![Mechanical overview of the radio unit [9].](https://www.researchgate.net/profile/Marko-Leinonen-2/publication/323220180/figure/fig4/AS:622644097454085@1525461329381/Mechanical-overview-of-the-radio-unit-9_Q320.jpg)
Content uploaded by Marko Leinonen
Author content
All content in this area was uploaded by Marko Leinonen on Aug 20, 2018
Content may be subject to copyright.
28 GHz Wireless Backhaul Transceiver Characterization
and Radio Link Budget
Marko E. Leinonen , Giuseppe Destino, Olli Kursu, Marko Sonkki, and Aarno P€
arssinen
Millimeter wave communication is one of the main
disruptive technologies in upcoming 5G mobile
networks. One of the first candidate applications,
which will be commercially ready by 2020, is wireless
backhaul links or wireless last mile communication.
This paper provides an analysis of this use-case from
radio engineering and implementation perspectives.
Furthermore, preliminary experimental results are
shown for a proof-of-concept wireless backhaul
solution developed within the EU-KR 5GCHAMPION
project, which will be showcased during the 2018
Winter Olympic Games in Korea. In this paper, we
verify system level calculations and a theoretical link
budget analysis with conductive and radiated over-
the-air measurements. The results indicate that the
implemented radio solution is able to achieve the target
key performance indicator, namely, a 2.5 Gbps data
rate on average, over a range of up to 200 m.
Keywords: Antenna array, Antenna pattern, Array
receiver, Millimeter wave radio, Noise figure, Radio
link budget.
I. Introduction
One of the key drivers for the development of the next
generation of communication systems, that is, 5G, is the
demand for communication data rates that are 10 times
higher than the current long term evolution (LTE;
20 Gbps and 10 Gbps peak rates for downlink and uplink,
respectively). A promising and concrete approach is to
exploit the large amount of available spectrum in the K
a
–
band (26.5 GHz–40 GHz) [1] to allow the transmission of
wide-band signals (from 100 MHz to 1 GHz bandwidth).
However, this necessitates the usage of large antenna
arrays with adaptive beamforming capabilities to combat
the path loss and sensitivity of the radio links to dynamic
obstructions. In [2], for instance, the results of a channel
measurement campaign at 28 GHz
1)
show how the line-
of-sight and multipath components of the radio channel
are highly affected by moving objects such as cars.
In the literature as well as prototype demonstrations,
only a few implementations of 28 GHz transceivers have
been proposed. For instance, in [3], a 28 GHz CMOS
direct conversion transceiver with eight antennas
integrated in the same package was proposed. However,
the transmission power is limited because the saturated
output power (Psat) of a single amplifier is 10.5 dBm. This
limits the conducted transmission power to 3 dBm with
LTE modulation. In [4], a 32-path transceiver was
presented with a 16 dBm Psat output power from each
path, which results in ~6 dBm conducted modulated signal
power with 10 dB peak-to-average ratio.
In this work, the focus is on the design and validation of
a radio frequency (RF) transceiver [5] operating in the
Manuscript received Oct. 10, 2017; revised Dec. 4, 2017; accepted Dec. 18, 2017.
Marko E. Leinonen (corresponding author, marko.e.leinonen@oulu.fi), Giuseppe
Destino (giuseppe.destino@oulu.fi), Olli Kursu (olli.kursu@oulu.fi), Marko Sonkki
(marko.sonkki@oulu.fi), and Aarno Pärssinen (Aarno.Parssinen@oulu.fi) are with
Centre for Wireless Communications, University of Oulu, Finland.
This is an Open Access article distributed under the term of Korea Open
Government License (KOGL) Type 4: Source Indication +Commercial Use
Prohibition +Change Prohibition (http://www.kogl.or.kr/info/licenseTypeEn.do).
1)
The frequency band at 28 GHz is not allocated for mobile services in Europe.
However, due to the strong opening from other countries (US, Korea, and Japan),
the European Commission recommends the 24.25 GHz–27.5 GHz as a pioneer
band for 5G above 24 GHz.
https://doi.org/10.4218/etrij.2017-0231 ©2018 pISSN: 1225-6463, eISSN: 2233-7326
89ETRI Journal, Volume 40, Number 1, February 2018
http://onlinelibrary.wiley.com/journal/10.4218/(ISSN)2233-7326
frequency band of 27.5 GHz28.5 GHz. This band
represents a frequency band partially endorsed in common
by the European, Korean, US, and Japanese Frequency
Spectrum Authorities, and it is the specification for the 5G
trials at the Winter Olympics in Korea [6].
From a design perspective, the proposed RF
implementation can be considered one of the first
complete solutions capable of achieving millimeter wave
(mmW) urban backhaul requirements, that is, medium-
long range coverage and robustness to dynamic changes
in the environment. In contrast to [3] and [4], which
are integrated circuit (IC) solutions, the proposed
implementation is able to produce a conductive power of
more than 30 dBm. Additionally, our solution can
produce 60 dBm effective isotropic radiated power
(EIRP), while the IC based transceivers provide 24 dBm
EIRP [3] and 28 dBm [4]. In addition, the proposed
solution supports higher multiple-input and multiple-
output (MIMO) capabilities.
Furthermore, from a validation and experimentation
viewpoint, we show how to circumvent the problem of
measuring the noise figure of an antenna array receiver.
More specifically, we propose a new noise figure
measurement method that is based on a modulation
accuracy measurement with conducted, or (preferably)
over-the-air (OTA) measurement. The OTA based method
has some advantages over traditional conductive noise
figure measurements when a high number of receiver
paths are measured simultaneously.
The conductive measurement setup requires complex
cabling, and noise figure measurement requires a narrow
measurement bandwidth to improve its accuracy.
