NYUSIM User Manual

by
Shihao Ju, Shu Sun, and Theodore. S. Rappaport

Version 3.1
Copyright 2016 - 2022
New York University and NYU WIRELESS

January 12, 2022

/
NYUSIM — User License
Copyright (c) 2016 - 2022 New York University and NYU WIRELESS

Permission is hereby granted, free of charge, to any person obtaining a copy

of this software and associated documentation files (the “Software”), to
deal in the Software without restriction, including without limitation the
rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
sell copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included
in all copies or substantial portions of the Software. Users shall cite NYU
WIRELESS publications regarding this work.

THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY

KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE
WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AU-
THORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAM-
AGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

i
NYUSIM Overview
Update: NYUSIM 3.1 January 12, 2022

NYU WIRELESS conducted millimeter-wave (mmWave) and sub-Terahertz

(THz) measurements from 2012 through 2020, having acquired a total of
over 1 TB of data, at frequencies from 28 to 142 GHz in various outdoor
and indoor environments such as urban microcell (UMi), urban macro-
cell (UMa), rural macrocell (RMa), and indoor hotspot for office (InH). The
measurements and analysis done in [1]–[21] led to the development of the
NYUSIM channel simulator. Built upon the NYUSIM version 2.01, the new
NYUSIM version 3.0 brings the capability of simulating wireless channels
in the indoor office scenario for carrier frequencies from 500 MHz to 150
GHz.

NYUSIM provides an accurate rendering of actual channel impulse responses

in both time and space, as well as realistic signal levels that were measured,
and may be utilized to support realistic physical layer and link layer simula-
tions such as those conducted in [8], [22]–[24]. The models and simulation
approach in NYUSIM involves the research of more than a dozen gradu-
ate and undergraduate students, and as of 2021, over 80,000 downloads of
NYUSIM have been recorded.

The current NYUSIM software package, version 3.0, extends the simula-
tion scenario from outdoor to indoor environments developed based on the
conventional drop-based statistical channel model. Previously, NYUSIM
version 1.x implemented the initial drop-based channel model for outdoor
scenarios for carrier frequencies from 0.5 GHz to 100 GHz [13], [25]. Then,
NYUSIM version 2.x implemented a spatial consistency-enabled channel
model with human blockage, and outdoor-to-indoor penetration loss mod-
eling components for outdoor scenarios [14]–[16]. The new NYUSIM ver-
sion 3.0 introduces the indoor scenario in the drop-based channel model
and allows the carrier frequency range for the indoor scenario from 0.5 GHz
to 150 GHz. Note that the frequency range from 100 to 150 GHz for out-
door scenarios and the spatial consistency mode for indoor scenario are not
available yet and will be added in future releases.

It is worth noting that in the 3GPP TR 38.901 outdoor channel model for
frequencies above 0.5 GHz [26]–[28], the number of clusters is unrealisti-
cally large. For example, in the UMi street canyon scenario, the number of
clusters in the line-of-sight (LOS) environment is as high as 12, and 19 in the
non-line-of-sight (NLOS) environment, which is not supported by the real-
world measurements at mmWave bands [1]–[4]. In contrast, in the outdoor
statistical spatial channel model (SSCM) implemented in NYUSIM [3], the
number of time clusters ranges from 1 to 6, and the mean number of spa-
tial lobes is about 2 and is upper-bounded by 5, which are obtained from
field observations and are much smaller than those in the 3GPP channel
model [8], [11], [13], [26].

Similarly, for the InH scenario in the 3GPP TR 38.901 indoor channel mod-
els, the number of clusters in the LOS and NLOS environments are set to 15

ii
and 19, respectively, and the number of subpaths per cluster is 20. In con-
trast, the measurements conducted in a typical office environment showed
that the maximum number of observed time clusters was 10 and 12 in the
LOS and NLOS scenarios. Thus, the NYUSIM SSCM models the number of
time clusters and the number of subpaths per cluster as random variables
with mean 5.4 and 6.3 for the NLOS scenario, respectively [18]. The upper
bounds of these channel statistics are determined from empirical measure-
ments [7], [17], [18]. The impractical number of clusters in the 3GPP channel
model is likely to result in a higher rank of mmWave channels, unrealistic
eigen-channel distributions, and thereby inaccurate spectral efficiency pre-
diction for 5G mmWave channels [8], [11].

Besides this user manual, five other files/folders are provided and available
for download from the website: http://wireless.engineering.nyu.edu/5g-
millimeter-wave-channel-modeling-software/. The other five files are listed
and explained as follows:

 NYUSIM — Channel Model Simulator Application: A standalone ap-

plication that contains a graphical user interface (GUI) for NYU’s chan-
nel simulator, which will be detailed in the first two sections of this
user manual. This application can run on computers with Windows
7 or higher and Mac operating systems (with or without MATLAB
installed).

 Base Code: The MATLAB code used in the current version of NYUSIM
to generate various channel parameters, which is created according
to the mmWave broadband SSCM presented in [3]. The “Base Code”
is applicable to arbitrary center carrier frequencies ranging from 500
MHz to 150 GHz, arbitrary RF bandwidths from 0 to 800 MHz, and
multi-antenna systems. Additionally, atmospheric attenuation effects,
which includes the attenuation induced by dry air (containing oxy-
gen), vapor, haze/fog, and rain, are also considered in the path loss
model in the “Base Code” (more details can be found in Section 3 of
this user manual).

 Time cluster - spatial lobe approach to cluster any measured or Ray

traced data: The MATLAB code implementing the time cluster - spa-
tial lobe (TCSL) algorithm used in the SSCM [3], [27]. Both outdoor
and indoor channel statistics were extracted using the TCSL algo-
rithm [18], [25].

 User License: The user license for NYUSIM, which is also shown on
the previous page of this user manual. This open-source license al-
lows anyone to use and modify the code provided copyright notices
and citations to the key papers of this work are provided. See the user
license for specific terms.

 Application Example Code: The example MATLAB code for possible

applications that make use of NYUSIM.

iii
New Features in Version 3.1
NYUSIM 3.1 enables simultaneous channel simulations with different trans-
mitter and receiver antenna polarizations. NYUSIM 3.0 and earlier can only
simulate one polarization, either co-polarization or cross-polarization, for a
single channel realization while NYUSIM 3.1 can simulate one polarization
(co- or cross-polarization), two polarizations (co- and cross-polarizations),
or four polarizations (namely, vertical-to-vertical (V-V), horizontal-to-horizontal
(H-H), vertical-to-horizontal (V-H), and horizontal-to-vertical (H-V) polar-
izations) for every single channel realization.

Cross-polarization discrimination (XPD) is defined as the signal power at-

tenuation (in dB) when the transmit antenna is vertically polarized and the
receive antenna horizontally polarized (i.e., cross-polarization) compared
to the case that both transmit and receive antennas are vertically or hori-
zontally polarized (i.e., co-polarization) including channel effect. XPD val-
ues over microwave and millimeter-wave frequencies at various indoor and
outdoor environments have been reported in the literature [2], [7], [18],
[29]–[37], showing that the XPD increases as the carrier frequency increases.
Thus, a linear function of frequency is used to fit the reported XPDs and cal-
culate the XPD over the NYUSIM supported frequency range (500 MHz to
150 GHz). In addition, the XPDs are found larger in LOS environments
than those in NLOS environments since the boresight path generally does
not have a depolarization effect. Two linear functions with different values
are used for the LOS and NLOS environments, respectively. A single XPD
is applied to the total received power instead of different XPDs to different
multipath components (MPCs) due to the limited available MPC-wise XPD
values in the literature, which makes the the polarization model succinct
yet valid.

Similarly, co-polarization discrimination (CPD) can be defined as the power

difference (in dB) between the V-V and H-H polarization settings or the V-H
and H-V polarization settings. The studies on the CPD are very limited [29],
[33], [38], [39] and are commonly regarded as negligible, which leads to a
simple model generating the H-H (H-V) power from the V-V (V-H) power
by adding a zero-mean Gaussian random variable with a small standard
deviation (1.6 dB used in NYUSIM 3.1).

Polarization has been widely used at mmWave frequencies as a degree of

freedom other than time, frequency, and space, to support multiple input
multiple output (MIMO) diversity and multiplexing [40]–[42]. The sub-
THz channels have been observed to be more sparse than mmWave chan-
nels [18], [20]. Thus, the dual-polarized transmit and receive antenna ar-
rays are necessary to deliver more signal energy and increase channel rank.
NYUSIM 3.1 can satisfy such simulation requirements to generate different
antenna polarization combinations for the same channel realization. More
details regarding the polarization modeling can be found in Section 3.9.

iv
Fixes and Updates

• NYUSIM 3.1 allows decimal inputs under the spatial consistency mode.
Specifically, the inputs for the correlation distances of shadow fading
and LOS/NLOS condition, update distance, moving distance, side
length for hexagon track can be decimals. Note that the correlation
distance of shadow fading, moving distance, and side length for
hexagon track need to be multiples of the update distance so that
the number of channel snapshots is an integer. Warnings related to
the user-specified inputs were added.

• NYUSIM 3.1 was compiled with the latest Matlab 2021b to tackle po-
tential system compatibility issues.

v
 Acknowledgements
We would like to thank Mr. Michael G. Cotton at the National Telecom-
munications & Information Administration (NTIA) for providing the ref-
erence and code on atmospheric attenuation characteristics at frequencies
below 1000 GHz, which have been used in the NYUSIM channel simulator
and this user manual.
This work is also supported by NYU WIRELESS Industrial Affiliates
and National Science Foundation (NSF) (Award Number: 1302336, 1320472,
1555332, 1702967, 1731290, and 1909206, 2037845).
The following NYU WIRELESS graduate students and faculty have con-
tributed to developing and supporting NYUSIM: Shihao Ju (lead), Shu Sun,
Yunchou Xing, Ojas Kanhere, George R. MacCartney, and Prof. Theodore
S. Rappaport.

vi
Revision History
1. Ver 3.1: January 12, 2022
Version 3.1 enables channel simulations with one polarization (co- or
cross-polarization), two polarizations (co- and cross-polarizations),
or four polarizations (namely, vertical-to-vertical (V-V), horizontal-
to-horizontal (H-H), vertical-to-horizontal (V-H), and horizontal-to-
vertical (H-V) polarizations) for every single channel realization. Note
that multiple-polarization (i.e., “Co/X-Pol” and “All-Pol”) options
are currently only enabled for the drop-based simulations. Consid-
ering the potentially massive number of output figures and files un-
der the spatial consistency mode, the options, “Co/X-Pol” and “All-
Pol”, will only generate co-polarized channels in the spatial consis-
tency simulations.
NYUSIM 3.1 allows users to input decimals for the correlation dis-
tances of shadow fading and LOS/NLOS condition, update distance,
moving distance, side length for hexagon track. Note that the corre-
lation distance of shadow fading, moving distance, and side length
for hexagon track need to be multiples of the update distance so that
the number of channel snapshots is an integer.

2. Ver 3.0: March 22, 2021

Version 3.0 introduces an indoor statistical channel model following
the mathematical framework of the existing NYUSIM outdoor chan-
nel modeling approach, which can simulate indoor channel impulse
responses from 500 MHz to 150 GHz with 0 Hz to 800 MHz RF band-
width. The indoor channel model was developed based on channel
statistics extracted from channel measurements conducted in an office
building at 28 GHz and 140 GHz [18].

3. Ver 2.01a: February 20, 2020

Version 2.01a modifies the example code “getChannelMatrix.m” in
the “Application Example Code” folder.
The multipath delay stored in the data “DirPDPInfo” is in the unit of
nanoseconds. A 10−9 was missing in Line 110 in the calculation of
the channel matrix to have the unit as second (s). In NYUSIM 2.01a,
the multipath delay is multiplied by 10−9 to have the correct unit as
seconds.

4. Ver 2.01: November 1, 2019

Version 2.01 provides two fixes in the function “getPowerInfo.m” and
the function “getLocalCIR.m”, respectively.

(a) The list of input parameters for the function “getPowerInfo.m” is

corrected to calculate large-scale path loss for the RMa scenario
in the spatial consistency mode.
(b) The unit inconsistency in the function “getLocalCIR.m” was fixed.
Because the default unit of the user-specified antenna spacing in

vii
 NYUSIM is wavelength, the equation implemented in NYUSIM
for calculating antenna array response vector is given by

a(φ) = [1, ej2πdλ cos(φ) , ej2π2dλ cos(φ) , ..., ej2π(N −1)dλ cos(φ) ] (1)

5. Ver 2.0: June 7th, 2019

Version 2.0 provides three new channel modeling components, spatial
consistency, human blockage, and O2I penetration loss.

(a) Spatial consistency indicates continuous and realistic channel evo-

lution along the user terminal (UT) trajectory in a local area.
To realize spatial consistency, spatially correlated large-scale pa-
rameters such as shadow fading, line-of-sight (LOS)/non-LOS
(NLOS) condition are generated, and time-variant small-scale
parameters such as angles, power, delay, phase of each MPC are
generated [43], [44]. Specifically, a geometry-based approach us-
ing multiple reflection surfaces is implemented to update spa-
tially correlated and time-variant angular information [45].
(b) A human blockage event usually causes a temporal shadowing
loss, which may last about several hundreds of milliseconds. A
typical blockage event can be divided into four stages, unshad-
owed, decay, shadowed, rising shown in Fig. 4 in [46], and a
four-state Markov model corresponding to four stages was pro-
posed to characterize blockage events [46], [47]. Such human
blockage shadowing loss model is implemented in NYUSIM 2.0
for both omnidirectional and directional channels.
(c) O2I penetration loss becomes more prominent at mmWave fre-
quencies as shown in measurements and models [30], [48]–[50].
Many modern buildings are constructed by concrete and have
infrared reflecting (IRR) glass, which induce a large penetration
loss when a mmWave signal is transmitted from outdoor to in-
door or vice versa [50]. A parabolic model for building penetra-
tion loss from [50], [51] is implemented in NYUSIM, which has
either a low loss or a high loss form, depending on the type of
building surface. The low loss model works for external build-
ing materials like standard glass and wood while the high loss
model works for external building materials like IRR glass and
concrete [50].

6. Ver 1.6.1: November 7th, 2018

Version 1.6.1 provides two major improvements to NYUSIM, based
on user feedback. The first improvement provides a complete MIMO
channel representation. The second improvement fixes a crash when
the RF bandwidth is set to be smaller than 800 MHz.

(a) Fix 1: In Version 1.6 and earlier versions, NYUSIM only used
the first TX antenna element with all RX antenna elements to
generate the MIMO channel impulse responses, where NYUSIM
should have used all TX antenna elements (NT X ) with all RX

viii
 antenna elements (NRX )when producing the MIMO channel im-
pulse responses. In other words, earlier versions of NYUSIM
only generated 1 x NRX complex voltages for a particular multi-
path component. However, for proper MIMO channel modeling,
NT X x NRX complex voltages of a particular multipath compo-
nent are needed. The MIMO channel function is improved to
generate a full MIMO impulse response (channel matrix) with
respect to the number of TX antennas and the number of RX an-
tennas. The parameter “CIR_MIMO” is expanded for all TX an-
tenna elements for actual MIMO implementations and adjusted
according to the user-specific RF bandwidth.
(b) Fix 2: In Version 1.6, a bug occurs when the RF bandwidth is set
to be less than 800 MHz. The time resolution of the system be-
comes more coarse as the bandwidth is set by the user to be nar-
rower than 800 MHz. Thus, fewer multipath components can
be resolved at narrower bandwidth compared to the 800 MHz
bandwidth. Thus, multipath components that arrive within a
time bin are vectorially summed, and the function “getNewPow-
erSpectrum.m” in v 1.6.1 will correctly generate the band-adjusted
power spectrum. AOA and AOD information retain 800 MHz
bandwidth resolution, even if the user decreases the RF band-
width.

7. Ver 1.6: December 15th, 2017

(a) In the base code package of NYUSIM, a new MATLAB script

with the name "NYUSIM_MainCode" is added which acts as the
main code of NYUSIM, such that users can obtain all the output
figures and data files by setting the input parameters in the main
code, and modify channel parameters and output files per users’
own needs, without opening the GUI. The meaning of each pa-
rameter is annotated in "NYUSIM_MainCode".
(b) For the LOS environment, the first cluster of multipath compo-
nents sometimes does not put the strongest multipath compo-
nent as the first arriving component. While strong reflections
could cause later arriving multipath components to be stronger
than the first arriving multipath component, generally it has been
found in the field that the first arriving multipath component
will always be the strongest for LOS conditions. Therefore, Ver-
sion 1.6 swaps (when necessary) the power level of the first ar-
riving multipath component with the strongest multipath com-
ponent in the first cluster of arriving components to ensure the
first arriving multipath component is the strongest for LOS. This
provides a realistic physical definition of the first arriving (short-
est delay) multipath component in LOS situations. Additionally,
for LOS environments, Version 1.5 and earlier did not keep track
of the boresight angles at the TX and RX, thus the AoD and AoA
of the LOS (strongest) multipath component may not be aligned
on boresight, as shown in the upper plot of Fig. 1. In Version 1.6,
the AoD and AoA of the LOS (strongest) multipath component

ix
F IGURE 1: Illustration of how the AoD and AoA of the LOS
(strongest) multipath component are aligned on boresight in LOS en-
vironments in NYUSIM Version 1.6.

are aligned on boresight for LOS environments, such that its az-
imuth AoD and azimuth AoA differ by 180◦ , and its elevation
AoA is the additive inverse of its elevation AoD, to match the
physics of on-boresight communication. The AoDs and AoAs
of all the other multipath components are shifted accordingly in
relation to the LOS multipath component, as illustrated in the
lower plot of Fig. 1. The "NYUSIM_MainCode" is documented
to show this, and this change was requested by industry users
who are performing 5G system simulations. No such changes
are made to NLOS environments.
(c) The PLE in the UMa NLOS scenario is changed from 3.2 to 2.9
to make it more accurate according to the data from extensive
measurements presented in [5]. This change is made inside the
GUI code that is inaccessible to users, but is also reflected in
"NYUSIM_MainCode" mentioned above that is available to users,
where the PLE is represented by the variable n as annotated in
the code itself.
(d) A new input parameter named "Distance Range Option" is added
to the GUI in Version 1.6, which has two options: standard (10-
500 m) (used in Version 1.5 and earlier versions) and extended
(10-10,000 m). The NYUSIM channel model predicts accurate
results for standard distance range (10-500 m). The extended
distance range beyond 500 m should be set with caution, since
NYUSIM is developed based on field measurements within 500
m, as explained in Section 3.1.1 of this user manual. The ex-
tended distance range is added based on users’ requests. For
the distance range no larger than 500 m, the dynamic range (i.e.,
largest possible path loss) is set to 190 dB in NYUSIM based

x
 on field measurement results [1], [2], [5], while for the distance
range beyond 500 m, the dynamic range is set to 220 dB.
(e) Another new input parameter named "Base Station Height" is
added to the GUI and output data files for the RMa scenario in
Version 1.6, which is utilized for calculating path loss in the RMa
scenario. This variable name is annotated in the "NYUSIM_MainCode"
itself.
(f) Both LOS and NLOS path loss models in the RMa scenario are
modified to incorporate the dependence on the base station height
based on Eqs. (21) and (22) in [12].
(g) For the RMa scenario, the number of time clusters and the num-
ber of spatial lobes are both set to one, and the maximum num-
ber of multipath components is set to two, based on our RMa
mmWave field measurements which show sparsity for rural sce-
narios [12].
(h) In the title or text in output figures, the scenario type (i.e., UMi,
UMa, or RMa) is added in addition to the environment type (i.e.,
LOS or NLOS).
(i) In the output AoD and AoA spectra, the multipath component
with the minimum received power is also shown as a line (it was
shown as a dot in the center in previous versions due to a dif-
ferent plotting setting). Furthermore, the view point of the AoD
and AoA spectra is changed in Version 1.6 to be along the z-axis,
and the legend position is adjusted to avoid overlapping with the
spectrum. For the path loss scatter plot, the x-axis is extended to
10 km even if the standard distance range is used.
(j) The Ricean K-factor is now calculated for every PDP, and is an
output parameter in the data file for each simulation run, where
the K-factor refers to the ratio of the power of the strongest mul-
tipath component in the PDP to the sum of powers of all the
other multipath components in the PDP [4]. For Version 1.6, in
LOS environments, this will be the ratio of the power of the first
arriving multipath component to the sum of powers of all later
arriving multipath components. This was suggested by industry
users.
(k) New warning messages are added that will occur when input
parameters are not reasonable or exceed their predefined ranges
shown on the GUI, as explained in detail in Section 2 of this user
manual.
(l) More references, such as [8] and [11], are added in the user man-
ual regarding the comparison between the 3GPP [26] and NYUSIM
channel models, and the significant impact of the choice of chan-
nel models on wireless system performance evaluation.
8. Ver 1.5: April 5th, 2017
(a) A more detailed overview about the mmWave measurements on
which NYUSIM has been built, the related publications, and key
differences/advantages of the NYUSIM channel model as com-
pared to the 3GPP TR 38.900 Release 14 channel model are added
in NYUSIM Overview on Page ii of this user manual.

xi
(b) The name of the output figure "Omni. PDP with Directional
Antenna Gains" is changed to "Directional PDP with Strongest
Power" to make it more accurate. Textual information on chan-
nel parameter values, such as frequency, environment, distance,
root-mean-square (RMS) delay spread, received power, path loss,
is added to the output figure "Directional PDP with Strongest
Power", in which the path loss equals the transmit power sub-
tracted by the transmitter (TX) and receiver (RX) antenna gains
and the received power.
(c) One additional output figure entitled "PathLossPlot" is added if
the user performs N (N ≥ 1) continuous simulation runs with
the same input parameters but with varying distances. The fig-
ure "PathLossPlot" illustrates a scatter plot of omnidirectional
path loss and directional path loss values for over the entire dis-
tance range generated from the N (N ≥ 1) continuous simu-
lation runs, along with the fitted path loss exponent (PLE) and
shadow fading standard deviation. Detailed information about
the figure "PathLossPlot" is provided in Section 2.2.1 of this user
manual.
(d) Four additional output data files, namely, "OmniPDPInfo.txt",
"DirPDPInfo.txt", "OmniPDPInfo.mat", and "DirPDPInfo.mat", are
added after N (N ≥ 1) continuous simulation runs with the
same input parameters are performed by the user. "Omni" de-
notes omnidirectional, while "Dir" stands for directional. De-
tailed descriptions about these four files were added in Section
2.2.2 of this user manual.
(e) The output files "BasicParameters.txt" and "BasicParameters.mat"
are generated only once for all N (N ≥ 1) continuous simula-
tion runs with the same input parameters, instead of being gen-
erated for each individual simulation run, since the parameters
contained in these two files remain the same for N (N ≥ 1) con-
tinuous simulation runs.
(f) Two parameter values in each of the two output files "BasicPa-
rameters.txt" and "BasicParameters.mat", i.e., omnidirectional re-
ceived power and omnidirectional path loss, are removed since
they are now included in the new output files "OmniPDPInfo.txt"
and "OmniPDPInfo.mat".
(g) Detailed explanation on how the directional PDP with strongest
power is generated is added in Section 2.2.1 of this user manual.
(h) Examples of the six output figures from NYUSIM are added in
Section 2.2.1 of this user manual.
(i) Error warnings are added for the cases where the input antenna
HPBW by the user exceeds the pre-defined range on the GUI.
Explanations are detailed in Section 2.1.5 of this user manual.
(j) Figs. 3.1 and 3.2 in Section 3.1 are modified where the humidity
is changed from 98% to 80%, and temperature is changed from
-20◦ C to 20◦ C, to reflect more realistic atmospheric conditions.

xii
 (k) A new chapter (Chapter 4) is created to present a variety of appli-
cations of NYUSIM. The original application section in Chapter
2 is also moved to Chapter 4.
(l) The MATLAB code "getChannelMatrix" and the associated ex-
planation in the NYUSIM application code package are updated
according to the new output data files.
(m) The root folder for users to select from to save output files is
changed from the C drive to the folder "Users".
(n) A new version of the NYUSIM software is created for the Mac
operating system, with all of the above updates implemented.
Installation instructions are provided in Section 1.

