Transcript Rate vs Range Test Methodology

Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Site-Specific Knowledge
for
Next Generation Wireless Networks
Prof. Ted Rappaport
Wireless Networking and Communications Group
Department of Electrical and Computer Engineering
The University of Texas at Austin
November 17, 2004
www.wncg.org
Submission
Slide 1
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Some Wireless Next Generation Activities
•
•
•
•
•
•
•
WiFi and 3G combination chipsets
UWB/Home Media Gateway
MiMo and OFMD-based modulation
Advanced Security
Cross Layer, Universal MAC
Integrated Multiband/colocated antennas
Mesh Networks/Site-specific Radio
Management
Submission
Slide 2
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Site Specific knowledge is needed in Next
Generation Networks
• We can substantially increase battery life, network
performance, enhance coexistence, reduce support
calls, and deploy no-fault wireless using “site specific”
knowledge
• PHY/MAC/Radio Resources of today will move to
baseband processing and digital “environmental map”
in each client
• Power vs. processing tradeoffs: RF power consumption
and Network Inefficiencies (today) versus baseband
processing and client’s environmental awareness (next
gen)
• Myriad new services, capabilities become viable
Submission
Slide 3
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Computing and device trends
• Vector graphics, 3-D processing capability evolving
naturally as part of microprocessor
• Multiple radios, frequency bands, applications, to
become part of PCs, phones, home media, enterprise
network products
• Memory costs and cost per MIPS decreasing
exponentially, at much faster rate than battery and RF
antenna/propagation breakthroughs
• History of wireless has not exploited
environmental/spatial knowledge in the network, yet
propagation depends solely on this!
Submission
Slide 4
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Wireless Technology and Semiconductor
ROADMAP
Year
Technology
Gate Width
(nm) [2]
Vdd Treshold
(V) [1]
1990
900 MHz
Cellular
1800 MHz 2G
Cellular
2.4 GHz
802.11b
5.8 GHz
802.11a
UWB
10 GHz
anticipated
BWA
800
5
Saturation
Current
(uA/[Vum]) [1]
150
275
2.5 – 1.8
200
130
2.5 – 1.8
300
100
1.8 – 1.5
375
90
60
1.8 – 1.5
1.5 – 1.2
400
500
30
1.2 - 0.9
650
1996
2001
2003
2004
2006
2010
Source:
1. Hotta, Imasao, Shoji Shukuri, and Koichi Nagasawa. “Trends of Semiconductor Techonology for
Total System Solutions.” http://www.hitachi.com/rev/1999/revapr99/r2_101.pdf.
2. http://phys.cts.nthu.edu.tw/workshop/tp5/20031204/T.%20F.%20Lei.pdf
Submission
Slide 5
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
ITRS Technology Nodes and Chip Capabilities
2001
2005
2010
2016
Microprocessor Speeds
(MHz)
1,684
5,173
11,511
28,751
Gate Length (nm)
65
32
18
9
DRAM Cost/bit (microcents)
7.7
1.9
.34
.042
DRAM memory size
512M
2G
8G
64G
Source: http://www.sia-online.org/downloads/itrs_2001.pdf
Submission
Slide 6
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
A paradigm shift – learning from
Qualcomm
•
•
•
•
Qualcomm changed the wireless world:
Narrowband radios became wideband radios
Tight RF filtering became sloppy RF filtering
Channel selection became a baseband
processing chore, not an RF/IF chore –plays to
Moore’s law
• Moving the processing to baseband enhanced
the network coordination/interoperability and
led to flexible upgrade path to data/3G
• Intellectual property enforcement
Submission
Slide 7
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Challenges Qualcomm faced
• Convincing carriers that CDMA improved
spectral efficiency, made network deployment
easier, increased users and revenue per MHz
• Convincing carriers to relearn how to design
and install base stations (no frequency
planning, but code offset planning and soft
handoff thresholds)
• End User has to wait 8 seconds for Qualcomm
phone to detect pilot and synch channels, 50 ms
speech coder delay, and immediate “hard
dropped” calls
Submission
Slide 8
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
A paradigm shift
Site-Specific propagation knowledge
• Site-specific knowledge will change the wireless world:
• MAC/PHY/QoS/applications will match the propagation
environment, instead of being rigid/iteratively implemented
• Channel selection, power level settings, and network
provisioning becomes a baseband processing chore, not an
RF/IF chore involving radio usage.
• Moving the processing to baseband enhances network
coordination/interoperability and leads to flexible
upgrades,interference mitigation, position location, 4G
• Intellectual property enforcement (Standards – sharing)
Submission
Slide 9
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Challenges for Site-specific adoption
• Convincing chip makers that networks perform
better with lower battery drain, plays to Moore’s
law if “environmental map” is digitized and
exploited
• Convincing OEM/ODM/ box makers that sitespecific network planning and management
reduces support calls, reduces user problems, and
enhances network performance and features
• Some site-specific data must be obtained at some
point
Submission
Slide 10
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Today: Network Deployment
• The need for site-specific prediction models
– Many consumers and IT professionals deploy WLAN by
trial and error due to limited awareness of antenna and
propagation issues. Poor experiences…..
– Models exist for signal-strength predictions, throughput
coverage, viable CAD software.
– Internet users and vendors are interested in application
throughput for many different user profiles.
– To manage interference, improve QoS, and end-user
quality, site-specific CAD design/deployment now being
used – large deployments starting to rely on CAD
– Eventually, this must become a commodity and brought
into networks for management of devices
Submission
Slide 11
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Network Coverage Software
Used by IT Admin./ Network Integrators
Submission
Slide 12
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Site-specific Prediction Models
• Predictions of signal strengths in buildings [Seidel,
Rappaport,1994], [Durgin et al,1998];
 
