Spectrum
Engineering
Advanced
MonteCarlo
Analysis
Tool
For
Optimisation of Radio Spectrum Usage
Artūras Medeišis
European Radiocommunications Office
Outline:
•
•
•
•
Spectrum sharing: means and rules
Worst-case vs. Statistical analysis
Monte-Carlo statistical analysis
SEAMCAT as spectrum optimisation tool
October 2006
2
Need for spectrum sharing:
• There are no more “empty” spectrum
• Proposed new systems have to find way of
“sharing” with some of existing systems
• Thus the need for spectrum engineering
and optimisation:
– to find which existing radio systems are
easiest to share with, and then
– determine the “sharing rules”
October 2006
3
Sharing methods:
• Spacing radio systems in frequency
– Using the gaps between existing channels
• Spacing geographically
– Using the gaps between intended deployment
areas (e.g. cities vs. rural areas)
• Time sharing
– Exploiting different work time (day vs. night)
• Working at different power levels
– E.g. “underlay” spectrum use by UWB
October 2006
4
Sharing implementation:
• Agile (cognitive) radio systems require
minimum sharing rules as they could be
adapting dynamically
– Simple example: finding free channel in a given
geographic area
• Traditional rigid-design radio system will
require precisely defined sharing rules
– Maximum transmit power, guard-bands to
existing systems, etc
October 2006
5
Defining the sharing rules:
• Analytical analysis, usually by worst-case
approach:
– Minimum Coupling Loss (MCL) method, to
establish rigid rules for minimum “separation”
• Statistical analysis of random trials:
– The Monte-Carlo method, to establish
probability of interference for a given realistic
deployment scenario
October 2006
6
MCL principle:
• The stationary worst-case is assumed
Wanted
Signal
Victim
Interferer
Dmin, or minimum frequency separation for D=0
– However such worst-case assumption will not
be permanent during normal operation and
therefore sharing rules might be unnecessarily
stringent – spectrum use not optimal!
October 2006
7
Monte-Carlo principle:
• Repeated random generation of interferers
and their parameters (activity, power, etc…)
Wanted
Signal
t=t0
Victim
t=ti
t=t1
Active
Interferer
Inactive
Interferer
– After many trials not only unfavourable, but also
favourable cases will be accounted, the resulting
rules will be more “fair” – spectrum use optimal!
October 2006
8
Monte-Carlo assumptions:
• User will need to define the distributions of
various input parameters, e.g.:
– How the power of interferer varies (PControl?)
– How the interferer’s frequency channel varies
– How the distance between interferer and victim
varies, and many others
• Number of trials has to be sufficiently high
(many 1000s) for statistical reliability:
– Not a problem with modern computers
October 2006
9
SEAMCAT:
• Spectrum Engineering Advanced MonteCarlo Analysis Tool
• An open software tool for analysing
various scenarios for co-existence of radio
systems
• Developed in co-operation by European
administrations and industry between
1997-2002, latest updated in 2005
• Free download from www.seamcat.org
October 2006
10
Using SEAMCAT:
• User defines a scenario, describing mutual
positioning of two systems, both in
geographical domain…
5 km
MS-Iti
Wti
Wr
BS-Vr
…as well as many other parameters:
October 2006
11
Scenario parameters:
•
•
•
•
•
Positioning of two systems in frequency
Powers
Masks
Activity
Etc.
October 2006
12
SEAMCAT event generator:
• Random generation of transceivers
• Link budget
• Signal values
October 2006
13
How event generator works:
• Succession of snapshots…
dRSS
iRSS
Snapshot#
WT
1) Calculate d, Ptx, GaTx, GaRx, L
Snapshot#
2) Calculate dRSSi
IT
WT
VR
1) Calculate d, Ptx, GaTx, GaRx, L
2) Calculate iRSSi
VR
1) Calculate d, Ptx, GaTx, GaRx, L
2) Calculate received signal, if PC, adjust Ptx
IT
WR
WR
October 2006
14
Result of simulations:
• Vectors for useful and interfering signals:
October 2006
15
Evaluating interference:
• By comparing signal instances:
- For each random event:
Desired signal value (dBm)
Interfering signal (dBm)
C/Itrial > C/Itarget?
Interference (dB)
Noise Floor (dBm)
- If C/Itriali >C/Itarget: “good” event
- If C/Itriali <C/Itarget: “interfered”
- Finally, after cycle of Nall events:
Overall Pinterference= 1- (Ngood/Nall)
October 2006
16
Result of simulations:
• One value of probability of interference,
• Or, as function of one of input parameters
October 2006
17
As a result:
• Deployment of radio systems would be
evaluated in realistic situations, so the
“averaging” of real-life interference could
be estimated with statistical precision
• Impact of input parameters may be
evaluated more precisely, therefore the
sharing rules will be more optimal
• As a result, all this leads to statistically
proven optimal exploitation of spectrum
October 2006
18
SEAMCAT-3 (2005):
• CDMA:
October 2006
19
SEAMCAT-3:
• Interference into CDMA as capacity loss:
October 2006
20
Conclusions:
• Sharing rules are important element of
spectrum optimisation process
• Unless some intelligent interference
avoidance is implemented in radio systems,
the careful choice of sharing conditions is
the only means for achieving successful
co-existence and optimal spectrum use
• Statistical tools, as SEAMCAT, could be a
simple yet powerful tool for such analysis
October 2006
21
Thank you!
• For further info:
– www.ero.dk/seamcat
– www.seamcat.org
– E-mail: [email protected]
October 2006
22