5G-PHOS D3.1 - 1st Generation of 64x64 MIMO Circuit and Design - FINAL PDF
5G-PHOS D3.1 - 1st Generation of 64x64 MIMO Circuit and Design - FINAL PDF
5G-PHOS D3.1 - 1st Generation of 64x64 MIMO Circuit and Design - FINAL PDF
Deliverable D3.1
Report on 1st generation 64x64
MIMO antenna mmWave circuit
design and development
Programme: H2020-ICT-2016-2
Project number: 761989
Project acronym: 5G-PHOS
Start/End date: 01/09/2017 – 31/08/2020
Author List:
Organization Author
Siklu Communication Yigal Leiba, Benny Versano, Elad Dayan
IMEC Xin (Scott) Yin, Guy Torfs, Nish Singh
FRAU Bogdan Sirbu
ICCS/NTUA Giannis Giannoulis
Christos Vagionas, George Kalfas, Agapi Mesodiakaki, Pavlos
AUTH
Maniotis, Marios Gkazianas, Nikos Pleros
Abstract: 5G-PHOS aims to develop and evaluate a converged Fiber Wireless (FiWi) 5G
broadband fronthaul/backhaul network for highly dense use cases based, among other
technologies, on mmWave massive MIMO technology. This deliverable reports the 1st
generation design of a 64x64 MIMO antenna and associated mmWave circuitry design
and development. The deliverable focuses on the design process and technology
associated with this antenna.
Copyright © 2017-2018 the 5G-PHOS Consortium. All rights reserved. This document may not be
copied, reproduced or modified in whole or in part for any purpose without written permission from
the 5G-PHOS Consortium. In addition to such written permission to copy, reproduce or modify this
document in whole or part, an acknowledgement of the authors of the document and all applicable
portions of the copyright notice must be clearly referenced.
Table of Contents
Abbreviations
BB Base Band
BS Base Station
BW Bandwidth
DC Direct Current
IC Integrated Circuit
LO Local Oscillator
MUX Multiplexer
NB Narrow Band
PA Power Amplifier
PC Personal Computer
RF Radio Frequency
1. Executive Summary
This deliverable describes the first design of a 64x64 MIMO antenna mmWave circuits. It
reviews the concept of a massive MIMO antenna, specifically one operating in the 60GHz
mmWave band, and its suitability for high density cellular communication. It then
proceeds to the reviewing of the entire system block diagram, which consists of modular
radiating tiles concentrated together using a platform consisting of several PCBs (TILE
FEED PCB, MIMO PCB, SPLIT PCB and Tile controller). The system block diagram also
specifies the detailed interfaces between these blocks and the connection with the optical
processing block.
Following the block diagram, detailed design information, as well as design
considerations are provided for the system blocks, PCBs and radiating elements. For
many of the blocks where simulation results are available, such results are presented.
An experimental performance evaluation of the Analog-Radio over Fiber (A-RoF) optical
data transmission on Intermediate Frequencies over Fiber (IFoF) and subsequent
demodulation of the I/Q signal constituents is described for the interfacing of the optical
fiber-fronthaul data stream with the analog electronic mixers of the mmWave antenna.
The deliverable also summarizes the block diagram and detailed design information for
the Narrowband Trans-Impedance Amplifiers (NB-TIA), which support the option for
direct conversion of a mmWave-modulated optical signal to an electrical one.
2. Introduction
2.3 Audience
This document is internal to the 5G-PHOS project consortium.
While all points above are quite effective in improving the BS sector throughput, the first
two points have mostly reached a point at which further improvement will have marginal
benefit on throughput in practical operation scenarios for traditional cellular spectrum
below 5GHz. The third point, however, is very promising, as channel bandwidth could be
increased by more than 10X by transitioning from frequencies below 5GHz to mmWave
frequencies in which spectrum is abundant. This transition to higher frequencies also
facilitates the second point (i.e. MIMO), since the increase in frequency enables the
antenna size to shrink by the same factor and significantly increases the number of
antennas that can possibly be integrated.
The second factor determining the cellular network capacity is the BS sector density. BS
sector density can be improved by increasing the density of BS sites and also by
increasing the number of BS sectors per BS site. Indeed, as cellular networks evolved,
the distance between the BSs in urban areas, where high capacity is required, has
decreased over the years from distances of >1000m to an order of 300m, currently.
Additionally, several BSs operating at different bands will typically be collocated at the
same site. The number of antennas per BS is typically set to 3 (i.e. 3 120 BS sectors).
