APPENDIX D
TECHNICAL SIMULATIONS
Document control
Intentionally blank
APPENDIX D
TECHNICAL SIMULATIONS
Table of contents
D.1 INTRODUCTION 1
D.2 STDMA/VDL MODE 4 PERFORMANCE SIMULATOR (SPS) 1
D.2.1 SPS overview 1
D.2.2 Simulation inputs and outputs 2
D.2.3 Modelling of physical layer effects 3
D.2.4 Operational scenarios modeled with SPS 3
D.2.5 Supplementary information 4
D.3 SIMULATIONS 4
D.3.1 VDL mode 4 performance in support of ADS-B in
high-density scenarios 4
D.3.1.1 Scenario descriptions 5
D.3.1.2 Technical assumptions 5
D.3.1.3 Results 6
D.3.1.3.1 Slot utilization 6
• Core European Scenario 6
• LA Basin Scenario 6
• Extended LA Basin scenario 7
D.3.1.3.2 MASPS requirements 7
• Effective update period and acquisition range 7
• ATC surveillance 10
D.3.1.3.3 Slot re-use ratio 12
D.3.1.3.4 Channel loading characteristics 14
• Optimal channel loading 14
• Robustness to overload and graceful degradation 14
• Effectiveness of the slot selection algorithm 16
D.3.2 SPS simulations of timing errors 16
D.3.2.1 Introduction 16
D.3.2.2 Technical context and conclusions 16
REFERENCES 19
Intentionally blank
D.1 INTRODUCTION
The STDMA/VDL Mode 4 Performance Simulator (SPS) has been used extensively
in the development and validation of VDL Mode 4 SARPs and in various international
projects to investigate the performance of VDL Mode 4 in support of ADS-B. For
instance, the SPS simulation tool has been used to;
• demonstrate compliance with VDL Mode 4 SARPs;
• analyse system behaviour under certain conditions such as signal
interference from CCI and ACI and the hidden terminal situation (see
Chapter 5 Section 5.9.3.1);
• analyse the effects of timing errors on link synchronisation and slot
separation;
• determine the expected performance with regard to message throughput
and error rates under various channel loadings;
• demonstrate compliance with ADS-B MASPS requirements;
• determine channel requirements to support ADS-B in different traffic
scenarios.
Technical simulations do not obviate the need for real world trials and
demonstrations. However, a simulation tool is fundamental for analysing system
behaviour without actually building the system, and allows alternative solutions to
be implemented with limited software modifications. Obviously, system behaviour
and performance in future traffic scenarios in which a large number of transmitters
are present can only be tested by simulations. It is not always possible to model a
full set of VDL Mode 4 and scenario properties, but with careful modelling adapted to
the particular objectives it is possible to arrive at conclusive and representative
results.
The Periodic Broadcast Reservation is the main protocol used for autonomous ADS-
B operations and accurately models the steady-state performance of VDL Mode 4 for
ADS-B. Only the Periodic Broadcast Reservation protocol was used in the
simulations outlined in this appendix.
The SPS was developed by the Swedish CAA with the objective of validating the
fundamental data link protocols and slot allocation features of STDMA. For the
purpose of enabling the SPS to be used more extensively in the validation of SARPs,
the EUROCONTROL Experimental Centre (EEC) developed an “Enhanced SPS” [3],
which supports incremental broadcast and random access protocols on top of the
periodic broadcast protocol supported by SPS1. EUROCONTROL’s main objective
was to gain confidence in the SPS as an appropriate tool to measure performance
and capacity of VDL Mode 4. A series of simulations were done using the enhanced
SPS under a variety of scenarios to exercise the newly added features, and to
compare with the standard SPS. The results were verified by theoretical analysis.
EUROCONTROL also made a thorough review of the SPS source code, and
identified the SARPs paragraphs which have been implemented in the SPS, those
which are not supported, and those where there are interpretation ambiguities.
• ADS-B (sync bursts) report period. This can be different for each user
(aircraft or ground station) and altered during the simulation. Each aircraft
may be specified to operate in autonomous or ground controlled mode.
• Slot reservation parameters like dither range, pre-announce times, slot
reservation hold times, and slot re-use ("Robin-Hood") related parameters.
