This document discusses putting spectrum allocation decisions on a more quantitative footing using risk-informed interference analysis. It outlines a four element method for risk-informed interference assessment that includes: (1) making an inventory of interference hazards, (2) defining consequence metrics, (3) assessing the likelihood and consequence of each hazard, and (4) aggregating the results. The document also provides a case study applying this method to analyze interference risks between meteorological satellite receivers and LTE mobile transmitters.

1. Silicon Flatirons Center, UC Boulder
Risk-informed Interference Analysis
Putting spectrum allocation decisions
on a more quantitative footing
Pierre de Vries
Senior Adjunct Fellow, Silicon Flatirons Center, UC Boulder
Presentation at TPRC 43
26 September 2015
v. 0.4

2. Context
Insatiable demand for spectrum rights leads to
– Squeezing services together ever more tightly
– Ever-tougher harm/benefit trade-offs
This entails regulatory judgments about harmful interference
– Informed by engineering, typically (unfortunately) worst case analysis
Most agencies nowadays complement worst case with quantitative risk analysis
(NRC, EPA, FAA, FDA, NASA, DHS, etc.) -- but not the FCC
Project goal
– Put harmful interference analysis on a more quantitative, statistical footing
– … in order to yield better decisions about allocations and rules
2

3. Worst case (deterministic extreme value) analysis
Does not represent reality accurately
– Most parameters that influence harm take a range of values
Over-conservative
– Provides too much protection; doesn’t serve public interest, nor economically
efficient
Can lead to false confidence that the resulting rules will avert harm
– There are many kinds of radio interference
– e.g. LightSquared/GPS: fixated on OOBE, but ABI was the real problem
BUT: Worst case is simple, and will continue to be used
– Quickly gives a black & white answer
3

4. Risk examples: Accidents
Accident on Mount Everest
– High likelihood, high consequence
– Overall risk: very high
Skydiving accident
– Low likelihood, high consequence
– Overall risk: moderate
Falling off a bicycle
– Low likelihood, low consequence
– Overall risk: low
4
Likelihood
Low Medium High
Consequence
High
Medium
Low
EverestSkydiving
Bicycling Unicycling

5. Some terminology
Risk
– The combination of likelihood and consequence for multiple hazards
– Kaplan and Garrick’s risk triplet: What can go wrong? How likely is it? What
are the consequences?
– There are other, complementary approaches: economics, psychology,
socio-cultural analysis
Quantitative risk assessment (QRA)
– Apply risk triplet using numerical estimates of likelihoods and consequences
Risk-informed interference assessment
– A systematic analysis of the likelihood and consequence of interference
hazards caused by the interaction between radio systems
5

6. Risk-informed interference assessment
Current scope
– Planning, e.g. allocation, rulemaking, waivers
– Not post-deployment operation (e.g. adjudication & enforcement, service provider ops)
– Leave aside cost/benefit analysis
Related work in spectrum (quantitative risk analysis generally: 30+ years of literature)
– Michael Marcus, IEEE-USA (2012): Noted MCL vs. stochastic modeling, flagged lack of FCC
policy
– Grunwald, Alderfer & Baker (TPRC 2014): Holistic analysis of 5 GHz Wi-Fi/Globalstar
interference
– Littman & De Vries (TPRC 2014): Lessons for FCC from use of QRA in nuclear regulation
– FCC TAC (2015): Noted value of risk assessment in allocation decisions, proposed a method
– Cui & Weiss (TPRC 2015): QoS and monetary risk for different kinds of spectrum sharing
6

7. Four element method
1. Make an inventory of all significant harmful interference hazard
modes
2. Define a consequence metric(s) to characterize the severity of
hazards
3. Assess the likelihood and consequence of each hazard mode
4. Aggregate the results to inform decision making
7

8. Applying the method: MetSat case study
8

9. Meteorological Satellite (MetSat) & LTE
1675–1710 MHz
– weather satellite receiving
earth stations
– NOAA and DoD
Focus on polar orbiting
satellites
– leave aside geostationary ones
1695–1710 MHz
– AWS-3 cellular mobile uplink
9

10. NTIA Fast Track, CSMAC WG-1 studies
Determine an interference
protection criterion (IPC):
aggregate i/f power not to be
exceeded in MetSat receiver
Assume a “sea” of LTE mobiles
Calculate the smallest radius
without mobiles that satisfies
IPC
Exclude/limit mobile ops within
this zone
Essentially worst case
10

11. 1. Make an inventory of hazards
Interference types
– Co-channel
– Out-of-band emission (OOBE)
– Adjacent band interference (ABI)
Types of interferer
– Point sources and aggregate
interference
– Unintentional and intentional
radiators
– Operators: well-meaning, ignorant or
malicious
Working taxonomy
– Interfering system (“transmitters”)
– Affected system (“receivers”)
– Coupling between transmitters &
receivers
11

12. Transmitter characteristics (LTE)
Cellular mobiles
– Transmit power per mobile, co-
channel and out-of-channel
– Frequency channel width
– Percentage loading of base station
– Location and density of mobiles
– Location and density of base
stations
12
CDF of total EIRP per scheduled mobile
Source: CSMAC (2013), Appendix 3-3

13. Receiver characteristics (MetSat)
13
Source: NTIA (2010) “Fast Track Report”

14. Transmitter-Receiver Coupling
– Propagation loss from
transmitter to receiver
– Additional losses
– Antenna heights: satellite
receiver and mobile transmitters
– Earth station antenna elevation
and azimuth
14
Propagation loss:
field measurements and fit
Source: Phillips, Sicker & Grunwald (2012)