Increasing the number of BS sectors per site (for BS operating in the same frequency
band) is limited by the fact that these sectors will interfere with each other if they use
the same frequency channel. The latter is typical in the recent cellular standards, which
aim towards a frequency reuse factor of 1.
The conclusion from the above is that increasing the capacity of the cellular network in
future cellular standards might become hard, if not impossible, as most of the
approaches that have been used in the past towards increasing the capacity seem to be
have reached a plateau. Fortunately, use of advanced photonics technology with massive
MIMO antennas and mmWave spectrum can provide another significant boost to cellular
network capacity.
Utilizing mmWave spectrum for cellular access has significant advantages in terms of
spectrum availability and reuse capability. Naturally, when using higher frequencies,
more spectrum is available, so that just by transitioning from operation at bands near
2GHz to bands near 60GHz, a factor of 30X in the available spectrum resource may be
gained. The size of the antenna elements at mmWave frequencies becomes much smaller
compared to that in frequencies below 5GHz, thus allowing us to pack large arrays of
antenna elements in a small space and enabling high order MIMO and beam-
forming/phased-array antennas. The benefit of deploying high order MIMO schemes on
the achievable capacity is evident when comparing to already existing wireless standards.
The use of beamforming has great potential for increasing the capacity of the system, as
beams directed towards the intended user focus the radio energy in the desired direction,
thus increasing the signal level on the intended user and decreasing the interference
level at unintended users. The use of beamforming enables the transmission of different
data to different recipients at the same time, thus practically enabling the increase of the
network capacity by a factor equal to the number of beams deployed in parallel.
Fig. 2. Attenuation by Oxygen, water vapor and fog at mmWave frequencies [1].
Considering all the factors that impact mmWave propagation, typical mmWave link
ranges can be calculated based on expected availability (which is derived from the
probability that rain rate exceeds a certain threshold, e.g. considering rain zone-E data
shown in Fig. 3 below, 99% availability means that we consider as unavailable the 1% of
the time in which the rain rate exceeds the threshold of 0.6mm/h, likewise, 99.99%
availability means that we consider as unavailable the 0.01% of the time in which the
rain rate exceeds the threshold of 22mm/h). The tables below show operation at 64GHz
and 60GHz, which approximately represent the minimal and maximal Oxygen absorption
within the 57-64GHz band. The example of typical range limitations calculations is based
on a typical 26cm diameter dish antenna used in this band and operation over a 500MHz
channel using 5dBm of transmit power.
Fig. 3. Link range and availability under various rain conditions at 64GHz.
Fig. 4. Link range and availability under various rain conditions at 60GHz.
The rain zone map which is used in conjunction with such tables to determine the link
range is shown in Fig. 5.
In order to alleviate the variation in link attenuation, mmWave equipment may deploy a
rate-adaptation algorithm, which would modify the bandwidth, code rate and modulation
used over the air interface in order to increase transmission robustness at the cost of
transmission throughput. As demonstrated in the table above, the link in this band may
retain connectivity for distances greater than about 1.4Km even under the worst weather
and propagation conditions met in European rain zones.
Parameter Description
The angular range (azimuth and elevation) across which the
Steering range
beam can be steered
Gain angular variation The variation in the gain of the antenna across is steering
range
Steering resolution The resolution by which the beam can be steered to a desired
direction within the steering range
Beam directivity The directivity of the beam when steered to a specific direction
Spurious side lobe The level of spurious side lobes generated by the steering
level operation
Antenna pattern The size of the widest chunk of spectrum which could be
coherence bandwidth steered simultaneously to the same direction
RF power efficiency The portion of the power that is injected to the antenna and is
eventually radiated
Steering speed The time duration it takes to change the beam direction
Physical dimension Physical size of the antenna (typically as compared to the
wavelength)
Architectural impacts Impacts of the antenna on the overall system design
Ease of construction The complexity of commercially manufacturing the antenna
Over the years, a large variety of techniques have been proposed to perform beam-
steering, however, here we shall focus in the method used for the 5G-PHOS project,
which is commonly referred to as beamforming.
Beamforming is probably the best known beam-steering method. The method is based on
the usage of an array of radiating elements which are fed by the signal to radiate, but
each element receives the signal at a specific power and phase, such that the combined
radiation from all array elements generates a coherent wave front at the desired
direction of propagation. The control of each individual radiating element’s amplitude and
phase can be performed either at baseband, IF or RF, and results in different complexity
of implementation and different performances ([2], [3]). Although multiple theoretical
techniques are available for “synthesizing” antenna patterns by selecting proper phase
and amplitude taper profiles [4], these techniques typically ignore physical hardware
constraints or assume ideal circuit models, making their actual implementation non-
trivial.