These are common for all users in the simulation.
• Slots per superframe, e.g. 4500 for 19.2 kbps radios. This is common for
all users in the simulation.
• Number of quarantined slots available for ground stations and aircraft
under ground-control.
• Aircraft data, i.e. the total number, their positions and altitudes, and their
trajectories. Aircraft may also be given randomly generated tracks, top
speed, maximum altitudes etc. Aircraft dynamics are not modelled, e.g.
turns are instantaneous.
• Ground station data, i.e. the total number and their positions. Ground
stations can be distributed in the scenario randomly or specifically. For
each ground station, the antenna height over sea level and position
reporting on/off can be set.
No terrain model is used, however earth curvature is taken into account for
transmission propagation calculations.
SPS does not accurately reflect the situation during the entry of new users (net
entry). The scenarios presented in this appendix, therefore, are based on the steady
state performance of VDL Mode 4 ADS-B.
1
EUROCONTROL has supported the development and validation process of the VDL Mode
4 SARPs and is considering the use of VDL Mode 4 for ADS-B applications, such as
air/ground surveillance, airborne separation assurance and airport surface movement control.
SPS generates results in a number of on-line plots and output files, for example:
• General log file;
• Kinematic log file;
• Transfer probability matrix;
• Transmission log file;
• Specific user performance;
• Histogram output file.
A fully implemented VDL Mode 4 ADS-B system would probably use additional
channels to provide more services. The configuration could for example include:
Only one VHF channel is modelled in the simulator. For an operational VDL Mode 4
ADS-B system, it is intended to use two channels for basic ADS-B. Each aircraft
reports at half the nominal rate on each channel. The impact of using two channels
can be determined with the simulator by modelling one channel at a time. In most
cases, the channels will have very similar performance.
D.3 SIMULATIONS
This section summarises important technical simulations using the SPS tool. The
conclusions described in this appendix only represent a subset of the full set of
results described in the source documents. They should be seen as illustrative
examples of what has been achieved by technical simulations.
A similar set of simulations was conducted within the TLAT (Technical Link
Assessment Team) activity in the Safe Flight 21 project for the purpose of comparing
the VDL Mode 4, Universal Access Transceiver (UAT) and Mode S data links. The
scenarios used in these simulations were developed to replicate projected traffic
volumes in the 2015-2020 timeframe and make use of further extensions to both the
LA (~2700 a/c) and the Core European (XCE with ~1940 a/c) scenarios. The
massive (and possibly unrealistic) traffic volumes in the TLAT scenarios called for
additional ADS-B channels compared to the two Global Signalling Channels capable
of supporting ADS-B in the simulations described below. The TLAT simulations are
not further addressed in this appendix. Reference is made to Appendix B to this
chapter in which the use of both Local Signalling Channels (LSC) and GSCs in
support of ADS-B and other data link applications is described.
The European scenario is different from the LA Basin scenarios. Whereas a very
high percentage of the traffic in the LA Basin is General Aviation (GA), the highest
density European areas generally contain more air transport (AT) aircraft. Because of
this, the altitude distribution is also different, with more aircraft at high altitudes in the
European scenario. Moreover, the traffic is spread over a much larger area in the
European scenario. This means that this scenario tests the re-use of slots in a wide
area. Slot re-use is an important aspect of VDL Mode 4.
D.3.13 Results
D.3.1.3.1 Slot utilisation
The slot utilisation obtained with the three different scenarios is summarised in Table
D-1.
Table D-1 Slot utilisation for the three scenarios
Note that ground station transmissions are included in the results presented in Table
D-1 (occupying approximately 1% of transmissions).
According to VDL Mode 4 validation documentation, update periods of 5 s in TMAs
and 10 s in en-route airspace shall be maintained with a 98% success probability per
message.
The percentage transmission in slots with a single message is not the same as the
probability of success per message but may be linked to it. It is important to
distinguish between slot re-use and slot collision for VDL Mode 4, i.e. between
intentional slot re-use, and unintentional slot reuse. Slot re-use does not necessarily
mean that a message is “lost”.
