Christopher R. Anderson, [email protected] , 410-293-6185 Propagation Measurements and Modeling Techniques for 3.5 GHz Radar-LTE Spectrum Sharing NSMA 5/15/2019
Christopher R. Anderson, [email protected], 410-293-6185
Propagation Measurements and Modeling Techniques
for 3.5 GHz Radar-LTE Spectrum Sharing
NSMA 5/15/2019
Overview
Simulated and emulated
interference between SPN-43
and LTE systems.
Disclaimer: The opinions expressed in this presentation are those of the presenter only, and do not necessarily
reflect the views of the United States Naval Academy, Department of the Navy, Department of
Defense, National Telecommunications and Information Administration, Department of Commerce,
or United States Government.
Measured interference from the
SPN-43 to LTE sysetms.
Propagation measurements and
modeling for 3.5 GHz sharing.
For more information on these topics
1. C. R. Anderson and G. D. Durgin, "Propagation measurements and
modeling techniques for 3.5 GHz radar-LTE spectrum sharing," 2017
XXXIInd General Assembly and Scientific Symposium of the International Union
of Radio Science (URSI GASS), Montreal, QC, 2017, pp. 1-4.
2. J. H. Reed et al., "On the Co-Existence of TD-LTE and Radar Over 3.5
GHz Band: An Experimental Study," in IEEE Wireless Communications Letters,
vol. 5, no. 4, pp. 368-371, Aug. 2016.
3. R. J. Achatz, "Interference protection criteria simulation," 2018 IEEE Radar
Conference (RadarConf18), Oklahoma City, OK, 2018, pp. 0473-0477.
4. R. J. Achatz, B. Bedford, “Interference protection criteria simulation,” NTIA
Technical Report, expected June 2019.
3.5 GHz Spectrum sharing as currently
envisioned. A one-slide review.
Allow PAL and GAA access to the 3.5
GHz on a non-interference basis with
current incumbents.
Environmental Sensing Capability
(ESC) device used to detect the
presence of Navy Radars.
A Spectrum Access Server (SAS)
must be able to protect incumbents
from interference from PAL and
GAA users.
Ensure operation without degrading
performance ( “free-for-all” at 2.4
GHz).
Key: Reliable Propagation Model
PAL
GAA
ESC
SPN-43 Propagation Measurement Scenario
Radar located at Webster Field Annex in
St. Inigoes, MD.
Two buildings on either side (~2 stories)
and tall pedestal immediately behind.
Radar Height: 26 ft.
Antenna Uptilt: 3º.
July 10 & Oct. 30, 2014. Nominal weather
both days.
Limiter
Herotek
LS0140
Tektronix
SA2500
Bandpass
FilterLNA
Miteq
AFS3
Dipole
Antenna
Dipole mag-mounted on msmt. vehicle.
Amplifier and filter were required to
record weak observed signals (net gain
30.8 dB).
Measurement results and visualization
Compared Against
Log-Distance Path Loss (baseline)
Extended Hata
Irregular Terrain Model
TIREM
GIS - Attenuation Factor Model (new)
Refined attenuation factor model using St.
Mary’s County GIS data.
Elevation Contour Plot – 10m resolution Example Terrain Profile & Diffraction Loss
Model Inputs:
Digital Elevation Maps (30m)
Saturated exponential diffraction
NLCD 2011 Land Use Classifications
Endpoint clutter loss
Log-Distance Propagation
Empirically determined slope
Model Outputs:
Path Loss Exponent
Diffraction Map
Clutter Loss Map
Linear optimization to
determine model parameters.
Finding the attenuation factors.
If we use a least-squares formulation, then path loss for the kth link can be written as:
Given K total measurements, we can express this in matrix form:
is a K x I matrix, where I is the total number of attenuation factors in the model. Each
entry represents the number of each factor present for that link.
is a 1 x I matrix of the attenuation losses for the model, and is the unknown we are solving for.
Since is not usually easily invertible, we use the pseudoinverse to solve:
Using the technique presented in: G. D. Durgin, T. S. Rappaport, and H. Xu, “Measurements and models for radio path loss and penetration loss in and
around homes and trees at 5.85 GHz,” IEEE Transactions on Communications, Vol. 46, No. 11, Nov. 1998, pp. 1484-1496.
The result minimizes the mean square error of the system, not necessarily for any one
specific attenuation type. Will be differences when compared to individual link analyses.
( ) ( ) ( )00 10
1
10 logL
dDiff i id
i
PL d PL d n L =
= + + +
1 e
clear
crit
d
d
Diff dL A−
= −
With diffraction Loss:
The GIS Attenuation Factor based
propagation model.
Net Path Loss:
1 e
0.5
0.25
T
c
d
d
Diff T
T
c
L D
D dB
d m
− = −
=
=
Diffraction Loss
Clutter Loss: dB adjustment for RX Endpoint only.