9. Ver 1.4: October 31st, 2016

(a) More operation steps were added on the top part of the simulator
GUI to better instruct users.
(b) Detailed explanations of the half-power beamwidth (HPBW) and
gain of the antenna array, and examples of how to create an an-
tenna pattern given the HPBW and gain, were added in Section
2.1.1 of this user manual.
(c) The .mat format of the output data files was provided in the
simulator, in addition to the existing .txt files. Descriptions of
the newly added files were provided in Section 2.2.2 of this user
manual. A new option was added on the GUI that allows users
to select the output file type (text file, MAT file, or both).
(d) A new set of output data file named “BasicParameters” was added
with both .txt and .mat formats available, which contains all the
input parameter values as shown on the GUI when running the
simulation and some relevant output parameters, as detailed in
Section 2.2.2.
(e) An example of the applications of the output data files was demon-
strated in Section 2.2.3, namely, calculating the condition number
and rank of multiple-input multiple-output (MIMO) orthogonal
frequency division multiplexing (OFDM) channel matrices, and
the corresponding MATLAB script was provided in the package
named “Application Example Code” available on the NYUSIM
downloading website.

10. Ver 1.3: August 19th, 2016

(a) Descriptions of the numbers of matrix columns and rows in the

output .txt files containing the complex channel coefficients (H),
phases (HPhases), and powers (HPowers) were corrected in Sec-
tion 2.2.2 of this user manual.
(b) The rural macrocell (RMa) path loss parameters in the close-in
reference distance (CI) path loss model have been slightly changed
in the simulator and Section 3.1 of this user manual, based on the
new mmWave rural path loss study in [9].
(c) The new frequency range of the simulator has been tested and is
usable from 500 MHz to 100 GHz.

xiii
11. Ver 1.2: August 2nd, 2016

(a) Four new input parameters, i.e., Polarization, Foliage Loss, Dis-
tance Within Foliage, and Foliage Attenuation, were added to the
input parameters on the NYUSIM GUI, the corresponding code
was added to the base code, and the corresponding descriptions
of the four new input parameters were added in Section 2 of the
user manual.
(b) The NYU CI RMa path loss model was added to the simulator
and user manual, and the option “RMa” was added to the “Sce-
nario” parameter on the GUI.
(c) More background regarding the creation and application of NYUSIM
was added to the section NYUSIM Overview, and two related ci-
tations were added to the reference list.
(d) A 32-bit version of NYUSIM was generated, in addition to the
existing 64-bit version.

12. Ver 1.1: July 16th, 2016

(a) Four new input parameters, i.e., Barometric Pressure, Humidity,

Temperature, and Rain Rate, were added to the input parameters
on the NYUSIM GUI, the corresponding code was added to the
base code, and the corresponding descriptions of the four new
input parameters were added in Section 2 of the user manual.

xiv
Contents

NYUSIM — User License i

NYUSIM Overview ii

New Features in Version 3.1 iv

Acknowledgements vi

Revision History vii

1 Installation of Software Package 1

1.1 System Requirements . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Operating System . . . . . . . . . . . . . . . . . . . . . 1
1.2 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2.1 Windows Operating System . . . . . . . . . . . . . . . 1
1.2.2 Mac OS X . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Basics 3
2.1 Graphical User Interface and Simulator Basics . . . . . . . . 3
2.1.1 Input Parameters . . . . . . . . . . . . . . . . . . . . . 4
2.1.2 Output Folder Selection . . . . . . . . . . . . . . . . . 12
2.1.3 Output File Type Selection . . . . . . . . . . . . . . . . 12
2.1.4 Operation of the GUI . . . . . . . . . . . . . . . . . . . 12
2.2 Output Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.1 Output Figures . . . . . . . . . . . . . . . . . . . . . . 13
— Drop-based Mode . . . . . . . . . . . . . . . . . . . 13
— Spatial Consistency Mode . . . . . . . . . . . . . . 19
2.2.2 Output Data Files . . . . . . . . . . . . . . . . . . . . . 19
— Drop-based Mode . . . . . . . . . . . . . . . . . . . 19
— Spatial Consistency Mode . . . . . . . . . . . . . . 24

3 Channel Model Supported by NYUSIM 27

3.1 Supported Scenarios . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Path Loss Model . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.1 Range Extension . . . . . . . . . . . . . . . . . . . . . 30
3.3 Wideband Temporal/Spatial Clustering Algorithm . . . . . 30
3.4 Indoor statistical channel model . . . . . . . . . . . . . . . . . 33
3.4.1 Large-scale Path Loss Model . . . . . . . . . . . . . . 33
3.4.2 Small-scale Channel Impulse Response Model . . . . 34
Input parameters of channel generation procedure . 34
Parameter frequency-dependency for the indoor sce-
nario . . . . . . . . . . . . . . . . . . . . . . . 34

xv
 3.5 Channel Bandwidth and Resolvable MPC . . . . . . . . . . . 37
3.6 Spatial Consistency Procedure . . . . . . . . . . . . . . . . . . 37
3.7 Outdoor to Indoor Penetration Loss Model . . . . . . . . . . 47
3.8 Human Blockage Shadowing Loss Model . . . . . . . . . . . 48
3.9 Polarization Model . . . . . . . . . . . . . . . . . . . . . . . . 52

4 Applications of NYUSIM 54
4.1 Application 1: MIMO Channel Condition Number . . . . . . 54
4.2 Application 2: Spectral Efficiency Comparison Between 3GPP
and NYUSIM Channel Models . . . . . . . . . . . . . . . . . 58
4.3 Application 3: Temporal Large-scale Path Loss Variation Due
to Spatial Consistency . . . . . . . . . . . . . . . . . . . . . . 59
4.4 Application 4: Human Blockage Shadowing Loss Simulations 61
4.5 Application 5: O2I Penetration Loss Simulations . . . . . . . 61
4.6 Application 6: Indoor Coverage Prediction . . . . . . . . . . 63

Bibliography 65

xvi
List of Figures

1 Illustration of how the AoD and AoA of the LOS (strongest)

multipath component are aligned on boresight in LOS envi-
ronments in NYUSIM Version 1.6. . . . . . . . . . . . . . . . x

1.1 Possible installation error for Mac system. . . . . . . . . . . . 2

2.1 Graphical User Interface (GUI) of NYUSIM 3.0 with four pan-
els: channel parameters, antenna properties, spatial consis-
tency parameters, human blockage parameters. . . . . . . . . 3
2.2 Example of a 3D AOD power spectrum generated from NYUSIM. 15
2.3 Example of a 3D AOA power spectrum generated from NYUSIM. 16
2.4 Example of an omnidirectional PDP generated from NYUSIM. 16
2.5 Example of a directional PDP with the strongest received
power generated from NYUSIM. "Ant." denotes antenna. . . 17
2.6 Example of the PDPs over different receive antenna elements
generated from NYUSIM. . . . . . . . . . . . . . . . . . . . . 17
2.7 Example of a scatter plot showing the omnidirectional and
directional path loss values generated from NYUSIM with
100 simulation runs for the 28 GHz UMi LOS scenario. n
denotes the pass loss exponent (PLE), σ is the shadow fading
standard deviation, "omni" denotes omnidirectional, "dir" rep-
resents directional, "dir-best" means the direction with the
strongest received power, "Ant." denotes antenna, "AZ" and
"EL" stand for azimuth and elevation, respectively. . . . . . . 18
2.8 A map of spatially correlated shadow fading with the BS and
UT locations. The map of spatially correlated shadow fading
is generated by filtering a map of independent shadow fad-
ing using an exponential function in (3.5). Shadow fading
[dB]∼N(0,4) in a UMi LOS scenario. T-R separation distance
is 100 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.9 A sample map of spatially correlated LOS/NLOS condition 20
2.10 A sample user track . . . . . . . . . . . . . . . . . . . . . . . . 21
2.11 A sample consecutive omnidirectional PDPs . . . . . . . . . 21
2.12 A sample consecutive directional PDPs . . . . . . . . . . . . 22

3.1 Propagation attenuation due to dry air, vapor, haze, and rain
at mmWave frequencies, with a barometric pressure of 1013.25
mbar, a relative humidity of 80%, a temperature of 20◦ C, and
a rain rate of 5 mm/hr [52]. . . . . . . . . . . . . . . . . . . . 28

xvii
3.2 Collective attenuation effects of dry air, vapor, haze, and rain
at mmWave frequencies, with a barometric pressure of 1013.25
mbar, a relative humidity of 80%, a temperature of 20◦ C, and
a rain rate of 5 mm/hr [52]. . . . . . . . . . . . . . . . . . . . 29
3.3 Example of an omnidirectional PDP with four multipath taps [3]. 32
3.4 Example of an AOA power spectrum with four multipath
taps [3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5 28 GHz indoor omnidirectional path loss scatter plot and
MMSE-fitted CI path loss model with distance for LOS and
NLOS scenarios [17]. . . . . . . . . . . . . . . . . . . . . . . . 33
3.6 Frequency-dependent PLE for the indoor LOS scenario. . . . 35
3.7 A 400 m x 400 m spatially correlated map for LOS/NLOS
condition in UMi scenario using a 2-D filter. The correlation
distance is 15 m. The granularity of the map is 1 m. The
heights of the BS and UT are 10 m and 1.5 m, respectively.
The LOS and NLOS condition transition may happen along
the UT trajectory according to the UT location in the map. . 40
3.8 UT movement and change of AOD in the azimuth plane.
The moving distance of the UT is approximated as the cor-
responding arc length with the BS as the center . . . . . . . . 41
3.9 3D geometry illustration of the coordination system for a LOS
component. AOD and ZOD from the BS and AOA and ZOA
from the UT are shown. . . . . . . . . . . . . . . . . . . . . . 42
3.10 UT movement and change of AOD in the azimuth plane in
NLOS scenarios. The mirror images of the UT location and
velocity direction φv are shown as UT’ and φvr . . . . . . . . 43
3.11 UT movement and change of AOD in the azimuth plane in
NLOS scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.12 Update of delay and phase based on the change of path length
in time interval ∆t. τ (t0 ), φv , and φAOA are known, τ (t1 ) can
be calculated using the law of cosine. . . . . . . . . . . . . . . 46
3.13 Four-state Markov model for unshadowed, decaying, shad-
owed, and rising states for modeling blockage events [78]. . 49
3.14 Five independently simulated Markov traces. . . . . . . . . . 51
3.15 The measured XPDs with linear fits for the LOS and NLOS
environments. . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.1 Empirical CDF of the condition number of channel matrices

for OFDM sub-carriers with different transmit and receive
antenna elements Nt and Nr for MIMO-OFDM channels in
one simulation run. . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Empirical CDF of the rank of channel matrices for OFDM
sub-carriers with different transmit and receive antenna ele-
ments Nt and Nr for MIMO-OFDM channels in one simula-
tion run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3 Spectral efficiency achieved by hybrid beamforming for a
256×16 system at 28 GHz with rectangular antenna arrays
at the transmitter and receiver with four RF chains. . . . . . 59

xviii
4.4 A map of spatially correlated shadow fading with the BS and
UT locations. The map of spatially correlated shadow fading
is generated by filtering a map of independent shadow fad-
ing using an exponential function in (3.5). Shadow fading
[dB]∼N(0,4) in a UMi LOS scenario. T-R separation distance
is 100 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.5 The UT moves in a partial hexagon track in a clockwise man-
ner. The track distance is 40 m, the side length of the hexagon
is 10 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.6 CDFs of simulated human blockage shadowing loss using
the four-state Markov model for directional channels with
RX antenna azimuth HPBWs (7°, 15°, 30°, and 60°. . . . . . . 61
4.7 An example of NYUSIM run with O2I penetration loss. . . . 62
4.8 Omnidirectional PDP output from a NYUSIM run with 15.0
dB O2I penetration loss. . . . . . . . . . . . . . . . . . . . . . 62
4.9 Directional PDP output from a NYUSIM run with 15.0 dB
O2I penetration loss. . . . . . . . . . . . . . . . . . . . . . . . 62
4.10 Scatter plot of the received powers and the average power
level of the received signals at distances from 5 to 50 m for
LOS and NLOS indoor directional channels. For the NLOS
environment, the average power level of received signals drops
below the receiver sensitivity beyond 35.8 m. . . . . . . . . . 64

xix
1. Installation of Software Package

1.1 System Requirements

1.1.1 Operating System
Minimum requirement: Windows 7 (or higher), or Macintosh operating
systems (Mac OS X)

The Windows version of NYUSIM runs on Windows operating systems

only, and the Mac version of NYUSIM runs on Mac OS X only.

1.2 Installation
The required program files of the NYUSIM software suite are provided by
NYU WIRELESS via Internet download at http://wireless.engineering.nyu.edu/5g-
millimeter-wave-channel-modeling-software/.

1.2.1 Windows Operating System

The Windows version of NYUSIM is delivered as an installer package for
Windows operating systems. All components required to run the appli-
cations are automatically installed after double clicking on the setup exe-
cutable (NYUSIM_V3_WIN.exe file).

1.2.2 Mac OS X
The Mac version of NYUSIM is delivered as an installer package for Mac
OS X. All components required to run the applications are automatically in-
stalled after double clicking on the setup executable (NYUSIM_V3_MAC.app
file).
One potential problem is that the recent Mac system (High Sierra 10.13
and later) introduced a new security feature called gatekeeper path ran-
domization. This security feature makes it difficult to run applications in-
stalled outside of the Mac App Store. If a popup window shows an error
during the installation as shown in Fig. 1.1, please execute the following
command in the terminal command line:

sudo xattr -d com.apple.quarantine NYUSIM_V3_MAC.app

1
 F IGURE 1.1: Possible installation error for Mac system.

For additional support, please contact [email protected].

2
2. Basics

2.1 Graphical User Interface and Simulator Basics

The screenshot in Fig. 2.1 shows the GUI of NYUSIM 3.0. The simulator
performs Monte Carlo simulations, generating samples of CIRs at specific
T-R separation distances. Note that the range of T-R separation is provided
by the user, and the actual T-R separation distance is uniformly selected
from the user-specified distance range. Two running modes, drop-based
mode and spatial consistency mode are made available from NYUSIM 2.0.
When the spatial consistency button is “on”, NYUSIM runs spatial consis-
tency procedure and generates successive and correlated CIRs along the
UT trajectory. When the spatial consistency button is “off”, NYUSIM runs
the drop-based model which is the same as previous versions of NYUSIM
and generates independent CIRs for different distances. Note that human
blockage module works for both drop-based mode and spatial consistency
mode. While earlier versions of NYUSIM provided outdoor channel mod-
els as well as outdoor-to-indoor penetration, NYUSIM 3.0 introduces the
indoor office scenario (only in the drop-based mode). The indoor scenario
can be selected under the “Scenario” drop-down list. The spatial consis-
tency feature for indoor scenario will be implemented in future release.

F IGURE 2.1: Graphical User Interface (GUI) of NYUSIM 3.0 with four
panels: channel parameters, antenna properties, spatial consistency
parameters, human blockage parameters.

3
2.1.1 Input Parameters
There are 49 input parameters to the channel simulator, which are grouped
into four main categories: Channel Parameters, Antenna Properties, Spatial
Consistency Parameters, and Human Blockage Parameters, as shown on the
GUI in Fig. 2.1. Note that two input parameters for O2I penetration loss
are put into the panel Channel Parameters.
The panel Channel Parameters contains 19 fundamental input parameters
about the propagation channel, as listed and explained below:

1. Scenario: a selectable parameter denoting the scenario. Four channel

scenario options, urban microcell (“UMi”), urban macrocell (“UMa”),
rural macrocell (“RMa”), and indoor hotspot (“InH”) are applicable.
The default scenario option is “UMi”. Selecting the “InH” option
will change the allowable carrier frequency range from 0.5-100 GHz
(for outdoor scenarios) to 0.5-150 GHz, distance range option from
“Standard (10-500m)” to “Indoor (5-50 m), the upper limit of base sta-
tion height from 150 m (for outdoor scenarios) to 3 m. Selecting the
“InH” option will also disable following options: “Rain Rate”, “Fo-
liage Loss”, “Distance Within Foliage”, “Foliage Attenuation”, “Out-
door to Indoor (O2I) Penetration Loss”, “O2I Loss Type”, and spatial
consistency mode.

2. Frequency (GHz): an editable parameter denoting the carrier frequency

in GHz. The default value is 28 (GHz), and it can be varied from 0.5
to 100 (GHz) with at most one decimal place for UMi, UMa, and RMa
scenarios. When “InH” is selected, the allowable carrier frequency
range will range from 0.5 to 150 (GHz).

3. RF Bandwidth (MHz): an editable parameter denoting the RF band-

width of the transmitted signal in MHz. The default value is 800 MHz,
and it can be varied from 0 to 800 MHz. As the simulator was devel-
oped from real-world measurements obtained with an RF bandwidth
of 800 MHz, it can only scale down from 800 MHz.

4. Distance Range Option: a selectable parameter denoting the distance

range. Three options, “Standard (10-500 m)”, “Extended (10-10,000
m)", and “Indoor (5-50 m)” are applicable. The default setting is “Stan-
dard (10-500 m)”. For distances less than 500 m, the dynamic range
(i.e., largest possible path loss) is set to 190 dB in NYUSIM based on
field measurement results [1], [2], [5], while for the distance range
beyond 500 m, the dynamic range is set to 220 dB. When “InH” is se-
lected, the “Distance Range Option” is fixed to be “Indoor (5-50 m)”
option.

5. Environment: a selectable parameter denoting the environment, ei-

ther line-of-sight (LOS) or non-line-of-sight (NLOS). The default set-
ting is LOS.

4
 6. Lower Bound of T-R Separation Distance (m): an editable parame-
ter denoting the smallest distance between the transmitter (TX) and
receiver (RX) in meters with at most one decimal place. The default
value is 10 m, and it can be varied from 10 m to 500 m for the standard
range (verified by extensive measurements by NYU), 10 m to 10 km
for the extended range, and 5 m to 50 m for the indoor range but no
more than the upper bound of the T-R separation distance.

7. Upper Bound of T-R Separation Distance (m): an editable parameter

denoting the largest distance between the TX and RX in meters with
at most one decimal place. The default value is 500 m, and it can
be varied from 10 m to 500 m for the standard range (verified by ex-
tensive measurements by NYU), and 10 m to 10 km for the extended
range, and 5 m to 50 m for the indoor range, but no less than the lower
bound of the T-R separation distance.

8. TX Power (dBm): an editable parameter denoting the transmit power

in dBm. The default value is 30 (dBm), and can be set to any value
ranging from 0 to 50 (dBm).

9. Base Station Height (m): an editable parameter denoting the base sta-
tion height in meters. The default value is 35 (m) [26], and can be set
to any value ranging from 10 to 150 (m) [12]. This base station height
is only applicable to RMa modeling and is ignored for UMi and UMa
scenarios. The maximum height of stations is 3 m for the InH scenario.

10. User Terminal Height (m): an editable parameter denoting the UT

height in meters. The default value is 1.5 (m) [26], and can be set to
any value ranging from 1 to 10 (m) above the ground [12].

11. Barometric Pressure: an editable parameter denoting the baromet-

ric pressure in mbar used in evaluating propagation path loss in-
duced by dry air. The default and typical value is 1013.25 mbar (mil-
libar) (i.e., nominal for sea level), and may range from 10−5 to 1013.25
(mbar) [52].

12. Humidity: an editable parameter denoting the relative humidity in

percentage used in evaluating propagation path loss induced by va-
por. The default value is 50 (%), and can be set to any number between
0 and 100 (%).

13. Temperature: an editable parameter denoting the temperature in de-

grees Celsius used in evaluating propagation path loss induced by
haze/fog. The default and typical value is 20 (◦ C), and may range
from -100 to 50 (◦ C) [52].

5
14. Rain Rate: an editable parameter denoting the rain rate in mm/hr
used in evaluating additional propagation path loss induced by rain.
The default value is 0 (mm/hr), and the typical range is 0 to 150
(mm/hr) [1]. This option is disabled for the InH scenario.