Where the slot re-use is small at 1 or 2%, and channel load is below 100%, as is the
case for the scenario simulation results in Table D-1, the slot re-use is likely to be
intentional and subjected to discrimination in most cases. Thus figures of 98 or 99%
of messages transmitted in slots with a single message mean that the VPCD
requirement of a 98% probability of success per message will be met whenever
physically possible.
2
Per GSC.
LA Basin Scenario
99.8% of messages are transmitted in slots with a single message. Thus again, a
98% probability of success per message will be achieved whenever physically
possible, and therefore this VPCD requirement is met for the LA Basin scenario.
40
35
Aid to Visual Conflict Separation Flight Path Deconfliction
30 Acquisition Aviodance and Assurance and Planning
Collision Sequencing (Head-on)
Avoidance
Update period (s)
25
24 24 24
MASPS 99th percentile
20
15
14
12 12 12
10 10
MASPS 95th percentile
7
6
5 5
3
0
3 10 20 40 90 120
Distance (nmi)
It should be noted that for flight path deconfliction planning, the 90 NM range is a
requirement while the 120 NM range is a “desirable” requirement only.
The MASPS update period and acquisition range requirements basically only apply
to the LA Basin scenario. Hence only a comparison of the MASPS requirements and
the simulation results with regard to effective update periods in the “basic” LA Basin
scenario are presented below. However, comments are provided on a comparison of
the MASPS requirements with the simulation results from the Extended LA Basin and
Core Europe scenarios. [1] provides full details.
In Figure D-3, the effective update periods at the 95th percentile obtained from LA
Basin scenario simulations with 5 and 7.5 s nominal update rates are plotted against
the MASPS 95% requirements. It should be noted that this scenario only extends up
to a radius of 60 NM and therefore data in the plots only applies to user pair
distances up to approximately 110 NM (reached by some aircraft after 30 minutes
slow flying).
12
Effective Update Period (s)
10
0
0 20 40 60 80 100 120 140
Distance (nmi)
The plot is the accumulated results for all messages and all users. All potential user
pairs are used to calculate a set of delivery probabilities per message, and stored in
the appropriate range bin (e.g. the 25 km range bin) for every message.
It can be seen from Figure D-3 that the MASPS 95% requirements are met for the LA
Basin scenario at distances greater than approximately 24 NM.
In Figure D-4, the effective update periods at the 99th percentile obtained from LA
Basin scenario simulations with 5 and 7.5 s nominal update rates are plotted against
the MASPS 99% requirements.
24
Effective Update Period (s)
12
0
0 20 40 60 80 100 120 140
Distance (nmi)
For the purpose of meeting the MASPS requirements it is sufficient to use the result
for the simulation performed at 66% channel loading, shown by the lower curve in
Figure D-4. This shows that the MASPS 99% requirements are met for the LA Basin
scenario at distances greater than approximately 5 NM.
When applied to the Extended LA Basin scenario, the MASPS 95% requirement was
met at distances greater than approximately 32 NM, while the more important 99%
requirement was met at distances greater than 15 NM. It was also concluded that the
delivery probability was very good with a data link load around 85%, that the
acquisition range considerably exceeds the range requirement in the ADS-B MASPS,
and that the 99% MASP requirement is met by a large margin at distances greater
than 20 NM.
When applied to the Core Europe scenario, the simulation results showed better
compliance than in the LA Basin scenarios with the ADS-B MASPS requirements on
effective update periods at shorter distances. This can be attributed to the fact that
there is a smaller number of users at low altitudes in the European airspace.
A comparison of the results between the Extended LA Basin and the Core Europe
scenarios is illustrated in Figure D-5. The figure also illustrates the better
performance at short ranges in the European scenario.
60
MASPS 99th percentile requirements
Core European nominal update period 5s (112%), 99th percentile
50 Extended LA Basin nominal update period 5s (116%), 99th percentile
Effective Update Period (s)
40
30
20
10
0
0 30 60 90 120 150 180 210 240 270
Distance (nmi)
Whereas the plots for the two scenarios were very similar at an 85% link load, the
performance at the illustrated 115% link load is better in the European scenario. The
reason for this is the large proportion of low-flying aircraft in the LA Basin scenario.