Additionally, the method is sensitive to receiver gain
ripple, and narrow-band signals are not realistic signals for
a wideband receiver. The effective isotropic noise figure
was recently proposed by [7], where the noise figure is
measured with a radiated noise temperature that requires a
narrow measurement bandwidth. Our proposed method is
based on a digitally modulated wideband OTA test signal.
The received signal is demodulated and the noise figure is
derived from the error vector magnitude (EVM) of the
received signal.
The paper is organized as follows. Section II presents an
overview of the radio architecture and presents the design
principles of mobile backhaul radio, including
assumptions with initial radio link budget calculations.
Section III focuses on the design and characterization of
the implemented antenna array. Section IV presents the
quantitative results of for the noise figure of the receiver
beam former, and Section V presents the resulting link
budget analysis, which includes realistic measured radio
performance figures. Finally, in Section VI, the
conclusions are drawn and remarks are made.
II. Design Requirements and Radio Architecture of
Wireless Backhaul System
The proposed mmW wireless backhaul system has been
designed to meet the following requirements [8]: (i) long
coverage (a few hundreds of meters), (ii) over 2 Gbps
data rate, (iii) MIMO transmissions, and (iv) adaptive
beamforming. We employ large antenna arrays to improve
coverage, RF beam steering, beam broadening abilities for
beam adaptation, and orthogonal frequency division
multiplexing (MIMO-OFDM), and high-order modulation
to achieve the desired data rate. Details on the actual
implementation architecture, system specifications [9], and
baseband configurations [10] are provided in Fig. 1,
Tables 1 and 2, respectively.
From an RF architecture viewpoint, Fig. 1 shows that
four antenna arrays are utilized for the transmission of
eight digital streams. This clearly indicates that, in pairs,
streams are aggregated in the RF domain and,
subsequently, the antenna and RF components need to
support wideband signals.
In addition, Table 2 shows that the target data rate
can be achieved with different configurations and, in
particular, with different numbers of streams. Because the
baseband is capable of 8 98 MIMO, carrier component
aggregation is also required and, the lower the modulation,
the larger the number of carrier components that need to
be aggregated.
A radio link budget based on the aforementioned
requirements and system design parameters are presented
in Table 3. This radio link budget is based on the
theoretical assumptions that the receiver has maximum
Radio card #1 with
16 signal paths and
antenna array #1
2 × 8 configuration
Radio unit 1
Radio card #2 with
16 signal paths and
antenna array #2
2 × 8 configuration
Radio card #3 with
16 signal paths and
antenna array #3
2 × 8 configuration
Radio unit 2
Radio card #4 with
16 signal paths and
antenna array #4
2 × 8 configuration
Baseband signal processing unit
Fig. 1. Block diagram of the full unit supporting 4 94 MIMO
transmission and reception.
90 ETRI Journal, Vol. 40, No. 1, February 2018
https://doi.org/10.4218/etrij.2017-0231
coherence gain, both transmission and reception antenna
arrays provide the maximum theoretical antenna gain,
the radio front end is implemented with a bandpass
preselection filter, the transmission power is 30 dBm, and
the noise figure of the receiver is 10 dB.
All these assumptions are addressed in other sections
in the paper. In Section V, a revised link budget is
presented.
III. Design of Receiver and Transmitter Radio
Solution
In this section, we focus on a description and analysis
of the implemented mmW radio transceiver [5], which is
designed for a time-division-duplex communication,
where the transmitting and receiving paths are not
simultaneously utilized, but separated by an RF switch.
The beamforming functionality of the radio solution is
implemented at RF with digitally controlled phase shifters
but without amplitude control. This is because amplitude
control of each transmission and reception signal path
would require a dedicated gain control component,
resulting in a complex circuit board design and calibration
control. The phase shifters are controlled by RF
beamformer logic and are used to modify the signal phase
at the input (or output) of each of the antenna elements.
The same phase shifters are used for the transmission and
reception of the signals. A signal distribution network that
combines and splits the RF signal is common for both
transmission and reception. A three-stage cascade of the
Table 2. Stream and modulation requirements to achieve the
(average) target rate of 2.5 Gbps.
Baseband performance
Modulation 4-QAM
(QPSK)
16-
QAM
64-
QAM
256-
QAM
Fast fourier
transform size 2,048 2,048 2,048 2,048
Channel bandwidth
(MHz) 100 100 100 100
Subcarrier spacing (kHz) 75 75 75 75
Modulation 4 16 64 256
Coding rate 0.85 0.85 0.85 0.85
UL/DL ratio (%) 50 50 50 50
Achievable-rate per
stream (Gbps) 0.08 0.16 0.24 0.32
Number of streams 32 16 11 8
Target rate (Gbps) 2.5 2.5 2.5 2.5
Table 3. Link budget calculation based on 64-QAM modulation
and 100 MHz channel bandwidth at 27.0 GHz.
Parameter Value Unit
ANumber of antenna elements or
antenna array gain
16 N/A
12.04 dB
B Gain of antenna element 10.72 dB
C SNRmin 25.20 dB
D SNRmin_coded 24.49 dB
E Transmitter (TX) EVM 26.00 dB
FRX SNR Requirement
(D –E on linear scale) 29.83 dB
G Noise density 174.00 dBm/Hz
H Thermal noise power over channel 94.17 dBm
I Receiver (RX) noise figure 10.00 dB
JSensitivity of one receiver
(F +H+I) 54.35 dBm
KConducted transmission power of
each transmission path 30.00 dBm
LAntenna gain transmitter
(A +B) 22.76 dBi
MAntenna gain receiver
(A +B) 22.76 dBi
N TX front end losses 5.00 dB
O EIRP (K +A–N+L) 59.76 dBm
P Link margin (O –J+M) 136.87 dB
Q Path loss coefficient 2.50 N/A
R Wavelength 11.1 mm
S Maximum distance 263.58 m
Table 1. Summary of system design targets for different
technology areas of proof-of-concept millimeter wave
radio solutions.