15. 2. Define consequence metric(s)
Corporate, aka operational: Cost, ability to complete the mission
– Receiver metrics not available (indications of high baseline
outage)
Service
– Availability: the percentage of time that the link margin is not met
– Quality: bit-error ratio
RF
– Fraction of interference-free margin consumed by interference
– Interfering signal power levels (IPC) to be exceeded no more than
20% or 0.0125% of the time
– Interference-to-noise ratio in the receiver
15

16. 3. Assess likelihood & consequence
Consider co-channel interference, a la Fast Track and CSMAC
WG-1
Calculate received interfering power for each exclusion radius
Determine the probability a given power is exceeded
1. Choose -121 dBm
interference
protection criterion
(IPC)
2. Select 80th percentile, so
that IPC is exceeded no
more than 20% of the time
3. Results in a 34
km protection radius
4. Vertical “slice”
through distribution
yields distribution of
interfering power at
the 34 km radius;
see next slide

17. Probability of exceeding IPC versus radius
Vertical slice through
range/interference
chart
Choose -121 dBm
interference protection
criterion (IPC)
Meets ITU-R SA.1026
long-term IPC:
– Below IPC more than
80% of time
– i.e. exceeded less
than 20% of the time 17
Satisfies -121 dBm IPC
at 34 km exclusion distance
Interference limit
met
> 80% of the time
Likelihoo
d
Consequence

18. 4. Aggregate risks
Non-RF: equipment failure,
operator error, …
Co-channel interference
– sunspot activity
– long-term low-level i/f
– short-term high-level i/f
Interference from adjacent band
– OOBE
– ABI
18
OOBE
(hypothetical)
ITU-R long-term criterion:
with 34 km exclusion, -121
dBm exceeded less than 20%
of time
ITU-R short-term criterion:
with 34 km exclusion, -118
dBm exceeded less than
0.0125% of time (hypothetical)
Likelihoo
d
Consequence

19. Conclusions
19

20. Benefits of risk-informed interference assessment
A common currency for comparing
– different interference mechanisms
– competing assessments
More comprehensive analysis
– increases the chances of identifying unexpected harmful interference
mechanisms
Objective information for decision makers
– balancing the benefits of a new service and its adverse technical impact
on existing services
20

21. Action items for regulators
Educate
– Get the community thinking & talking via papers, workshops and consultations
– Develop know-how through lectures and in-house training
Set a good example
– Quantify likelihoods and consequences in own findings
– Request (ideally, require) disclosure and analysis of likelihood & consequence in
filings
Start small, but start soon
– Changing the culture is going to take a long time
– Pilot approach on low risk/impact proceedings, e.g. waivers for services at fixed
locations
21

22. Action items for the Executive & Legislature
Oversight
– Make risk-informed assessment an oversight requirement
– Don’t fall for nightmare scenarios
– Support and encourage regulators that use risk-informed interference
assessments
Require spectrum regulators to do risk analysis
– US: Extend existing requirements for risk and cost/benefit analysis (cf.
Executive Orders, OMB directives) to independent agencies
22

23. Summary
Risk analysis considers the likelihood-consequence combinations for
multiple hazards, and complements a “worst case” analysis
This will yield better spectrum allocation decisions
Four element method: (1) inventory hazards; (2) define metrics; (3)
assess likelihood & consequence; (4) aggregate
Adding this to the toolkit requires culture change, so start small – but
start soon
23

24. We are not able in life to avoid risk
but only to choose between risks
Stanley Kaplan & John Garrick (1981)
24

25. Backup
25

26. Risk Chart
26
Likelihood
Qualitative
descriptors
Rare Unlikely Possible Likely Certain
Quant
scales
Determined case by case
Consequence
Very High Severity Determinedcaseby
case
High Severity
Medium Severity
Low Severity
Very Low Severity

27. Deterministic methods and worst case analysis
Deterministic methods: evaluate risk in terms of a predetermined set of causes characterized
by single-valued parameters
– potentially interfering transmitter operating at a fixed distance
– at a fixed transmit power
– to a specific receiver
– single-valued path loss
“Worst” case: parameters take extreme values
– transmitter at closest distance
– maximum allowed transmit power
– the least interference-resistant receiver on the market
– direct path without any intervening obstructions
27
Deterministic method
doesn’t necessarily
entail using extreme
values
(but usually does)

28. Utility of extreme value analysis
Extreme value analysis can be useful
– if worst case assumptions show that a particular hazard doesn’t
pose a risk, it can be omitted from subsequent analysis
– if best case assumptions indicate that a hazard poses risks even
in favorable circumstances, further analysis is needed
Typically, though, worst case is used “illogically”
– worst case assumptions showing harm used to justify further
analysis, and perhaps even determine rules
28

29. Interference analysis: a schematic
– Interfering (transmitting)
system
– Affected (receiving)
system
– Coupling between
transmitters & receivers
29
Affected System
Characteristics
Interfering System
Characteristics
Affected System
Locations
Interfering System
Locations
Coupling
Characteristics
Likelihood and
Consequence
Metrics
Risk
Assessment

30. Probability of exceeding IPC versus radius
Horizontal slice through
range/interference chart
As radius increases,
probability that IPC will
met increases
Choose -121 dBm
interference
Meeting IPC 80% of time
means exceeded 20%
of the time: ITU-R
SA.1026 criterion
30
34 km
exclusion
distance
Interference
limit met > 80%
of the time