The large number of radiating elements in a massive MIMO antenna (which does not lead
to excessive antenna sizes in mmWave frequencies) also allows the combination of
beam-steering with traditional MIMO spatial multiplexing [5], the latter being introduced
for frequencies below 6GHz (and already incorporated into LTE, LTE-Advanced). It is,
perhaps, unfortunate that traditional MIMO signal processing is also sometimes referred
to as “beamforming” by the communications community; it should be emphasized that
the traditional MIMO “beamforming” used to create multiple spatial streams and the
“beamforming” beam-steering method (via power, phase variations) mentioned in the
previous paragraph are distinct methods, based on different types of signal processing
and serving different purposes. However, these methods can co-exist in actual mmWave
implementations, in which case the general term “beamforming” is used to refer to them.
Different beamforming system architectures (analog, digital, hybrid [6], [7]), depending
on the number of available RF chains, precoder blocks and number of spatial streams
have been proposed, and their performance/complexity tradeoff investigated [8].
Table 2 describes the antenna performance parameters for beamforming at mmWave
frequencies.
Parameter Description
Assuming an array composed of identical elements, the
steering range is defined by the radiation power envelope of
the radiating element. Typical radiation elements (patches,
Steering range
dipoles, slots) can easily reach steering range in excess of 90.
Steering range in excess of 180 would typically require
arranging the elements on a non-planar surface
Gain angular variation Typically follows the radiation power envelope of the radiating
element
Steering resolution Derived from the resolution of controlling the radiating
elements phase
Beam directivity Assuming equal power at the radiating elements, it is roughly
proportional to 10log10(N) where N is the number of radiating
elements
Spurious side lobe level Strongly depends on the number of radiating elements as well
as the distance between them and the power tapering applied
to the elements
Antenna pattern Depends on the way the phase differences are implemented.
coherence bandwidth When implemented by true time delay the bandwidth is
limited by the bandwidth of the radiating elements
RF power efficiency Potentially very high, depending on the exact implementation.
Power may be lost in the feeding network, especially if split
multiple times. In case of using multiple power amplifiers to
reduce the amount of power, splitting loss may be low
Steering speed As fast as the power and phase to the radiating elements can
be varied. Typically in mmWave implementations that would
be within nanoseconds
Physical dimension May be very compact. Typical radiating element distance is
/2 which is 2.5mm at 60GHz, so an array of 4x8 elements
yielding about 25dBi gain would roughly be 20mm x 10mm in
size
Architectural impacts In mmWave implementations, radiating elements are typically
fed by a dedicated PA. This enables radiating significantly
more power than a single PA serving all elements.
Ease of construction Radiation directly from a mmWave chip is hard to implement
due to the size of the chip as compared to the size of the
array. Typically, the radiating elements are implemented as
part of the package, facilitating tight integration between the
chip and the radiating elements. Loss and phase errors in the
feeding network should be carefully observed
Master
SPLIT 16X1
PCB OBFN
To FlexBox
RF wave-front
(beam ) at 60GHz
Frequency
Tile CTRL
Generator
Up to 16 x MIMO PCB
The system, however, is flexible and a variable number of MIMO PCBs may be connected
to support various data rates and parallel beam combinations. Note that each MIMO PCB
is dedicated to one direction of transmission in order to ensure possibility of FDD
operation with adequate separation between the TX and RX channels, so a minimum
system implementation would require to have two MIMO PCBs. Fig. 7 below shows the
modular structure of the antenna as it connects to the FlexBox, being able to support a
variable number of beams and data rates.
Master Master
OBFN OBFN
MIMO
2x
MIMO MIMO
PCB
MIMO MIMO
PCB MIMO
PCB
PCB
PCB MIMO
PCB
MIMO
PCB
MIMO
PCB
Up to
MIMO
PCB
16 x MIMO
PCB
FlexBox FlexBox
MIMO
PCB
MIMO
PCB
MIMO
PCB
MIMO
PCB
MIMO
PCB
MIMO
PCB
MIMO
PCB
MIMO
16 x mmwave BLOCK
PCB
MIMO
PCB
PCB
Fig. 7. 64x64 antenna modular structure (up to 400Gbps when fully loaded)
The MIMO PCB is composed of several sub-blocks whose names are shown in Fig. 8
below.
mmwave BLOCK
8x TILE
Optical 8X1
TILE FEED
convert OBFN
PCB PCB
To
SPLIT
PCB
MIMO PCB
The basic building block is the TILE PCB which contains the basic mmWave array. Eight
TILE PCB units are fed by one TILE FEED PCB, which is responsible for multiplexing the
data, control and frequency reference signals together and allocating them to each tile.
