power considering the separate PA measurement results [23].
The loss is close to the value of the on-chip feeder simulation
result in Fig. 5. The phase noise of the 77-GHz output signal
measured at the waveguide output is 109.1 dBc/Hz at 1-MHz
offset frequency. The phase noise of the 7.7-GHz reference
signal is 132.3 dBc/Hz at the same offset. The dc power
consumption of the Tx module is 264.4 mW.

B. Rx Module
The Rx module is shown in Fig. 16. The chip is also directly
mounted on the waveguide aperture. The other side of the aper-
ture and the metal cover are the same as those of the Tx module.
The reference signal, the dc bias, and the IF signals are dis-
tributed using the FR-4 PCB. The dimension of the module is 5
cm 4.5 cm 1.4 cm.
For gain measurement, the waveguide input port of the Rx
module is driven by the W-band signal from the ME7838A
vector network analyzer (VNA) and the external signal gener-
ator provides 7.7-GHz reference with 0-dBm input power. The
IF signal is measured using the Agilent N9030A signal analyzer.
For comparison, the Rx chip without the on-chip feeder is mea-
sured using on-wafer probing. The measurement results of the
Rx module with and without the on-chip feeder are shown in
Fig. 17(a). The module achieves 13-dB gain at 77 GHz, which
is 2.1 dB lower than that of the chip without the feeder. The input
Fig. 18. In-door range detection measurement results. (a) Experiment environ-
return loss measured at the waveguide aperture shows frequency ment. (b) Measured IF spectrum with the target 10.5 m away.
up-shift and is 6 dB at 77 GHz, as shown in Fig. 17(b).
The impedance mismatch will lead to additional 0.8-dB loss
compared to the well-matched condition. Taking into account power with 50- waveguide termination at the RF input. The
the mismatch loss, insertion loss of the feeder itself is close to measured system NF is 41 dB at 1 MHz. The estimated flicker
the simulation result in Fig. 5. The Rx gain with different RF fre- corner frequency is larger than 10 MHz and so the system suf-
quencies also shows the frequency shift. As shown in Fig. 17(c), fers from large flicker noise. The dc power consumption of the
the module has 14.8-dB peak gain at around 79 GHz and 3-dB Rx module is 203 mW. A comparison of the performance with
BW of 4.5 GHz for RF input. The Rx NF is roughly estimated published single-channel CMOS transceivers (TRxs) and Tx for
using the spectrum analyzer by measuring the average noise 77-GHz FMCW automotive radar is given in Table I.

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
3744 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 11, NOVEMBER 2015

expected. The signal power was 73 dBm only with the aper-
ture antenna. When the high gain horn antenna was attached
to the Rx module, the signal power was increased by 15 dB to
58 dBm. Though the in-door environment is not anechoic for
accurate comparison, the result may confirm that the gain of the
simple aperture antenna is higher than 5 dBi, as explained in
Section II.
The range detection in the outdoor environment was per-
formed where a sedan served as a target, as shown in Fig. 19(a).
An external 1:1 transformer with 30-kHz–70-MHz BW was
used to combine the differential IF output signals for the
single-ended input of the signal analyzer. The measured IF
signal and the corresponding signal-to-noise ratio (SNR) are
shown in Fig. 19(b). It was clearly observed that the frequency
of the echo signal from the target changed as the car was
moving. As specified in [26], 16-dB SNR was used for deter-
mining the maximum detectable range. We measured the echo
signal with the SNR above 16 dB at 284 kHz with 1-kHz reso-
lution BW and this corresponds to the distance of 85.2 m. The
other signals in Fig. 19(b) remained unchanged as the target
was moving and therefore appeared to be reflected from the
standing-still objects on the sides of the road. The roll-off of
the noise floor spectrum around dc is due to the BW limitation
of the IF transformer.