15. Polarization: a selectable parameter denoting the polarization rela-

tion between the TX and RX antennas or antenna arrays. The default
setting is Co-Pol (co-polarization), and can be changed to X-Pol (cross-
polarization), Co/X-Pol (co- and cross-polarization), and all-Pol (V-V,
H-H, V-H, and H-V polarization). The cross-polarization discrimi-
nation (XPD) can vary from 5 dB to 27 dB [53]–[55], depending on
the frequency and environment. In this simulator, for Co-Pol, no ex-
tra loss will be added to the path loss, while an extra scenario- and
frequency-dependent loss will be added to the path loss for X-Pol due
to polarization mismatch based on the measurement results in [55].
In addition, a zero-mean Gaussian random variable will be added to
generate the H-H (H-V) power from the V-V (V-H) power. For more
detailed background, please refer to Section 3.9.

16. Foliage Loss: a selectable parameter indicating whether or not fo-

liage loss will be considered in the simulation. The default setting
is “No” (which implies foliage loss will not be considered), and can
be changed to “Yes” (which means foliage loss will be considered).
This option is disabled for the InH scenario.

17. Distance Within Foliage: an editable parameter representing the dis-

tance in meters that the transmitted signal travels within foliage. The
default value is 0, and can be set to any non-negative number no
larger than the lower bound of the T-R separation distance. This op-
tion is disabled for the InH scenario.

18. Foliage Attenuation: an editable parameter denoting the propagation

loss induced by foliage in dB/m. The default value is 0.4 (dB/m)
based on the measurement results in [55], and can be set to any value
between 0 and 10 (dB/m). For more detailed background, please refer
to [55]. This option is disabled for the InH scenario.

19. Number of RX Locations: an editable parameter denoting the number

of RX locations. It can be any positive integer number. The default
value is 1, and can be set to any integer from 1 to 10,000.

20. Outdoor to Indoor (O2I) Penetration Loss: a selectable parameter de-

noting that a UT is in an indoor or outdoor scenario. Two options,
“Yes” and ”No”, are provided. The default value is “No”. If “Yes” is
selected, additional building penetration loss will be added to the to-
tal received power before small-scale parameters are generated [51],
[56]. If “No” is selected, penetration loss is not considered. This op-
tion is disabled for the InH scenario.

6
 21. O2I Loss Type: a selectable parameter denoting the O2I loss type. Two
option, “Low Loss” and “High Loss”, are provided. The “Low Loss”
model corresponds to low loss building materials such as standard
glass and wood [50], [51]. The “High Loss” corresponds to high loss
building materials such as infrared reflecting (IRR) glass and concrete
[50], [51]. This option is disabled for the InH scenario.

The panel Antenna Properties contains 12 input parameters related to the TX

and RX antenna arrays, as listed and explained below:

1. TX Array Type: a selectable parameter denoting the TX antenna array

type. The default setting is ULA (uniform linear array), and can be
changed to URA (uniform rectangular array).

2. RX Array Type: a selectable parameter denoting the RX antenna array

type. The default setting is ULA, and can be changed to URA.

3. Number of TX Antenna Elements Nt : an editable parameter denot-

ing the total number of TX antenna elements in the array. The default
value is 1, and can be set to any integer from 1 to 128.

4. Number of RX Antenna Elements Nr : an editable parameter denot-

ing the total number of RX antenna elements in the array. The default
value is 1, and can be set to any integer from 1 to 64.

5. TX Antenna Spacing (in wavelength): an editable parameter denot-

ing the spacing between adjacent TX antennas in the array in terms of
the carrier wavelength. The default value is 0.5, and can be set to any
positive number with up to one decimal place from 0.1 to 100.

6. RX Antenna Spacing (in wavelength): an editable parameter denot-

ing the spacing between adjacent RX antennas in the array in terms of
the carrier wavelength. The default value is 0.5, and can be set to any
positive number with up to one decimal place from 0.1 to 100.

7. Number of TX Antenna Elements Per Row Wt : an editable parameter

denoting the number of TX antennas in one dimension when the TX
Array Type is ULA or URA, which should divide the number of TX
antenna elements. The default value is 1.

8. Number of RX Antenna Elements Per Row Wr : an editable parameter

denoting the number of RX antennas in one dimension when the RX
Array Type is ULA or URA, which should divide the number of RX
antenna elements. The default value is 1.

7
 9. TX Antenna Azimuth HPBW (degrees): an editable parameter denot-
ing the azimuth half-power-beamwidth (HPBW) of the TX antenna
(array) in degrees. The default value is 10◦ , and can be set to any
value from 7◦ to 360◦ (since the smallest azimuth HPBW of the anten-
nas used in the measurements for the simulator was 7◦ ).

10. TX Antenna Elevation HPBW (degrees): an editable parameter de-

noting the elevation HPBW of the TX antenna (array) in degrees. The
default value is 10◦ , and can be set to any value from 7◦ to 45◦ (since
the smallest elevation HPBW of the antennas used in the measure-
ments for the simulator was 7◦ ).

11. RX Antenna Azimuth HPBW (degrees): an editable parameter denot-

ing the azimuth HPBW of the RX antenna (array) in degrees. The
default value is 10◦ , and can be set to any value from 7◦ to 360◦ .

12. RX Antenna Elevation HPBW (degrees): an editable parameter de-

noting the elevation HPBW of the RX antenna (array) in degrees. The
default value is 10◦ , and can be set to any value from 7◦ to 45◦ .

It is worth noting that the HPBW in the input parameters is for the entire
antenna array, instead of for each antenna element, when the number of an-
tenna elements is more than one at the TX and/or RX. Conventionally, the
HPBW of an antenna array is a function of the number of antenna elements
and the antenna spacing, but in this simulator these three parameters (i.e.,
the HPBW, number of antenna elements, and antenna spacing) can be inde-
pendently specified by the user, since there may be a wide range of beam-
forming approaches (e.g., digital, analog, hybrid), in all of which different
individual antenna element types (e.g., patch antennas, vertical antennas,
horns) may be used. To make the simulator as general as possible, we spec-
ify the number of individual antenna elements (without specifying the gain
or combining losses thereof) and the total array HPBW. Note that the an-
tenna pattern and implementation details are only very loosely defined to
specify the gain in the particular pointing (maximum gain) direction.

Since some antenna elements may have more gain or loss depending on
the specific array type (e.g., ULA); that is, someone may choose to fabri-
cate their individual array with more individual antenna element gain than
another person will, and losses vary with fabrication process (e.g., on chip
antenna elements have more loss when a lens is not used than if a lens is
used). Instead of dealing with the myriad antenna fabrication and connec-
tion details needed to make an array — there are numerous variables —
we simply allow the total HPBW and the number of elements to be speci-
fied, without specifying the individual antenna element gain. For example,
assuming there are four elements in an antenna array, where each element
has 6 dB gain because they are patch or Yagi/directional or horn antennas
for each element, then this array can have the same overall HPBW as a 16-
element array with unity gain dipole or vertical elements.

8
Without specifying the specific antenna elements and their interconnected
characteristics, only the HPBW is specified while the beam pattern is left
undefined. Consequently, one has the freedom to implement an antenna
pattern of their choice for system simulations, which may include inter-
ference from signals outside of the main gain direction. Here are some
heuristic approaches of constructing one’s own antenna pattern based on
the specified HPBW, where the antenna gains in the following equations
are all relative to an isotropic antenna.

• The antenna pattern employed in [3] has the following form:

G02 −βφ2
G(θ, φ) = max(G0 e−αθ ) ,
100
4ln(2) 4ln(2) 41253η (2.1)
α= 2 , β = 2 , G0 =
θ3dB φ3dB θ3dB φ3dB

where (θ, φ) denote the azimuth and elevation angle offsets from
the boresight direction in degrees, G0 is the maximum directive gain
(boresight gain) in linear units, (θ3dB , φ3dB ) represent the azimuth
and elevation HPBWs in degrees, (α, β) are parameters that depend
on the HPBW values, and η = 0.7 is a typical average antenna effi-
ciency.

• The antenna pattern used in [6] has the expression below:

f (φ, θ) = G sinc2 (a · sin(φ)) cos2 (φ)

 
(2.2)
· sinc2 (b · sin(θ)) cos2 (θ)
 

where φ and θ represent the azimuth and elevation angles with re-
spect to the antenna boresight, respectively, f (φ, θ) denotes the radia-
tion power density at the azimuth angle φ and elevation angle θ, G is
the antenna boresight gain, and a and b are functions of the azimuth
(AZ) and elevation (EL) HPBWs of the horn antenna, respectively, i.e.,
   
2 HPBWAZ 2 HPBWAZ 1
sinc a · sin( ) cos = (2.3)
2 2 2
   
HPBWEL HPBWEL 1
sinc2 b · sin( ) cos2 = (2.4)
2 2 2

• The radiation pattern of a sectored cell site antenna was employed

in [57], where the azimuthal radiation pattern is modeled as a cardioid
given by [57]
π
r(θ) = α[1 + sin(θ + )] (2.5)
2
where r is the gain of the antenna at azimuth angle θ from its maxi-
mum lobe and is a scaling factor. The elevation radiation pattern is an
ellipse with the base station at a focus point [57]:

x2 y 2
+ 2 =1 (2.6)
a2 b

9
 • A sectored antenna pattern model was introduced in [58], where con-
stant directivity gains are assumed for the main lobe and the side lobe.
The panel Spatial Consistency contains 10 input parameters related to the
spatial consistency implementation, as listed and explained below: (Spatial
consistency mode will be disabled for indoor channel simulation)
1. Correlation Distance of Shadow Fading (5-60 m): an editable param-
eter denoting the correlation distance of shadow fading. The default
value is 10 m. Correlation distance of shadow fading, also known as
“local area” determines the length of a channel segment. Correlation
distance is the distance beyond which the auto-correlation value of a
large-scale parameter (LSP) falls below 0.5 [14]. Within the correla-
tion distance, LSPs are considered spatially correlated during the UT
movement. If a UT moves beyond the correlation distance, the LSPs
are considered independent and are randomly generated again based
on the pre-defined probabilistic distributions [15]. Specifically, a new
channel will be initialized and new CIRs will be independently regen-
erated when a UT moves beyond the correlation distance of shadow
fading.

2. Correlation Distance of LOS/NLOS Condition (5-60 m): an editable

parameter denoting the correlation distance of LOS/NLOS condition.
Correlation distance of LOS/NLOS condition is used to generate spa-
tially correlated LOS/NLOS values, which means a UT that moves
in a local area should always experience same visibility condition
(LOS or NLOS) [14], [15]. Specifically, the visibility condition does
not change within a channel segment, but may change by reading the
spatially correlated map based on the UT location when the next seg-
ment starts.

3. UT Track Type: a selectable parameter denoting the UT track type.

Two options, “Linear” and “Hexagon” are provided to the user. The
default option is “Linear” [16].

4. Track Distance (1-100 m): an editable parameter denoting the track

length of a UT trajectory. Note that the input track distance for the
“Hexagon” track should not be larger than the perimeter of the hexagon.
Realistic CIR generation cannot be guaranteed out of the limit. The
track distance is divided into multiple channel segments based on the
correlation distance of shadow fading.

5. Update Distance (<1 m): an editable parameter denoting the distance

interval between two channel snapshots. The default value is 1 m.
Decreasing the update distance causes more CIRs to be generated by
NYUSIM with the separation between each CIR generated by NYUSIM
equal to the update distance. Note that the update distance should
not be set too small because it is computationally expensive. Update
distances of 0.1, 0.5, and 1 m are suggested.

6. Moving direction (0°-360°): an editable parameter denoting the ini-

tial UT moving direction. The default value is 45°. Both “Linear”
and “Hexagon” track require an initial moving direction. For a linear
track, the moving direction does not change during the simulation.

10
 For a hexagon track, the moving direction changes based on the track
distance and side length of the hexagon.

7. UT Velocity (1 - 30 m/s): an editable parameter denoting the UT ve-

locity. The default value is 1 m/s, close to the typical human walking
speed. The UT velocity in concert with the update distance determine
the update time which is the time interval between two consecutive
CIRs.

8. Side Length: an editable parameter denoting the side length of the

hexagon track. This parameter only works for the “Hexagon” track.
The default value is 10 m.

9. Orientation: a selectable parameter denoting the orientation of the

hexagon track. This parameter only works for the “Hexagon” track.
Two options, “Clockwise” and “Counter”, are provided.

10. Segment Transitions: a selectable parameter denoting whether the

smooth transitions between two segments. Two options, “Yes” and
”No” are provided. With “Yes” selected a post-processing is done
to “connect” generated channel segments via cluster birth and death.
“No” generates independent channel segments, with channel snap-
shots are spatially correlated in each channel segment [16].

The panel Human Blockage contains 6 input parameters related to the

human blockage shadowing loss due to a person near the mobile phone
(UT), as listed and explained below. Human blockage can be turned “on”
or “off”. As explained in Section 3, the length of a Markov trace is 20 s long,
and the time resolution is 1 ms.

1. Default Settings for Human Blockage: a selectable parameter. Two

options, “Yes” and “No”, are provided. If the user chooses “Yes”, the
transition rates and average mean attenuation are implemented based
on a linear fit to the data given in [59], and changes to fit the work in
[59] as the RX antenna azimuth HPBW changes. If the user chooses
“No”, the transition rates and mean attenuation are free to edit for
user-specified preferences, in case other Markov model parameters
are preferred or discovered in the future [47], [59].

2. Mean Attenuation: an editable parameter denoting the mean attenu-

ation of blockage events (SEmean ). The default value is 14.4 dB.

3. Trans. Rate from Unshadow to Decay (1/s): an editable parameter

denoting the transition rate from unshadowed state to decay state.
The default value is 0.20 when the RX antenna azimuth HPBW is 10°.

4. Trans. Rate from Decay to Shadow (1/s): an editable parameter de-

noting the transition rate from decay state to shadowed state. The
default value is 8.08 when the RX antenna azimuth HPBW is 10°.

5. Trans. Rate from Shadow to Rise (1/s): an editable parameter denot-

ing the transition rate from shadowed state to rise state. The default
value is 7.85 when the RX antenna azimuth HPBW is 10°.

11
 6. Trans. Rate from Rise to Unshadow (1/s): an editable parameter de-
noting the transition rate from rise state to unshadowed state. The
default value is 6.70 when the RX antenna azimuth HPBW is 10°.

2.1.2 Output Folder Selection

To the right of the above four category panels on the GUI (see Fig. 2.1),
there is an option named “Select a Folder to Save Files”. This option allows
users to select a folder to save the output files (e.g., figures, data files) from
the simulator. The default path is the disk that contains the current running
folder. Users can select a folder inside the default path by clicking on the
desired paths/folders.

2.1.3 Output File Type Selection

To the right of the folder selection option on the GUI, there is an option
named “Output File Type”. This option allows users to select a file type for
the output data files from the simulator. The default type is Text File, and
can be switched to MAT File, and Both Text and MAT File.

2.1.4 Operation of the GUI

Five basic steps, as shown on the top left panel on the GUI, need to be
executed to run the channel simulator:

1. To begin (reset) the simulator, click Start (Reset)

2. Set your input parameters below

3. Select a folder to save files

4. Click Run

5. To run another simulation, click Reset, and repeat Steps 2-4

The meanings of these steps are detailed as follows.

When opening the GUI for the first time, users shall click on the yellow but-
ton “Start” to initiate the simulator and enable the input of the GUI. Next,
users can set and/or select the input parameters according to their own
needs; the simulator will execute the default values for each parameter if
there is no input from the user. Then, users shall select a folder to save all
the output files by clicking on the desired paths/folders under the option
“Select a Folder to Save Files”. Finally, users shall click on the light green
button “Run” to run the simulations.

While the simulator is running, a progress bar will pop up at the bottom left
of the GUI, which informs users of the progress of the simulations. After
the simulation is complete, the progress bar will disappear automatically,
and the five figures generated from the first simulation run will pop up on
the screen.

To initiate another set of simulation runs (e.g., with different input param-
eter values), users shall click on the dark green button “Reset”, after which

12
all the input parameters from the previous simulation run will be set as the
default input values.

The simulator can be closed at any time by clicking the red button “Exit”
on the bottom right of the GUI.

If the input parameter exceeds the predefined range shown on the NYUSIM
GUI or is not reasonable/logical, an error message will pop up to let the
user either reset the parameter on GUI or modify the NYUSIM source code
without using the GUI to meet users’ own needs. For example, if the input
number of TX antenna elements per row Wt is larger than the input number
of TX antenna elements Nt , a corresponding error message will pop up.
Users shall click "OK" on the error bar, then click the dark green button
"Reset" on the GUI to reset the input parameter, and the following steps are
identical to those described above.

2.2 Output Files

Since Version 2.0, NYUSIM provides two running simulation modes, drop-
based mode and spatial consistency mode, which depends on that the spa-
tial consistency feature is “on” or “off”. The default output figures and files
in the drop-based mode are the same as the figures and files output from
older version of NYUSIM (e.g. Version 1.6 and earlier). The output figures
and files for the drop-based mode and the spatial consistency mode are in-
troduced separately as follows.

2.2.1 Output Figures

— Drop-based Mode
The drop-based mode assumes random and independent channel simula-
tions for any T-R separation distance. An ensemble of random, indepen-
dent channel responses is easily generated in NYUSIM for a fixed T-R sepa-
ration distance or a range of T-R separation distances using the drop-based
mode. A simulation run may produce hundreds or thousands of channel
impulse responses for different T-R separation distances.
For each simulation run, five figures will be generated and stored that
are based on the particular results of the simulation that is being run, and
an additional figure of path loss scatter plot will be generated and stored af-
ter N (N ≥ 1) continuous simulation runs with the same input parameters
are complete. Note that the T-R separation distance is not an input param-
eter, but, instead, the lower bound and upper bound of the T-R separation
distance are input parameters, such that the actual T-R separation distance
will vary automatically among different simulation runs. Regardless of the
number of simulation runs (RX locations), the five figures generated from
the first simulation run, as well as the last figure generated for N (N ≥ 1)
continuous simulation runs with the same input parameters, will pop up on
the screen for visual purposes. The contents of those figures are as follows:

• Three-dimensional (3D) AoD power spectrum, as illustrated in Fig. 2.2.

• 3D AoA power spectrum, as shown in Fig. 2.3.

13
• A sample omnidirectional PDP, as displayed in Fig. 2.4. Some fun-
damental information such as the frequency, environment, T-R sepa-
ration distance, RMS delay spread, omnidirectional received power,
omnidirectional path loss, and PLE [1], [2] is displayed on the PDP
plot. The lower limit of the y-axis denotes the noise threshold (i.e.,
the minimum received power of each resolvable multipath compo-
nent) determined by the transmit power, dynamic range of our mea-
surement system (180 dB), and a 10 dB SNR, i.e., the threshold equals
the transmit power in logarithmic scale minus 170 dB.

• A sample directional PDP with strongest power, where directional an-

tenna gain patterns are implemented at the TX and/or RX, as depicted
in Fig. 2.5. This figure is generated by allowing users to implement
arbitrary directional antenna patterns (gains and HPBWs) in an om-
nidirectional PDP, since directional antennas/antenna arrays will be
utilized at the TX and/or the RX in a realistic mmWave communica-
tion system to provide gains to compensate for the higher free space
path loss at mmWave frequencies. To obtain the directional PDP with
the strongest received power, NYUSIM searches for the best point-
ing angle out of all possible pointing angles, using the specified an-
tenna details (i.e., azimuth and elevation HPBWs of TX and RX an-
tennas) after first generating the omnidirectional PDP, such that the
pointing angle of the TX and RX are found that gives the strongest
received power. The TX/RX antenna gain pattern is calculated by
NYUSIM using Eq. (2.1) in this user manual by employing the az-
imuth and elevation HPBWs of TX and RX antennas specified by the
user on the GUI. This feature makes the channel simulator more valu-
able as it shows how a PDP will look like in a channel with direc-
tional antennas/antenna arrays used at the communication link end.
On the directional PDP figure, relevant channel and antenna param-
eters, such as the frequency, environment, T-R separation distance,
directional RMS delay spread, directional received power, directional
path loss, directional PLE, and TX and RX antenna HPBWs and gains,
are also displayed, where the directional path loss equals the transmit
power plus TX and RX antenna gains, minus the directional received
power [2], [6], [7].

• A series of PDPs over each receive antenna element obtained using

Eq. (3) of [60], as shown in Fig. 2.6, where the antenna array type,
number of antenna elements, and antenna element spacing are speci-
fied on the GUI by the user.

• A path loss scatter plot entitled "PathLossPlot" generated after N (N ≥

1) continuous simulation runs with the same input parameters, as
illustrated in Fig. 2.7. This figure shows omnidirectional path loss
and directional path loss values for over the entire distance range
generated from the N (N ≥ 1) continuous simulation runs, along
with the fitted PLE and shadow fading standard deviation using the
minimum-mean-square-error (MMSE) method [5], [7]. In the legend
of the figure "PathLossPlot", n denotes the pass loss exponent (PLE),

14
 F IGURE 2.2: Example of a 3D AOD power spectrum generated from
NYUSIM.

σ is the shadow fading standard deviation, "omni" denotes omnidi-

rectional, "dir" represents directional, and "dir-best" means the direc-
tion with the strongest received power. For producing the directional
path loss at each RX location, NYUSIM searches for all possible point-
ing angles in increments of the azimuth and elevation HPBWs of the
TX/RX antenna specified by the user on the GUI after first generating
the omnidirectional PDP. The TX/RX antenna gain pattern is calcu-
lated by NYUSIM using Eq. (2.1) in this user manual based on the
azimuth and elevation HPBWs of TX and RX antennas specified by
the user on the GUI. The directional path loss is equal to the transmit
power plus the TX and RX antenna gains, minus the directional re-
ceived power [1], [2], [61], [62]. For generating Fig. 2.7, the antenna
azimuth and elevation HPBWs are set to 10.9◦ and 8.6◦ , respectively,
at both the TX and the RX, to match the antenna HPBWs used in the
28 GHz measurements [1], [2]. The simulated PLE and shadow fading
standard deviation values agree well with the measured results pre-
sented in Table V and Table VIII of [2]. Directional path loss and di-
rectional PLE will always be larger (i.e., a directional channel is more
lossy) than the omnidirectional case, because the directional antenna
will spatially filter out many multipath components due to its direc-
tional pattern, such that the RX receives fewer multipath components
hence less energy, thereby the directional path loss is higher after re-
moving the antenna gain effect from the received power [2], [7].
If the generated path loss in a simulation run exceeds the corresponding dy-
namic range, i.e., if there are no detectable multipath components, then it
will be shown on the output figures that "No Detectable Multipath Compo-
nents above the Threshold of XXX dBm", where the threshold value equals
the transmit power in dBm minus the dynamic range in dB for that simula-
tion run, such as -190 dBm. Note that no path loss data points will be shown

15
F IGURE 2.3: Example of a 3D AOA power spectrum generated from
NYUSIM.

F IGURE 2.4: Example of an omnidirectional PDP generated from

NYUSIM.

