AIR TRAFFIC ORGANIZATION
Future Communications Study Technology
Assessment Team: Outcome of Detailed
Technology Investigations
Presented at ICAO ACP WGC Meeting,
Brussels, Belgium
September 19, 2006
Prepared by:
ITT/Glen Dyer, Tricia Gilbert
NASA/James Budinger
Briefing Outline
• Overview
• L-Band Modeling
–
–
–
–
–
L-Band Channel Modeling
L-Band Cost Modeling
P34 Modeling
LDL Modeling
Interference Modeling
• SATCOM Availability Modeling
• C-Band Modeling
2
Overview
• Detailed analysis of all the short listed technologies
against all of the evaluation criteria is prohibitively
expensive
• In general, each technology has an area of concern that
warrants detailed investigation
– Focus of L-Band investigations was to
• Define a channel model that could be used for common
characterization of waveform performance in A/G channel
• Define a framework for specifying the infrastructure costs associated
with an L-Band system
• Analyze recommended technologies (P34 and LDL) performance with
common channel model and potential to interfere with incumbent
users of the band
– Focus of Satellite Modeling was availability
– Focus of C-Band Modeling was airport surface performance
3
L-band Channel Modeling
•
•
–
σ0
Tx
1
rSR1
rTR
Mountain 1
Rx
rSR2
rTS2
rTSk
rSRk
dA2
σ02
dAk
σ0k
Mountain 2
Mountain k
106°30’ W
•
dA1
rTS1
107°30’ W
•
A literature search revealed that while
many channel models exist for the
terrestrial channel in close proximity to
L-Band, there had been no previous
activity to develop a channel model that
characterizes the L-Band A/G channel.
Most standardization bodies consider it
best practice to test candidate
waveform designs against carefully
crafted channel models that are
representative of the intended user
environment
As a consequence of these
considerations, a simulation was
developed to characterize the A/G
channel at L-Band
For modeling purposes, a severe
channel (from a delay spread
perspective) was considered
39°30’ N
+RCAG
Figures show the model context
38°45’ N
4
L-Band Channel Modeling
Methodology Overview
•
Methodology used for generating
power delay profiles:
– A series of concentric oblate
spheroids was generated using the
Tx & Rx locations as the focal
points
• The semi-minor axis for each
successive spheroid was increased
by a fixed increment
– The contour of terrain trapped
between two successive spheroids
was used to calculate multipath
dispersion for a particular time
delay
• Each contour consisted of a set of
terrain points that represented
potential scatterers
• Ray-tracing was used to determine
Specular and diffuse multipath
5
L-Band Channel Modeling
Methodology Details
6
L-Band Channel Modeling –
Suggested Channel Model
• Specified model for a terminal area is shown in table
• Extension to larger distance can be found using:
   d e A
0
– where e = 0.6337, στ0= 0.1 μs and A = 6 dB
Tap #
Delay (µs)
Power (lin)
Power (dB)
Fading
Process
Doppler
Category
1
0
1
0
Ricean
Jakes
2
1.6
0.0359
-14.5
Rayleigh
Jakes
3
3.2
0.0451
-13.5
Rayleigh
Jakes
4
4.8
0.0689
-11.6
Rayleigh
Jakes
5
6.4
0.0815
-10.9
Rayleigh
Jakes
6
8.0
0.0594
-12.2
Rayleigh
Jakes
7
9.6
0.0766
-11.2
Rayleigh
Jakes
7
L-Band Channel Modeling –
Predicted RMS Delay Spreads
•
•
•

RMS

RMS

RMS
= 0.1 μs for average 1 km distance from transmitter
in mountainous terrain (simulated)
= 1.4 μs for average 64 km distance from transmitter
in mountainous terrain (simulated)
= 2.5 μs for 160 km aircraft-tower separation
distance (extrapolated)
8
L-Band Cost Modeling – Process for
Determining Service Provider Cost
Derive number of required radio sites
Develop link
budget
Infer
communication
distance
Derive required
radio sites
for US coverage
No
Develop
radio site
configuration
Determine the
availability
Meets
requirements?
