slides - iNETS - RWTH Aachen University

Illumina(ng the Road from Engineering and Policy
to Radio Regula(on
Ljiljana Simić, Petri Mähönen, Marina Petrova, RWTH Aachen University, Germany
J. Pierre de Vries, Silicon FlaBrons Centre, University of Colorado, USA
TPRC 2012
h;p://ssrn.com/abstract=2031656

ObjecBve of our work
• To understand the interplay between engineering,
policies, and radio regulaBon
• Our focus is mainly on radio spectrum allocaBon and use
• More specifically:
• There is a tussle, paraphrasing David D. Clark
• However, the tussle in radio communicaBons is markedly
different both historically and today when compared to
Internet development
• The underlying quesBon is how to efficiently allocate
scarce radio spectrum, and how sensible decisions can
be made by a regulator

Simplified (engineering) view of the problem
• Policies are developed by policy makers with
a reasonable understanding of technology
• Engineers develop efficient systems based on
policy constraints
• Stakeholders will tussle with the regulators &
market place to acquire spectrum licenses
• Market forces and ‘adversary’ compeBBon will
ensure (nearly) efficient allocaBon

QuesBons
• So if the previous is true:
• Why are there failures in the market place?
• Why are the efficiency claims between stakeholders
so widely different?
• Why do different processes end up at widely different
outcomes?
• GSM (success)
• Wi-­‐Fi (success)
• HiperLAN, WiMAX, …
• Is it hit & miss with unpredictable markets, or could the
engineering community help to streamline the process?

þ It’s Complicated
Gedng away from Cargo Cult Engineering
and Oversimplified Success Stories

RecommendaBons in 20 seconds
• Transparency
• Embrace N-­‐dimensionality
• Disclosure of interests
• Systems perspec(ve
• Technology Circle
• Fail-­‐Safe designed into regulaBon

From 1-­‐dimensional design space…
• The complexity of wireless systems has not been usually
considered transparently and explicitly
• There is a natural tendency in wireless community to try to
characterize efficiency using a single or very few quanBtaBve
numbers (and markeBng departments do not help us):
• Spectral efficiency: bps/Hz
• Maximum aggregate throughput: bps
• Power consumpBon: bps/W

…to N-­‐dimensional design space
• Reality tends to bite!
• “Everything should be as simple as possible,
but not simpler”, A. Einstein
• Wireless systems are complex:
• IniBal assumpBons & policy constraints
will affect strongly the performance metrics
• Single metric is usually nearly useless, e.g.
throughput (“bandwidth”) can be highly different if you ask
average, cell-­‐edge, or maximum theoreBcal value:
Antenna
config.
Average
spectrum
efficiency
(bps/Hz/cell)
Cell edge spectrum
efficiency
(bps/Hz/cell/user)
Average spectrum efficiency
(bps/Hz/cell) by a vendor es(mate
UPLINK
1x2
2x4
1.2
2.0
0.04
0.07
1.39
2.25
DOWNLI
NK
2x2
4x4
2.4
3.7
0.07
0.12
2.26
3.46

TVWS Wi-­‐Fi as an example
• TVWS someBmes sold as “super Wi-­‐Fi”
or “Wi-­‐Fi on steroids”
• Certainly true from a certain perspecBve …
• … but different perspecBve can also be provided
• Note: the different conclusions can be reached from
the engineering perspec(ve without excessively using policy
or commercial arguments
L. Simić, M. Petrova, P. Mähönen, "Wi-­‐Fi, but not on Steroids: Performance Analysis of a Wi-­‐Fi like Network
Opera?ng in TVWS under Realis?c Condi?ons,” in Proc. IEEE ICC, Otawa, 2012.