Main radio technology areas
Antenna Radio solution Baseband
Phased-array
(16 94 radiators)
with/without
antenna transmit
array
26.5 GHz–29.3 GHz Wideband 1 GHz
signal bandwidth
Structure 8 92RF
beamformer –with
292 antenna
subarray in each,
linearly polarized
Operational band
at the Olympics:
26.5 GHz–27.5 GHz
898 MIMO-
OFDM (symmetric
in uplink and
downlink)
Maximum gain
22.7 dBi (sim.),
full array
4 RF beam formers
(phase shift based)
Max bandwidth:
89100 MHz
carrier components
N/A
Digital phase shift
control, branch
enabled, gain
control
Modulation up to
64-QAM
(quadrature
amplitude
modulation)
91Marko E. Leinonen et al.
http://onlinelibrary.wiley.com/journal/10.4218/(ISSN)2233-7326
power-splitting network is used to minimize network
insertion losses when one side of the radio card supports
eight antenna elements. A block diagram of the radio card
is shown in Fig. 2, and a mechanical overview of the
structure of one radio unit is presented in Fig. 3 with two
integrated radio cards [9]. A mixer is used to upconvert or
downconvert the RF signal to an intermediate frequency
(IF) that is used by the digital front end.
1. Radio System Design of Transmitter and Receiver
On the transmit path, the signal is up-converted from
the 1 GHz6 GHz IF frequency to the final transmis-
sion frequency of 26 GHz30 GHz using a commercial
subharmonically pumped 21 GHz–31 GHz GaAs mixer
with an integrated local oscillator LO-amplifier.
Subharmonic pumping alleviates the requirement for
the LO-generation because the final RF-frequency is
29LO +IF, which means that the synthesizer can
operate at less than half of the final frequency. A three-
stage Wilkinson power divider network is used to divide
the signal to eight antenna modules. Finally, the signal is
amplified before the power divider network using two
gallium nitrate (GaN) power amplifiers (PAs). Each
antenna branch contains a digitally controlled passive
phase shifter that can be controlled with 11.25°steps.
The last stage of power amplification is done with a GaN-
based PA, namely the TGA2595 [11]. This specific
amplifier can provide up to 30 dB gain, its output third-order
intercept point is 48.5 dB, and its power added efficiency
(PAE) is 24%. However, when using higher order QAM
schemes, 10 dB or more backoff is required to achieve the
required EVM which reduces the PAE significantly.
In this system, special attention is paid to the heat
dissipation of the PA. In fact, the thermal plot shown in
Fig. 4 indicates that high temperatures can be reached in
the operational mode. The implemented solution is to
solder each PA on a Cu coin that perforates the printed
wire board (PCB) material and is directly attached to a
heat sink on the bottom side of the PCB.
In the receiver path, one of the most important
components is the first low noise amplifier (LNA). Here, we
used a gallium arsenide (GaAs) monolithic microwave
integrated circuit operating on the 22 GHz–38 GHz band.
The LNA has a noise figure of 2.5 dB and a gain of 19 dB.
Receiver gain is controlled by a digital attenuator with
5-bit resolution that is located between the two LNAs
following the power-combining network. The receiver
path contains a commercial logarithmic power detection
circuit that acts as a received signal strength indicator for
the combined output.
2. Radio Control Software
As illustrated in Fig. 3, each radio unit contains an
auxiliary (digital) module for a microcontroller unit. The
reason is to perform RF calibrations, beamforming and
LO gen
Data signals from and to the baseband
LO gen
Data signals from and to the baseband
Fig. 2. Block diagram of the proof-of-concept radio unit
supporting two 2 98 antenna element arrays.
Fig. 4. Thermal simulation of the PA.
PWR input
Processor control
Ref Clk in/out
TX/RX signals
Nucleo
processor
board
Antenna
array
Antenna
array
RF-PCB
RF-PCBAUX-PCB
Fig. 3. Mechanical overview of the radio unit [9].
92 ETRI Journal, Vol. 40, No. 1, February 2018
https://doi.org/10.4218/etrij.2017-0231
beam alignment procedures thus easing the need for
control signaling to the baseband.
The effect of the finite resolution of phase shifter was
simulated, and the result is shown in Fig. 5. The mean
error of beam steering is 0.229°between 84°to 84°,
when a step size of 0.5°is used in the simulation. The
quantized result is 90°between 84°and 90°and 84°and
90°, leading to a 5°error. Phase calibration between
radio paths is implemented with a look-up table based on
measurement results for the phase shifter control words.
Figures 6 and 7 show two types of multilevel beam
patterns supported by the implemented mmW transceiver.
The first one is based on antenna element deactivation and
provides beams with relatively high sidelobes but a
smooth beam shape. The second one is based on the
subarray broadening deactivation technique proposed in
[12]. In contrast to the deactivation approach, this method
provides lower sidelobes and maximum emission power,
but it has a beam shape with ripples. In addition, as shown
in [12], it also improves the probability of beam alignment
under many channel propagation conditions.
The measured raise time of the phase shifter is 30 ns
and the raise time of the RF switch is 20 ns; thus, they
do not limit the speed of beam alignment. The beam
alignment procedure has not been fully implemented at the
time of writing.
IV. Implemented Antenna Array Solution with
Measurement Results
In Fig. 8, a prototype antenna with a 2 98antenna
matrix is shown. The array consists of 16 sub-arrays which
each include a 2 92 patch antenna element, oriented in
BF gain (dB)
Root beam with 16 isotropic elements
Azimuth angle (˚)
10
5
0
–5
–10
–100 –80 –60 –40 –20 0 20 40 60 80 100
Level0
Level1
Level2
Level3
Level4
Fig. 7. SubArray. Example of multi-level beampattern using
subarray broadening and antenna deactivation.