Clutter Loss
( ) ( ) ( )00 10
1
10 logL
dDiff i id
i
PL d PL d n L =
= + + +
Note: Negative clutter loss does not imply a gain. It is simply less measured loss than
the baseline model predicted.
Numerical analysis of the accuracy of the
three major proposed propagation models.
Existing models are extremely poor at predicting RSS of our dataset.
However, they were not designed to meet the conditions of our scenario.
Log Distance anchor point at 80 meters at ground level.
eHata anchored at 1 km Avg. RSS, does not meet 30m TX height. Implemented
from NTIA TR-15-517, ported from Fortran.
ITM/TIREM from NRL Builder Implementation w/ theoretical antenna patterns.
GIS Model incorporates derived clutter losses.
( )0PL d
Comparison with model derived from ITS
3.5 GHz measurements in Denver and San Diego.
Base Propagation Model
Barren Rural Rural
Forest
Suburban Suburban
Forest
Urban Dense
Urban
µ
dB
µ
dB
µ
dB
µ
dB
µ
dB
µ
dB
µ
dB
ITM (Raw) N/A — — — — — —
ITM + Clutter N/A -1.7 — 11.9 — 5.2 4.5
Log Distance N/A -1.5 — 5.1 — 2.5 2.7
Free-Space + Diffraction N/A 17.2 — 21.2 — 19.3 20.4
Base Propagation Model
Mean Error RMSE PL Exp. ΔIDH
dB dB dB / m
ITM (Raw) -6.4 20.1 — —
ITM + Clutter 0.0 18.8 — —
Log Distance 4.5 13.6 2.53 0.094
Free-Space + Diffraction 3.2 12.9 2.0 0.079
ITS measurements result in an ~10 dB higher endpoint clutter loss than the
SPN-43 measurements.
Ldiff is ITU P.526 double knife-edge diffraction.
Investigating impact of radar pulse on LTE
Performance.
Site 1
Site 3,3A
Site 2
Setup 3 test links using a R&S EnB Emulator and handset.
Environments were clear, cluttered, and forested.
Distances were 1.2 – 4.0 km from the Radar.
-8.0 dBm UE transmit power to emulate femtocell links.
Swept LTE frequency to investigate effects of Radar pulse.
System Configuration
Site Distance
from Radar
(km)
Obstruction Radar Ant.
Elevation
(degrees)
Radar Power
Density
(dBm/cm2)
1 1.19 None +3 -17.1
2 3.98 Forest clutter +3 -84.6
3 1.26 Forest +3 -57.4
3A 1.26 Forest 0 -14.8
Downlink Uplink [1] Uplink [2,3] Uplink [3A]
Modulation 16 QAM QPSK QPSK QPSK
LTE Bandwidth (MHz) 10 10 10 10
# Res. Block 50 50 10 10
TX Power/Block (dBm/15 kHz) -42.8 -35.8 -29 -1
Total TX Power (dBm) -15.0 -8.0 -8.0 +20
RX Noise Floor (dBm) -126 -81 -81 -81
Results
Downlink BLER vs. Freq Offset
Downlink Throughput vs. Freq Offset
Uplink BLER vs. Freq Offset
Uplink Throughput vs. Freq Offset
Notes
Radar spectrum is asymmetric – causes asymmetric performance.
Site 2 and 3A are consistently the best performing, likely because of local geometry.
IPC emulation and simulations by NTIA/ITS.
Measurement setup and configuration.
Simulation of SPN-43 and LTE link operating simultaneously.
LTE configured in FDD mode, 10 MHz BW, 22.5 dB SNR, 50 Mbps.
SPN-43 operated such that and at baseline.
SPN-43 prf was set at exactly 1.0 kHz (pessimistic).
LTE was configured with 0 HARQ (vs. 8 in a nominal deployment).
Updated results in new NTIA Tech Report that more closely match a
nominal LTE configuration.
0.9dP =510faP −=
Throughput Performance and
comparison to our field test results.
Downlink
Throughput
vs. Freq Offset
Approximately the region where
our test was operating.
Of note: The ITS simulations (configured for worst-case) mostly agreed
with our measured results. The ITS lab measurements (interference to a
random RB) were more optimistic.
SPN-43 interference predictions.
For measured performance, 10
test targets were generated
along a radial for every
rotation. In 20 rotations, a
tester counts the number of
visible targets.
Stark contrast between the simulated Pd and Pfa and the measured Pd and Pfa.
Pd and Pfa were based on subjective judgment when a number of false alarms are
present on the screen.
Noticeable impact at -12 dB INR, 6 dB below the current -6 dB INR protection.
Observations
Field measurements, lab emulations, and simulations have demonstrated
the ability of LTE to co-exist with the SPN43.
The current “classic” models (e.g., very tall transmitter, very large area
coverage, path over land) are not a good match for small-cell 3.5 GHz
deployments.
The measured data we’ve analyzed thus far demonstrates a site-specific
nature to clutter losses along with a potentially high variability.