16
F IGURE 2.5: Example of a directional PDP with the strongest re-
ceived power generated from NYUSIM. "Ant." denotes antenna.

F IGURE 2.6: Example of the PDPs over different receive antenna ele-
ments generated from NYUSIM.

17
F IGURE 2.7: Example of a scatter plot showing the omnidirectional
and directional path loss values generated from NYUSIM with 100
simulation runs for the 28 GHz UMi LOS scenario. n denotes the
pass loss exponent (PLE), σ is the shadow fading standard deviation,
"omni" denotes omnidirectional, "dir" represents directional, "dir-
best" means the direction with the strongest received power, "Ant."
denotes antenna, "AZ" and "EL" stand for azimuth and elevation, re-
spectively.

18
on the path loss scatter plot as long as there are no detectable multipath
components in the omnidirectional PDP, regardless of the directional PDP.
Users may modify this in "NYUSIM_MainCode.m" per your own needs.

— Spatial Consistency Mode

The spatial consistency mode assumes a UT travels a specific path, and
generates correlated and consecutive channel impulse responses for suc-
cessive sampling points on the path. A series of spatially correlated and
time-variant channel responses are easily generated in NYUSIM under the
spatial consistency mode. A simulation run may produce hundreds or
thousands of closely-spaced channel impulse responses based on the UT
trajectory and velocity.
For each simulation run, five figures will be generated and stored that
are based on the particular results of the simulation that is being run. These
five figures will pop up on the screen for visual purposes. The contents of
those figures are as follows.
• A map of spatially correlated shadow fading with the UT and BS lo-
cations and UT track is output as shown in Fig. 2.8. Some basic in-
formation such as frequency, environment, T-R separation distance,
standard deviation of shadowing fading, track distance, velocity, is
displayed in the figure.
• A map for spatially correlated LOS/NLOS condition using scenario-
specific LOS probability is output as illustrated in Fig. 2.9, which indi-
cates that all locations in a local area experience the same propagation
condition (LOS or NLOS). Some basic information such as frequency,
environment, T-R separation distance, standard deviation of shadow-
ing fading, track distance, velocity is displayed in the figure.
• User track: The part of the map for spatially correlated shadow fading
along which the UT moves as illustrated in Fig. 2.10. The moving
direction and track distance are shown in this figure.
• Consecutive omnidirectional PDPs along the user trajectory are dis-
played in Fig. 2.11 to show the power variation and delay drifting
of MPCs, where directional antenna gain patterns are implemented
at the TX and/or RX. Some fundamental information such as fre-
quency, environment, T-R separation distance, track distance, velocity
is shown in the PDP plot.
• Consecutive directional PDPs with the strongest received power along
the user trajectory are displayed in Fig. 2.12 to show the power varia-
tion and delay drifting of MPCs. Some fundamental information such
as frequency, environment, T-R separation distance, track distance,
velocity is shown in the PDP plot.

2.2.2 Output Data Files

— Drop-based Mode
For each simulation run, five sets of .txt files and five corresponding .mat
files are generated, namely, “AODLobePowerSpectrumn_Lobex.txt”,

19
F IGURE 2.8: A map of spatially correlated shadow fading with the BS
and UT locations. The map of spatially correlated shadow fading is
generated by filtering a map of independent shadow fading using an
exponential function in (3.5). Shadow fading [dB]∼N(0,4) in a UMi
LOS scenario. T-R separation distance is 100 m.

F IGURE 2.9: A sample map of spatially correlated LOS/NLOS con-

dition

20
 F IGURE 2.10: A sample user track

F IGURE 2.11: A sample consecutive omnidirectional PDPs

21
 F IGURE 2.12: A sample consecutive directional PDPs

“AODLobePowerSpectrumn.mat”, “AOALobePowerSpectrumn_Lobex.txt”,
“AOALobePowerSpectrumn.mat”, “OmniPDPn.txt”, “OmniPDPn.mat”,
“DirectionalPDPn.txt”, “DirectionalPDPn.mat”, “SmallScalePDPn.txt”, and
“SmallScalePDPn.mat”, where n denotes the nth RX location (i.e., nth simu-
lation run), and x represents the xth spatial lobe. After N (N ≥ 1) continu-
ous simulation runs with the same input parameters are complete, another
three .txt files and three corresponding .mat files are produced, i.e., "Basic-
Parameters.txt", "BasicParameters.mat", "OmniPDPInfo.txt", "OmniPDPInfo.mat",
"DirPDPInfo.txt", and "DirPDPInfo.mat".

Each text file “AODLobePowerSpectrumn_Lobex” is associated with the

output figure of 3D AOD power spectrum, and contains five parameters
(columns) of each resolvable multipath component in an AOD spatial lobe,
which are listed and explained below.
1. pathDelay (ns): an array containing the absolute propagation time
delays of all resolvable multipath components in nanoseconds (ns).
2. pathPower (mWatts): an array containing the received powers of all
resolvable multipath components in mWatts.
3. pathPhase (rad): an array containing the phases of all resolvable mul-
tipath components in radians.
4. AOD (degree): an array containing the azimuth AODs of all resolv-
able multipath components in degrees.
5. ZOD (degree): an array containing the ZODs of all resolvable multi-
path components in degrees.
Note that inside the corresponding .mat file “AODLobePowerSpectrumn”
is a structure containing the lobe matrices, each of which is composed of
five columns as described above.

Each text file “AOALobePowerSpectrumn_Lobex” is associated with the

output figure of 3D AOA power spectrum, and contains five parameters

22
(columns) of each resolvable multipath component in an AOA spatial lobe,
which are listed and explained below.

1. pathDelay (ns): an array containing the absolute propagation time

delays of all resolvable multipath components in nanoseconds (ns).

2. pathPower (mWatts): an array containing the received powers of all

resolvable multipath components in mWatts.

3. pathPhase (rad): an array containing the phases of all resolvable mul-

tipath components in radians.

4. AOA (degree): an array containing the azimuth AOAs of all resolv-

able multipath components in degrees.

5. ZOA (degree): an array containing the ZOAs of all resolvable multi-

path components in degrees.

Note that inside the corresponding .mat file “AOALobePowerSpectrumn”

is a structure containing the lobe matrices, each of which is composed of
five columns as described above.

Each .txt and .mat file “OmniPDPn” is associated with the output figure of
omnidirectional PDP, and contains two columns: the first column denotes
the propagation time delay in nanoseconds, and the second column repre-
sents the received power in dBm.

Each .txt and .mat file “DirectionalPDPn” is associated with the output fig-
ure of omnidirectional PDP with directional antenna gain patterns imple-
mented, and contains two columns: the first column denotes the propa-
gation time delay in nanoseconds, and the second column represents the
received power in dBm.

Each .txt and .mat file “SmallScalePDPn” is associated with the output fig-
ure of the series of omnidirectional PDPs over RX antenna elements, and
contains three columns: the first column denotes the receiver antenna sep-
aration in terms of number of wavelengths, the second column is the prop-
agation time delay in nanoseconds, and the third column represents the
received power in dBm. Note that the noise power is set to -150 dBm for
visual purpose.

The text file "BasicParameters.txt" and the .mat file "BasicParameters.mat"

subsume all the input parameter values as shown on the GUI when run-
ning the simulation.

The text file "OmniPDPInfo.txt" and the .mat file "OmniPDPInfo.mat" con-
tain five columns where each column represents a key parameter for each
of the N omnidirectional PDPs from N continuous simulation runs. The
parameters are listed and explained below.

1. T-R Separation Distance (m)

2. Received Power (dBm): omnidirectional received power in dBm

23
 3. Path Loss (dB): omnidirectional path loss in dB

4. RMS Delay Spread (ns): omnidirectional RMS delay spread in nanosec-

ond (ns)

5. Ricean K-factor (dB): ratio of the strongest power of the multipath

component (the first arriving multipath component in LOS) to the
sum of powers of the other multipath components [4], converted from
linear to the dB scale.

The text file "DirPDPInfo.txt" and the .mat file "DirPDPInfo.mat" contain
11 columns where each column represents a key parameter for each of the
directional PDPs from N continuous simulation runs, where the same kind
of parameters from each simulation run are cascaded in the same column.
The parameters are listed and explained below.

1. Simulation run number

2. T-R Separation Distance (m)

3. Time Delay (ns): absolute propagation time delay of each resolvable

multipath component in ns

4. Received Power (dBm): received power of each resolvable multipath

component in dBm without antenna gains

5. Phase (rad): phase of each resolvable multipath component in radians

6. Azimuth AoD (degree): azimuth AoD of each resolvable multipath

component in degrees

7. Elevation AoD (degree): elevation AoD of each resolvable multipath

component in degrees

8. Azimuth AoA (degree): azimuth AoA of each resolvable multipath

component in degrees

9. Elevation AoA (degree): elevation AoA of each resolvable multipath

component in degrees

10. Path Loss (dB): directional path loss obtained by aligning the TX/RX
antenna boresight on the AoD/AoA of each resolvable multipath com-
ponent. The directional path loss equal the transmit power plus the
TX/RX antenna boresight gains, minus directional received power.

11. RMS Delay Spread (ns): directional RMS delay spread in ns for each
directional PDP

— Spatial Consistency Mode

For each simulation run, five sets of .txt files and five corresponding .mat
files are generated, namely, “OmniPDP_Snapx.txt(/.mat)”, “DirectionalPDP_Snapx.txt(/.mat)”,
“OmniPDPInfo.txt(/.mat)”, “DirPDPInfo.txt(/.mat)”, and “BasicParameters.txt(/.mat)”.
x represents the index of a channel snapshot.
Each text file “OmniPDP_Snapx” contains seven parameters (rows) of
all resolvable MPCs (columns), which are listed and explained below.

24
 1. pathDelay (ns): The first column represents delays of MPCs in ns.

2. pathPower (dBm): The second column represents powers of MPCs

with omnidirectional antennas at TX and RX in dBm.

3. pathPhase (rad): The third column represents phases of MPCs in rad.

4. AOD (degree): The fourth column represents AODs of MPCs in de-

gree.

5. ZOD (degree): The fifth column represents ZODs of MPCs in degree.

6. AOA (degree): The sixth column represents AOAs of MPCs in degree.

7. ZOA (degree): The seventh column represents ZOAs of MPCs in de-

gree.

Each text file “DirectionalPDP_Snapx” contains seven parameters (rows)

of all resolvable MPCs (columns) of the xth channel snapshot, which are
listed and explained below.

1. pathDelay (ns): The first column represents delays of MPCs in ns.

2. pathPower (dBm): The second column represents powers of MPCs

with user-specified directional antenna patterns at TX and RX in dBm.

3. pathPhase (rad): The third column represents phases of MPCs in rad.

4. AOD (degree): The fourth column represents AODs of MPCs in de-

gree.

5. ZOD (degree): The fifth column represents ZODs of MPCs in degree.

6. AOA (degree): The sixth column represents AOAs of MPCs in degree.

7. ZOA (degree): The seventh column represents ZOAs of MPCs in de-

gree.

The text file “OmniPDPInfo” contains five columns where each column
represents a key parameter for each of omnidirectional PDPs from all simu-
lated channel snapshots. Each row corresponds to a channel snapshot. The
parameters are listed and explained below.

1. T-R Separation Distance (m)

2. Received Power (dBm): omnidirectional received power in dBm

3. Path Loss (dB): omnidirectional path loss in dB

4. RMS Delay Spread (ns): omnidirectional RMS delay spread in nanosec-

ond (ns)

5. Ricean K-factor (dB): ratio of the strongest power of the multipath

component (the first arriving multipath component in LOS) to the
sum of powers of the other multipath components [4], converted from
linear to the dB scale.

25
 The text file “OmniPDPInfo” contains five columns where each column
represents a key parameter for each of directional PDPs from all simulated
channel snapshots, where the same kind of parameters from each channel
snapshot are cascaded in the same column. The number of rows of each
channel snapshot corresponds to the number of MPCs of that channel snap-
shot. The parameters are listed and explained below.

1. Snapshot index

2. T-R Separation Distance (m)

3. Received Power (dBm): directional received power in dBm

4. Path Loss (dB): omnidirectional path loss in dB

5. RMS Delay Spread (ns): omnidirectional RMS delay spread in nanosec-

ond (ns)

The text file “BasicParameters” contains all user-input parameters such

as channel parameters, antenna properties, spatial consistency parameters,
and human blockage parameters.

26
3. Channel Model Supported by NYUSIM

The broadband statistical spatial channel model (SSCM) [3] developed by

NYU is used in NYUSIM as a base. Spatial consistency procedure [14], [15],
[45] and human blockage model [47], [59] developed by NYU, and the O2I
penetration loss model in [50], [51] are used as additional features.

3.1 Supported Scenarios

Earlier versions included outdoor urban microcell (UMi), urban macrocell
(UMa) and rural macrocell (RMa) scenarios, spatial consistency, human
blockage, and outdoor-to-indoor penetration loss modeling components
[13], [16]. With the release of NYUSIM 3.0, we have added the drop-based
channel models for indoor office scenario developed based on the conven-
tional drop-based statistical channel model [17], [18]. NYUSIM version 1.x
implemented the initial drop-based channel model for outdoor UMi, UMa,
and RMa scenarios including for carrier frequencies from 0.5 GHz to 100
GHz [13]. Then, NYUSIM version 2.x implemented a spatial consistency-
enabled channel model with human blockage, and outdoor-to-indoor pen-
etration loss modeling components for outdoor scenarios [16]. The current
NYUSIM version 3.0 can simulate the indoor hotspot (InH) scenario and
extends the carrier frequency range of 0.5 GHz to 150 GHz for the InH sce-
nario [18]. Section 3.4 introduces the path loss model and the small-scale
multipath channel model for the indoor scenario.

3.2 Path Loss Model

The close-in free space reference distance (CI) path loss model with a 1 m
anchor point, with an extra attenuation term due to various atmospheric
attenuation factors [52], is employed in NYUSIM, which is expressed as [2],
[5], [63]:
 
CI d
PL (f, d)[dB] =FSPL(f, 1 m)[dB] + 10nlog10 + AT[dB] + χCI
σ ,
d0
where d ≥ d0 m
(3.1)

where f denotes the carrier frequency in GHz, d is the 3D T-R separation

distance, n represents the path loss exponent (PLE), d0 denotes the free
space reference distance in meters, which is set to 1 m in the NYUSIM
channel model [5]. Users can change d0 to some value other than 1 m in
NYUSIM_MainCode.m, but d0 should not exceed 5 m to guarantee free
space propagation within d0 . AT is the attenuation term induced by the

27
 10-3
0.015 1.5

(dB/m) - Dry Air

(dB/m) - Vapor
0.01 1

0.005 0.5

0 0
20 40 60 80 100 20 40 60 80 100
f (GHz) f (GHz)
10-6 10-3
6 4
(dB/m) - Haze, Fog

(dB/m) - Rain
3
4
2
2
1

0 0
20 40 60 80 100 20 40 60 80 100
f (GHz) f (GHz)

F IGURE 3.1: Propagation attenuation due to dry air, vapor, haze, and
rain at mmWave frequencies, with a barometric pressure of 1013.25
mbar, a relative humidity of 80%, a temperature of 20◦ C, and a rain
rate of 5 mm/hr [52].

atmosphere, χCIσ is a zero-mean Gaussian random variable with a standard

deviation σ in dB, and FSPL(f, 1 m) denotes the free space path loss in dB
at a T-R separation distance of 1 m at the carrier frequency f :

4πf × 109
 
FSPL(f, 1 m)[dB] = 20log10 = 32.4[dB] + 20log10 (f ) (3.2)
c

where c is the speed of light, and f is in GHz. The term AT is characterized

by:
AT[dB] = α[dB/m] × d[m] (3.3)
where α is the attenuation factor in dB/m for the frequency range of 1 GHz
to 100 GHz, which includes the collective attenuation effects of dry air (in-
cluding oxygen), water vapor, rain, and haze [52]. d is the 3D T-R separation
distance as in (3.1).

Fig. 3.1 illustrates example propagation attenuation values due to dry air,
vapor, haze, and rain at mmWave frequencies from 1 GHz to 100 GHz, with
a barometric pressure of 1013.25 mbar, a relative humidity of 80%, a temper-
ature of 20◦ C, and a rain rate of 5 mm/hr, while the collective attenuation
effects of these four main natural absorbers are displayed in Fig. 3.2. These
results were obtained and reproduced from five reported controlled exper-
iments on atmospheric attenuation [52].

Note that the CI model inherently has an intrinsic frequency dependency

of path loss already embedded within the FSPL term. A useful property
of (3.1) is that 10n describes path loss in dB in terms of decades of distances
beginning at 1 m (making it very easy to compute power over distance in

28
 0.018

0.016

(dB/m) - Collective Attenuation

0.014

0.012

0.01

0.008

0.006

0.004

0.002

0
10 20 30 40 50 60 70 80 90 100
f (GHz)

F IGURE 3.2: Collective attenuation effects of dry air, vapor, haze, and
rain at mmWave frequencies, with a barometric pressure of 1013.25
mbar, a relative humidity of 80%, a temperature of 20◦ C, and a rain
rate of 5 mm/hr [52].

one’s mind).

The CI path loss model is based on fundamental principles of wireless prop-

agation, dating back to Friis and Bullington, where the PLE parameter of-
fers insight into path loss based on the environment, having a PLE value of
2 in free space (as shown by Friis) and a value of 4 for the asymptotic two-
ray ground bounce propagation model (as shown by Bullington). Standard-
izing to a reference distance of 1 m makes comparisons of measurements
and models simple, and provides a standard definition for the PLE, while
enabling intuition and rapid computation of path loss [5].

Compared with the existing alpha-beta-gamma (ABG) path loss model used
in 3GPP/ITU channel models [5], [7], [64], the CI model has an identical
mathematical form, while offering intuitive appeal, better model parameter
stability, and better prediction performance over a vast range of microwave
and mmWave frequencies, distances, and scenarios, with fewer parame-
ters [5].

In the current version of NYUSIM, the PLE and shadow fading standard
deviations for UMi, UMa, and RMa scenarios are displayed in Table 3.1.
The CI path loss model is employed for UMi and UMa scenarios, while for
the RMa scenario, the CIH model (CI model with a height-dependent PLE)
is adopted, as given by Eqs. (21) and (22) in [12]. The RMa PLE values in
Table 3.1 are for a base station height of 35 m. For RMa, the base station
height provided by the user on the GUI is used for the CIH path loss
model, while other scenarios do not use base station height at all. Note
that although the channel parameters for the UMi scenario are also used for

29
the UMa and RMa (except for the path loss model parameters) scenarios in
this simulator, users can adjust the UMa and RMa channel parameters in
the source code according to their own needs.

TABLE 3.1: Omnidirectional PLEs and shadow fading standard de-

viations for different scenarios in NYUSIM [5]. The RMa PLEs are for
a base station height of 35 m [12].

Scenario PLE Shadow Fading Standard Deviation (dB)

LOS 2 4.0
UMi
NLOS 3.2 7.0

LOS 2 4.0
UMa
NLOS 2.9 7.0

LOS 2.31 1.7

RMa
NLOS 3.07 6.7

3.2.1 Range Extension

Extensive mmWave propagation measurements conducted by NYU WIRE-
LESS have shown that in dense urban environments, mmWave signals can
cover around 200 m cell radius even under NLOS conditions [1], [2], [63],
and is likely to reach 500 m in lightly populated urban and suburban areas.
Furthermore, recent RMa propagation measurements at 73 GHz demon-
strated over 10 km coverage range in clear weather [12]. To make NYUSIM
cater for more users and wider applications, the maximum allowable T-R
separation distance is extended from 500 m to 1 km in NYUSIM v1.6, by re-
moving all the lower bounds on received power (including cluster power,
subpath power, and lobe power), or equivalently, upper bounds on path
loss, assuming there exits such a virtual receiver that can detect very low
received power.

For LOS environments, calculated path loss beyond 500 m is still accurate
using the NYUSIM path loss models (e.g., CI and CIH [5], [12]) since they
are applicable to over 10 km distances. For NLOS environments, however,
the NLOS path loss models employed in NYUSIM may not be accurate for
distances larger than 500 m since they were developed for ranges within
500 m [5], thus caution should be given when setting the distance beyond
500 m for UMi or UMa NLOS scenarios. The dynamic range for multipath
components in the extended range is extended to 220 dB from the default
value of 190 dB used for distances no larger than 500 m.

3.3 Wideband Temporal/Spatial Clustering Algorithm

The SSCM given here utilizes time clusters (TC) and spatial lobes (SL) to
model the omnidirectional CIR and corresponding joint AOD/AOA power
spectra, which have been used successfully in modeling mmWave chan-
nels [3]. Time clusters are composed of multipath components traveling
close in time, and that arrive from potentially different angular directions in
a short propagation time window. Spatial lobes represent main directions of

30
arrival (or departure) where energy arrives over several hundred nanosec-
onds. This SSCM structure is motivated by field measurements, which have
shown that multiple paths within a time cluster can arrive at unique point-
ing angles, detectable due to high gain directional antennas, and this feature
has not been modeled in current 3GPP and WINNER models. These def-
initions de-couple the time and space dimensions by extracting temporal
and spatial statistics separately. The definition of time cluster here consid-
ers multipath components traveling close in time, but that can arrive from
many angular directions, whereas current 3GPP and WINNER models as-
sume that subpaths belonging to a cluster travel along the same propaga-
tion path, but arrive at the same time delay over a certain AOA angular
spread.