VI. CONCLUSION
This work has proposed 77-GHz CMOS Tx and Rx chipsets
Fig. 19. Outdoor range detection measurement results. (a) Experiment envi- equipped with the on-chip waveguide feeders. Using the de-
ronment. (b) Measured IF spectrum with the target 85 m away. signed chipsets, a radar system can be simply implemented
without complex mm-wave packaging issues. Though the
single-channel application is demonstrated, the number of
V. SYSTEM RESULTS channels for array radar systems and data TRxs can be easily ex-
panded with good isolation between channels. For short-range
The radar system is tested by the range detection measure- applications, a simple aperture array on a conducting wall
ment. The Rx and Tx modules are located about 50 cm above can configure array TRxs using the proposed on-chip feeder.
the ground surface considering the height of the installation in When the large antenna gain is essential, a waveguide slot
the car. For the FMCW implementation, an accurate chirp ref- array antenna directly fed by the proposed on-chip feeder can
erence signal generation is required. An external module com- be a solution. Cost and mass-production issues to implement
posed of the Analog Device AD9958 direct digital frequency the multi-channel system using metallic waveguide slot array
synthesizer (DDFS) and the 7.7-GHz Hittite HMC764LP6CE antennas are expected to be solved by virtue of the recent
PLL operating in the integer mode is used to obtain the refer- development of cost competitive slot-array antenna systems
ence chirp signal with 50-MHz BW and 1-ms ramp time. The using diffusion bonding [27], electroforming [28] or completely
DDFS generates FMCW waveform at 50 MHz and the signal PCB-based post-wall slot array systems with aperture feeding
is used as the reference for the 7.7-GHz PLL. The final trans- capability [29]. Wide BW nature of the on-chip feeder can also
mitting FMCW signal has 500-MHz BW and 1-ms ramp time, be utilized for high-speed data TRxs in mm-wave applications.
which leads to the range resolution of 30 cm. The PLL phase
noise is 117.8 dBc/Hz at 1-MHz offset and the 77-GHz Tx ACKNOWLEDGMENT
output achieves 97.6-dBc/Hz phase noise. For the Rx mea- The chip fabrication and computer-aided design (CAD) tools
surement, the signal analyzer was used as an IF detector for the used in this work were supported by the IDEC.
test.
At first, in-door measurement was performed using one of the
REFERENCES
differential IF output signals of the Rx, as shown in Fig. 18(a).
[1] J. Hasch, E. Topak, R. Schnabel, T. Zwick, R. Weigel, and C. Wald-
The target was the wall 10.5 m away from the radar and the beat schmidt, “Millimeter-wave technology for automotive radar sensors in
frequency should be 35 kHz accordingly. Although multi-tar- the 77 GHz frequency band,” IEEE Trans. Microw. Theory Techn., vol.
gets were detected, as shown in Fig. 18(b), we can determine 60, no. 3, pp. 845–860, Mar. 2012.
[2] Y. Kawano, T. Suzuki, M. Sato, T. Hirose, and K. Joshin, “A 77 GHz
the echo signal from the target by analyzing the frequency and transceiver in 90 nm CMOS,” in Proc. IEEE Int. Solid-State Circuits
the power level. The highest power was observed at 35 kHz, as Conf., Feb. 2009, pp. 310, 311a–311.

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
CUI et al.: 77-GHz FMCW RADAR SYSTEM USING ON-CHIP WAVEGUIDE FEEDERS IN 65-nm CMOS 3745