Yes
Specify radio site
architecture
Derive
required equipment
per radio site
Other costs
(e.g cost of telco)
Derive
Deployment Costs
L-Band Cost
Estimating
Process
9
L-Band Cost Modeling – Rules &
Assumptions
• Assumptions
– L-Band system provides coverage to either the continental Unites States
or to core Europe
– Coverage is above FL 180
– System Availability of Provision meets COCR requirements for Phase II
En-route services (sans Auto-Execute)
– Cost elements considered are
• Research and Development
– System Design and Engineering
• Investment
– Facilities
– Equipment
• Operations and Maintenance
– Telecommunications
– Other costs (personnel, utilities, etc.)
10
P34 Modeling – OPNET Simulation
Configuration that
was simulated was
the fixed-network
equipment (FNE) to
mobile radio (MR).
The MR to MR and
repeater modes
were not simulated.
The modeled
configuration aligns
with the P34
“concept of use”.
The custom OPNET
development
included modeling of
the P34 PHY, MAC,
LLC and SN Layers.
11
P34 Modeling – OPNET Results
•
•
The figures show the response time of
the P34 simulation to the offered load
for each of the transmitted messages
It seems that sub-network latencies
over P34 protocols (SNDCP, LLC CP,
LLC UP, MAC) meet COCR latency
requirements
–
Note outliers
Some startup outliers, but 95% is under
0.7 seconds
12
P34 Modeling – Validation of
Receiver Model
•
•
The P34 Scaleable Adaptive
Modulation (SAM) physical layer
interface was modeled by
developing a custom application
using C code
The transmitter was implemented
as detailed in the specification for
the 50 kHz channel using QPSK
modulation
The receiver implementation was
tested against known results
– Top figure is from Annex A of
TIA-902.BAAB-A
– Bottom figure shows simulation
results for AWGN and the HT200
channel model
QPSK BER
100
10
1
BER (%)
•
0.1
HT200
AWGN
0.01
0.001
0
5
10
15
20
25
30
35
40
45
50
Es/No (dB)
13
P34 Modeling – Investigation of
Coding Gain
•
From the previous results, it was
unclear if satisfactory performance
was being achieved in the mobile
fading channel
– Needed to know what a raw BER
of 3*10-3 translated to after coding
•
P34 SAM uses a system of
concatenated Hamming codes.
The basic scheme is shown in the
top figure
– Simulated the rate ½ coding by
concatenating two Hamming
coders and a block interleaver
•
Coding gain is shown in bottom
figure
– 3*10-3 raw BER is approximately
10-5 coded BER
14
P34 Modeling – Predicted
Performance
•
The A/G channel was simulated
using a two tap model
– Tap 1 was modeled as Rician, with a
K-factor of 18 dB, unity gain, Jakes
Doppler Spectrum
– Tap 2 was modeled as Rayleigh,
with a 4.8 s delay, -18 dB average
energy, Jakes Doppler
•
• Initial simulations indicate good performance
can be achieved in the aeronautical channel
(primarily a consequence of the strong LOS
component of the received signal)
• These are initial results and are still being
validated
The mobile velocity was taken to be
0.88 mach
– COCR gives this as the maximum
domestic airspeed based on Boeing
777 maximum speed of 0.88 mach
•
P34 tuned frequency was taken to
be 1024 MHz
– Maximum Doppler shift - 1022 Hz
•
The predicted P34 performance is
quite good for K factors greater than
four
15
LDL Modeling – Validation of
Receiver Model
To validate simulation,
compare simulation
results with theory
– The theoretical curve is
the performance of
binary CPFSK with
coherent detection using
n = 5, and
h = 0.715 [Proakis]
– Model uses the same
traceback length
(n = 5) and modulation
index
(h = 0.715)
Theory vs. Simulation
1
0.1
BER
•