The TCSL approach implements a physically-based clustering scheme (e.g.,

the use of a fixed inter-cluster void interval representing the minimum
propagation time between likely reflection or scattering objects) derived
from field observations, and can be used to extract time cluster and spa-
tial lobe statistics for any ray-tracing or measurement data sets [3]. The
time-partitioning methodology delineates the beginning and end times of
each time cluster, using a 25 ns minimum inter-cluster void interval. Se-
quentially arriving multipath components that occur within 25 ns of each
other are assumed to belong to one time cluster. The 3D spatial distribution
of received power was reconstructed from the 28 GHz and 73 GHz LOS and
NLOS directional received powers by linearly interpolating adjacent power
level segments in azimuth and elevation with a 1◦ resolution, and extract-
ing 3D spatial angular statistics. A -10 dB threshold below maximum peak
power in the 3D power spectrum is used in both LOS and NLOS environ-
ments, where all power segments below this threshold are disregarded for
further processing. Fig. 3.3 illustrates an omnidirectional PDP, where there
are four multipath taps which are grouped into two time clusters with expo-
nentially decaying amplitudes. The four multipath taps are then grouped
into two AOA spatial lobes, as shown in Fig. 3.4.

31
 F IGURE 3.3: Example of an omnidirectional PDP with four multipath
taps [3].

F IGURE 3.4: Example of an AOA power spectrum with four multi-

path taps [3].

It is noteworthy that for the RMa scenario, the number of time clusters and
the number of spatial lobes are both set to one, and the maximum number
of multipath components is set to two, based on our RMa mmWave field
measurements [12]. The base station height is only used for RMa in the
CIH path loss model but not other scenarios.

32
 TABLE 3.2: Measured PLE and SF for 28 and 140 GHz in LOS and
NLOS scenarios.

PLE, SF [dB] 28 GHz 140 GHz

LOS 1.2, 1.8 1.8, 2.9
NLOS 2.7, 9.7 2.7, 6.6

3.4 Indoor statistical channel model

3.4.1 Large-scale Path Loss Model
Path loss models describe the distance-power law that the received power
decreases exponentially with distance and are commonly used in the pre-
diction of signal strength and cell range. A popular path loss model, close-
in free space reference distance (CI) path loss model with 1 m reference
distance, is given by 3.1.
PLE n indicates that the power decays by 10n dB per decade of dis-
tance beyond 1 m [65]. Fig. 3.5 shows that PLE for 28 GHz LOS and NLOS
scenarios via minimum mean square error (MMSE) fitting are 1.2 and 2.8,
respectively. 1.2 is derived from 2 MetroTech dataset for 28 GHz LOS case,
which is lower than 1.3-1.9 found in the literature [66]–[68]. To verify this
low PLE, we conducted LOS measurements in another office building, 370
Jay, at 28 GHz. The resulting PLE for 370 Jay dataset is also 1.2, suggesting
that the power attenuates much slower (12 dB per ten meters) than values
reported in the literature, which might be attributed to the strong waveg-
uide effect of long and narrow corridors in the indoor environment. The
PLE and the standard deviation of shadow fading (SF) measured at 28 and
140 GHz in LOS and NLOS scenarios are given in Table 3.2.

F IGURE 3.5: 28 GHz indoor omnidirectional path loss scatter plot

and MMSE-fitted CI path loss model with distance for LOS and
NLOS scenarios [17].

33
3.4.2 Small-scale Channel Impulse Response Model
Input parameters of channel generation procedure
The outdoor and indoor scenarios share the same set of input parameters
for channel coefficient generation procedure, but the probabilistic distri-
bution of several parameters such as the number of time clusters, cluster
subpaths, and spatial lobes are different. The number of time clusters and
cluster subpaths for NLOS environment is no longer uniformly distributed.
Instead, the number of time clusters for the indoor environments is Poisson
distributed, and the number of cluster subpaths is modeled as a composite
distribution of a discrete exponential (DE) distribution and a δ-function, as
given in Table 3.3.

Parameter frequency-dependency for the indoor scenario

For the indoor scenario, two sets of channel measurements (28 GHz and 140
GHz) are available for developing the NYUSIM channel model across the
mmWave and sub-Terahertz bands. Distinct channel characteristics mea-
sured at 28 and 140 GHz (e.g., the number of time clusters and cluster sub-
paths) motivated us to make generated channels frequency-variant by us-
ing frequency-dependent input parameters listed in Table 3.3. Since only
two center frequencies (i.e., 28 GHz and 140 GHz) were measured, a linear
interpolation is used to generate channel parameters between 28 and 140
GHz. Note that the allowable user-specified carrier frequency of NYUSIM
ranges from 0.5 GHz to 150 GHz for the indoor scenario. Thus, the input
parameters for frequencies below 28 GHz are set to be equal to the values
at 28 GHz; the input parameters for frequencies above 140 GHz are set to
be equal to the values at 140 GHz. The relation is given by

 p(28) , f 6 28
p(140)−p(28)
p(f ) = f+ 5p(28)−p(140) , 28 < f < 140 (3.4)
 140−28 4
p(140) , f > 140

where p(f ) denotes an input parameter at frequency f (GHz).

A special case is the PLE for the indoor LOS scenario due to the low
value of the measured PLE. The measured PLE in the 2 MetroTech building
for 28 GHz LOS case is 1.2 [17] while the reported PLEs in the literature
for the indoor LOS scenario at 28 and 60 GHz are 1.7 [66] and 1.88 [67],
respectively. To verify the measured PLE in the 2 MetroTech building, we
conducted another set of indoor LOS measurements in the 370 Jay office
building, which showed that the resulting PLE is also 1.2 [17]. Such low
PLE is attributed to strong waveguide effect in the corridor.
We also found the measurement results at 1.5 GHz from [69], which
showed that the PLE for the indoor scenario was about 1.8. Thus, it is
not sensible to apply a single-slope function for frequencies from 0.5 to 150
GHz; however, a dual-slope function is used specifically for the indoor LOS
scenario. Without loss of generality, we assume another frequency reference
point at 1 GHz with PLE 1.8. Similar to the interpolation applied above, an-
other linear interpolation between 1 and 28 GHz is adopted to make the
simulated PLE more realistic. Fig. 3.6 shows the frequency-dependent PLE
values for the indoor LOS scenario.

34
F IGURE 3.6: Frequency-dependent PLE for the indoor LOS scenario.

35
 TABLE 3.3: I NPUT PARAMETERS FOR CHANNEL COEFFICIENT GENERATION PROCEDURE [18]

Channel Parameters 28 - 73 GHz UMi [25] 28 - 140 GHz InH [18]

# Time clusters N N ∼ DU(1, Nc ) N ∼ Poisson(λc )
# Cluster subpaths Mn Mn ∼ DU(1, Ms ) Mn ∼ (1 − β)δ(Mn ) + DE(µs )
τn00 ∼ Exp(µτ ) or Logn(µτ , στ )
∆τn = sort(τn00 ) − min(τn00 )
Cluster delay τn (ns) (
0 ,n = 1
τn =
τn−1 + ρMn−1 ,n−1 + ∆τn + MTI , n = 2, ...N
 1+Xn
1
ρm,n = × (m − 1) ,
Intra-cluster delay ρm,n (ns) Bbb ρm,n ∼ Exp(µρ )
m = 1, 2, ..., Mn , n = 1, 2, ..., N
τn Zn
Pn0 = P̄0 e− Γ 10 10 ,
P0
Cluster power Pn (mW) Pn = PN n × Pr [mW ],
0
k=1 Pk
Zn ∼ N (0, σZ ), n = 1, 2, ..., N
ρm,n Um,n
−
Π0m,n = Π̄0 e γ 10 10 ,
Π0m,n
Subpath power Πm,n (mW) Πm,n = PMn × Pn [mW ],
0
k=1 Πk,n
Um,n ∼ N (0, σU ), m = 1, 2, ..., Mn
SP phase ϕ (rad) Uniform(0, 2π)
LAOD ∼ min{Lmax , max{1, Poisson(µAOD )}} LAOD ∼ DU(1, LAOD,max )
# Spatial lobes L
LAOA ∼ min{Lmax , max{1, Poisson(µAOA )}} LAOA ∼ DU(1, LAOA,max )
φi ∼ U (φmin , φmax ), φmin = 360(i−1) , φmax = 360i
L , i = 1, 2, ..., L
SL mean angle φi , θi (o ) L
θi ∼ N (µl , σl )
i ∼ DU[1, LAOD ], j ∼ DU[1, LAOA ]
(∆φi )m,n,AOD ∼ N (0, σφ,AOD )
SP angle offset ∆φi , ∆θi
(∆θi )m,n,ZOD ∼ N (0, σθ,ZOD )
w.r.t φi , θi (o )
(∆φj )m,n,AOA ∼ N (0, σφ,AOA )
(∆θj )m,n,ZOA ∼ N (0, σθ,ZOA )
3.5 Channel Bandwidth and Resolvable MPC
The temporal resolution of wireless channels depend on the channel band-
width. More specifically, the time resolution is the inverse of one half of the
RF bandwidth. For example, 800 MHz channel bandwidth corresponds
to 2.5 ns time resolution, which means two MPCs within 2.5 ns cannot
be resolved by the receiver. Two extreme cases of bandwidth are single-
frequency continuous wave and ideally infinite bandwidth, which respec-
tively correspond to infinite and zero time resolution. Thus, fewer MPCs
can be resolved at narrower bandwidth. When MPCs arrive within a time
bin, these MPCs are vectorially summed. The functions, "getNewPower-
Spectrum.m" and "getLocalCIR.m", are used to generate bandwidth-adjusted
power spectrum for SISO and MIMO channels, respectively. Note that in
the angular domain, the angular information of the first arriving MPC in a
set of MPCs, which falls in one time bin (cannot be resolved) will be used
as the angles of the combined MPC.

3.6 Spatial Consistency Procedure

In NYUSIM 3.0, spatial consistency is only implemented for outdoor sce-
narios. Spatial consistency indicates continuous and realistic channel evo-
lution along the UT trajectory in a local area. To realize spatial consis-
tency, spatially correlated large-scale parameters such as shadow fading,
LOS/NLOS condition are generated, and time-variant small-scale param-
eters such as angles, power, delay, phase of each MPC are generated [14],
[15]. The correlation distance of LSPs precisely characterizes the concept of
“a local area”, which also confines the length of a channel segment (where
a channel segment is 10-15 m long) [15]. In a channel segment, the channels
are considered highly correlated and updated using spatial consistency pro-
cedure. A channel segment can be divided into several channel snapshots.
The distance between two channel snapshots is the update distance (e.g. 1
m), which means the channel coefficients are updated for every 1 m incre-
ment along a traveled path.

Spatially Correlated Map of Shadow Fading: LSPs defined in the 3GPP

38.901 model are delay spread, angular spread, Rician K factor, SF, and
LOS/NLOS condition [70]. The two LSPs explicitly used in the channel
generation procedure in NYUSIM is SF and LOS/NLOS condition. Spa-
tially correlated SF and LOS/NLOS condition values are generated in the
same way. For SF, a 2-dimensional (2-D) grid map is generated to contain
values of spatially correlated SF in a simulated area. The granularity of the
map is set to be 1 m, which means the distance between two neighboring
grid points is 1 m. SF is modeled as a log-normal random variable with
zero mean and σ dB standard deviation as shown in (3.1). The map of SF is
initialized by assigning an i.i.d normal distributed random variable at each
grid. A 2-D exponential filter is applied to the map, which is given by [14]:
p
p2 + q 2
h(p, q) = exp(− ), (3.5)
dco

37
 p
where p and q are coordinates with respect to the center of the filter. (p2 + q 2 )
represents the distance to the center of the filter. dco is the correlation dis-
tance of SF. L is the length of the filter, which is set as 8-fold dco since the
correlation is negligible beyond 4-fold dco . Applying this 2-D filtering, the
correlated values in the map is calculated by:
XX
Mc (i, j) = h(p, q)M (i − p + 1, j − q + 1) (3.6)
p q

where Mc is the correlated map and M is the initialized independent map.

i and j are the coordinates of grid points in the map. A map of spatially
correlated shadow fading over a 200 m x 200 m area is shown in Fig. 2.8.
The correlation distance of shadow fading in a UMi LOS scenario is set to
be 10 m [15], [45]. The UT moved 40 m in a partial hexagon track, which is
also illustrated in Fig. 2.8. Shadow fading varies from -10 dB to 10 dB in a
continuous manner. Similar shadow fading values are observed at closely
spaced locations whereas independent values for close locations are always
used in the drop-based model.

Spatially Correlated Map of LOS/NLOS Condition: LOS/NLOS condi-

tion is important to evaluate the wireless communication systems, espe-
cially at mmWave frequencies. LOS/NLOS condition will determine path
loss exponent (PLE) and variance of shadow fading, and the transition be-
tween LOS and NLOS condition will cause a drastic change in the received
power when a user moves. LOS propagation can provide more reliable per-
formance in mmWave communications as compared to NLOS propagation.
Generally, LOS/NLOS condition does not change when a user moves in a
local area (within the correlation distance of LOS/NLOS condition).
By generating a map of spatially correlated LOS/NLOS condition, a UT
will experience the same visibility condition in a local area. In each chan-
nel segment, the visibility condition is kept constant, which indicates the
large-scale parameters are spatially consistent. When a UT moves from one
channel segment to the next channel segment, and the visibility condition
is changed according to the UT location in the map, different values of pa-
rameters for LOS or NLOS are used to initialize the channel coefficients for
the new channel segment. In other words, LOS/NLOS transition can be
realized by the spatially correlated map.
Similar with the spatially correlated shadow fading, a 2-D grid map is
also generated for LOS/NLOS condition. The same 2-D double-side expo-
nential function is used. The length of filter coefficient is 8-fold correlation
distance of LOS/NLOS condition (e.g. 50 m in UMi and UMa scenario [70])
divided by the granularity (e.g. 1 m), since the correlation coefficient of
the filter is close to zero beyond fourfold correlation distance. A 2-D geo-
graphical grid map is generated, the size of which depends on the distance
between BS and UT and the UT trajectory. The granularity of the map is
determined such as 1 m. An i.i.d standard Gaussian random variable is as-
sociated to each grid point. Then, the 2-D grid map is filtered by the 2-D
exponential function and have spatially correlated Gaussian random vari-
ables. A transformation from Gaussian distribution to uniform distribu-
tion is required to generate spatially correlated uniform random variables,

38
which is given by
ũ = g −1 (ṽ) = Fũ−1 (Fṽ (ṽ)), (3.7)
where ũ and ṽ are the spatially correlated uniform and Gaussian random
variable, respectively. Fũ and Fṽ are the CDF of the uniform distribu-
tion and Gaussian distribution, respectively. Using cumulative distribution
function (CDF) approach, the relation is given by
!
1 ṽ
ũ = 1 + erf( p ) , (3.8)
2 (2)

where erf(x) is the error function and given by

x
t2
Z
1
p exp(− )dt, (3.9)
−∞ (2π) 2

The method of deciding the LOS/NLOS condition at a certain location

is to compare a uniformly distributed random variable to LOS probability
at that location. There are several LOS probability models available in the
literature such as 3GPP 38.901 model [70], 5GCM model [51], and METIS
model [71]. In NYUSIM, NYU squared model for LOS probability is applied
for UMi and UMa scenario [72], and 3GPP 38.901 model for LOS probability
is used for RMa scenario [70]. The LOS probability model used in NYUSIM
for three scenarios is given in Table 3.4.

TABLE 3.4: LOS probability models in the UMi, UMa and RMa sce-
nario [72]

Scenarios LOS probability models (distances are in meters) Parameters

d1 = 22 m
UMi [72] PrLOS (d) = (min(d1 /d, 1)(1 − exp(−d/d2 )) + exp(−d/d2 ))2
d2 = 100 m
PrLOS (d) = ((min(d1 /d, 1)(1 − exp(−d/d2 )) + exp(−d/d2 )) (1 + C(d, hU T )))2
where, 
0, hU E < 13 m
C(d, hU T ) =  hU T −13 1.5 d1 = 20 m
UMa [72] 
10
g(d), 13 m ≤ hU T ≤ 23 m
d2 = 160 m
and, (
0, d ≤ 18 m
g(d) =
(1.25 × 10−6 ) · d3 exp(−d/150), d > 18 m
(
1, d < 18 m
RMa [70] PrLOS (d) = d−10
exp(− 1000 ), d > 10 m
Note : d is the 2D Euclidean distance between BS and UT.

Thus, a UT at a certain location under LOS condition or NLOS condi-

tion is determined by comparing correlated value ũ to the LOS probability
P rLOS (d) (
LOS ũ <= PrLOS (d)
Condition = (3.10)
NLOS ũ > PrLOS (d)
A map of spatially correlated LOS/NLOS condition in UMi scenario
using LOS probability equations in Table 3.4 shown in Fig. 3.7. The “white”
represents LOS whereas the “black” represents NLOS. The size of a UMi
cell for mmWave communication systems is typically 200 m [72]. Thus, the
simulated area is set to be 400 m x 400 m. The correlation distance is 15 m

39
[73]. The granularity of the map is 1 m. The heights of the BS and UT are 10
m and 1.5 m, respectively.

F IGURE 3.7: A 400 m x 400 m spatially correlated map for

LOS/NLOS condition in UMi scenario using a 2-D filter. The cor-
relation distance is 15 m. The granularity of the map is 1 m. The
heights of the BS and UT are 10 m and 1.5 m, respectively. The LOS
and NLOS condition transition may happen along the UT trajectory
according to the UT location in the map.

Small-scale Parameters: Small-scale parameters in the NYUSIM channel

model are angles (i.e. AOD, ZOD, AOA, ZOA), delay, power, phase for each
MPC. A geometry-based approach using multiple reflection surfaces origi-
nated from [74] is derived to update the angles for each NLOS MPC. First,
the update equations for LOS component is derived based on geometry.
For NLOS components, the mirror images of a UT and the UT velocity are
calculated by using multiple reflection surfaces. Then, the NLOS compo-
nents can be updated using the same equations as the LOS component with
mirrored UT location and velocity. The delay and phase of each MPC is up-
dated simply based on the path length change using Law of cosines [16],
[45]. The power of each MPC is updated by redistributing cluster powers
and MPC powers in each cluster following the same way that these powers
were initialized using the cluster time excess delay and intra-cluster time
excess delay, explained in [25].

LOS case: To derive the update equations of four angles (AOD, ZOD,
AOA, and ZOA) of UT at time t along the UT trajectory, we assume that the
UT only has azimuthal movement for simplicity, which means θv = 90◦ .
The UT location at time t0 is [75]
 
d2D cos(φAOD (t0 ))
 d2D sin(φAOD (t0 ))  , (3.11)
hU E

40
where d2D is the projection of the T-R separation distance r in the azimuth
plane. The UT location at time t is [75]
 
d2D cos(φAOD (t0 )) + vt cos(φv )
 d2D sin(φAOD (t0 )) + vt sin(φv )  (3.12)
hU E

Then, the accurate AOD at time t can be given by [75]

d2D sin(φAOD (t0 )) + vt sin(φv )

φAOD (t) = arctan( ). (3.13)
d2D cos(φAOD (t0 )) + vt cos(φv )

The problem for this accurate AOD update shown above is computa-
tionally expensive. Considering that the moving distance (less than a me-
ter) is much smaller than the separation distance between TX and RX (r),
the change of r can be ignored. Then, the moving distance (or trajectory)
of the UT can be approximated as the corresponding arc length with the
BS as the center as shown in Fig. 3.8. The change of AOD, ∆φAOD can
be calculated by the central angle of the arc. Thus, it is assumed that the
track distance in each channel snapshot is set to be within a meter, which,
usually, is much smaller than the distance (r) between the BS and UT in out-
door environments such as UMi and UMa. The angle deviation or change
is expected to be small, and user track within 1 m can be regarded as a short
arc centering at BS with radius r. Thus, a linear approximation method is
valid to update time-variant angles, which is given by [70], [74]

φangle (tk ) = φangle (tk−1 ) + Sangle · ∆t, (3.14)

where “angle” is an index for AOD, ZOD, AOA, and ZOA. tk and tk−1

𝐵𝑆 𝑥
Δ𝜙𝐴𝑂𝐷

d2𝐷 (𝑡0 )

d2𝐷 (𝑡) 𝑈𝑇(𝑡0 )

𝒗∥
𝒗⊥
𝒗
𝑈𝑇(𝑡)

F IGURE 3.8: UT movement and change of AOD in the azimuth plane.

The moving distance of the UT is approximated as the corresponding
arc length with the BS as the center

are two successive time instances, and the interval between two successive
time instances is ∆t. Sangle is the linear changing rate for “angle”.

41
 F IGURE 3.9: 3D geometry illustration of the coordination system for
a LOS component. AOD and ZOD from the BS and AOA and ZOA
from the UT are shown.