BF gain (dB)
Root beam with 16 isotropic elements
Azimuth angle (˚)
10
5
0
–5
–10
–100 –80 –60 –40 –20 0 20 40 60 80 100
Level0
Level1
Level2
Level3
Level4
Fig. 6. DEACT. Example of a multilevel beam pattern using
only antenna deactivation.
Absolute error (˚)
6
Beem steering error due to quantization
4
5
3
2
1
0
–100 –50 050 100
Steering angle (˚)
Fig. 5. Effect of phase shifter resolution to beam steering
accuracy.
Fig. 8. Photograph of a 2 98 sub-array antenna prototype with
measurement cables connected to the antenna connectors.
93Marko E. Leinonen et al.
http://onlinelibrary.wiley.com/journal/10.4218/(ISSN)2233-7326
45°polarization. Thus, the total number of elements in
the array matrix is 64, and the size of the array is
90 934 mm
2
(W9L). The physical dimensions of a
single sub-array are 10.7 910.7 91.6 mm
3
(L9W9
H), and the size of the sub-array was driven by the sizes of
the radio components available on the open market. The
structure includes RF connectors called SMPM and a metal
cover over the feed network to prevent backward radiation.
The total form-factor of the planar antenna array is
9.6 92.4 cm
2
. The presented radiation pattern simulation
and measurement results are based on the first version of
the antenna array, where the goal of 10 dB impedance
bandwidth was set for 26.65 GHz to 27.50 GHz. The target
for the second version of the antenna array is to fulfill the
bandwidth from 24.25 GHz to 29.20 GHz, and it will
present two orthogonal polarizations components (45°).
In Fig. 9, simulated and OTA measured beam patterns of
the full 2 98 antenna matrix are presented. The azimuth
plane (horizontal plane) results are shown because the
horizontal direction is emphasized more than the elevation,
as the beams are narrower in the azimuth plane. The
radiation patterns are measured element by element and the
results are summed in postprocessing. The side lobe levels
of the antenna array are 15 dB below the maximum gain,
which agrees well with simulation results. The measured
maximum gain of the antenna array is approximately
20 dBi, which is close to the simulated value of 21.5 dBi.
The following OTA measurements for the radio solution
were performed with a 1 98 antenna array configuration
and, thus, the maximum antenna gain was reduced by
3 dBs to 17 dBi.
V. Quantification of the Receiver Beamformer
Noise Figure
The noise figure of a radio receiver is how much the
receiver increases noise above the thermal noise level
when a signal passes through it. The noise figure of the
receiver limits the level of signal that can be received.
Thus, measuring the noise figure is one of the most
important tasks in radio receiver characterization.
1. System-Level Calculation of the Receiver Array
Noise Figure
The noise figure for any receiver can be defined using
the signal-to-noise (SNR) ratios of the input and output of
the receiver
SNRoutput ¼SNRinput NF;(1)
where SNR
output
is the SNR in the output of the receiver,
SNR
input
is the SNR in the input of the receiver, and NF is
the noise figure of the receiver (all expressed in dB).
The noise figure for a phased array receiver is more
complicated to define than for a single path because it
depends on the definition of the input signal-to-noise ratio.
One approach is to analyze the antenna array as a single
path receiver if the first LNA has high enough gain [13].
Multiple receiver paths can improve the SNR of the
received signal. In a phased array receiver, the received
signals from the antenna array elements are rotated into
the same phase to facilitate coherent signal combination.
This is equivalent to steering the combined antenna
pattern to the direction of the incoming signal. Noise from
antenna elements has a random phase, which means that
the noise is combined incoherently. That is, it is combined
as root mean square power. The SNR of an array receiver
can be ideally improved by Ntimes or 10log
10
(N)dB
with respect to single-receiver SNR, where Nis the
number of receiver paths. This gain is referred to as the
coherence gain. A high number of receiving antennas in
the array could mathematically lead to a situation where
the calculated noise figure of the receiver is negative on
logarithm scale, which is not physically feasible [14]. This
dB
25
10
0
–10
–20
–30
–90 –60 –40 –20 0 20 40 60 90
θ/˚
dB
25
10
0
–10
–20
–30 –90 –60 –40 –20 0 20 40 60 90
θ/˚
(a)
(b)
Fig. 9. Radiation patterns of 2 98 antenna array matrix: (a)
simulated and (b) measured results at 26.5 GHz in the
azimuth plane.
94 ETRI Journal, Vol. 40, No. 1, February 2018
https://doi.org/10.4218/etrij.2017-0231
is compensated for by a correction factor that keeps the
noise figure above zero [14].
In practical implementations, the coherence gain will be
lower than the theoretical value because of component
variations like IC manufacturing and packaging process
variations. In addition, the layout has an impact on the
radio performance because, in practice, it is not possible to
make the signal lines perfectly similar and the layout will
not be perfectly symmetrical.
If the gain of an individual receiver path Nis denoted by
G
N
and the input power conducted into the receiver path
from antenna element Nis P
in
, then the combined output
signal in the output of receiver P
out
is
Pout ¼Pin G1þG2þþGN
ðÞ:(2)
If all receiver paths have equal gains G
1
=G
2
=G
N
, then
the output is
Pout ¼Pin G1þG1þþG1
ðÞ¼PinG1N;(3)
where Nis the number of receiver paths. If the receiver
paths have unequal gains and G
1
is the maximum gain,
then the output is
Pout ¼PinG11þG2=G1þþGN=G1
ðÞ
¼PinG1X
N
k¼1
1þGk=G1
ðÞ\PinG1N:(4)
Hence, the coherence gain achievable with unequal gains
is lower than that with equal gains depending on the gain
ratios. Similarly, with OTA measurements, if the antenna
elements in the array have unequal antenna gains, the
power conducted from the antenna follows (4). In the
OTA case, P
in
is the radiated input signal level and G
k
is
the antenna gain of the antenna element.