Using a modulation of 0.715
minimizes probability of error for
binary CPFSK [Schonhoff 1976]
0.01
0.001
0.0001
0.00001
0
2
4
6
8
10
12
14
16
SNR (dB)
Theory
Simulation
16
LDL Modeling – Investigation of
Coding Gain
– Changing the
modulation index
from 0.715 to 0.6
pushes the BER
curve out ~1 dB
– The Reed-Solomon
(72,62) code
provides a coding
gain of 3-4 dB in the
expected region of
operation
BER for h=0.6 & RS Coding
1
0.1
0.01
BER
• A modulation index of
0.715 was required to
validate the model with
published results, but
LDL calls for a
modulation index of
0.6
In order for the RS code to provide a substantial coding
gain, the raw BER must be less than 10-2 and ideally, it
should be less than 2*10-3
0.001
0.0001
0.00001
0.000001
0
2
4
6
8
10
12
14
16
SNR (dB)
Sim (h=0.6)
Sim w/RS (h=0.6)
17
LDL Modeling – Predicted
Performance
• The plot below shows the system
performance of LDL in the presence both
AWGN and the L-Band Channel Model
Non-Coherent (Limiter/Discriminator) CPFSK
1
0.1
Theory (Coherent)
0.01
BER
• The LDL channel
model is a conservative
model that introduces
an irreducible error
floor to system
performance
• Based on the results of
this model, LDL will
require channel
equalization to mitigate
the effects of the
Air/Ground
Aeronautical Channel
in L-Band
DISC-LIM/AWGN
DISC-LIM/AWGN/Rayleigh
0.001
DISC-LIM/AWGN/L-Band Channel
0.0001
0.00001
0
2
4
6
8
10
12
14
16
18
20
SNR (dB)
18
Interference Modeling UAT
Performance
1.00E-01
•
•
1.00E-02
Probabilty of BER
•
The top chart provides a collection
of BER curves for varying degrees
of LDL Interference into UAT signal
The bottom chart provides a
collection of BER curves for
varying degrees of P34
Interference into UAT signals
From the curves, it would appear
that a C/I ratio between 12 and 15
dB is required for minimum
degradation to the UAT receiver
LDL has slightly better
performance than P34 in terms of
not interfering with UAT receivers
UAT without Interference
C/I = 5 dB
1.00E-03
C/I = 8 dB
C/I = 10 dB
C/I = 12 dB
C/I = 15 dB
1.00E-04
1.00E-05
10
11
12
13
14
15
16
17
18
19
20
Eb/No, dB
1.00E-01
UAT without Interference
C/I = 7 dB
C/I = 8 dB
C/I = 10 dB
1.00E-02
C/I = 12 dB
Probabilty of BER
•
C/I = 15 dB
1.00E-03
1.00E-04
1.00E-05
10
11
12
13
14
15
16
17
18
19
20
Eb/No, dB
19
Interference Modeling Mode S
Performance
Probability of correct preamble
detection curves
1.05
– Based on an algorithmic
assumption to declare preamble
detection of
94% correlation
100% correlation
•
Probabilty of Correct Preamble Detection
•
Probability of false preamble
detection curves
1
0.95
P34 Interferer (C/No = 73 dB)
LDL Interferer (C/No = 73 dB)
0.9
Note:
Conversion from C/No to C/N within necessary
bandwidth can be done as follows:
C/N within necessary bandwidth =
C/No+10log10(2)-10*log10(4000000)
0.85
0.8
5
6
7
8
9
10
11
12
13
14
15
C/I, dB
1
1.05
LDL Interferer (C/No = 73 dB)
Probabilty of Correct Preamble Detection
Probabilty of False Preamble Detection
P34 Interferer (C/No = 73 dB)
0.1
Note:
Conversion from C/No to C/N within necessary
bandwidth can be done as follows:
C/N within necessary bandwidth =
C/No+10log10(2)-10*log10(4000000)
0.01
0.001
1
0.95
P34 Interferer (C/No = 77 dB)
LDL Interferer (C/No = 77 dB)
0.9
Note:
Conversion from C/No to C/N within
necessary bandwidth can be done as follows:
C/N within necessary bandwidth =
C/No+10log10(2)-10*log10(4000000)
0.85
0.8
0.0001
5
6
7
8
9
10
C/I, dB
11
12
13
14
15
5
6
7
8
9
10
11
12
13
14
15
C/I, dB
20
SATCOM Availability Modeling
Overview
• Two satellite service architectures with AMS(R)S
frequency allocations were selected for consideration in