The TILE FEED PCB and the 8 TILE PCB units jointly compose one mmWave block. The
SPLIT PCB can support up to 16 such mmWave blocks and provides them with the control
and frequency reference signals. The data signals come from the optical conversion
block, which is connected to the 8x1 OBFN block responsible for phase shifting of the
incoming optical data signals in order to control the array’s beam pointing direction.
RFIC
LO
32 x IF
channels
cable
MUX
PA Control
Table 3 summarizes the key design goals for the TILE PCB.
LO1
DATA#0
TILE LO2
Triplexer
PCB
MUX CTRL
8:1
CFG
8 tiles
DATA#7
TILE
Triplexer
PCB
The TILE FEED PCB takes care to keep the trace lengths for the relevant LO and data
signals as equal as practically possible so as not to introduce unnecessary phase
differences between TILE PCB units.
Optical
input
CFG
CTRL
LO2 in
TILE FEED
TILE PCB
PCB LO1 in
CFG
CTRL
LO2 in
TILE FEED
TILE PCB
PCB LO1 in
CFG
#0 #15
CTRL
Tile controller
7.8GHz (LO2)
X16
X16 X16 X16
LO1 _T
LO2_T CTRL_T CFG_T
TX,RX
outputs I/O outputs
outputs
Phase
Phase
Phase
Phase
Phase
Phase
Phase
Phase
Phase
Phase
Phase
Phase
shift
shiftPhase
shift
shiftPhase
Phase
shiftPhase
shift
shift
shift
shift
shift
shift
shift
shift
shift
shift
shift x16
X16
MUX
logic buffers for
Split Split 16:1
SPI interface
Split
CFG_C
CTRL_C
Table 5 below summarizes the characteristics and signals of the SPLIT PCB.
CTRL
PCI card Multiplexed
LO/CTRL
with BB Diplexer
RFIC LO2
Table 6 below summarizes the characteristics and signals of the SPLIT PCB.
Table 6: TILE controller PCB characteristics signals
Parameter Data
LO2 frequency 7.2 – 8GHz
CTRL frequency >300MHz
CFG signal SPI three wire 5.0V digital signal
Multiplexed LO/CTRL Analog, DC – 8GHz
USB USB 5V digital signal
Optics PCB
vertical option
SPLIT PCB
vertical option
TILE FEED
PCB
TILE FEED
PCB
TILE FEED
tiles PCB
horizontal
option
5.1.2 Layout
The TILE PCB is responsible for combining multiple radiating elements into a large array.
The arrangement of the elements themselves in the tile level determines the tile’s
capability and the packing of multiple tiles together determines the array capability.
Regarding the array packing consideration, the target is to make the tile as small as
possible, so that tiles can be packed close together thus minimizing array grating lobes.
However, since the tile holds also ICs for RF and DC and needs some mechanical support
as well as provisioning for heat dissipation, some area must be allocated for these needs
as well, thus setting some additional consideration for the size of the tile.
The TILE PCB layout and dimensions are shown in Fig. 21 below.
30mm
16.75mm
Fig. 22. TILE PCB front (radiating elements) and back (feeding wires).
5.2.2 TX chain
The TX chain block diagram is shown in Fig. 23 below with signal levels noted (in dBm,
unless noted otherwise) along the transmission chains. The different chains are each
marked with a different color and are all merged to one after the triplexer block. The
data signal is coming in around a 5GHz carrier and up-converted to around 15GHz IF
frequency (the exact frequency is the RF frequency divided by 4). The LO frequency is
around 7.5GHz (the exact frequency is the IF frequency divided by 2). There is a splitter
for each LO that splits it 8 to 1 (for the 8 chains), and likewise an 8 to 1 multiplexer for
the control signal.