If one or more antennas or receiver paths have higher
gains than the others, then the available coherence gain
may be reduced significantly.
An RF system-level analysis of the cascaded noise figure
and cascaded gain of the implemented radio solution is
shown in Fig. 10 [5]. The noise figure calculated for the
receiver array with eight receiver paths is 8.0 dB, and the
noise figure for one receiver path is 21.0 dB. An image of
the radio solution is shown in Fig 11, which is half of the
final radio board implementation, as shown in Fig. 2.
2. Measurement of the Receiver Array Noise Figure
The noise figure of a receiver can be measured in
different ways, and one conductive measurement method
is to feed a signal with a known SNR into an input of the
receiver and measure the SNR at the output of the
receiver. The two signals are then compared. Another
conductive measurement method is based on measuring
the gain of a receiver with a continuous wave (CW)
signal. After the gain measurement, the thermal noise level
at the output of the receiver is measured when the input is
terminated with a 50 Ωstandard load. The noise level is
measured with a narrow measurement bandwidth to
increase the sensitivity of the measurement. This method
is shown in Figs. 12(a) and (b).
A block diagram of the measurement system for the
proposed OTA noise figure measurement based on
signal quality is shown in Fig. 12(c). A test signal is
transmitted OTA, the received signal is demodulated,
and the noise figure is calculated from the received
signal EVM or SNR.
The modulated signal that is used in the OTA noise
figure measurement is a standard digital wideband
dB
40
35
30
25
20
15
10
5
0
–5
–10
–15
–20
Cascaded gain (max attn) Cascaded gain (min attn) Cascaded INF (max attn) Cascaded INF (min attn)
Fig. 10. Cascaded signal level and linearity block level analyses
for 8 receiver paths [5].
Fig. 11. Image of the proof-of-concept 28-GHz band radio
transceiver board supporting an 1 98 antenna array
[5].
95Marko E. Leinonen et al.
http://onlinelibrary.wiley.com/journal/10.4218/(ISSN)2233-7326
modulated signal, which is transmitted via a known
reference antenna to the antenna array receiver.
Photographs of the OTA measurement setups are shown
in Figs. 13(a) and (b). The measured antenna array is
placed in front of the absorber material and the measured
radio unit is behind the absorber. This arrangement
prevents coupling of the OTA signal directly with the
radio receiver signal paths. OTA measurements were
performed at a 2-m distance, which is the far-field region
of the receiving antenna array.
The received output signal level Pout
RX
in the OTA
measurement is measured directly from the output of the
receiver and can be calculated as follows.
PoutRX ¼PTX ILTX þGantTX ILpath
þX
N
i¼1
GantRX
iILRX
iþGpathRX
i;(5)
where P
TX
is the transmission power of the test signal,
IL
TX
is the cable loss between the signal generator and the
transmission antenna, Gant
TX
is the transmission antenna
directivity and radiation efficiency, IL
path
is the free space
loss of the signal at the test frequency, GantRX
iis the
antenna directivity and radiation efficiency of antenna i,
ILRX
iis the cable loss between the antenna and receiver
path iand GpathRX
iis the power gain of receiver iwith the
modulated wideband test signal.
The modulation accuracy of the received signal is
measured with a modulation analyzer, which can measure
the output power level and EVM. EVM is a metric of
digital modulation accuracy, and it depends on the SNR of
a signal. EVM can be defined for both received and
transmitted signals, and radio system specifications state
the requirements for the transmission EVM levels. EVM is
defined according to the following equation [15].
EVM ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
NPN1
n¼0Ierr n½
2þQerr n½
2
qPreff
;(6)
where nis a symbol index, Nis the number of symbols,
I
err
=I
ref
I
meas
is the error of signal Ifrom the reference
point and Q
err
=Q
ref
Q
meas
is the error of signal Q
from the reference point. The EVM has the following
direct relationship to the received signal SNR [16]:
SNR ¼PoutRX Noise ¼20 log10
EVM
100 %
:(7)
The power gains of the receiver paths of the receiver
array with a 16-QAM modulated 100-MHz wide test
signal with a 100-MHz step size over the operation
frequency range were measured, and the results are shown
in Fig. 14. A significant gain ripple of almost 15 dB can
be seen over the operational frequency range on all
receiver paths. The gain variation among signal paths is
caused by component variations, but is mainly due to
manufacturing defects of in the wire bonding of the phase
shifter components.
The noise figure of the receiver can be calculated at a
fixed EVM value by comparing the noise level of the SNR
of the EVM measurement to thermal noise level of a
modulation bandwidth. More accurate noise figures can be
Narrow band gain
measurement of receiver
RF signal generator
CW tone at
25 GHz–29 GHz
Ports 1…8
Phase rotations
4 GHz
Spectrum analyzer
(e.g. RBW = 10 kHz)
Measurement equipment:
Arbitrary waveform generator: M8190A
RF signal generator: E8267D
Spectrm analyzer: UXA N9040B
Vector signal analyzer: UXA N9040B
Narrow band noise
measurement of receiver
50 Ω
terminations
Ports 1…8
Phase rotations
4 GHz
RF amplifer with
known gain and NF
Spectrum analyzer
(e.g. RBW = 10 kHz)
OTA noise figure
measurement based on
EVM measurement
Digital modulation
signal generation SW
Arbitrary waveform
generator
(e.g. BW = 100 MHz)
RF signal generator
25 GHz–29 GHz
OTA
Antenna
array
Ports 1 …8
RF test board
receiver mode
4 GHz
Vector signal
analyser
(e.g. BW = 100 MHz)
(a) (b) (c)
(e.g. CW signal)
Fig. 12. Noise figure measurement system configurations for (a)
CW gain measurement, (b) thermal noise measurement,
and (c) radiated modulated signal.