this availability analysis
– Inmarsat-4 SwiftBroadband service
– Iridium communication service
• Calculated availability of these architectures was
contrasted with the calculated availability of a generic
VHF terrestrial communication architecture
– Data communications architecture based on existing infrastructure
21
SATCOM Availability Modeling
Approach
• Utilized SATCOM availability
analysis model described in
RTCA DO-270
– Defines availability fault-tree to
permit individual
characterization and evaluation
of multiple availability
elements
– Organized into two major
categories
• System Component Failures
• Fault-Free Rare Events
– Model is useful for comparing
architectures and was used for
this study
Communications
Unavailable for >TOD
System Component
Failures
OR
Fault-Free Rare
Events
OR
Ground Station
Equipment
Failure Event
Satellite
Control
Equipment
Failure Event
RF Link Event
Capacity
Overlaod
Event
Aircraft
Station
Failure Event
Interference
Event
Satellite
Failure
Event
Scintillation
Event
22
SATCOM Availability Modeling
Summary Results
•
Summary –
– Limiting factors for availability are as follows:
• SATCOM systems:
– Satellite equipment failures and RF link effects
– Capacity Overload (Iridium)
– Interference (Iridium)
• VHF Terrestrial communication systems:
– RF link events
System Component Failures
Ground Control Aircraft Satellite
Station Station Station
~1
~1
~1
0.9999
0.99997
~1
~1
0.99
0.99999
N/A
~1
N/A
RF
Link
0.95
0.995
0.999
Fault-Free Rare Events
Capacity
Interference Scintillation
Overload
~1
~1
~1
1
0.996
~1
-2
~1
N/A
Inmarsat
Iridium
VHF
Terrestrial
Notes:
1. Iridium Capacity Overload availability of AES to SATCOM traffic is essentially one (1) (for both ATS
only and ATS & AOC). No steady-state can be achieved for SATCOM to AES traffic.
2. Terrestrial Capacity Overload availability is for VHF-Band reference architecture business case; for LBand Terrestrial Capacity Overload availability would be essentially one (1).
23
C-Band Modeling – 802.16e
Transmitter Model
PN Sequence
Generator
Reed Solomon Coding
PN Sequence
Generator
Full_BW_TestVector
Read in Data from
MATLAB WS
Bit to Integer
Converter
XOR
Integer to Bit
Converter
Data
Randomizer
Convert Integers
to Bits
Zero Pad
Convert Bits
to Bytes
RS Encoder
Zero Pad to
Code Word Size
RS Encode
These blocks model the data randomization
process
U U(E)
Puncture Code
Integer to Bit
Converter
Convert Bytes
to Bits
Modulator
Convolutional
Encoder
Convolutional
Coding
••
••
Puncture
Puncture
General
Block
Interleaver
Bit to Integer
Converter
General Block
Interleaver
Bit to Integer
Converter
General
QAM
General QAM
Modulator
Create OFDM
Symbols
Create OFDM
Symbols
TxSignal
Subcarrier Mapping
(as shown on p. 444
of specification)
Model Info
This
Created by: Glen Dyer
Thisisisthe
thedeveloped
developedmodel
modelfor
forthe
the802.16
802.16OFDM
OFDMTransmitter
Transmitter
Created
date: Sun Mar 19 14:11:35 2006
The
802.16
standard
defines
the
following
elements
for
OFDM
transmitter
implementation
TheDOC
802.16 standard defines the following elements for OFDM transmitter
implementation
Modified
by: dye27622
Modified date: Sat Jun 10 15:38:27 2006
• • Bit
BitScrambling
Scrambling
Model Version Number: 1.6
Text
• • Concatenated
ConcatenatedPunctured
PuncturedReed-Solomon
Reed-Solomonand
andPunctured
PuncturedConvolutional
ConvolutionalEncoding
Encoding
• • Bit
Interleaving
Bit Interleaving
• • Adaptive
AdaptiveModulation
Modulation
• • OFDM
Symbol
OFDM SymbolCreation
Creation
24
C-Band Modeling – 802.16e
Receiver Model
IEEE 802.16 OFDM