To obtain expressions for Sangle , the definitions of four angles (AOD,

ZOD, AOA, ZOA) and the GCS for LOS component is shown in 3.9. rRX is
the spherical unit vector with AOD (φAOD ) and ZOD (θZOD ), which is given
by [76]  
sin θZOD (t) cos φAOD (t)
rRX (t) = −r(t) sin θZOD (t) cos φAOD (t) , (3.15)
cos θZOD (t)
where r(t) is the T-R separation distance at time t. The RX velocity vector
is equal to the derivative of rRX (t), which is given by [76]
 
r(t) sin(θZOD (t)) cos(φAOD (t))
  d  r(t) sin(θZOD (t)) sin(φAOD (t)) 
vx r(t) cos(θZOD (t))
drRX (t)
v =  vy  = = , (3.16)
dt dt
vz

    
vx sin(θZOD ) cos(φAOD ) r cos(θZOD ) cos(φAOD ) −r sin(θZOD ) sin(φAOD ) ṙ
vy  =  sin(θZOD ) sin(φAOD ) r cos(θZOD ) sin(φAOD ) r sin(θZOD ) cos(φAOD ) · θ̇ZOD 
vz cos(θZOD ) −r sin(θZOD ) 0 φ̇AOD
(3.17)
By calculating the inverse of the above matrix [76],
    
ṙ cos(φAOD ) sin(θZOD ) sin(φAOD ) sin(θZOD ) cos(θZOD ) vx
 θ̇ZOD  =  cos(φAOD ) cos(θZOD )/r sin(φAOD ) cos(θZOD )/r − sin(θZOD )/r · vy  ,
 
φ̇AOD − sin(φAOD )/(r sin(θZOD )) cos(φAOD )/(r sin(θZOD )) 0 vz
(3.18)

42
where θ̇ZOD and φ̇AOD are SZOD and SAOD . Linear changing rates for four
angles for a LOS component are given by [76]

vy cos(φAOD ) − vx sin(φAOD )
SAOD =
r sin(θZOD )
vx cos(φAOD ) cos(θZOD ) + vy cos(θZOD ) sin(φAOD ) − vz sin(θZOD )
SZOD =
r
vy cos(φAOA ) − vx sin(φAOA )
SAOA =
r sin(θZOA )
vx cos(φAOA ) cos(θZOA ) + vy cos(θZOA ) sin(φAOA ) − vz sin(θZOA )
SZOA =
r
(3.19)

F IGURE 3.10: UT movement and change of AOD in the azimuth

plane in NLOS scenarios. The mirror images of the UT location and
velocity direction φv are shown as UT’ and φvr

NLOS case: For NLOS components, there is no direct path between

TX and RX. MPCs are reflected, diffracted, or scattered at least once. A
multiple reflection surfaces method is first proposed here to update angles
of NLOS components, which is motivated by a simple case from [74], where
the MPC is only reflected once before it reaches the UT. By obtaining the re-
lation between the actual UT location and the mirror image of the actual
UT location, the mirror image of the UT velocity vector can be calculated
using the image theory. Then, the NLOS component is transformed to a
virtual LOS component, and the angles of NLOS components can be up-
dated using the update equations for the LOS component. More details are
explained below.
An illustration of the single reflection of a NLOS MPC is shown in Fig.
3.10. The reflection surface is assumed to be perpendicular to the azimuth
plane and has a reflection surface angle, φRS [74]. φRS is within [0,π]. The
mirror image of the UT location and the UT velocity vector with respect
to the reflection surface is also shown in Fig. 3.10. From the view of BS,
UT is equivalent to moving at the velocity of vr at the location U T 0 . Note
that U T 0 is in the LOS condition from the BS. Thus, angles can be updated
using equations for the LOS component with assuming that U T 0 moves at

43
the velocity of vr . Then, φvr and φAOA are given by

φAOA = 2φRS + π − φAOD

(3.20)
φvr = 2φRS − φv

F IGURE 3.11: UT movement and change of AOD in the azimuth

plane in NLOS scenarios.

Then, we consider that a NLOS MPC is reflected twice before it reaches

the UT, which is illustrated in Fig. 3.11, φvr and φAOA are given by

φAOA = 2φRS2 − 2φRS + φAOD = ∆RS + φAOD

(3.21)
φvr = 2φRS2 − 2φRS + φv = ∆RS + φv

where φRS1 and φRS2 are the angles of the first and second reflection sur-
faces with respect to the x axis. Further, a general relation with M reflection
surfaces between φAOA and φAOD , and between φvr and φv can be given by
M
X
M
φAOA = (−1) φAOD + 2 (−1)i φRSi + M π = ∆RS + (−1)M φAOD + M π
i=1
M
X
φvr = (−1)M φv + 2 (−1)i φRSi = ∆RS + (−1)M φv
i=1
(3.22)

where φRSi is the angle of the ith reflection surface with respect to the x
axis, where the reflection surfaces are assumed to be perpendicular to the
x-y plane (ground). It can be seen from (3.22) that there are possibilities of
the relation between φAOD and φAOA . Same is the relation between φr and
φvr , which depends on the number of reflection surfaces a NLOS MPC goes
through. If it is an odd number of reflection surfaces, (3.20) is applied; if it
is an even number of reflection surfaces, (3.21) is applied. Since φAOD and
φAOA are known, φvr can be calculated based on ∆RS and φv . Using φvr
instead of φv , the changing relation for four angles for NLOS component

44
are given by

SAOD = B · SAOA
(3.23)
SZOD = −SZOA

where B can be -1 (even reflections)or 1 (odd reflections) with 50 % and

50 % probability, which indicates that the probabilities that a NLOS MPC
experiences an odd or an even number of MPCs are assumed to be equal.
Finally, the update equations for AOD, ZOD, AOA, and ZOA are summa-
rized as
φangle (tk ) = φangle (tk−1 ) + Sangle (tk − tk−1 ), (3.24)
where Sangle for a LOS component are given by

vy cos(φAOD ) − vx sin(φAOD )
SAOD =
r sin(θZOD )
vx cos(φAOD ) cos(θZOD ) + vy cos(θZOD ) sin(φAOD ) − vz sin(θZOD )
SZOD =
r
vy cos(φAOA ) − vx sin(φAOA )
SAOA =
r sin(θZOA )
vx cos(φAOA ) cos(θZOA ) + vy cos(θZOA ) sin(φAOA ) − vz sin(θZOA )
SZOA = ,
r
(3.25)

and Sangle for NLOS components are given by

vr,y cos(φAOD ) − vr,x sin(φAOD )

SAOD =
l sin(θZOD )
vr,x cos(φAOD ) cos(θZOD ) + vr,y cos(θZOD ) sin(φAOD ) − vr,z sin(θZOD )
SZOD =
l
vr,y cos(φAOA ) − vr,x sin(φAOA )
SAOA =B
l sin(θZOA )
vr,x cos(φAOA ) cos(θZOA ) + vr,y cos(θZOA ) sin(φAOA ) − vv,z sin(θZOA )
SZOA = ,
l
(3.26)

Note that r is the T-R separation distance while l is the actual path length
for a certain NLOS component, as shown in Fig. 3.10 and Fig. 3.11. It is
impossible to acquire the actual path length using geometry without deter-
ministic reflection surface information. Thus, the path length is calculated
by multiplying the absolute time delay generated from the NYUSIM chan-
nel model with the speed of light.

Delay, Power, and Phase Update: The update of the delay of each
MPC uses reflection surfaces based on geometry. Basically, the change of
the delay is the change of the path length divided by the speed of light, as
shown in Fig. 3.12. Using law of cosines, the update of the delay can be
given by

∆l
τ (t) = τ (t0 ) +
p c (3.27)
τ (t) = (v∆t) + (cτ (t0 ))2 − 2v(∆t) · cτ (t0 ) · cos(φv − φAOA )
2

45
 F IGURE 3.12: Update of delay and phase based on the change of path
length in time interval ∆t. τ (t0 ), φv , and φAOA are known, τ (t1 ) can
be calculated using the law of cosine.

where ∆l is the change of the path length, and ∆t is the update time. τ (t) is
the absolute time delay at time t. φv is the UT moving direction in azimuth
plane, and φAOA is the AOA of a MPC received at the UT.
Similarly, the update of the phase of each MPC is also based on the
change of the path length. The update of the phase can be given by

2π∆l
ϕ(t) = ϕ(t0 ) + ∆l · k = ϕ(t0 ) + (3.28)
λ
where ∆l is the change of the path length given in (3.27). k is the wavenum-
ber depending on the wavelength, k = 2π/λ. Note that the initial phase
generation of MPCs is assumed to be i.i.d, and uniform between 0 and 2π
since each MPC experiences a different scattering environment, thus arriv-
ing at arbitrary AOA spatial lobe. However, the update of the phase should
add the phase change to the initialized phase, but is not generated ran-
domly again.
The update of the power of each MPC is based on the updated delay
information. Using the same generation procedure for cluster powers and
cluster subpath powers in [25], the powers of MPCs are re-distributed based
on the updated delay. Generation of cluster powers Pn (mW) is given by
[25]
τn Zn
Pn0 = P̄0 e− Γ 10 10
P0
Pn = PN n × Pr [mW ] (3.29)
0
i=1 Pi
Zn ∼ N (0, σZ )

where P̄0 is the average power in the first arriving MPC, Γ is the cluster de-
cay time constant, and Zn is a lognormal random variable with 0 dB mean
and STD σZ . N is the number of time clusters. The actual cluster powers
Pn are calculated by ensuring the sum of Pn to be equal to the total received
power Pr . Generation of subpath powers Πm,n in each cluster is given by

46
[25]
ρm,n Um,n
−
Π0m,n = Π̄0 e γ 10 10

Π0m,n
Πm,n = PMn × Pn [mW ] (3.30)
Π 0
j=1 j,n
Um,n ∼ N (0, σU )

where Π̄0 is the average power in the first received intra-cluster subpath, γ
is the subpath delay time constant, and Um,n is a lognormal random vari-
able with 0 dB mean and STD σU . Mn is the number of subpaths in the nth
cluster.
In a word, the cluster powers are modeled as an exponential function
over time excess delay. The cluster subpath powers in each cluster are
also modeled as an exponential function over intra-cluster time excess de-
lay [25]. By re-distributing cluster powers and subpath powers, the channel
model with spatial consistency can model the power variation along the UT
movement. Considering the cluster powers and subpath powers still fol-
low exponential decay, the power of each MPC will not change drastically
within the update distance (e.g. 10 m), which matches the observations in
[15].
The update procedure for SSPs of each MPC which are angles, delay,
power, phase is described above. A geometry-based approach using reflec-
tion surfaces is applied to update time-variant angles (AOA, AOD, ZOA,
ZOD). The delay and phase is updated based on the change of path length,
which is calculated using reflection surfaces. Cluster powers are re-distributed
using the total received power with respect to the time excess delay of clus-
ters, and cluster subpath powers are re-distributed using cluster powers
with respect to the intra-cluster time excess delay of subpaths.
LSPs and SSPs have been continuously updated in each channel seg-
ment, while the initial channel coefficients of channel segments are inde-
pendently generated. Therefore, a smooth transition procedure is applied
to “connect” channel segment by cluster birth and death. The power of one
old cluster ramps down and one new cluster ramps up [77]. Since the num-
ber of time clusters in two channel segments may be different, the power of
one cluster can ramp up or down individually. Note that cluster birth and
death only happen to one time cluster in a channel snapshot.

3.7 Outdoor to Indoor Penetration Loss Model

A simple parabolic model was first proposed in [50], which fits the mea-
sured building penetration loss with standard glass and IRR glass. The
parabolic model is given by [51]

BPL[dB] = 10log10 (A + B · fc2 ) + N (0, σP2 ) (3.31)

where fc is the carrier frequency. A = 5, B = 0.03, and σP = 4.0 for low-loss

buildings (with standard glass). A = 10, B =5, and σP = 6.0 for high-loss
buildings(with IRR glass). Most O2I penetration loss measurements set the
incident angle as 0 degree (perpendicular to the building wall). However,

47
the losses have been observed to increase by up to 15 to 20 dB as the inci-
dence angle becomes more larger (parallel to the building wall).

3.8 Human Blockage Shadowing Loss Model

Measurement-based two-state and four-state Markov models for human
blockage shadowing loss were proposed in [78]. The human blockage mea-
surements were conducted by precious researchers at NYU WIRELESS in
an open square in downtown Brooklyn. Measurements captured the effects
of large moving crowds in an urban setting for a peer-to-peer communica-
tion links at mmWave (73 GHz) [78]. Three types of horn antennas were
used to study the effects of antenna beamwidth on the blockage event. The
antenna gain and azimuth and elevation half power beamwidth (HPBW)
are given in Table. 3.5. TX and RX antennas were boresight-aligned and re-
mained fixed during the measurements. In an observation window of 135
s, people crossed the LOS link, and 40,800 PDP snapshots were recorded
with a 3.3 ms time separation between two consecutive PDPs. Details of
the measurement setup and hardware specifications are provided in [78]. A
135 s long time-variant received power trace was obtained with 135 s long,
which was then used to determine the probability of blockage occurrences
and the statistics of the shadowing loss of each blockage event. Note that
aforementioned human blockage measurements are on peer-to-peer level,
and are not representative of a BS-UT setting. However, T-R separation dis-
tances were usually taken to be greater than 40 m, BSs were on the horizon
and thus were at similar heights relative to the UT. Thus, it is reasonable to
extend blockage statistics obtained from peer-to-peer measurements to the
BS-UT setting. The knife edge diffraction (KED) model is another method
of modeling the blockage effect, which is not implemented in NYUSIM [79].

TABLE 3.5: A NTENNA S PECIFICATIONS IN THE URBAN ENVIRON -

MENT MEASUREMENTS [80]

Antenna Index Antenna 1 Antenna 2 Antenna 3

TX&RX Antenna Gain 27 dBi 20 dBi 9.1 dBi
TX&RX Az./El. HPBW 7°/7° 15°/15° 60°/60°

A simple two-state Markov model can be used to characterize shad-

owed and unshadowed states for a communication link in the presence of
pedestrian induced variations in the received signal strength [80]. A better
four-state piece wise linear approach can also be used to characterize block-
age with the following states: unshadowed, shadowed, a decaying period
from unshadowed to shadowed, and a rising period from shadowed to un-
shadowed, which is given by [78], [81]
 SEmean

 rdecay · t, for 0 ≤ t ≤ rdecay
SEmean
tD − SErrise

 SEmean , for rdecay ≤t≤ mean
SE(t) = (3.32)


SEmean − rrise · t, fortD − SErrise
mean
≤ t ≤ tD

0, otherwise

48
where SEmean is the mean attenuation (in dB) of a blockage event, which
is calculated over the interval [ 13 tD , 23 tD ]. tD is the blockage event time du-
ration in ms, which is the time interval between the between the last zero-
crossing before and the first zero-crossing after the rapid signal fade. rdecay
is the decay rate of the signal strength in dB/ms, which is the slope of the
line segment between the last zero-crossing before the rapid fading and the
first SEmean -crossing in the rapid fading. rrise is the rising rate of the sig-
nal strength in dB/ms, which is the slope of the line segment between the
first zero-crossing after the rapid fading and the last SEmean -crossing in the
rapid fading.
Base on the piece wise function in (3.32), a four-state Markov model
corresponding to four stages was proposed to characterize blockage events
[80], [82]. The corresponding transition matrix is shown in Fig. 3.13 [78]
and is given by
 
1 − pdecay pdecay 0 0
 0 1 − pshad pshad 0 
  (3.33)
 0 0 1 − prise prise 
punshad 0 0 1 − punshad

F IGURE 3.13: Four-state Markov model for unshadowed, decaying,

shadowed, and rising states for modeling blockage events [78].

Using the four-state Markov model, transition rates with three antenna
settings (7°, 15°, 60°) are calculated based on measurements and given in
Table V in [78]. While the work in [80] offers values for the Markov model
for three different antenna beamwidths, more measurements with differ-
ent beamwidths and different shadowing scenarios will be needed to more
accurately parameterize the four-state Markov model for various applica-
tions that may emerge in the future. For this reason, NYUSIM offers both a
default setting as a linear fit of different beamwidths based on [78], but also
allows the user to input different Markov model parameters for different
human shadowing models that may be developed from future measure-
ments. The simple linear fit of transition rates with respect to the RX an-
tenna azimuth HPBW in degrees is adopted in the NYUSIM channel model

49
and given by

λdecay = 0.2
λshad = 0.065 ∗ HPBW(◦ ) + 7.425
(3.34)
λrise = 0.05 ∗ HPBW(◦ ) + 7.35
λunshad = 6.7

λ is the transition rate from one state to the next state. The transition prob-
ability p = λ · t, and t is the time resolution, which is set to be 1 ms in
NYUSIM. Using this four-state Markov model given transition rates, a ar-
bitrarily long Markov trace (20 s used in NYUSIM) is generated with time
resolution 1 ms, which has 20,000 points in total. Each point on the trace
is assigned a state (unshadowed, shadowed, decay, or rise). The blockage
fade duration, decay time, and rising time are determined. The attenua-
tion of all the points in the shadowed state are the same (SEmean ). The
attenuation of the points in the decay state is decreasing from 0 to SEmean
linearly, and the attenuation of the points in the rising state is increasing
from SEmean to 0 in a linear manner.
Note that the mean attenuation of a blockage event (SEmean ) is still un-
known. SEmean is modeled as a log-normal random variable with the av-
erage mean attenuation µA and the variance σA . The average mean atten-
uation µA depends on the RX antenna azimuth HPBW in degrees, and a
simple analytical expression is given in [78]:
 
180
µA [dB] = 10log10 b + (3.35)
HPBW(◦ )

where b is a constant, and HPBW is the RX antenna azimuth HPBW in de-

grees, which can be from 1° to 180°. In NYUSIM, b = 9.8 and σA = 0.31 are
obtained by curve fitting the measured data with three antenna settings.
For different antenna settings, σA is always assumed to be 0.31, which can
be characterized by a better model based on more measurements in the fu-
ture. Admittedly, the measurements are limited and may not provide a
accurate relation between the transition rate and mean attenuation and an-
tenna beamwidth. More measurements with different antenna settings are
required to accurately parameterize the four-state Markov model.
NYUSIM can generate omnidirectional and directional channels, for ei-
ther a drop realization in the drop-based mode or a channel snapshot in
the spatial consistency mode. For omnidirectional channels generated from
NYUSIM, antenna HPBW is not applicable, the spatial lobe width, instead,
is used to calculate transition rates and mean attenuation. Spatial lobes
represent main directions of arrival (or departure) where energy may ar-
rive over several hundreds of nanoseconds. A spatial lobe is modeled as
a Gaussian distribution, with a variance equal to the lobe angular spread.
The largest value of lobe angular spread obtained from measurements was
11.5° in a LOS scenario [25]. The typical width of a spatial lobe is less than
11.5°×6 = 69°with 99.7% probability, which is close to the 60° HPBW an-
tenna used in the measurements [80]. Thus, it is valid to assume that all
MPCs in this spatial lobe will experience the same amount of shadowing
loss when a blockage event happens in the direction within the range of a
spatial lobe. Based on field measurements, the number of main directions

50
of arrival (AOA spatial lobes nAOA ) and the number of main directions of
departure (AOD spatial lobes nAOD ) are between 1 and 5. The generated
MPCs are randomly assigned to both an AOA and an AOD spatial lobes.
In other words, MPCs in an AOD spatial lobe may belong to different AOA
spatial lobes, and vice versa. Thus, there are nAOA · nAOD lobe combinations
in the channel. For each lobe combination, the human blockage is generated
independently.
In addition, a directional channel realization output from NYUSIM is
generated by applying the user-specified antenna pattern in the strongest
MPC direction to the omnidirectional channel realization. The applied an-
tenna pattern emulates the horn antenna pattern used in the mmWave field
measurements [48], and can be calculated by the user-specified antenna az-
imuth and elevation HPBW. For directional channels generated from NYUSIM,
a blockage event will only induce the same shadowing loss to the MPCs
within the RX antenna HPBW.
3GPP 38.901 proposed 5 potential independent blockers for each path
[70]. In NYUSIM, it is assumed that m independent blockage events (uni-
formly distributed between 1 and 5) may occur for each spatial lobe. m
independent blockage events correspond to m independent Markov traces.
m Markov traces are superimposed to obtain a total loss trace. The actual
shadowing loss for a spatial lobe is attained by randomly selecting a time
instance t0 over the total loss trace. An example of 5 independent Markov
traces running the drop-based mode in NYUSIM is shown in Fig. 3.14.

F IGURE 3.14: Five independently simulated Markov traces.

51
3.9 Polarization Model
Dual-polarized transmitter and receiver antenna arrays have become a stan-
dard setup for mmWave wireless systems in 5G to deliver more signal en-
ergy and increase channel rank for higher-order MIMO diversity and multi-
plexing. Correspondingly, the channel simulators are expected to generate
different polarization combinations for a single channel realization at the
same time. Therefore, NYUSIM 3.1 enables simultaneous multiple polar-
ization realizations for channel simulations from 500 MHz to 150 GHz in
indoor and outdoor environments. The adopted model for polarization is
described as follows.
Cross-polarization discrimination (XPD) is defined as the signal power
attenuation (in dB) when the transmit antenna is vertically polarized and
the receive antenna horizontally polarized (i.e., cross-polarization) com-
pared to the case that both transmit and receive antennas are vertically or
horizontally polarized (i.e., co-polarization) including channel effect. XPD
values over microwave and millimeter-wave frequencies at various indoor
and outdoor environments have been reported in the literature [2], [7], [18],
[29]–[37], showing that the XPD increases as the carrier frequency increases.
Thus, a linear function of frequency is used to fit the reported XPDs and cal-
culate the XPD over the NYUSIM supported frequency range (500 MHz to
150 GHz), which is given by

XPD (dB) = k · f (GHz) + b, (3.36)

where k and b are the slope and intercept, respectively. In addition, the
XPDs are found larger in LOS environments than those in NLOS environ-
ments since the boresight path generally does not have a depolarization
effect. Two linear functions with different values are used for the LOS and
NLOS environments. Here we apply a single XPD to the total omnidirec-
tional received power instead of applying different XPDs to different mul-
tipath components (MPCs) due to the limited available MPC-wise XPD val-
ues in the literature, which makes the polarization model succinct yet valid.
The used XPD values and linear fits for the LOS and NLOS environments
are shown in Fig. 3.15.

F IGURE 3.15: The measured XPDs with linear fits for the LOS and
NLOS environments.

Similarly, co-polarization discrimination (CPD) can be defined as the

52
power difference (in dB) between the V-V and H-H polarization settings or
the V-H and H-V polarization settings. The studies on the CPD are very
limited [29], [33], [38], [39], showing that the V-V polarization usually has
a slightly larger received power than the H-H polarization when walls are
the primary reflectors in the environment. However, the difference in the
overall received power is typically negligible (within 1 dB) [29], [33], [38],
[39]. One exception is the tunnel scenario, where the particular confined
waveguide shape causes the CPD variation between -20 dB and 80 dB with
a mean of 5.4 dB [38]. NYUSIM focuses on the common indoor and outdoor
environments. Thus the CPD is modeled as a zero-mean Gaussian random
variable with a small standard deviation (1.6 dB used in NYUSIM 3.1) [29],
[33], [38].
NYUSIM 3.1 allows users to use one polarization (co- or cross-polarization),
two polarizations (co- and cross-polarization), or four polarizations (vertical-
to-vertical (V-V), horizontal-to-horizontal (H-H), vertical-to-horizontal (V-
H), and horizontal-to-vertical (H-V) polarizations) for every single channel
realization. The output figures and files are named with the specified po-
larization.