[3] T. Mitomo, N. Ono, H. Hoshino, Y. Yoshihara, O. Watanabe, and I. [25] W. L. Chan and J. R. Long, “A 58–65 GHz neutralized CMOS power
Seto, “A 77 GHz 90 nm CMOS transceiver for FMCW radar applica- amplifier with PAE above 10% at 1-V supply,” IEEE J. Solid-State
tions,” IEEE J. Solid-State Circuits, vol. 45, no. 4, pp. 928–937, Apr. Circuits, vol. 45, no. 3, pp. 554–564, Mar. 2010.
2010. [26] M. I. Skolnik, Introduction to Radar Systems. New York, NY, USA:
[4] J. Lee, Y.-A. Li, M.-H. Hung, and S.-J. Huang, “A fully-integrated McGraw-Hill, 2001.
77-GHz FMCW radar transceiver in 65-nm CMOS technology,” IEEE [27] J. Hirokawa, M. Zhang, and M. Ando, “94 GHz fabrication of a
J. Solid-State Circuits, vol. 45, no. 12, pp. 2746–2756, Dec. 2010. slotted waveguide array antenna by diffusion bonding of laminated
[5] T.-N. Luo, C.-H. E. Wu, and Y.-J. E. Chen, “A 77-GHz CMOS auto- thin plates,” in IEEE Sensors, Christchurch, New Zealand, Oct. 2009,
motive radar transceiver with anti-interference function,” IEEE Trans. pp. 907–911.
Circuits Syst. I, Reg. Papers, vol. 60, no. 12, pp. 3247–3255, Dec. 2013. [28] D.-Y. Kim, Y. Lim, H.-S. Yoon, and S. Nam, “High-efficiency W-band
[6] J. Park, H. Ryu, K.-W. Ha, J.-G. Kim, and D. Baek, “76–81-GHz electroforming slot array antenna,” IEEE Trans. Antennas Propag., vol.
CMOS transmitter with a phase-locked-loop-based multichirp modu- 63, no. 4, pp. 1854–1857, Apr. 2015.
lator for automotive radar,” IEEE Trans. Microw. Theory Techn., vol. [29] T. Shijo, K. Hashimoto, S. Obayashi, and H. Shoki, “Single-layer
63, no. 4, pp. 1399–1408, Apr. 2015. slotted post-wall waveguide array with compact feed-line structures
[7] K. Okada et al., “A 64-QAM 60 GHz CMOS transceiver with for 77 GHz automotive radar,” in Proc. Int. Electromagn. Theory
4-channel bonding,” in Proc. IEEE Int. Solid-State Circuits Conf., Symp., Hiroshima, Japan, May 2013, pp. 159–161.
Feb. 2014, pp. 346–347.
[8] A. Margomenos, “A comparison of Si CMOS and SiGe BiCMOS tech-
nologies for automotive radars,” in IEEE Silicon Monolithic Integr.
Circuits RF Syst. Top. Meeting, 2009, pp. 1–4.
[9] E. Laskin, M. Khanpour, R. Aroca, K. W. Tang, P. Garcia, and S. P.
Voinigescu, “95 GHz receiver with fundamental frequency VCO and
static frequency divider in 65 nm digital CMOS,” in Proc. IEEE Int.
Solid-State Circuits Conf., San Francisco, CA, USA, Feb. 2008, pp. Chenglin Cui (S’11) received the B.S. degree in
180–181. telecommunication engineering from the Nanjing
[10] R. Berenguer, G. Liu, A. Akhiyat, K. Kamtikar, and Y. Xu, “A University of Posts and Telecommunications,
117 mW 77 GHz receiver in 65 nm CMOS with ladder struc- Nanjing, China, in 2009, the M.S. degree in elec-
tured tunable VCO,” in Proc. Eur. Solid-State Circuits Conf., tronic, electrical, and computer engineering from
Sep. 2010, pp. 494–497. Sungkyunkwan University, Suwon, Korea, in 2011,
[11] I. Gresham et al., “A compact manufacturable 76–77-GHz radar and is currently working towards the Ph.D. degree
module for commercial ACC applications,” IEEE Trans. Microw. in electronic, electrical, and computer engineering at
Theory Techn., vol. 49, no. 1, pp. 44–58, Jan. 2001. Sungkyunkwan University.
[12] S.-C. Yen and T.-H. Chu, “An th-harmonic oscillator using an His current research interest is in millimeter-wave
-push coupled oscillator array with voltage-clamping circuits,” CMOS integrated circuit design for automotive radar
in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2003, vol. 3, pp. systems.
2169–2172.
[13] N. Mazor and E. Socher, “Analysis and design of an X-band-to-W-
band CMOS active multiplier with improved harmonic rejection,”
IEEE Trans. Microw. Theory Techn., vol. 61, no. 5, pp. 1924–1933,
Seong-Kyun Kim (S’08–M’14) received the
May 2012.
B.S., M.S., and Ph.D. degrees from the College
[14] Z. Chen and P. Heydari, “An 85–95.2 GHz transformer-based injec-
of Information and Communication Engineering,
tion-locked frequency tripler in 65 nm CMOS,” in IEEE MTT-S Int.
Sungkyunkwan University, Suwon, Korea, in 2007,
Microw. Symp. Dig., May 2010, pp. 776–779.
2009, and 2013, respectively.
[15] S.-K. Kim, C. Choi, C. Cui, B.-S. Kim, and M. Seo, “A W-band signal
He is currently a Postdoctoral Research Fellow
generation using -push frequency multipliers for low phase noise,”
with the Department of Electrical and Computer En-
IEEE Microw. Wireless Compon. Lett., vol. 24, no. 10, pp. 710–712,
gineering, University of California at Santa Barbara,
Oct. 2014.
Santa Barbara, CA, USA. His research interests
[16] C.-C. Wang, Z. Chen, and P. Heydari, “W-band silicon-based fre-
include RF and millimeter-wave integrated circuits
quency synthesizers using injection-locked and harmonic triplers,”
for wireless communications and radar systems.
IEEE Trans. Microw. Theory Techn., vol. 60, no. 5, pp. 1307–1320,
May 2012.
[17] A. Jentzsch and W. Heinrich, “Theory and measurements of flip-chip
interconnects for frequencies up to 100 GHz,” IEEE Trans. Microw.
Theory Techn., vol. 49, no. 5, pp. 871–878, May 2001. Reem Song (S’02–M’06) received the Ph.D. degree
[18] K. Kang et al., “A 60-GHz OOK receiver with an on-chip antenna in electrical engineering from the University of
in 90 nm CMOS,” IEEE J. Solid-State Circuits, vol. 45, no. 9, pp. Southern California, Los Angeles, CA, USa, in 2006.
1720–1731, Sep. 2010. From 2007 to 2010, she was with Skyworks So-
[19] Y.-C. Ou and G. M. Rebeiz, “On-chip slot-ring and high- lutions, Inc., Thousand Oaks, CA, USA, where she
gain horn antennas for millimeter-wave wafer-scale silicon sys- was a Senior Engineer involved in the development
tems,” IEEE Trans. Microw. Theory Techn., vol. 59, no. 8, pp. of power amplifiers for cellular applications. Since
1963–1972, Aug. 2011. 2014, she has been with Sungkyunkwan University,
[20] L. Samoska et al., “A submillimeter wave HEMT amplifier module Suwon, Korea, where she performs research on mil-
with integrated waveguide transitions operating above 300 GHz,” limeter-wave circuits, antennas, and systems.
IEEE Trans. Microw. Theory Techn., vol. 56, no. 6, pp. 1380–1388,
Jun. 2008.
[21] M. Urteaga et al., “InP HBT integrated circuit technology for terahertz
frequencies,” in IEEE Compound Semicond. IC Symp., 2010, pp. 1–4.
[22] S. C. Cripps, RF Power Amplifiers for Wireless Communications. Jae-Hoon Song (S’12) received the B.S. de-
Boston, MA, USA: Artech House, 1999. gree in semiconductor systems engineering from
[23] S.-K. Kim, C. Cui, S. Nam, and B.-S. Kim, “A low-power 77 Sungkyunkwan University, Suwon, Korea, in 2010,
GHz transceiver for automotive radar system in 65 nm CMOS the M.S. degree in electrical engineering and com-
technology,” in Proc. Asia–Pacific Microw. Conf., Nov. 2013, puter science from Seoul National University, Seoul,
pp. 236–238. Korea, in 2012, and is currently working toward the
[24] S.-K. Kim, C. Cui, G. Huang, S. Kim, and B.-S. Kim, “A 77 GHz low Ph.D. degree at Seoul National University.
LO power mixer with a split self-driven switching cell in 65 nm CMOS His research interests include CMOS implementa-
technology,” IEEE Microw. Wireless Compon. Lett., vol. 22, no. 9, pp. tion of millimeter-wave radar.
480–482, Sep. 2012.