PR (d )  PT  GT  GR  PL(d )  20 log 10  
 4 
PLd   10n log
10
d  a X
a
b X
b
• Throughput prediction models [He01], [Ra00]
Submission
Slide 13
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Extensive measurements to validate
site-specific throughput
• Sites: Three restaurants (Schlotzsky’s deli)
• Apparatus: laptops, IEEE-802.11b
wireless network interface cards (NICs): Cisco
and ORiNOCO
• Throughput Measuring software: LANFielder
(Wireless Valley Inc.), Iperf, Wget (FTP)
• Measurements conducted outside of normal
business hours
• Measurement Scenarios: 1. single user; 2. multiple
users
Submission
Slide 14
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Single-user Measurement Platform
Submission
Slide 15
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
The Guadalupe Restaurant
Partition description
Color
Attenuat
ion (dB)
Glass doors and
windows
Red
5.26
Concrete block walls
Dark gray
6.83
Wooden partitions
Light blue
4.70
Short counters
Light gray
0.50
Submission
Slide 16
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
The Northcross Restaurant
Partition description
Color
Attenua
tion
(dB)
Glass doors and windows
Red
5.65
Concrete block walls
Dark
gray
8.39
Wooden partitions
Light
blue
0.59
Short counters
Light
gray
1.84
Metallic racks
Yellow
7.47
Tree
Green
0.10
Submission
Slide 17
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
The Parmer Restaurant
Partition description
Color
Attenuati
on (dB)
Glass doors and
windows
Red
2.00
Concrete block walls
Blue
5.10
Wooden partitions
Yellow
3.48
Short counters
Light gray
0.50
Stony pillars
Purple
1.50
Thin pillars
Green
3.00
Submission
Slide 18
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Multi-user Measurement Platform
Submission
Slide 19
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Multi-user Measurement Applications and Tools
Client
Server
Computer
Dell C640 & HP Omnibook
Compaq N600c
OS
Windows XP
Windows XP
NIC
Cisco & ORiNOCO
N/A
FTP
Wget
IIS
LANFielder
LANFielder Client
LANFielder Server
Iperf
Iperf Client
Iperf Server
SNR
LANFielder & netstumbler
N/A
Submission
Slide 20
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
11 locations (Guadalupe)
Submission
Slide 21
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Two Throughput Models that relate sitespecific SNR to Throughput
• The piecewise model
Tmax
, if

T 
 Ap  SNR  SNR0  , if
SNR  SNRc (dB)
SNR  SNRc (dB)
• The exponential model
SNRc (dB) 