Fig. 13. Measurement setups: (a) modulated signal measurement
system with reference antenna and (b) EVM OTA
measurement with line-of-sight.
96 ETRI Journal, Vol. 40, No. 1, February 2018
https://doi.org/10.4218/etrij.2017-0231
calculated by curve fitting the EVM measurement results
with an EVM curve with varying SNR values, which can
be calculated from (7). The noise component of SNR can
be calculated as
Noise ¼174 dBm=Hz þ10 log10 BWðÞþNF;(8)
where BW is the modulation bandwidth of the test signal
in Hz and NF is the noise figure of the receiver. The noise
figure can be calculated using (6), (7), and (8) when the
EVM is measured as a function of the output power of the
receiver. A theoretical EVM curve with a known NF can
be calculated in a similar manner. The noise figure of the
receiver array can be calculated by minimizing the mean
squared error (MSE) between the EVM measurement
results and the calculated theoretical EVM curve. The
noise figure of the receiver is given by the best fit of the
EVM curves.
EVM curve fitting for the measurement results and noise
figure modeling was performed around the 10-dB signal
range in the noise-limited signal region. In the noise-limited
region of EVM curve, the noise level or noise figure of the
receiver is the main contributor to EVM performance.
Other EVM contributors such as the gain ripple at the
measurement band, LO-phase noise, and carrier leakage are
designed to be the main EVM contributors at higher signal
levels. The noise-limited EVM region of the measured
receiver is at signal levels below 40 dBm. An example of
curve fitting for measured and theoretical EVM curves for
receiver path 6 is shown in Fig. 15, and the MSE of the
curve fitting is shown in Fig. 16.
The coherence gain of the receiver was measured using
the OTA method with a 100-MHz wide test signal. A
common receiver signal path creates a constant noise floor
at the output of the receiver and thus the SNR of the
received signal can be improved by activating more
receiver paths. The calculated output signal levels based
on the conducted gain measurements are shown with solid
and dotted lines in Fig. 17. The markers in Fig. 17
indicate OTA measurement results, and a good match
between the calculated and measured signal levels can be
Receiver power gain (dB)
RF frequency (GHz)
RXl
RX2
RX3
RX4
RX5
RX6
RX7
RX8
15
10
5
0
–5
–10
–15
–20 25.5 26.0 26.5 27.0 27.5 28.0 28.5 29.0
Fig. 14. Small signal gains of receiver paths of the array receiver
measured with modulated 100 MHz test signal with
OTA.
EVM (%)
Input signal power (dBm)
20
18
16
14
12
10
8
6
4
2
0
–60 –55 –50 –45 –40 –35
RX6
EVM based on NF model
Fig. 15. Noise figure modeling using MSE in the noise-limited
EVM region.
0.35
0
19.0
0.30
0.25
0.20
0.15
0.10
0.05
19.5 20.0 20.5 21.0 21.5
Noise figure model of RX6 (dB)
MSE of EVM (%)
Fig. 16. Best fit of the noise figure model for receiver path 6.
Output level at receiver output (dBm)
Frequency (GHz)
–10
–15
–20
–25
–30
–35
25.0 25.5 26.0 26.5 27.0 27.5 28.0 28.5 29.0
Power sum of conducted RX1..RX8
Coherent sum conducted RX1..RX8
Power combined OTA measurements
Power combined calculated OTA results
Best angle to OTA result
Receiver output levels: calculated and measured results
Fig. 17. Measured and calculated RF output signal levels with
an eight-receiver path array.
97Marko E. Leinonen et al.
http://onlinelibrary.wiley.com/journal/10.4218/(ISSN)2233-7326
seen. Phase rotation is applied to the received signal to
maximize the output signal level for the best-angle OTA
results. The small difference between the measured and
calculated output signal levels can be attributed to signal
level variation during the OTA measurements. The mean
value of the difference at 26.5 GHz is 0.1 dB with a
1.0 dB standard deviation, at a 37.0 dBm average
conducted input signal level after each receiving antenna
element is used.
The theoretical coherence gain for eight receiver paths
with equal gains is indicated by a blue solid line and the
maximum available coherence gain based on the measured
signal gains of the receiver array is shown by a red dotted
line in Fig 18. Black circles indicate the coherence gain
when the best OTA phase-rotated signal is compared to
the measured power-combined OTA signal levels. Green
crosses show the coherence gain if the best phase-rotated
OTA results are compared to the calculated power-
combined individual signal levels. A good fit between the
calculated and measured coherence gains can be seen. The
coherence gain of the receiver is reduced when the gain of
the receiver peaks. Moreover, the average measured
coherence gain over the frequencies is 5.5 dB.
Individual receiver path noise figure measurement
results based on different conducted receiver noise
measurement methods are shown in Table 4. It can be
seen that the single-point EVM value and curve fitting-
based noise figure results are almost equal. The
narrowband noise figure measurement results deviate from
the wideband results because of the ripple seen at the
narrowband gain measurements.
Multiple receiver path noise figure measurements are
shown in Table 5. Both conducted and OTA-based noise
figure measurement results are shown, and a good match
between the OTA and conducted results is seen when two
receivers are measured. When two receivers have an almost
equal gain and signals can be combined with the same
phase, the noise figure is improved by 6 dB. The activation
of the second receiver boosts the signal level by 3 dB and
an additional improvement of almost 3 dB comes from the
coherence gain. It was not possible to simultaneously
measure more receiver paths using conductive
measurements because of the limited number of laboratory
power splitters and cables. Thus, the OTA measurement
was used to measure the noise figure of the full array. The
noise figure based on the curve-fitting method matches the
radio system level calculations. A single point EVM
measurement would yield a noise figure result for the full
eight receiver array that was too optimistic.