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.
3746 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 11, NOVEMBER 2015

Sangwook Nam (S’87–M’88–SM’11) received the Byung-Sung Kim (S’96–A’98–M’03) received

B.S. degree from Seoul National University, Seoul, the B.S., M.S., and Ph.D. degrees in electronic
Korea, in 1981, the M.S. degree from the Korea engineering from Seoul National University, Seoul,
Advanced Institute of Science and Technology Korea, in 1989, 1991, and 1997, respectively.
(KAIST), Seoul, Korea, in 1983, and the Ph.D. de- In 1997, he joined the College of Information
gree from The University of Texas at Austin, Austin, and Communication Engineering, Sungkyunkwan
TX, USA, in 1989, all in electrical engineering. University, Suwon, Gyeonggi-Do, Korea, where he
From 1983 to 1986, he was a Researcher with is currently a Professor. In 2013, he was a Visiting
the Gold Star Central Research Laboratory, Seoul, Researcher with the University of California at Santa
Korea. Since 1990, he has been a Professor with the Barbara. His research interests include high-fre-
School of Electrical Engineering and Computer Sci- quency device modeling and RF/millimeter-wave
ence, Seoul National University. His research interests include analysis/design CMOS integrated circuit design.
of electromagnetic (EM) structures, antennas, and microwave active/passive
circuits.

Authorized licensed use limited to: National Taiwan University. Downloaded on September 25,2021 at 09:15:45 UTC from IEEE Xplore. Restrictions apply.