Tmax
 SNR0 (dB)
Ap
 Ae  SNR  SNR0 
T  Tmax 1  e
Submission
Slide 22

Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Cisco card data
Guadalupe
Parmer
Northcross
Submission
All three restaurants
Slide 23
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Cisco card data (spatial average)
Guadalupe
Parmer
Northcross
Submission
All three restaurants
Slide 24
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
For General In-Building
Environments
•
•
•
•
•
•
Spatial Average
All Three Restaurants
Cisco card
Exponential model
Scales to 3 different apps.
Also see 802.11-04-1473-00-000t
Tmax
(Mbps)
Ae
(dB-1)
SNR0
(dB)
μ (Mbps)
σ
(Mbps)
R(%)
Iperf
5.26
0.069
5.39
0
0.88
76.4
Wget
4.47
0.0747
11.0
0
0.615
90.9
LANFielder
1.76
0.113
8.25
0
0.295
81.1
Submission
Slide 25
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Blind Throughput Predictions for a New
Environment using Site Specific map
• Predicted RSSI in dBm
– Use [Se94,Du98] models, auto-tuning implemented in sitespecific prediction tool LANPlanner by Wireless Valley
• The ambient noise level in dBm
– Perform a quick calibration test in the new environment
(typical value: -90 dBm)
• Mapping SNR to throughput for different apps
– Determine Tmax by back-to-back calibration tests; use Ae
and SNR0 of foregoing results
Submission
Slide 26
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Performing Tests in WNCG
• Noise is -90 dBm
• Tmax for LANFielder was
calibrated as 2.403 Mbps
• Reading the table, Ae is
0.113 dB-1, and SNR0 is
8.25 dB