16-QAM Modulation
Rate 1/2 Concatenated Coding
Extract Data
Symbols
Data
Extract Data
from OFDM Symbol
16QAM
Demodulator
General
Block
Deinterleaver
Demodulate
Deinterleave
Reed Solomon De-coding
XOR
Integer to Bit
Converter
U(E) U
Selector
U(E) U
Convert to Bits
Select Info Bytes
Insert Zero
Err RS Decoder
Unipolar to
Bipolar
Converter
Decoder expects
Un-do Convolutional
ones and minus ones
Puncturing
Integer-Output
RS Decoder
Terminator
-376
z
Delay
PN Sequence
Generator
Un-do Randomization
Zero Pad1
These blocks invert the data randomization
process
••
••
U(E) U
Bit to Integer
Converter
Re-order
Bytes
Convert Bits to
Bytes
-478
z
Delay - Compensate for
Viterbi Decoding
Viterbi Decoder
Decoder inserts delay of 34
Model Info
Receiver
This
Dyer
by: Glen
Created
Receiver
OFDM
802.16OFDM
the802.16
forthe
modelfor
developedmodel
thedeveloped
Thisisisthe
2006
16:43:33
Sat Jun 10that
date:
Created
are
operations
the
invert
must
implementation
receiver
The
by: dye27622 that are
Modified
operations
The receiver implementation must invert the
Modified date: Sat Jun 10 17:05:10 2006
defined
the Model Version Number: 1.0
includingthe
transmitter,including
thetransmitter,
forthe
definedfor
• • Bit
Scrambling
BitScrambling
• • Concatenated
Punctured
andPunctured
Solomanand
ReedSoloman
PuncturedReed
ConcatenatedPunctured
Convolutional
Encoding
ConvolutionalEncoding
• • Bit
Interleaving
BitInterleaving
• • Adaptive
Modulation
AdaptiveModulation
Symbol
• • OFDM
Creation
OFDM SymbolCreation
25
C-Band Modeling – Model Validation
802.16 OFDM 16-QAM Modulation Simulated BER Performance
••
••
••
The
exercised
wasexercised
simulationwas
developedsimulation
Thedeveloped
to
compared
and
AWGN
against
against AWGN and compared to
published
purposes
validationpurposes
forvalidation
resultfor
publishedresult
(uncoded)
raw
the
shows
slide
This
BER
This slide shows the raw (uncoded)BER
against
simulation
our
of
performance
performance of our simulation against
theoretical
results
theoreticalresults
contrast,
For
the
senseofofthe
getaasense
andtotoget
For contrast,and
achieved
the
afterthe
BERafter
theBER
gain,the
codinggain,
achievedcoding
is
decoding
Solomon
Reed
and
Viterbi
Viterbi and Reed Solomon decoding is
also
shown
alsoshown
26
C-Band Modeling – Results
••
••
••
Finally,
Finally,an
anapproximation
approximationtotothe
theOhio
Ohio
University
suggested
airport
channel
University suggested airport channel
models
modelswas
wasmade,
made,and
and802.16
802.16was
was
evaluated
against
this
model
evaluated against this model
The
Thechannel
channelmodel
modelwas
wasfor
foraalarge
large
airport
in
the
Non-LOS
region
airport in the Non-LOS region
The
Thecurves
curvesshow
showexpected
expectedperformance
performance
for
various
maximum
Doppler
for various maximum Dopplershifts,
shifts,and
and
represent
802.16
performance
from
a
represent 802.16 performance from a
virtual
virtualstandstill
standstillthrough
throughexpected
expected
velocities
in
the
movement
velocities in the movementarea
area
27
Action Request
• The ACP Working Group is invited to consider the
technology investigation activities described in this paper,
and provide comments if desired
• It is recommended that the ACP Working Group consider
the A/G channel model that is presented in this paper and
adopt it for the evaluation of candidate technologies for
the Future Radio System
• It is recommended that the ACP Working Group consider
the cost modeling approach that is presented in this paper
and adopt it for the evaluation of candidate technologies
for the Future Radio System
28