LO1-AMP
LO1-split
LO1
15 25 0-3
DATA#0 (TX)
-15..-13 -7..0 0
To 32-tile Triplexer
-17
-18 -7 LO2
-18 -3 - 0
0.5Vpp LO2-split
MUX 0.5Vpp CTRL
8:1
CFG
3.3V
5.2.3 RX chain
The RX chain is very similar to the TX chain except that it requires significantly more
amplification. Its block diagram is shown in Fig. 24 below with signal levels noted (in
dBm, unless noted otherwise) along the receive chains. The different chains are each
marked with a different color and are all merged to one after the triplexer block. The
data signal is coming at the IF frequency around 15GHz (the exact frequency is the RF
frequency divided by 4) and down-converted to 5GHz. The LO frequency is around
7.5GHz (the exact frequency is the IF frequency divided by 2). There is a splitter for each
LO that splits it 8 to 1 (for the 8 chains), and likewise an 8 to 1 multiplexer for the
control signal. The RX signal comes in considerably weaker than the TX signal, so there is
a need for about 45dB of amplification to restore it to the level desired by the optical
card. A passive attenuation pad is also provided in order to tune the gain more closely if
such a need arises. The RX chain also requires a filter to block the undesired outputs of
the mixer from getting to the output (in the TX chain, the same function is performed by
the triplexer).
LO1-AMP
LO1-split
LO1
15 25 0-3
RX-AMP (45dB)
DATA#0 (RX)
-22..-25 -31..-39 -1..+9 0-4
ATTN
from 32-tile Triplexer
-17
-20 -7 LO2
-3 - 0
LO2-split
MUX 0.5Vpp CTRL
8:1
CFG
3.3V
5.2.4 Layout
The annotated layout for the TX chain is shown in Fig. 25 below.
Triplexer
LO1 signal
from LO1
splitter
Input
signal
Mixer
DC generation
block
DC generation
Output block
signal
RX filter
Amplification chain (3 x 15dB)
Mixer
LO1 signal
from LO1
Triplexer
splitter
5.2.5 DC section
The DC section of the PCB is designed to feed the RF chains and the Tile PCB. It is
protected against polarity reversal and regulated. It is recommended to work with a 6.5V
voltage source. The relevant schematic is shown in Fig. 27 below.
POWER:
10GHz AMPs: 5v* 0.220A=1.1W
5GHz AMPs: 12* 0.057A* 5v=3.42W
TILE PCBs =8* 3.3* 0.6=15.84W
Tot: ~21W
Port 1
CAP
(LO2) Port 2
(MUXED)
Port 3
(CTRL)
The simulated performance for the first diplexer is shown in Fig. 29 below.
Fig. 29. Simulated performance for diplexer between LO2 and CTRL signals.
The second diplexer is more complicated and consists of two microwave printed filters.
This diplexer multiplexes the output of the first diplexer (roughly DC to 8GHz) with the IF
signal (around 15GHz). The diplexer layout is shown in Fig. 30 below.
Port 1 Port 2
(LO2 + CTRL) (MUXED)
Port 3
(LO1)
The simulated performance for the second diplexer is shown in Fig. 31 below.
Fig. 31. Simulated performance for diplexer between LO1, LO2+CTRL signals.
The last filter is the RX image rejection filter placed between the mixer and amplifiers in
the RX chain. The mixer is designed to provide rejection starting a little after the desired
pass band which is up to 6GHz. The layout and simulated performance is shown in Fig.
32 and Fig. 33 below.
Port 1 Port 2
Waveguides
Wire Bond
TripleX
InP
Si Submount
MIMO PCB
Fig. 34. Schematic of the assembly concept for MIMO PCB.
Input/
Signal name Description Level Connector
output
LO1_T (TX,RX) LO signal, 9.5-11GHz Output 0 to 3dBm 16 x SMA
LO2_T LO signal, 7.2-8GHz Output -3 to 0dBm 16 x SMA
CTRL_T CTRL signal, 230MHz Bi-Dir 0dBm 16 x SMA
16 x USB type-A
CFG_DATA_T Configuration, SPI data, digital Bi-Dir 5V CMOS
receptacle
16 x USB type-A
CFG_CLK_T Configuration, SPI clock, digital Output 5V CMOS
receptacle
16 x USB type-A
CFG_STB_T Configuration, SPI strobe, digital Output 5V CMOS
receptacle
LO1_C (TX,RX) LO signal, 9.5-11GHz Input configurable 2 x SMA
LO2_C LO signal, 7.2-8GHz Input -3 to 0dBm SMA
CTRL_C CTRL signal, 230MHz Bi-Dir 0dBm SMA
USB type-B
CFG_DATA_C Configuration, SPI data, digital Bi-Dir 5V CMOS
receptacle
USB type-B
CFG_CLK_C Configuration, SPI clock, digital Input 5V CMOS
receptacle
USB type-B
CFG_STB_C Configuration, SPI strobe, digital Input 5V CMOS
receptacle
Fig. 36 shows the SPLIT PCB design which is currently being fabricated. There is an input
LO signal around 10GHz and an input LO signal around 7.8GHz, which are split in 16
outputs for the 16 TILE FEED PCBs so that the entire system uses coherent frequency
sources. It also has one control input signal which is also split in 16 signals for the TILE
FEED PCBs so that only one controller can be used.