TVWS Wi-­‐Fi as an example
Maximum coverage range [m]
350
300
250
200
150
100
50
TVWS (630 MHz)
2.4 GHz ISM band
5 GHz unlicensed band
Wi-Fi
but not
on steroids!
Average AP downlink rate [Mbps]
30
25
20
15
10
5
0
0 5 10 15 20 25 30
Transmission power [dBm]
TVWS (630 MHz)
2.4 GHz ISM band
5 GHz unlicensed band
0
0 5 10 15 20 25 30
Transmission power [dBm]
Average spectral efficiency [bit/s/Hz]
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
TVWS (630 MHz)
2.4 GHz ISM band
5 GHz unlicensed band
0
0 5 10 15 20 25 30
Transmission power [dBm]

Average AP downlink rate [Mbps]
30
25
20
15
10
5
So which frequency is king?
outdoor urban, P tx
= 20 dBm
indoor urban, P tx
= 20 dBm
Average spectral efficiency [bit/s/Hz]
1.5
1
0.5
indoor urban, P tx
= 20 dBm
outdoor urban, P tx
= 20 dBm
0
0 0.63 2.4 5
Frequency [GHz]
0
0 0.63 2.4 5
Frequency [GHz]
Average band spectral efficiency [bits/s/Hz]
1.5
1
0.5
indoor urban, P tx
= 20 dBm
outdoor urban, P tx
= 20 dBm
…all of them!
0
0 0.63 2.4 5
Frequency [GHz]

Embracing N-­‐dimensionality
• Instead of hiding the complex N-­‐dimensionality
of systems, we should transparently describe it
• State clearly assump(ons
• Provide many plots and even visualizaBon tools to show issues
• Any single 2-­‐D slice out of full N-­‐dimensional space can be misleading
Cell size (m)
Spectral efficiency (bits/s/Hz)
Spectral efficiency (bits/s/Hz/m 2 )
Frequency (Hz)
Frequency (Hz)
Frequency (Hz)

Disclosure of interests
• Regulators need to judge the credibility
of technical evidence supplied by stakeholders:
• Complexity of design space forces the regulator
to assign relaBve importance and correctness of
technical arguments for compeBng soluBons
• Current regulatory process aims at this, but is it efficient enough?
• Very litle independent amica curae help available
• Adversaries are mostly blocks (in technical sense): Open vs. Cellular etc.
• We can learn from different regulatory approaches, e.g. FDA:
• FDA has insBtuted a process requiring disclosure
of financial interests in order to alert its staff to potenBal
bias in submited evidence and reliability of clinical data

Enforcing a systems view
• Both quanBtaBve and qualitaBve consideraBons exists
• …and should be a part of the decision making process
• neither is superior & should not try to quanBfy qualitaBve arguments
• Engineering community is not parBcularly good with qualitaBve
arguments, nor at combining qualitaBve & quanBtaBve analysis
into a single coherent systems view
• Need to have framework to support and enforce systems view

Technology Circle
• Moving from OSI network layer stack to Technology Circle
• TC Bes in engineering, regulaBon, and business models
network stack
(OSI layers 1-­‐7)
2
3
5
4 6
7
1
0
8
‟Layer Eight”
business models
& social prac

Fail-­‐safe engineering & regulaBon
• Any complex system is beyond exact predicBons
• SensiBvity to iniBal values and assumpBons
• Difficult to know all the dynamics
• Markets and technology evolve over the Bme
• Do not try to make it perfect, instead incorporate
fail-­‐safe mechanisms into regulaBon and systems engineering
• Resilience of decisions: revisited & corrected as necessary
• Alloca

Conclusions
• Do not hide complexity, embrace it openly
• Require N-­‐dimensional analysis &
full disclosure of assumpBons
• Engage engineers in the process
• Require transparency in analysis, but also in conflicts of interest
• DeclaraBon of interests required or incenBvized
• Support and require systems view in processes and analysis
• Educate future engineers on policy & regulatory issues
• Educate & provide support for policy makers on engineering & regulaBon
• Support regulators with engineering analysis & explicit policy frameworks
• RegulaBon and policies themselves may need to also consider
how to build-­‐in fail-­‐safe procedures