VI. Link-Budget with Measured Parameters
In this section, we analyze the radio link budget shown
in Table 3 with realistic radio performance values from
previous sections. It was shown that noise figure of the
198 radio receiver is 8.0 dB, which corresponds well
with RF system calculations. The noise figure of the full
298 array receiver should be 3 dB better than the 1 98
Gain (dB)
Frequency (GHz)
10
Equal gains in receiver paths
Max coherence gain based measurements
Calculated coherence result
Measured coherence individual path OTA results
Available and OTA measured coherence gains
9
8
7
6
5
4
3
2
1
0
25.0 25.5 26.0 26.5 27.0 27.5 28.0 28.5 29.0
Fig. 18. Measured coherence gains of an eight receiver path
array.
Table 4. Conducted noise figures of individual receivers at
26.5 GHz obtained by different measurement methods.
Individual receiver path noise figures (dB)
Receiver
array paths RX1 RX2 RX3 RX4 RX5 RX6 RX7 RX8
MSE EVM
curve fitting 22.5 35.6 20.3 24.8 30.0 21.3 22.1 31.5
Single 4%
point EVM
result
23.1 36.1 20.1 25.1 30.1 21.0 22.3 31.5
10-kHz noise
measurement 16.2 27.8 14.3 17.8 13.4 15.0 16.9 25.5
Table 5. Conducted and OTA noise figures at 26.5 GHz.
Receiver array path noise figures (dB)
RX3 RX6 RX3 +R6 All RX
paths on
Conducted MSE EVM
curve fitting 20.3 21.3 14.0 N/A
Conducted single 4%
point EVM result 20.1 21.3 14.0 N/A
OTA MSE EVM
curve fitting 17.2 20.2 14.5 8.0
OTA single 4% point
EVM result 17.1 19.3 14.1 6.7
98 ETRI Journal, Vol. 40, No. 1, February 2018
https://doi.org/10.4218/etrij.2017-0231
array noise figure, and the 8.0 dB figure used is a
pessimistic value for the full array at row I.
The measured average coherence gain of the receiver
over the frequencies is 5.5 dB, which is 3.5 dB lower than
the theoretical value for eight receivers. The antenna gain
of 16 receivers with measured antenna gain and coherence
gain is 20 dBi 3.5 dB =16.5 dBi and this is used for
row M. The antenna gain of 16 transmitters was measured
as 20 dBi and this value is used for row L.
Front-end losses were decreased to 2.0 dB at row N
because filtering is located close to the mixer. Transmission
power can be adjusted with an attenuator and a conducted
power of 30 dBm is used for analysis at row K.
The radio link margin for the measured values is
132.94 dB, which is 3.93 dB lower than the original values.
The expected link distance is reduced from 263.6 m to
183.53 m, which is close to the 200-m requirement for
a 2.5 Gbps data rate with a 100-MHz wide signal with
64-QAM modulation. However, note that we used a
rather pessimistic path loss coefficient of 2.5 in our
calculations.
VII. Conclusions
We implemented a mobile backhaul phased array
transceiver and an antenna array based on 5G high data-
rate requirements. We introduced a new noise figure
measurement method based on OTA EVM measurement,
which is easier to use than conducted noise figure
measurements for antenna array receivers. Measured results
obtained using the proposed OTA noise figure measurement
method match well with the traditional conducted noise
figure measurement results. A radio link analysis based on
the measured radio performances of the proof-of-concept
transceiver and antenna array, indicates that our system can
reach one of the key 5G system requirements of a 2.5-Gpbs
average data speed for a range of 200 m.
Acknowledgements
The research leading to these results has received
funding from the European Union H2020 5GPPP under
grant n. 723247 and supported by the Institute for
Information & communications Technology Promotion
(IITP) grant funded by the Korea government (MSIP)
(No.B0115-16-0001, 5GCHAMPION).
References
[1] F. Boccardi, R.W. Heath, A. Lozano, T.L. Marzetta, and
P. Popovski, “Five Disruptive Technology Directions for
5G,”IEEE Commun. Mag., vol. 52, no. 2, Feb. 2014,
pp. 74–80.
[2] R.J. Weiler, M. Peter, W. Keusgen, K. Sakaguchi, and F.
Undi, “Environment Induced Shadowing of Urban
Millimeter-Wave Access Links,”IEEE Wirel. Commun.
Lett., vol. 5, no. 4, Aug. 2016, pp. 440–443.
[3] H. Kim et al., “A 28 GHz CMOS Direct Conversion
Transceiver with Packaged Antenna Arrays for 5G Cellular
System,”In IEEE Radio Frequency Integr. Circuits Symp.,
Honolulu, HI, USA, June 4–6, 2017, pp. 69–72.
[4] B. Sadhu et al., “A 28 GHz 32-Element Phased-Array
Transceiver IC with Concurrent Dual Polarized Beams
and 1.4 Degree Beam-Steering Resolution for 5G
Communication,”In IEEE Int. Solid-State Circuits Conf.,
San Francisco, CA, USA, Feb. 5–9, 2017, pp. 128–129.
[5] G. Destino et al., “System Analysis and Design of
mmW Mobile Backhaul Transceiver at 28 GHz,”In Eur.
Conf. Netw. Commun., Oulu, Finland, June 12–15, 2017,
pp. 1–5.
[6] M. Mueck et al., “5G CHAMPION—Rolling out 5G in
2018,”In IEEE Globecom Workshops, Washington, DC,
USA, Dec. 4–8, 2016, pp. 1–6.
[7] K. Khalaf et al., “A 60 GHz 8-Way Phased Array Front-
End with TR Switching and Calibration-Free Beamsteering
in 28 nm CMOS,”In Eur. Solid State Conf., Leuven,
Belgium, Sept. 11–14, 2017, pp. 203–206.