T  2.403 1  e0.113SNR8.25
Submission
Slide 27
Ted Rappaport, WNCG, Univ of Texas

Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Site-specific RF Network Management
DESIGNED
RF REMEDIATION / RECONFIGURATION w/SITE SPECIFIC
Submission
Slide 28
DEPLOYED
SSID
COVERAGE
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Deployed Network Coverage
Cube-farm has no coverage in the deployed network
due to human deployment error or “bad” equipment
Submission
Slide 29
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Deployed Network Coverage- Autonomous Network
Management using Site-specific knowledge
AP01 is automatically reconfigured using digitized map at switch; cube-farm
now has coverage in the deployed network
Submission
Slide 30
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Home and Enterprise Network Management
System using Site-specific knowledge
• How does it work?
– User spends approximately 30 - 60 seconds inputting basic site-specific
information into a GUI
– Software uses site-specific algorithms on a digital map to determine coverage
areas and optimal equipment positions/configurations within the environment;
digitizes finalized infrastructure map and pushes to clients
– Devices share site-specific knowledge and measured responses through the
network to monitor, control, and diagnose changing RF conditions.
– Unless desired, the end user never needs to interact with the software beyond
the initial network setup stages and added infrastructure – everything else is
automated behind the scenes (power levels, handoff, auto-reconfig. with new
nodes).
– Hidden node problem, next door neighbor is diagnosed and controlled much
more reliably using site-specific knowledge
Submission
Slide 31
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Alternate Embodiments: Embedded Network–
Centralized Hub, AP, Clients
– Embedded software on a centralized network appliance (e.g., media gateway,
hub, switch, etc.) and/or on APs or clients. Leverage site-specific information
stored locally on the device to make informed decisions regarding network
configurations. Site specific knowledge shared with clients.
• How does it work?
– Site-specific information regarding the environment and network
infrastructure is downloaded to the embedded software
• Embedded software may be pre-loaded on the device or downloaded from the
Home NMS
– The embedded software monitors network and radio activity it sees in the
environment
– As events occur that negatively impact network performance, the embedded
software can independently analyze the event in the context of the overall
network and can respond quickly with device configuration changes that are in
the best interests of the overall network
Submission
Slide 32
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
Embodiment of Embedded Network Software in Clients
• Embedded software runs on clients either as services in the
operating system, as part of a device driver, or directly
integrated onto the hardware in some fashion
• Why do we need it?
– This technology places intelligence in the hands of the client devices, with
greatest power concern and in closest contact to end-users
– Site-specific knowledge, combined with Moore’s law in processing power,
allows mobile devices to know where, when, and how to properly manage its
power, and applicability, while improving overall network performance.
– Memory and CPU requirements scale to allow this to be viable in next one to
three years
– Ties in with intelligent infrastructure, security, new services
– Site-specific knowledge of the client offers ultimate intelligence for
communication. Why God gave us eyes, why we like maps in new cars
Submission
Slide 33
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
QOS in a Hybrid Environment
AP2:
Ch. 1, 30 mW
802.11g
Client #1:
Ch 1, -59 dBm
41 dB SIR
24 Mbps
AP1:
Ch. 6
1 mW
802.11g
Association
TV:
Ch 6
-45 dBm
55 dB SIR
54 Mbps
Submission
.
AP1 lowers its power levels to a minimum in order to avoid serving
distant clients who can be served by AP2. Client PDA stays with AP2
Slide 34
Ted Rappaport, WNCG, Univ of Texas
and has good service. TV on AP1 retains good service.
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
QOS in a Hybrid Environment
Client #1:
Ch 1, -42 dBm
28 dB SIR
1 Mbps
Association
AP1:
Ch. 1
30 mW
802.11g
TV:
Ch 6, -35 dBm
36 dB SIR
11 Mbps
Submission
AP2:
Ch. 1
30 mW
802.11g
Without site-specific NMS, client associates with AP1
because AP1 offers higher power levels, but interferes with
TV on same channel, reduces bandwidth of TV streaming
video, and experiences its own reduced bandwidth.
Slide 35
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
The Site-Specific Revolution….
Coming to Next Generation Networks
• Theoretical formulations for quantifiable data, metrics,
and tradeoffs for semiconductor baseband, RF, software,
site-specific traffic, and power overhead are needed, but
are emerging.
• Computing power is evolving to allow “electronic maps” to
be exploited in devices for new wireless devices
• This is an entirely new and unexploited dimension to MAC
and PHY – and is cross-layer processing unlike previous
solutions in the wireless world
• Broad scale market adoption is likely, and IEEE should
begin studying and standardizing this concept
• Why did God give us eyes, and why do we like cars with
navigation systems in them – they make us more efficient
Submission
Slide 36
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
References
• [Ch04] Jeremy Chen, “Site Specific Network Throughput modeling,” M.S.
Thesis, Summer 2004, WNCG, University of Texas at Austin
• [Na04] Chen Na, Jeremy Chen, T.S. Rappaport, “Public WLAN Traffic
statistics and throughput prediction,” Electronics Letters, Sept. 13, 2004
• [He01] B. E. Henty, T. S. Rappaport, “Throughput Measurements and
Empirical Prediction Models for IEEE 802.11b Wireless LAN (WLAN)
Installations”, ECE Dept., Virginia Tech technical report, MPRG 01-08,
2001
• [Ra00] T. S. Rappaport, B. Henty, and R. Skidmore, “System and method
for design, tracking measurement, prediction and optimization of data
communication networks,” pending U.S. and International Patents.
• [Du98] G. Durgin, T. S. Rappaport, and H. Xu, “Measurements and models
for radio path loss and penetration loss in and around homes and trees at
5.85 Ghz,” IEEE Transactions on Communications, vol. 46, no. 11, pp.
1484–1496, November 1998.
• [Se94] S. Y. Seidel and T. S. Rappaport, “Site-specific propagation
prediction for wireless in-building personal communication system design,”
IEEE Transactions on Vehicular Technology, vol. 43, no. 4, pp. 879–891,
1994.
Submission
Slide 37
Ted Rappaport, WNCG, Univ of Texas
Nov 2004
doc.: IEEE 802.11-04-1478-00-0wng
References (II)
• [He03] M. Heusse et al. “Performance Anomaly of
802.11b”, INFOCOM 2003
• [Bi00] G. Bianchi, “Performance Analysis of the IEEE
802.11 Distributed Coordinated Function,” IEEE JSAC,
vol. 18, pp. 535-547, Mar. 2000
• [Ch03] P. Chatzimisios et al, “Influence of channel BER on
IEEE 802.11 DCF,” Electronics Letters, vol. 39, no. 23, pp.
1687–1689, November 2003.
• [Ga03] S. Garg et al, “An experimental study of
throughput for UDP and VoIP traffic in IEEE 802.11b
networks,” IEEE WCNC, 2003
• [Va02] A. Vasan et al, “An empirical characterization of
instantaneous throughput in 802.11b WLANs,” U of
Maryland tech report
Submission
Slide 38
Ted Rappaport, WNCG, Univ of Texas