53
4. Applications of NYUSIM

The output figure and data files generated from NYUSIM can be used in
various ways based on users’ needs, e.g., to simulate channel impulse re-
sponses for mmWave systems, to investigate MIMO performance, etc. Here
we present five example applications for how NYUSIM may be used.

4.1 Application 1: MIMO Channel Condition Number

First we show an example of how to obtain the condition number of a
MIMO channel by making use of the output data files "BasicParameters.mat"
and "DirPDPInfo.mat", assuming OFDM modulation is utilized. This appli-
cation also shows the wide range of parameters that a user may adjust in
NYUSIM2.01.

The condition number is defined as the ratio of the largest to smallest sin-
gular value in the singular value decomposition of a matrix, and is a met-
ric to characterize the quality of MIMO channels in the context of wireless
communications [63], [83]–[85]. The condition number will be high (e.g.,
over 20 dB) if the minimum singular value is close to zero, and will be 0
dB if singular values are equal. Physically, a small condition number value
(e.g., below 20 dB) indicates good orthogonality of different sub-channels (a
sub-channel usually has a distinct spatial direction), and the channel gains
are comparable in different spatial directions. The rank of a matrix is the
dimension of the vector space generated (or spanned) by its columns (or
rows) [86], and it determines how many data streams can be multiplexed
over the channel in the context of MIMO communications [83], [87], [88].
The condition number is related to the rank of a matrix: a low condition
number usually corresponds to a high rank and vice versa; the matrix has
full rank (the highest rank) when the condition number is equal or close to
0 dB (the lowest theoretical condition number).

Let’s look at the condition number of a MIMO channel matrix for a single
sub-carrier in an OFDM system. As described in Section 2.2.2, the output
data files "BasicParameters.mat" and "DirPDPInfo.mat" contain paramount
parameters of each resolvable multipath component, which will be useful
in generating the MIMO channel coefficient for an OFDM sub-carrier. Take
ULAs at both the transmitter and receiver for example, the equation for
generating such a channel coefficient is provided below, which is adapted

54
from Eq.(2) in [89]:
X
hm,k (f ) = αm,k,p ejΦm,k,p e−j2πf τm,k,p e−j2πdT m sin(φm,k,p ) e−j2πdR k sin(ϕm,k,p )
p
(4.1)
where hm,k (f ) denotes the MIMO channel coefficient between the mth trans-
mit antenna and the k th receive antenna for the sub-carrier f , p represents
the pth resolvable multipath component, α is the amplitude of the channel
gain, Φ denotes the phase of the multipath component, τ represents the
time delay, dT and dR are the antenna element spacing at the transmitter
and receiver, respectively, while φ and ϕ denote the azimuth angle of de-
parture and angle of arrival, respectively. All of the above parameters can
be extracted from the files "BasicParameters.mat" and "DirPDPInfo.mat".
For each sub-carrier f in a MIMO-OFDM system, there exists an Nt × Nr
channel matrix H whose elements are hm,k (f ), where m = 1, ..., Nt and
k = 1, ..., Nr . The condition number of H can be obtained consequently.

Using the above mentioned approach, and the input parameters on the
NYUSIM GUI with the following values:

• Distance Range Option: Standard (10-500 m)

• Frequency: 28 GHz

• RF bandwidth: 800 MHz

• Scenario: UMi

• Environment: LOS

• Lower Bound of T-R Separation Distance: 100 m

• Upper Bound of T-R Separation Distance: 100 m

• TX Power: 30 dBm

• Base Station Height: 35 m (This parameter is ignored by the program

because it is the UMi scenario, and not the RMa scenario.)

• Barometric Pressure: 1013.25 mbar

• Humidity: 50%

• Temperature: 20◦ C

• Rain Rate: 0 mm/hr

• Polarization: Co-Pol

• Foliage Loss: No

• Number of RX Locations: 100

• TX Array Type: ULA

• RX Array Type: ULA

• Number of TX Antenna Elements Nt : 2

55
 • Number of RX Antenna Elements Nr : 2

• TX Antenna Spacing: 0.5 wavelength

• RX Antenna Spacing: 0.5 wavelength

• Number of TX Antenna Elements Per Row Wt : 2

• Number of RX Antenna Elements Per Row Wr : 2

• TX Antenna Azimuth HPBW: 10◦

• TX Antenna Elevation HPBW: 10◦

• RX Antenna Azimuth HPBW: 10◦

• RX Antenna Elevation HPBW: 10◦

and assuming the frequency interval between adjacent sub-carriers is 500

kHz, which corresponds to 800 MHz/500 kHz = 1600 sub-carriers, we per-
form 100 simulation runs (i.e., set the number of RX locations to 100) to
emulate 100 random MIMO channel realizations with the input parameters
described above. Then the following changes are made to the four input
parameters below with all the other input parameter values remaining the
same:

• Number of TX Antenna Elements Nt : 3

• Number of RX Antenna Elements Nr : 3

• Number of TX Antenna Elements Per Row Wt : 3

• Number of RX Antenna Elements Per Row Wr : 3

Fig. 4.1 illustrates the empirical CDF of the condition number of channel
matrices for OFDM sub-carriers with the above two sets of input param-
eters in one of the 100 simulation runs. The empirical CDF of the corre-
sponding rank of the channel matrices are plotted in Fig. 4.2, where the
rank here is defined as the number of singular values of the channel ma-
trix that are larger than σm /1000, where σm is the maximum singular value
of the channel matrix. By this definition of rank, we are essentially ignor-
ing the singular values that are 30 dB smaller than the maximum singular
value, which is physically reasonable since the singular value is a measure
of the channel gain along its corresponding singular vector and it makes lit-
tle sense to consider the direction with an extremely small channel gain. It is
apparent from Fig. 4.1 that the condition numbers of the individual OFDM
sub-carriers for a 3×3 MIMO channel is about 18 dB larger compared to the
2 × 2 case on average, and the relatively large condition number of the 3 × 3
channel matrix may stem from the fact that the matrix is rank deficient, as
evident from Fig. 4.2, in which about 96% of the channel matrices have a
rank of 2 instead of 3 (full rank).

Based on the results from the 100 random simulation runs, it is found that
the average median value of the condition numbers of the channel matri-
ces for individual sub-carriers over the 100 simulation runs is around 13 dB
(i.e., the difference in dB between the largest and smallest singular value of

56
 100

80
Empirical CDF (%)

60

40

20
Nt=2, Nr=2
Nt=3, Nr=3
0
0 10 20 30 40 50
Condition Number (dB)
F IGURE 4.1: Empirical CDF of the condition number of channel ma-
trices for OFDM sub-carriers with different transmit and receive an-
tenna elements Nt and Nr for MIMO-OFDM channels in one simu-
lation run.

100

80
Empirical CDF (%)

60

40

20
Nt=2, Nr=2
Nt=3, Nr=3
0
1 2 3 4
Rank
F IGURE 4.2: Empirical CDF of the rank of channel matrices for
OFDM sub-carriers with different transmit and receive antenna el-
ements Nt and Nr for MIMO-OFDM channels in one simulation run.

57
the channel matrix is 13 dB), and the average matrix rank is 2. When the
numbers of TX and RX antenna elements are both changed to 3, the average
median value of the condition number increase to 31 dB, with a mean rank
of roughly 2, indicating that two spatial streams can be sent simultaneously
using the spatial multiplexing technique in this case.

The Matlab code for extracting channel coefficients based on Eq. (4.1), gen-
erating the channel matrix for each sub-carrier frequency, and calculating
the condition number and rank in the example above is contained in the
package named “Application Example Code” available on the NYUSIM
downloading website.

4.2 Application 2: Spectral Efficiency Comparison Be-

tween 3GPP and NYUSIM Channel Models
As mentioned in NYUSIM Overview on Page ii of this user manual, the
3GPP TR 38.900 Release 14 channel model [64] contains unrealistic num-
ber of clusters (e.g., 19 clusters for UMi NLOS) and up to 20 rays per clus-
ter, which is excessively large and not borne out by measurements, and is
likely to result in incorrect predictions on MIMO system performance such
as spectral efficiency (SE). in this subsection, we use the 3GPP TR 38.900
Release 14 channel model [64] and NYUSIM channel model to analyze and
compare the spectral efficiency for mmWave MIMO channels.

Let us assume a single-cell single-user MIMO system without multiple-

access interference, operating at 28 GHz with an RF bandwidth of 100 MHz
in the UMi scenario. The base station (BS) is equipped with 256 antenna ele-
ments comprising a uniform rectangular array (URA) with cross-polarized
elements, which can reduce the hardware size and make better use of mul-
tipath components with different polarizations since the cross-polarization
discrimination ratio is 8-9 dB in both models, where the antenna pattern on
Page 23 of [64] is adopted. The user equipment (UE) has 16 antenna ele-
ments constituting a URA with cross-polarized omnidirectional elements.
Signal-to-noise ratios (SNRs) are fixed at certain values to investigate the
SE achieved by the 3GPP [64] and NYUSIM models described in [3] and
Section 3 of this user manual. Two hundred random channel realizations
using random distances ranging from 10 to 435 m were performed for each
channel model, where the NYUSIM code runs about 30 times faster than
the 3GPP code.

Fig. 4.3 illustrates the SE achieved by the hybrid beamforming algorithm

proposed in [90] for a 256×16 mmWave system at 28 GHz with rectangular
planar arrays at the transmitter and receiver for various numbers of data
streams Ns , using the 3GPP and NYUSIM models. As shown by Fig. 4.3,
the SE generated by the 3GPP channel model is just slightly smaller than
that yielded by the NYUSIM channel model for Ns = 1, but much larger
for Ns = 4. For instance, for an SNR of 20 dB and Ns = 4, the 3GPP SE (40
bits/s/Hz) is about 13 bits/s/Hz greater than the NYUSIM SE (about 27

58
 40
3GPP, 1 stream

Spectral Efficiency [bits/s/Hz]

35 NYU, 1 stream
3GPP, 4 streams
30 NYU, 4 streams

25

20

15

10

0
-30 -20 -10 0 10 20
SNR [dB]
F IGURE 4.3: Spectral efficiency achieved by hybrid beamforming for
a 256×16 system at 28 GHz with rectangular antenna arrays at the
transmitter and receiver with four RF chains.

bits/s/Hz). This is likely because NYUSIM yields one or two strong dom-
inant clusters and much weaker non-dominant clusters, while the 3GPP
model has less focused directional energy than what realistically exists. The
results indicate that the 3GPP channel model is optimistic when predicting
diversity and the achievable SE at mmWave frequencies, while NYUSIM
provides realistic channel parameters and SE predictions due to the use
of extensive real-world measurement data at mmWave frequencies, thus
could help avoid system errors inherent with legacy modeling approaches.

4.3 Application 3: Temporal Large-scale Path Loss Vari-

ation Due to Spatial Consistency
Spatial consistency is used for simulating the changing channel conditions
of a user as it moves in a local area, or may be used to generate correlated
channels for closely-spaced users (as opposed to independent responses
generated from the drop-based mode). As an example of spatial consis-
tency mode in NYUSIM, time-variant large-scale path loss based on the UT
trajectory is calculated using the map of spatially correlated shadow fad-
ing and UT and BS locations. A map of spatially correlated shadow fading
over a 200 m x 200 m area is shown in Fig. 4.4. The correlation distance of
shadow fading in a UMi LOS scenario is set to be 10 m [16], [44]. The UT
moved 40 m in a partial hexagon track, which is also illustrated in Fig. 4.4.
Shadow fading varies from -10 dB to 10 dB in a continuous manner. Simi-
lar shadow fading values are observed at closely spaced locations whereas
independent values for close locations are always used in the drop-based
model.

59
 F IGURE 4.4: A map of spatially correlated shadow fading with the BS
and UT locations. The map of spatially correlated shadow fading is
generated by filtering a map of independent shadow fading using an
exponential function in (3.5). Shadow fading [dB]∼N(0,4) in a UMi
LOS scenario. T-R separation distance is 100 m.

The time-variant path loss is shown in Fig. 4.5 [16], [45]. It can be seen
that the path loss varies rapidly about 16 dB with independent shadow
fading values. However, the path loss varies smoothly with spatially cor-
related shadow fading, which corresponds to the UT trajectory. The path
loss increases first and then drops as the UT first moves away from the base
station and then comes back. Spatially correlated shadow fading supports
a more realistic link budget analysis and cellular coverage prediction.

F IGURE 4.5: The UT moves in a partial hexagon track in a clockwise

manner. The track distance is 40 m, the side length of the hexagon is
10 m.

60
4.4 Application 4: Human Blockage Shadowing Loss
Simulations
To investigate the human blockage effect on directional channels generated
from NYUSIM, four sets of TX and RX antennas with HPBW 7°, 15°, 30°,
and 60° are selected in a UMi scenario at 73 GHz. The T-R separation dis-
tance ranges from 100 m to 500 m. 1000 simulations (i.e. 1000 UT locations)
are performed for each TX and RX antenna setting. A comparison of CDFs
of shadowing loss for directional channels in the UMi NLOS scenario using
four sets of TX and RX antennas with HPBW 7°, 15°, 30°, and 60° is shown
in Fig. 4.6 [16], [45]. UTs equipped with narrower HPBW antennas are more
likely to experience severe blockage shadowing loss. 31% of UTs equipped
with 7° HPBW antenna experience more than 15 dB shadowing loss.

F IGURE 4.6: CDFs of simulated human blockage shadowing loss us-

ing the four-state Markov model for directional channels with RX
antenna azimuth HPBWs (7°, 15°, 30°, and 60°.

4.5 Application 5: O2I Penetration Loss Simulations

The O2I feature will be turned on if the option “Outdoor to Indoor (O2I)
Penetration Loss” is “Yes” under the drop-based mode, as shown in Fig.
4.7.This application is used for simulating the channel responses for users
inside buildings.
The simulated O2I penetration loss value can be read from the omnidi-
rectional and directional PDPs, as shown in Fig. 4.8 and Fig. 4.9. There is
an additional 15.0 dB O2I penetration loss added to the total path loss.

61
F IGURE 4.7: An example of NYUSIM run with O2I penetration loss.

F IGURE 4.8: Omnidirectional PDP output from a NYUSIM run with

15.0 dB O2I penetration loss.

F IGURE 4.9: Directional PDP output from a NYUSIM run with 15.0
dB O2I penetration loss.

62
 Apart from the example demonstrated above, NYUSIM can also find
applications in other types of analysis, such as the design of antenna ar-
rays [91] and bit error simulation [22], [92].

4.6 Application 6: Indoor Coverage Prediction

The prediction of the coverage of an indoor WiFi hotspot is crucial for the
deployment of large WiFi networks. Due to the increasing path loss within
the first meter of propagation distance at sub-THz frequencies, the cell size
shrinks and the ultra dense network is required to provide sufficient link
margin. Here we show an example of using NYUSIM 3.0 to predict the typ-
ical coverage of a sub-THz indoor wireless system. The input parameters
on the NYUSIM are set with the following values:

• Frequency: 140 GHz

• RF bandwidth: 800 MHz

• Scenario: InH

• Distance Range Option: Indoor (5-50 m)

• Environment: LOS and NLOS

• Lower Bound of T-R Separation Distance: 5 m

• Upper Bound of T-R Separation Distance: 50 m

• TX Power: 10 dBm

• Base Station Height: 2.5 m

• Polarization: Co-Pol

• Number of RX Locations: 100 in LOS and 100 in NLOS

• TX Array Type: URA

• RX Array Type: URA

• Number of TX Antenna Elements Nt : 16

• Number of RX Antenna Elements Nr : 4

• TX Antenna Spacing: 0.5 wavelength

• RX Antenna Spacing: 0.5 wavelength

• Number of TX Antenna Elements Per Row Wt : 4

• Number of RX Antenna Elements Per Row Wr : 2

• TX Antenna Azimuth HPBW: 10◦

• TX Antenna Elevation HPBW: 10◦

• RX Antenna Azimuth HPBW: 30◦

• RX Antenna Elevation HPBW: 30◦

63
We assume the receiver sensitivity of a mobile receiver under the WLAN
compliance is -82 dBm. Fig. 4.10 shows the scatter plot of the received pow-
ers and the average power level of the received signals at distances from 5 to
50 m for indoor directional channels in both LOS and NLOS environments.
Antenna arrays providing narrow beams with HPBW beamwidths of 10°
and 30° are used at the transmitter and receiver, respectively, which can
provide good directional gains and eliminate interference from communi-
cation links of other hotspots or mobile devices. The TX and RX beams are
pointed in the boresight direction in the LOS environment while the beams
are pointed in the strongest reflection direction in the NLOS environment.
Fig. 4.10 suggests that the receiver maintains a sufficient SNR level for dis-
tances even beyond 50 m in the LOS environment. However, in the NLOS
environment, the received power drops below the receiver sensitivity (-82
dBm) at the separation distance of 35.8 m when the TX and RX antennas are
pointed to the strongest reflection direction, indicating the potential usage
of multi-beam antenna combing [93] and reconfigurable intelligent surfaces
[94] to extend the signal coverage.

F IGURE 4.10: Scatter plot of the received powers and the average
power level of the received signals at distances from 5 to 50 m for
LOS and NLOS indoor directional channels. For the NLOS environ-
ment, the average power level of received signals drops below the
receiver sensitivity beyond 35.8 m.

64
Bibliography

[1] T. S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G. N.

Wong, J. K. Schulz, M. Samimi, and F. Gutierrez, “Millimeter wave
mobile communications for 5G cellular: It will work!” IEEE Access,
vol. 1, pp. 335–349, 2013. [Online]. Available: http://ieeexplore.
ieee.org/document/6515173/.
[2] T. S. Rappaport, G. R. MacCartney, Jr., M. K. Samimi, and S. Sun,
“Wideband millimeter-wave propagation measurements and chan-
nel models for future wireless communication system design (Invited
Paper),” IEEE Transactions on Communications, vol. 63, no. 9, pp. 3029–
3056, Sep. 2015. [Online]. Available: http://ieeexplore.ieee.
org/document/7109864/.
[3] M. K. Samimi and T. S. Rappaport, “3-D millimeter-wave statistical
channel model for 5G wireless system design,” IEEE Transactions on
Microwave Theory and Techniques, vol. 64, no. 7, pp. 2207–2225, Jul.
2016. [Online]. Available: http://ieeexplore.ieee.org/document/
7501500/.
[4] ——, “Local multipath model parameters for generating 5G millimeter-
wave 3GPP-like channel impulse response,” the 10th European Confer-
ence on Antennas and Propagation (EuCAP 2016), Apr. 2016. [Online].
Available: http://arxiv.org/abs/1511.06941.
[5] S. Sun et al., “Investigation of prediction accuracy, sensitivity, and pa-
rameter stability of large-scale propagation path loss models for 5G
wireless communications,” IEEE Transactions on Vehicular Technology,
vol. 65, no. 5, pp. 2843–2860, May 2016. [Online]. Available: http :
//ieeexplore.ieee.org/document/7434656/.
[6] S. Sun, G. R. MacCartney, M. K. Samimi, and T. S. Rappaport, “Syn-
thesizing omnidirectional antenna patterns, received power and path
loss from directional antennas for 5G millimeter-wave communica-
tions,” in 2015 IEEE Global Communications Conference (GLOBECOM),
Dec. 2015, pp. 1–7. [Online]. Available: http://arxiv.org/abs/
1511.07271.
[7] G. R. MacCartney, Jr., T. S. Rappaport, S. Sun, and S. Deng, “Indoor of-
fice wideband millimeter-wave propagation measurements and chan-
nel models at 28 and 73 GHz for ultra-dense 5G wireless networks,”
IEEE Access, vol. 3, pp. 2388–2424, 2015. [Online]. Available: http:
//ieeexplore.ieee.org/document/7289335/.

65
 [8] T. S. Rappaport, S. Sun, and M. Shafi, “Investigation and comparison
of 3GPP and NYUSIM channel models for 5G wireless communica-
tions,” in 2017 IEEE 86th Vehicular Technology Conference (VTC-Fall),
Sep. 2017, pp. 1–5. [Online]. Available: https://arxiv.org/pdf/
1707.00291.pdf.
[9] G. R. MacCartney, Jr., S. Sun, T. S. Rappaport, Y. Xing, H. Yan, J.
Koka, R. Wang, and D. Yu, “Millimeter wave wireless communica-
tions: New results for rural connectivity,” in All Things Cellular’16, in
conjunction with ACM MobiCom, Oct. 2016. [Online]. Available: https:
//arxiv.org/abs/1608.05384.
[10] “Investigation of prediction accuracy, sensitivity, and parameter sta-
bility of large-scale propagation path loss models from 500 MHz to
100 GHz,” [Online]. Available: http://wireless.engineering.
nyu.edu/presentations/NTIA-propagation-presentation-
JUNE-15-2016_v1-3.pdf.
[11] T. S. Rappaport, S. Sun, and M. Shafi, “5G channel model with im-
proved accuracy and efficiency in mmWave bands,” IEEE 5G Tech Fo-
cus, vol. 1, no. 1, Mar. 2017. [Online]. Available: https://5g.ieee.
org/tech-focus/march-2017/5g-channel-model.
[12] G. R. MacCartney and T. S. Rappaport, “Rural macrocell path loss
models for millimeter wave wireless communications,” IEEE Jour-
nal on Selected Areas in Communications, vol. 35, no. 7, pp. 1663–1677,
Jul. 2017. [Online]. Available: http://ieeexplore.ieee.org/
stamp/stamp.jsp?tp=&arnumber=7914696.
[13] S. Sun, G. R. MacCartney, and T. S. Rappaport, “A novel millimeter-
wave channel simulator and applications for 5G wireless commu-
nications,” in 2017 IEEE International Conference on Communications
(ICC), May 2017, pp. 1–7. [Online]. Available: https : / / arxiv .
org/pdf/1703.08232.pdf.
[14] S. Ju and T. S. Rappaport, “Simulating motion - incorporating spatial
consistency into the NYUSIM channel model,” 2018 IEEE 88th Vehic-
ular Technology Conference Workshops, pp. 1–6, Aug. 2018.
[15] ——, “Millimeter-wave extended nyusim channel model for spatial
consistency,” 2018 IEEE Global Communications Conference (Globecom),
pp. 1–6, Dec. 2018.
[16] S. Ju, Y. Xing, O. Kanhere, and T. S. Rappaport, “A Millimeter-Wave
Channel Simulator NYUSIM with Spatial Consistency and Human
Blockage,” IEEE 2019 Global Communications Conference, pp. 1–6, Dec.
2019.
[17] ——, “3-D Statistical Indoor Channel Model for Millimeter-Wave and
Sub-Terahertz Bands,” IEEE 2020 Global Communications Conference,
pp. 1–7, Dec. 2020.
[18] S. Ju, Y. Xing, O. Kanhere, and T. S. Rappaport, “Millimeter wave and
sub-Terahertz spatial statistical channel model for an indoor office
building,” IEEE Journal on Selected Areas in Communications, Special
Issue on TeraHertz Communications and Networking, pp. 1–15, Second
Quarter 2021.