[8] G. Destino et al., “Deliverable D2.2 5GCHAMPION Key
Performance Indicator and Use-Cases Defined and
Specification Document Written,”5GCHAMPION Project,
2017, Accessed Sept. 21, 2017. http://www.5g-champion.e
u/Documents/5GCHAMPION_D2.2_Final.pdf
[9] E.C. Strinati et al., “Deliverable D1.2 First Period Report,”
5GCHAMPION Project, 2017, Accessed Sept. 21, 2017.
http://www.5g-champion.eu/Documents/5GCHAMPION_
D1.2_Final.pdf
[10] W. Tavernier et al., “Deliverable D2.1 5GCHAMPION
Architecture, API, and Interface Document,”
5GCHAMPION Project, 2017, Accessed Sept 21, 2017.
http://www.5g-champion.eu/Documents/5GCHAMPION_
D2.1_Final.pdf
[11] Qorvo, “27.5 –31 GHz 9 W GaN Power Amplifier,”TGA
2595 Data Sheet, Rev. A, 2014.
[12] Z. Xiao, T. He, P. Xia, and X.G. Xia, “Hierarchical
Codebook Design for Beamforming Training in Millimeter-
Wave Communication,”IEEE Trans. Wirel. Commun., vol.
15, no. 5, May 2016, pp. 3380–3392.
[13] J.J. Lee, “G/T and Noise Figure of Active Antennas,”IEEE
Trans. Antennas Propag., vol. 41, no. 2, Feb. 1993, pp.
241–244.
[14] A. Natarajan et al., “A Fully-Integrated 16-Element Phased-
Array Receiver in SiGe BiCMOS for 60-GHz
99Marko E. Leinonen et al.
http://onlinelibrary.wiley.com/journal/10.4218/(ISSN)2233-7326
Communications,”IEEE Solid State J., vol. 46, no. 5, May
2011, pp. 1059–1075.
[15] National Instruments, “Modulation Error Ratio (MER) and
Error Vector Magnitude (EVM),”3652EN, 2014.
[16] Keysight, “What is the Relationship between the Error Vector
Magnitude and the Signal to Noise Ratio for a 64-QAM
signal?”2012, Accessed Sept. 21, 2017. http://www.keysight.
com/main/editorial.jspx?ckey=847674&id=847674&nid=
-11143.0.00&lc=eng&cc=UG
Marko E. Leinonen received his MSc and
Licentiate in Technology degrees in
electrical engineering from the University
of Oulu, Finland, in 1996 and 2002,
respectively. From 1994 to 2012 he was
with Nokia Mobile Phones, Oulu, Finland,
working in various positions with radio
engineering and technology management. From 2006 to 2007,
he was a senior engineering manager in Bangalore, India. From
2012 to 2016, he was a master developer with Ericsson, Oulu,
Finland. Since 2017, he has been with the Centre for Wireless
Communications, University of Oulu, Finland, where he is
currently a project manager. His research interests include
wireless radio systems and quality topics in radio engineering.
He holds more than 30 granted international patent families
concentrating on radio engineering.
Giuseppe Destino received his DrSc
degree from the University of Oulu in
2012, and his MSc (EE) degrees
simultaneously from the Politecnico di
Torino, Italy and University of Nice,
France in 2005. Currently, he is working
as an Academy of Finland postdoctoral
researcher as well as project manager of national and
international projects at the Centre for Wireless
Communications, University of Oulu, Finland. His research
interests include wireless communications, millimeter wave
radio access technologies, especially algorithms for channel
estimation, hybrid beamforming, and positioning. He has served
as a member of the technical program committee of IEEE
conferences.
Olli Kursu received his MSc and DrSc
degrees in electrical and electronics
engineering from the University of Oulu,
Finland, in 2006 and 2015, respectively.
Currently, he is working as a postdoctoral
researcher at the Centre for Wireless
Communications, University of Oulu. His
research interests include mmW, RF, analog, and mixed signal
circuit design for wireless communication systems.
Marko Sonkki received his MSc in
electrical engineering from the Department
of Electrical and Information Engineering.
University of Oulu, Finland, in 2004. In
2013, he received his DrSc in radio
telecommunications engineering from the
University of Oulu. The topic of his
dissertation was wideband and multi-element antennas for
wireless applications focusing on antenna design based on
spherical and characteristic modes theories. He is currently a
post-doctoral researcher with the Centre for Wireless
Communications, University of Oulu. His current research
interests include the design and analysis of wideband antennas,
wideband multimode and full-duplex antennas, MIMO and
diversity systems, and antenna array design, including millimeter
waves. He is also a deputy manager at the University of Oulu
Research Institute Japan –CWC Nippon, and a visiting
researcher at the Universitat Polit
ecnica de Val
encia, Valencia,
Spain.
Aarno P€
arssinen received his MSc,
Licentiate in Technology, and DrSc
degrees in electrical engineering from the
Helsinki University of Technology,
Finland, in 1995, 1997, and 2000,
respectively. From 1994 to 2000, he was
with the Electronic Circuit Design
Laboratory, Helsinki University of Technology working on
direct conversion receivers and subsampling mixers for wireless
communications. From 2000 to 2011, he was with the Nokia
Research Center, Helsinki, Finland. From 2011 to 2013, he was
at Renesas Mobile Corporation, Helsinki, Finland working as a
distinguished researcher and RF research manager. From
October 2013 to September 2014, he was an associate technical
director at Broadcom, Helsinki, Finland. Since September 2014,
he has been with the Centre for Wireless Communications,
University of Oulu, Oulu, Finland, where he is currently a
professor. His research interests include wireless systems and
transceiver architectures for wireless communications with
special emphasis on RF and analog integrated circuit and system
design. He has authored and co-authored one book, one chapter
of a book, more than 50 international journal and conference
papers. He also holds several patents.
100 ETRI Journal, Vol. 40, No. 1, February 2018
https://doi.org/10.4218/etrij.2017-0231