However, I would advocate against “dump everything into a deep learning
network and let it figure out the rest.” There’s still a need for a physics
foundation and knowledgeable propagation engineers.
Conclusions
Accurate, Reliable, Robust propagation models are vitally
important to the envisioned operation of spectrum sharing
systems.
A simple GIS-based Attenuation Factor model produced a 3-
20 dB improvement in RMSE vs. current models used in the
band.
IPC evaluation of LTE and SPN-43 under simulation,
emulation, and field measurements appear to agree – but
suggest careful selection of the IPC criteria.
Questions?
Backup Slides
Overview of the measurement system for a
new measurements-based model.
Signal Generator
Power Amplifier
Rubidium
Oscillator
Signal
Analyzer
GPS Receiver
& Signal monitor
Rubidium
Oscillator
RF
out
RF
outRF
In
Transmitting
Antenna
RF
In
Receiving
AntennaGPS
Antenna
RF In
TX Antenna installed on
18m extensible mast.
RX Antenna installed on
measurement vehicle,
3m above ground level.
CW Tone, 1755 or 3550 MHz
TX power: +47 dBm.
TX Gain: 7.7 dBi.
RX Gain: 2.2 dBi.
RX MDS: -120 dBm.
Max Distance: 18 km
Record raw I/Q samples as
time series.
Average over 1.0 sec to
eliminate small-scale fading.
Calibrated on NIST open-air
test site.
Antenna pattern effects are
extracted from measured RSS.
Detailed description of measurement system,
location, and procedures
Limiter
Herotek
LS0140
Tektronix
SA2500
Bandpass
FilterLNA
Miteq
AFS3
Dipole
Antenna
Simple dipole mag-mounted on roof of
measurement van.
Amplifier and filter were required to
record weak observed signals (net gain
30.8 dB).
Receiver in Amplitude vs. Time mode.
Each msmt. captured ~20 msec of data in
Max Hold mode with GPS coords.
Both stationary and moving (<55 MPH).
Raw data was postprocessed via Python
scripts, then loaded into Matlab for
analysis and visualization.
Antenna pattern translated to Matlab
from measured data.
SPN-43 Propagation Measurement Scenario
SPN-43 Operating Parameters
Tuning Range 3500 – 3650 MHz
Pulse Repetition Rate 889 (±20) µs
Pulse Width 0.9 (±0.15) µs
TX Power (Max) 850 (±150) kW
Bandwidth 1.3 (±0.3) MHz
Antenna Gain 32 dBi
Polarization Horizontal or Circular
Beamwitdth (3 dB) 3º
Rotation Rate 15 RPM (4 sec)
Radar located at Webster Field Annex in
St. Inigoes, MD.
Two buildings on either side (~2 stories)
and tall pedestal immediately behind.
Radar Height: 26 ft.
Antenna Uptilt: 3º.
July 10 & Oct. 30, 2014. Nominal weather
both days.
100m Digital Elevation Map (m) Terrain Blockage Distance (m)
Saturated exponential diffraction model using
digital elevation maps.
1 e
clear
crit
d
d
Diff dL A−
= −
Diff
d
clear
crit
L
A
d
d
Total Diffraction Loss (dB)
Total Diffraction loss (dB) [1.48 dB]
Clearance distance below terrain (m)
Critical distance (m) [2.0 m]
cleard
TX
Illustration of the matrix computation
Inte
rdeci
le H
eig
hts
Rura
l
Rura
l Fore
st
Suburb
an F
ore
st
Urb
an
Dense
Urb
an
Suburb
an
Solving minimizes the mean square error with respect to
the difference between measurements and models.
Can sometimes be instructive to illustrate which attenuation factors have the
biggest impact on the propagation loss.
Technique can be used for model tuning and optimization if attenuation loss
for a particular category is fixed a priori.
To understand clutter “gain”, consider first
Two-Ray Propagation.
Consider the following scenario with grazing angle conditions:20 t rh h
d
th
rh
d
The LOS and Reflected signal will combine destructively at the receiver.
Path Loss will follow the “Two-Ray Model” where:
Note 1: Essentially Log-Distance with n = 4.0.
( ) ( )1040log Other TermsPL d d= +
What might be causing negative clutter loss?
Hypothesis based on Two-Ray model.
Hypothesis 1: Direct path is blocked, but lateral wave propagates over the canopy. Results
provides enhancement relative to 2-Ray (n = 4) Propagation.
Hypothesis 2: Direct path is blocked by foliage/canopy, with no lateral wave. Propagation is only
through ground reflection; no destructive interference as in 2-Ray model.
First: What’s causing the RSS spread? A
closer look at the < 1 km distance.
Low height of Radar +
surrounding building/forest
leads to diffractive and
shadowing losses at
certain AoA/AoD.
Clear sight line to Radar
leads to very high RSS.
Result is a very high
spread of RSS inside the
first ~ 1.0 km radius.