66
[19] Y. Xing and T. S. Rappaport, “Spectrum coexistence between active
and passive systems in space and on the ground at frequencies above
100 ghz (invited),” in submission to IEEE Communications Letters, pp. 1–
5, Feb. 2021.
[20] Y. Xing, T. S. Rappaport, and A. Ghosh, “Millimeter wave and sub-
thz indoor radio propagation channel measurements, models, and
comparisons in an office environment (invited),” in submission to IEEE
Communications Letters, pp. 1–5, Feb. 2021.
[21] Y. Xing and T. S. Rappaport, “Propagation Measurements and Path
Loss models for sub-THz in Urban Microcells,” submitted to 2021 IEEE
International Conference on Communications (ICC), pp. 1–5, Jun. 2021.
[22] V. Fung, T. S. Rappaport, and B. Thoma, “Bit error simulation for pi/4
DQPSK mobile radio communications using two-ray and measurement-
based impulse response models,” IEEE Journal on Selected Areas in
Communications, vol. 11, no. 3, pp. 393–405, Apr. 1993. [Online]. Avail-
able: http://ieeexplore.ieee.org/document/219546/.
[23] J. C. Liberti and T. S. Rappaport, “Analysis of CDMA cellular radio
systems employing adaptive antennas in multipath environments,”
in IEEE 46th Vehicular Technology Conference, vol. 2, Apr. 1996, pp. 1076–
1080. [Online]. Available: http://ieeexplore.ieee.org/document/
501477/.
[24] J. Lota, S. Sun, T. S. Rappaport, and A. Demosthenous, “5G uniform
linear arrays with beamforming and spatial multiplexing at 28, 37,
64, and 71 GHz for outdoor urban communication: A two-level ap-
proach,” IEEE Transactions on Vehicular Technology, vol. 66, no. 11, pp. 9972–
9985, Nov. 2017. [Online]. Available: http://ieeexplore.ieee.
org/document/8012543/.
[25] M. K. Samimi and T. S. Rappaport, “3-D millimeter-wave statistical
channel model for 5G wireless system design,” IEEE Transactions on
Microwave Theory and Techniques, vol. 64, no. 7, pp. 2207–2225, Jul.
2016.
[26] 3GPP, “Study on channel model for frequencies from 0.5 to 100 GHz,”
3rd Generation Partnership Project (3GPP), TR 38.901 V14.1.1, Jul.
2017. [Online]. Available: https://portal.3gpp.org/desktopmodules/
Specifications/SpecificationDetails.aspx?specificationId=
3173.
[27] S. Sun, T. S. Rappaport, M. Shafi, P. Tang, J. Zhang, and P. J. Smith,
“Propagation models and performance evaluation for 5G millimeter-
wave bands,” IEEE Transactions on Vehicular Technology, vol. 67, no. 9,
pp. 8422–8439, Sep. 2018.
[28] S. Sun, T. S. Rappaport, M. Shafi, and H. Tataria, “Analytical frame-
work of hybrid beamforming in multi-cell millimeter-wave systems,”
IEEE Transactions on Wireless Communications, vol. 17, no. 11, pp. 7528–
7543, Nov. 2018.

67
[29] T. Manabe, K. Sato, H. Masuzawa, K. Taira, T. Ihara, Y. Kasashima,
and K. Yamaki, “Polarization dependence of multipath propagation
and high-speed transmission characteristics of indoor millimeter-wave
channel at 60 ghz,” IEEE Transactions on Vehicular Technology, vol. 44,
no. 2, pp. 268–274, 1995.
[30] Y. Xing, O. Kanhere, S. Ju, and T. S. Rappaport, “Indoor Wireless
Channel Properties at Millimeter Wave and Sub-Terahertz Frequen-
cies,” IEEE 2019 Global Communications Conference, pp. 1–6, Dec. 2019.
[31] Y. Xing and T. S. Rappaport, “Propagation measurement system and
approach at 140 ghz–moving to 6g and above 100 ghz,” in 2018 IEEE
Global Communications Conference (GLOBECOM), Dec. 2018, pp. 1–6.
[32] A. Maltsev et al., “Impact of polarization characteristics on 60-GHz
indoor radio communication systems,” IEEE Antennas and Wireless
Propagation Letters, vol. 9, pp. 413–416, May 2010.
[33] J.-M. Molina-Garcia-Pardo, J.-V. Rodriguez, and L. Juan-Llacer, “Po-
larized indoor mimo channel measurements at 2.45 ghz,” IEEE Trans-
actions on Antennas and Propagation, vol. 56, no. 12, pp. 3818–3828,
2008.
[34] A. Jagannatham and V. Erceg, “Mimo indoor wlan channel measure-
ments and parameter modeling at 5.25 ghz,” in IEEE 60th Vehicular
Technology Conference, 2004. VTC2004-Fall. 2004, vol. 1, 2004, 106–110
Vol. 1.
[35] P. Zhang, J. Li, H. Wang, H. Wang, and W. Hong, “Indoor small-scale
spatiotemporal propagation characteristics at multiple millimeter-wave
bands,” IEEE Antennas and Wireless Propagation Letters, vol. 17, no. 12,
pp. 2250–2254, 2018.
[36] D. Solomitckii, M. Allén, D. Yolchyan, H. Hovsepyan, M. Valkama,
and Y. Koucheryavy, “Millimeter-wave channel measurements at 28
ghz in digital fabrication facilities,” in 2019 16th International Sympo-
sium on Wireless Communication Systems (ISWCS), 2019, pp. 548–552.
[37] M. Khalily, S. Taheri, S. Payami, M. Ghoraishi, and R. Tafazolli, “In-
door wideband directional millimeter wave channel measurements
and analysis at 26 ghz, 32 ghz, and 39 ghz,” Transactions on Emerging
Telecommunications Technologies, vol. 29, no. 10, e3311, 2018.
[38] D. He, B. Ai, K. Guan, Z. Zhong, B. Hui, J. Kim, H. Chung, and I. Kim,
“Channel measurement, simulation, and analysis for high-speed rail-
way communications in 5g millimeter-wave band,” IEEE Transactions
on Intelligent Transportation Systems, vol. 19, no. 10, pp. 3144–3158,
2018.
[39] X. Liu, X. Yin, and G. Zheng, “Experimental investigation of millimeter-
wave mimo channel characteristics in tunnel,” IEEE Access, vol. 7,
pp. 108 395–108 399, 2019.
[40] S. Sun, T. S. Rappaport, R. W. Heath, A. Nix, and S. Rangan, “Mimo
for millimeter-wave wireless communications: Beamforming, spatial
multiplexing, or both?” IEEE Communications Magazine, vol. 52, no. 12,
pp. 110–121, 2014.

68
[41] 3GPP, “Technical specification group radio access network; study on
channel model for frequencies from 0.5 to 100 GHz (Release 16),” TR
38.901 V16.0.0, Oct. 2019.
[42] A. Maltsev et al., “Channel models for IEEE 802.11ay,” doc.: IEEE
802.11-15/1150r9, May 2016.
[43] S. Ju and T. S. Rappaport, “Simulating motion - incorporating spatial
consistency into the NYUSIM channel model,” 2018 IEEE 88th Vehic-
ular Technology Conference Workshops, pp. 1–6, Aug. 2018.
[44] ——, “Millimeter-wave extended NYUSIM channel model for spatial
consistency,” 2018 IEEE Globecom, pp. 1–6, Dec. 2018.
[45] S. Ju, “Channel modeling and channel simulation for fifth-generation
and beyond millimeter-wave wireless communications,” Ph.D. dis-
sertation, NEW YORK UNIVERSITY, TANDON SCHOOL OF ENGI-
NEERING, May 2019.
[46] G. R. MacCartney, T. S. Rappaport, and S. Rangan, “Rapid fading due
to human blockage in pedestrian crowds at 5G millimeter-wave fre-
quencies,” in 2017 IEEE Global Communications Conference, Dec. 2017,
pp. 1–7.
[47] G. R. MacCartney et al., “Millimeter-wave base station diversity for
5G coordinated multipoint (CoMP) applications,” IEEE Transactions
on Wireless Communications, May 2019.
[48] T. S. Rappaport et al., “Millimeter Wave Mobile Communications for
5G Cellular: It Will Work!” IEEE Access, vol. 1, pp. 335–349, May 2013.
[49] H. Zhao et al., “28 GHz millimeter wave cellular communication mea-
surements for reflection and penetration loss in and around buildings
in New York city,” in 2013 IEEE International Conference on Communi-
cations (ICC), Jun. 2013, pp. 5163–5167.
[50] K. Haneda et al., “5G 3GPP-like channel models for outdoor urban
microcellular and macrocellular environments,” in 2016 IEEE 83rd Ve-
hicular Technology Conference (VTC2016-Spring), May 2016, pp. 1–7.
[51] Aalto University, AT&T, BUPT, CMCC, Ericsson, Huawei, Intel, KT
Corporation, Nokia, NTT DOCOMO, New York University, Qual-
comm, Samsung, University of Bristol, and University of Southern
California, “5G channel model for bands up to 100 GHz,” 2016, Oct.
21. [Online]. Available: http://www.5gworkshops.com/5GCM.
html.
[52] H. J. Liebe, G. A. Hufford, and M. G. Cotton, “Propagation modeling
of moist air and suspended water/ice particles at frequencies below
1000 GHz,” AGARD Conference Proceedings 542, May 1993. [Online].
Available: http : / / www . its . bldrdoc . gov / publications /
2670.aspx.
[53] J. J. A. Lempianen, J. K. Laiho-Steffens, and A. F. Wacker, “Experi-
mental results of cross polarization discrimination and signal corre-
lation values for a polarization diversity scheme,” in IEEE 47th Ve-
hicular Technology Conference, vol. 3, May 1997, pp. 1498–1502. [On-
line]. Available: http : / / ieeexplore . ieee . org / document /
605617/.

69
[54] H.-l. Xiao, S. Ouyang, and Z.-p. Nie, “The cross polarization discrim-
ination of MIMO antennas at mobile station,” in International Con-
ference on Communications, Circuits and Systems (ICCCAS), May 2008,
pp. 203–206. [Online]. Available: http : / / ieeexplore . ieee .
org/document/4657759/.
[55] T. S. Rappaport and S. Deng, “73 GHz wideband millimeter-wave fo-
liage and ground reflection measurements and models,” in 2015 IEEE
International Conference on Communication Workshop (ICCW), Jun. 2015,
pp. 1238–1243. [Online]. Available: https : / / arxiv . org / pdf /
1509.00436.
[56] K. Haneda et al., “Indoor 5G 3GPP-like channel models for office and
shopping mall environments,” in 2016 IEEE International Conference
on Communications Workshops (ICCW), May 2016, pp. 694–699.
[57] T. S. Rappaport and R. A. Brickhouse, “A simulation of cellular sys-
tem growth and its effect on urban in-building parasitic frequency
reuse,” IEEE Transactions on Vehicular Technology, vol. 48, no. 1, pp. 286–
294, Jan. 1999. [Online]. Available: http : / / ieeexplore . ieee .
org/document/740107/.
[58] T. Bai, A. Alkhateeb, and R. W. Heath, “Coverage and capacity of
millimeter-wave cellular networks,” IEEE Communications Magazine,
vol. 52, no. 9, pp. 70–77, Sep. 2014. [Online]. Available: http : / /
ieeexplore.ieee.org/document/6894455/.
[59] G. R. MacCartney, Jr., T. S. Rappaport, and S. Rangan, “Rapid fading
due to human blockage in pedestrian crowds at 5g millimeter-wave
frequencies,” in 2017 IEEE Global Communications Conference (GLOBE-
COM), Dec. 2017, pp. 1–7.
[60] A. Alkhateeb and R. W. Heath, “Frequency selective hybrid precod-
ing for limited feedback millimeter wave systems,” IEEE Transactions
on Communications, vol. 64, no. 5, pp. 1801–1818, May 2016. [Online].
Available: http://ieeexplore.ieee.org/document/7448873/.
[61] A. I. Sulyman, A. T. Nassar, M. K. Samimi, G. R. Maccartney, T. S. Rap-
paport, and A. Alsanie, “Radio propagation path loss models for 5G
cellular networks in the 28 GHz and 38 GHz millimeter-wave bands,”
IEEE Communications Magazine, vol. 52, no. 9, pp. 78–86, Sep. 2014.
[Online]. Available: http://ieeexplore.ieee.org/document/
6894456/.
[62] A. I. Sulyman, A. Alwarafy, G. R. MacCartney, T. S. Rappaport, and A.
Alsanie, “Directional radio propagation path loss models for millimeter-
wave wireless networks in the 28-, 60-, and 73-GHz bands,” IEEE
Transactions on Wireless Communications, vol. 15, no. 10, pp. 6939–6947,
Oct. 2016. [Online]. Available: http://ieeexplore.ieee.org/
document/7522613/.
[63] T. S. Rappaport, R. W. Heath, Jr., R. C. Daniels, and J. N. Murdock,
Millimeter Wave Wireless Communications. Pearson/Prentice Hall 2015.
[64] 3GPP, “Study on channel model for frequency spectrum above 6 GHz,”
3rd Generation Partnership Project (3GPP), TR 38.900 V14.2.0, Dec.
2016. [Online]. Available: http://www.3gpp.org/DynaReport/
38900.htm.

70
[65] T. S. Rappaport et al., “Wideband millimeter-wave propagation mea-
surements and channel models for future wireless communication
system design (Invited Paper),” IEEE Transactions on Communications,
vol. 63, no. 9, pp. 3029–3056, Sep. 2015.
[66] J. Ko, S. U. Lee, Y. S. Kim, and D.-J. Park, “Measurements and analy-
ses of 28 GHz indoor channel propagation based on a synchronized
channel sounder using directional antennas,” Journal of Electromag-
netic Waves and Applications, vol. 30, no. 15, pp. 2039–2054, Jun. 2016.
[67] Hao Xu, V. Kukshya, and T. S. Rappaport, “Spatial and temporal char-
acteristics of 60-GHz indoor channels,” IEEE Journal on Selected Areas
in Communications, vol. 20, no. 3, pp. 620–630, Apr. 2002.
[68] L. Rubio et al., “Wideband propagation channel measurements in an
indoor office environment at 26 GHz,” in 2019 IEEE International Sym-
posium on Antennas and Propagation and USNC-URSI Radio Science Meet-
ing, Oct. 2019, pp. 2075–2076.
[69] A. Saleh and R. A. Valenzuela, “A statistical model for indoor multi-
path propagation,” IEEE Journal on Selected Areas in Communications,
vol. 5, no. 2, pp. 128–137, Feb. 1987.
[70] 3GPP, “Technical specification group radio access network; study on
channel model for frequencies from 0.5 to 100 GHz (Release 15),”
3rd Generation Partnership Project (3GPP), TR 25.104 V15.4.0, Sep.
2018. [Online]. Available: http://www.3gpp.org/DynaReport/
25104.htm.
[71] METIS, “METIS Channel Model,” METIS2020, Deliverable D1.4 v3,
Jul. 2015.
[72] T. S. Rappaport et al., “Overview of millimeter wave communications
for fifth-generation (5G) wireless networks - with a focus on propaga-
tion models,” IEEE Transactions on Antennas and Propagation, vol. 65,
no. 12, pp. 6213–6230, Dec. 2017.
[73] F. Ademaj, M. K. Mueller, S. Schwarz, and M. Rupp, “Modeling of
spatially correlated geometry-based stochastic channels,” in 2017 IEEE
86th Vehicular Technology Conference (VTC-Fall), Sep. 2017, pp. 1–6.
[74] Z. Wang, H. Liu, S. Xu, X. Bu, and J. An, “A diffraction measurement
model and particle filter tracking method for RSS-based DFL,” IEEE
Journal on Selected Areas in Communications, vol. 33, no. 11, pp. 2391–
2403, Nov. 2015.
[75] Y. Wang, Z. Shi, L. Huang, Z. Yu, and C. Cao, “An extension of spatial
channel model with spatial consistency,” in 2016 IEEE 84th Vehicular
Technology Conference (VTC-Fall), Sep. 2016, pp. 1–5.
[76] 3GPP, “Discussion on procedure a for spatially-consistency ut mobil-
ity modeling,” ZTE, TDOC R1-1707267, May 2017.
[77] P. Kyösti et al., “WINNER II channel models,” European Commis-
sion, IST-WINNER, D1.1.2 V1.2, Feb. 2008. [Online]. Available: http:
/ / projects . celticinitiative . org / winner + /WINNER2 -
Deliverables/.

71
[78] G. R. MacCartney, Jr. and T. S. Rappaport, “A flexible millimeter-
wave channel sounder with absolute timing,” IEEE Journal on Selected
Areas in Communications, vol. 35, no. 6, pp. 1402–1418, Jun. 2017.
[79] G. R. MacCartney, Jr. et al., “Millimeter-wave human blockage at 73
GHz with a simple double knife-edge diffraction model and exten-
sion for directional antennas,” in 2016 IEEE 84th Vehicular Technology
Conference (VTC2016-Fall), Sep. 2016, pp. 1–6.
[80] G. R. MacCartney, Jr. and T. S. Rappaport, “Study on 3GPP rural
macrocell path loss models for millimeter wave wireless communica-
tions,” in 2017 IEEE International Conference on Communications (ICC),
May 2017, pp. 1–7.
[81] M. Jacob and T. Kürner, “Measurement campaign at 60 GHz in the
living room environment at TUBS,” doc.: IEEE 802.11-11-09-0743-00-
00ad r0, Jul. 2009.
[82] G. R. MacCartney et al., “Millimeter-wave base station diversity for
5g coordinated multipoint (comp) applications,” IEEE Transactions on
Wireless Communications, May 2019.
[83] Q. H. Abbasi, H. E. Sallabi, E. Serpedin, K. Qaraqe, and A. Alomainy,
“Condition number variability of ultra wideband MIMO on body
channels,” in 2016 International Workshop on Antenna Technology (iWAT),
Feb. 2016, pp. 167–169. [Online]. Available: http://ieeexplore.
ieee.org/document/7434833/.
[84] R. W. Heath and D. J. Love, “Multimode antenna selection for spatial
multiplexing systems with linear receivers,” IEEE Transactions on Sig-
nal Processing, vol. 53, no. 8, pp. 3042–3056, Aug. 2005. [Online]. Avail-
able: http://ieeexplore.ieee.org/document/1468498/.
[85] X. Lu, J. Wu, X. Huang, W. Li, J. Lu, and Z. Gong, “An improved
semi-orthogonal user selection algorithm based on condition number
for multiuser MIMO systems,” China Communications, vol. 11, no. 13,
pp. 23–30, Supplement 2014. [Online]. Available: http://ieeexplore.
ieee.org/document/7022522/.
[86] N. Bourbaki, Elements of Mathematics, Algebra I. Hermann 1974.
[87] M. Matthaiou, D. I. Laurenson, and C. X. Wang, “Reduced complex-
ity detection for ricean MIMO channels based on condition number
thresholding,” in 2008 International Wireless Communications and Mo-
bile Computing Conference, Aug. 2008, pp. 988–993. [Online]. Available:
http://ieeexplore.ieee.org/document/4600070/.
[88] S. Sun, T. S. Rappaport, R. W. Heath, A. Nix, and S. Rangan, “MIMO
for millimeter-wave wireless communications: Beamforming, spatial
multiplexing, or both?” IEEE Communications Magazine, vol. 52, no. 12,
pp. 110–121, Dec. 2014. [Online]. Available: http://ieeexplore.
ieee.org/document/6979962/.
[89] A. Adhikary, E. A. Safadi, M. K. Samimi, R. Wang, G. Caire, T. S. Rap-
paport, and A. F. Molisch, “Joint spatial division and multiplexing
for mm-wave channels,” IEEE Journal on Selected Areas in Communi-
cations, vol. 32, no. 6, pp. 1239–1255, Jun. 2014. [Online]. Available:
http://ieeexplore.ieee.org/document/6823686/.

72
[90] O. E. Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. Heath, “Spa-
tially sparse precoding in millimeter wave MIMO systems,” IEEE
Transactions on Wireless Communications, vol. 13, no. 3, pp. 1499–1513,
Mar. 2014. [Online]. Available: http://ieeexplore.ieee.org/
document/6717211/.
[91] R. B. Ertel, P. Cardieri, K. W. Sowerby, T. S. Rappaport, and J. H. Reed,
“Overview of spatial channel models for antenna array communica-
tion systems,” IEEE Personal Communications, vol. 5, no. 1, pp. 10–22,
Feb. 1998. [Online]. Available: http://ieeexplore.ieee.org/
document/656151/.
[92] B. Thoma, T. S. Rappaport, and M. D. Kietz, “Simulation of bit error
performance and outage probability of pi/4 DQPSK in frequency-
selective indoor radio channels using a measurement-based channel
model,” in IEEE Global Telecommunications Conference (GLOBECOM),
Dec. 1992, 1825–1829 vol.3. [Online]. Available: http://ieeexplore.
ieee.org/document/276698/.
[93] S. Sun et al., “Millimeter wave multi-beam antenna combining for 5G
cellular link improvement in New York City,” in 2014 IEEE Interna-
tional Conference on Communications (ICC), Jun. 2014, pp. 5468–5473.
[94] I. Yoo, M. F. Imani, T. Sleasman, H. D. Pfister, and D. R. Smith, “En-
hancing capacity of spatial multiplexing systems using reconfigurable
cavity-backed metasurface antennas in clustered mimo channels,” IEEE
Transactions on Communications, vol. 67, no. 2, pp. 1070–1084, 2019.

73