5.5.2 Layout
The annotated layout for the TILE controller is shown in Fig. 36 below.
DC input
PC
signal LO2
input amplifier
Diplexer
Testing block
CTRL output U421 USB converter
Output
1.6V 1.6V 1.6V Matching
VBias_PD I IBias_2 IBias_3
Bias_1
The simulated characteristics of the RFIC are shown in the figures below. The NB-TIA
RFIC consumes 34.4 mW from a 1.6 V supply while delivering +4 dBm to the load with
amplification and noise figure exceeding specifications and ample matching.
Fig. 41. Optimum noise input impedance (real part) of the NB-TIA.
Fig. 45. DSP-assisted Analog Mobile Fronthaul architecture and signal flow, with
orange color indicating the evaluated A-RoF optical transmission and conversion
requires the use of an external Local Oscillator (LO) at IF frequency. The first stage of
characterization measurements includes the experimental verification of demodulator
process which is presented below. The evaluation performance of I/Q modulator will be
carried out through the use of new broadband analog RF modules that are going to be
integrated within the Flexbox as described in the last part. Fig. 46 shows the I/Q
demodulator module, commercially available from Analog Devices, that was employed in
these experimental measurements. It should be mentioned that the I/Q baseband
outputs were evaluated through the single-ended outputs.
Fig. 48. Constellation diagrams and eye diagrams of the IQ demodulator outputs
in back-to-back configuration for symbol rates of 250Mbuad, 500Mbaud and 750
Mbaud.
Fig. 50. Constellation diagrams and eye diagrams of the IQ demodulator outputs
by varying the roll-off factor of pulse shaping filter
Fig. 51. Constellation diagrams and eye diagrams of the IQ demodulator outputs
by varying the IF frequency of the modulated carrier
8. References
[1] J. Preissner, “The influence of the atmosphere on passive radiometric measurements”,
AGARD Conference on Millimeter and Submillimeter Wave Propagation and Circuits AGARD
Conference reprint 245, (1978).
[2] F. Meng, C. Boon, S. Xu, et al., “Millimeter-wave IC design techniques for beam-forming
applications”, IEEE International Workshop on Electromagnetics: Applications and Student
Innovation Competition (iWEM), 2016.
[3] M. Fakharzadeh, M. R. Nezhad-Ahmadi, B. Biglarbegian, et al., “CMOS Phased Array
Transceiver Technology for 60 GHz Wireless Applications”, IEEE Transactions On Antennas
and Propagation, vol. 58, no. 4, Apr. 2010
[4] C. Balanis, “Antenna Theory: Analysis and Design”, 4th edition, Wiley, 2016.
[5] S. Sun, T. Rappaport, R. Heath et al., “MIMO for Millimeter-Wave Wireless Communications:
Beamforming, Spatial Multiplexing, or Both?” IEEE Communications Magazine, December
2016.
[6] W. Roh, J.-Y. Seol, J. Park et al., “Millimeter-Wave Beamforming as an Enabling Technology
for 5G Cellular Communications: Theoretical Feasibility and Prototype Results”, IEEE
Communications Magazine, February 2014.
[7] N. Song, H. Sun and T. Yang, “Coordinated Hybrid Beamforming for Millimeter Wave Multi-
User Massive MIMO Systems”, IEEE Global Communications Conference (GLOBECOM),
December 2016.
[8] A. Roze, M. Crussiere, M. Helard, C. Langlais, “Comparison between a Hybrid Digital and
Analog Beamforming system and a Fully Digital Massive MIMO System with Adaptive
Beamsteering Receivers in Millimeter-Wave Transmissions”, International Symposium on
Wireless Communications Systems (ISWCS), Sep. 2016.
[9] ADL 5380, Quandrature demodulator from Analog Devices
[10] 3GPP, TS 38.104 V15.0.0, Table 9.6.2.3-1, Dec. 2017
[11] N. Argyris et al., “DSP enabled Fiber-Wireless IFoF/mmWave link for 5G Analog Mobile
Fronthaul”, in Proc. of 2018 IEEE 5G World Forum, 9-11 July 2018, Santa Clara, CA, USA