40% of the traffic volume in the LA Basin (out to 60 NM radius) is below 4,000 ft. In
an operational ADS-B implementation, a large proportion of these aircraft could
operate on LSCs.
ATC Surveillance
The Core Europe scenario was selected to illustrate simulation of ATS surveillance in
this appendix. Refer to [1] for details of the other scenarios. 98% of the transmitted
ADS-B messages were received by at least one of the five ground stations. See
Figure D-6.
100% 97%
90%
83%
80%
73%
70%
60%
% messages
60%
50%
40% 37%
30%
20%
10%
0%
1 2 3 4 5
Minmum number of ground stations receiving ADS-B messages
3% of messages were not received by any ground station. This was either because
there was a slot collision – two messages transmitted in the same slot resulting in
garbling at the ground station – or because the aircraft was out of line of sight of all
ground stations.
There was considerable redundancy in the ground network with over 60% of the
messages being received by four of the five ground stations. However, for two
reasons this is a lower redundancy than achieved with the LA scenarios:
• the separation between ground stations is greater than in the LA Basin
scenarios, because the Core Europe scenario covers a wider area;
• no attempt was made to optimise the position of the ground stations.
The ADS-B MASPS requires that at least 98% of the transmitted ADS-B be received
by at least one ground station. As shown in Figure D-6, this requirement is not met
with the five-station configuration tested. In order to investigate whether more ground
stations would allow the MASPS requirement to be met, the simulation was re-done
using four ground stations. The results are shown in Figure D-7.
99%
100%
90% 87%
80% 76%
70% 67%
% messages
60% 55%
50% 46%
40%
30%
24%
20%
10%
3%
0.9%
0%
1 2 3 4 5 6 7 8 9
Minmum number of ground stations receiving ADS-B messages
Figure D-7 shows that 99% of messages were received by at least one ground
station in the nine-station scenario, and therefore that 9 ground stations are sufficient
to enable the MASPS 98% reception requirement to be met.
Station D
Station B
y
x
Station C
Station A
0.99 RR 2 (6dB)
RR 3 (10dB)
RR 4 (12dB)
RR 8 (18dB)
0.97 RR 10 (20dB)
Delivery probability
0.95
0.93
0.91
0.89
0 25 50 75 100 125 150 175 200 225 250 275 300
Distance (km)
Figure C-9 shows that with the re-use ratio reduced to 2 there is an improvement in
the delivery probability up to a distance of about 200 km that equates to an increase
in range of 30-40 km. Reducing the re-use ratio in an overloaded scenario gives an
aircraft a better chance of broadcasting to other aircraft nearby because it is then
able to re-use slots from a greater proportion of the more distant aircraft.
An increase in re-use ratio decreases delivery probability. For a re-use ratio of 8-10
we see a major decrease up to 50km. The scenario size was not big enough get
accurate data above 50 km since there were very few user pairs 10x50km from each
other. It is however expected that the re-use ratio has most effect on the short-range
performance and that long-range performance is limited by propagation effects (i.e.
BER rather than slot re-use).
The VDL Mode 4 protocol allows stations to share slots when they are unable to find
free slots for transmission. The slot selection rules embodied in the Mode 4 standard
make possible slot sharing between pairs of distant aircraft, in which case the
transmissions of each aircraft can be heard by a reduced set of aircraft located close
to the aircraft. As the channel becomes more highly loaded, it becomes possible for
more than two aircraft to share a slot. There is little reduction in performance for
channel loads of up to 85-90%. At higher loads there is a “graceful” degradation in
simulated performance as an increasing number of transmissions share slots. See
Chapter 5 Section 5.5.6.1.
The slot sharing issue is investigated in more detail below.
Figure D-10 Transmissions subject to slot sharing within LOS for the Core
Europe scenario
9000
8000 >=3
2
7000 1
6000
Messages
5000
4000
3000
2000
1000
0
VAL1011 (86%) Random (100%) Random (120%) VAL1010 (171%)
Scenario
At channel loads of 85%, most slots are occupied by single transmissions. This
confirms that below a channel load of between 85% and 90%, the system is limited
by propagation (BER) effects only. At higher channel loading, there is an increasing
number of dual and triple occupancy slots as stations are forced to share slots. This
results in the decrease in performance at longer range, since the transmissions in
shared slots will only be received by closer range aircraft.
Note that the VDL Mode 4 protocol favours selection of garbled slots before selection
of single occupancy slots. Hence a high proportion of single occupancy slots is
maintained even at high load and there is a relatively large population of triple (or
greater) occupancy slots (i.e. garbled slots become more garbled).
Figure D-11 illustrates transmissions subject to slot sharing within LOS for the LA
Basin scenario.
Figure D-11 Transmissions subject to slot sharing within LOS for the LA Basin
scenario
9000
8000 8
7
7000 6
5
6000 4
3
Messages
5000 2
1
4000
3000
2000
1000
0
LA1006 (66%) LA1001 (100%) LA1080 (199%)
Scenario
The results for the LA Basin show a high proportion of garbled slots at high loading. It
is believed that this is attributable to the small geographical range of the LA Basin
scenario.
As noted above, the VDL Mode 4 slot selection protocol will choose garbled slots
rather than slots with single occupancy. To understand the effect illustrated in Figure
D-11, it is necessary to look at the slot selection mechanism in more detail.
Stations that are close to one transmitter in a double occupancy slot have a high
chance of decoding the transmission because of favourable CCI conditions and
hence will perceive the slot as a single occupancy slot. This slot therefore has a
reduced chance of being selected for a transmission from a new station. Hence there
is a tendency to preserve dual occupancy slots that, because the reservation streams
move from time to time, have a good chance of “decaying” to become single
occupancy slots. Hence there is a strong generation mechanism for single
occupancy slots. This mechanism is responsible for maintaining the proportion of
single occupancy slots in the Core Europe scenario.
However, in the LA Basin scenario, all aircraft are located relatively close to each
other and there is a reduced probability that a dual occupancy slot will be regarded
as a single occupied slot by close range stations. Hence the mechanism to maintain
the supply of dual occupancy slots, and hence single occupancy slots, is reduced
and slots therefore tend to become more garbled, with a reduction in the number of
single occupancy slots.
Simulations were conducted using the SPS modified as required to enable the effects
of timing errors to be studied. A number of different error types were implemented to
reflect all the potential sources of error as follows:
• Random errors representing the inherent inaccuracy of a time source;
• Progressively increasing errors representing loss of synchronisation with a
primary time source;
The simulations focused primarily on the Core Europe scenario as representing the
most demanding requirements both in terms of traffic density and in air traffic control
applications that would make use of VDL Mode 4.
For the Core Europe scenario, the simulations of random errors provided an
indication of the effects of errors in the primary time source. Although the magnitude
of this error was very large for some simulations, in many cases it only had a small
effect on the effective update period. It should be noted that the magnitude of the
random error will, in the majority of cases, be much smaller than the maximum
possible magnitude and the system would be expected to be more tolerant of this
type of error.
All of the simulations that were carried out for the Core Europe scenario were
completed for the Atlantic scenario and it was found to be very tolerant of timing
inaccuracy due to the low traffic density. Similarly, simulations of the Ground
scenario showed a significant tolerance to timing errors due to low traffic densities
and the short ranges.
A further error was simulated which considered the effect of timing inaccuracies in
distant transmitters blocking the transmissions from nearby transmitters. This was
found to reduce the effective update period at shorter ranges (up to 40 NM).
The simulations indicate that the VDL Mode 4 system is very tolerant of timing
inaccuracies. The robustness is primarily dependent upon data link load. At low
loading levels, it is extremely robust. It can also be concluded that it is more robust at
shorter range. This is an important characteristic as the primary concern is likely to
be in ensuring that reports are received from the closest transmitters when the
system is an a degraded mode.
It can also be concluded that, for the scenarios considered, the timing requirements
specified in the draft SARPs will not result in any loss of performance due to timing
inaccuracy.
As an appendix to the Timing error report an assessment was made of how timing
errors might affect the XCE (1942 a/c) scenario. During these simulations channel
management was used involving two GSCs and two RSCs. The results of the
substantially larger scenario indicate a high robustness of the system. The load of
each channel was around 90% compared to the medium density scenario, in which
the load was 112%, due to fewer channels.
REFERENCES: