Accurate Physical Layer Modeling for Realistic
Wireless Network SimulationThe Need for Validation
Ruben Merz1 Cigdem Sengul1 Mustafa Al-Bado2
1Deutsche Telekom Laboratories, TU Berlin
2FG INET, TU Berlin
Workshop on ns-3 (WNS-3) / March 2nd, 2009
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 1 / 15
The Tension Between Accuracy and Complexity
Simulator objectives
Short simulation time
Accurate simulation results
Wireless network simulator
Short simulation time = simplified models
Lack of validation with real environments
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 2 / 15
We do Need Simulators. Hence Validation
Why do we need simulators
Testbed: not necessary flexible, easily configurable
Example: protocol validation before deployement, traces replaying fordebugging
The Need for Validation
Trustworthy wireless simulations: usable and dependable simulator
Understand what the (valid) assumptions are
Accuracy/networking point view: what must be modeled precisely, whatcan be dropped
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 3 / 15
A Typical 802.11a/b/g TransceiverWistron CM9 mini-PCI http://www.ipmedia.cz/img.asp?attid=655
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 4 / 15
A Typical 802.11a/b/g TransceiverWistron CM9 mini-PCI http://www.ipmedia.cz/img.asp?attid=655
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 4 / 15
A Typical 802.11a/b/g TransceiverWistron CM9 mini-PCI http://www.ipmedia.cz/img.asp?attid=655
Atheros5211/5111
McFarland et al. “A 2.4 & 5 Ghz dual band 802.11 wlan supporting data rates to 108 mb/s”, GaAs IC Symposium, 2002
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 4 / 15
A Typical 802.11a/b/g TransceiverWistron CM9 mini-PCI http://www.ipmedia.cz/img.asp?attid=655
Atheros5211/5111
McFarland et al. “A 2.4 & 5 Ghz dual band 802.11 wlan supporting data rates to 108 mb/s”, GaAs IC Symposium, 2002
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 4 / 15
A Typical RF Frontend
RF transceiver Digital baseband
BPF
RX
TX
Transmitter LPF
Base
ban
d a
nd
MA
C
Receiver LPF
Offset control
8 MHz
I
Gain control
Frequencysynthesizer
DAC
QDAC
IADC
QADC
Meng at al. “Design and implementation of an all-CMOS 802.11a wireless LAN chipset”, IEEE Communications Magazine, 2003
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 5 / 15
A Typical RF Frontend
Meng at al. “Design and implementation of an all-CMOS 802.11a wireless LAN chipset”, IEEE Communications Magazine, 2003
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 5 / 15
A Typical Baseband Engine
Pipelinecontrol
I and Q fromanalog
front-end
Rx gainto analogfront-end
Rx datato MAC
FIR
FIR
RemoveDC
offsetRotate
Frequencylock
Symboltiming
Channelestimate
andtracking
FFT ViterbiChannelcorrect
Deinterleave
Signaldetect
and AGC
ADC
Auto-correlate
Meng at al. “Design and implementation of an all-CMOS 802.11a wireless LAN chipset”, IEEE Communications Magazine, 2003
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 6 / 15
A Typical Baseband Engine
Meng at al. “Design and implementation of an all-CMOS 802.11a wireless LAN chipset”, IEEE Communications Magazine, 2003
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 6 / 15
What do We Have in ns-3.3
PRX(k) > γED SNIR(k, t) =Sk(t)
P
m 6=k m,t+NthPERR(k) = f(SNIR(k, t))
rand > PERR(k)
What’s missing?
Hardware specifics: RF frontend, non-standard transmission modeMultiple transmission channelsAntenna modeling / Multiple AntennasPropagationPacket detection and timing acquisition (synchronization) / Capture
Is the existing model sufficient/accurate? How to model/validate?What are the features that affects accuracy?
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 7 / 15
What do We Have in ns-3.3
PRX(k) > γED SNIR(k, t) =Sk(t)
P
m 6=k m,t+NthPERR(k) = f(SNIR(k, t))
rand > PERR(k)
What’s missing?
Hardware specifics: RF frontend, non-standard transmission modeMultiple transmission channelsAntenna modeling / Multiple AntennasPropagationPacket detection and timing acquisition (synchronization) / Capture
Is the existing model sufficient/accurate? How to model/validate?What are the features that affects accuracy?
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 7 / 15
What do We Have in ns-3.3
PRX(k) > γED SNIR(k, t) =Sk(t)
P
m 6=k m,t+NthPERR(k) = f(SNIR(k, t))
rand > PERR(k)
What’s missing?
Hardware specifics: RF frontend, non-standard transmission modeMultiple transmission channelsAntenna modeling / Multiple AntennasPropagationPacket detection and timing acquisition (synchronization) / Capture
Is the existing model sufficient/accurate? How to model/validate?What are the features that affects accuracy?
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 7 / 15
Accuracy Issue 1: Hardware Specifics
Power control granularity Output power spectrum not constant
5160 5180 5200 5220 5240 5260 5280 5300 5320 5340 5360−100
−90
−80
−70
−60
−50
−40
−30
−20
−10
Frequency (MHz)
Pow
er
Spectr
al D
ensity (
dB
m/1
.8M
Hz)
txpower=1
txpower=6
txpower=12
txpower=18
txpower=24
txpower=30
txpower=36
txpower=99
noise
MADWIFI, CM9 mini-PCI
Parameters dependent on transmission mode:output power, sensitivityProprietary features: antenna diversity
Kowalik et al., “Practical Issues of Power Control in IEEE 802.11 Wireless Devices”, 2008
Cheng et al., “Adjacent Channel Interference in Dual-radio 802.11a Nodes and Its Impact on Multi-hop Networking”, Globecom 06
Zargari et al., “A Single-Chip Dual-Band Tri-Mode CMOS Transceiver for IEEE 802.11a/b/g Wireless LAN”, IEEE Journal of Solid-Circuits, 2004
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 8 / 15
Accuracy Issue 2: Multiple Transmission Channels
5160 5180 5200 5220 5240 5260 5280 5300 5320 5340 5360−100
−90
−80
−70
−60
−50
−40
−30
−20
−10
Frequency (MHz)
Pow
er
Spectr
al D
ensity (
dB
m/1
.8M
Hz)
txpower=1
txpower=6
txpower=12
txpower=18
txpower=24
txpower=30
txpower=36
txpower=99
noise
0 5 10 15 20 25 30 35 400
2
4
6
8
10
12
Signal to Interference Ratio (dB)
UD
P T
hro
ug
hp
ut
(Mb
ps)
5220MHz 18dBm
5240MHz 18dBm
5240MHz 0dBm
Field Experiment Results
Channels are not orthogonal: ACI
Cheng et al., “Adjacent Channel Interference in Dual-radio 802.11a Nodes and Its Impact on Multi-hop Networking”, Globecom 06
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 9 / 15
Accuracy Issue 3: Antenna Pattern, RF Frontend
30 40 50 60 70 80 90100 200 3000
200
400
600
800
1000
1200
1400
Distance in Logarithmic Scale (meter)
Thro
ughput
(kbps)
Horizontal to Horizontal Elevated
Horizontal to Horizontal
Vertical to Vertical
Horizontal to Horizontal Cross!Polarized
Antennas: need orientation
Cheng et al., “Performance Measurement of 802.11a Wireless Links from UAV to Ground Nodes with Various Antenna Orientations”, ICCCN 06
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 10 / 15
Accuracy Issue 4: Channel Propagation
Propagation is a random process
Environment specific: indoor/outdoorFrequency specific
Channels are not symmetricSpace and time correlation
!"#
$" !# %#&" &' ((
)' ""
&##
!
"!!
#!!
$!!
%!!
&!!
'!!
!(!)"
!)"(!)#
!)#(!)$
!)$(!)%
!)%(!)&
!)&(!)'
!)'(!)*
!)*(!)+
!)+(!),
!),("
&
Reddy, Riley, “Measurement–Based Physical Layer Modeling for Wireless Network Simulations”, MASCOTS 07
Vyas et al., “Characterization of an IEEE 802.11a Receiver using Measurements in an Indoor Environment”, Globecom 06
Kurth et al., “Multi-Channel Link-level Measurements in 802.11 Mesh Networks”, 2006
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 11 / 15
Accuracy Issue 5: Synchronization and Capture,Automatic Gain Control (AGC)
Sync depends on SNR
Performance of sync and decodingnot equivalent
With several transmitters: captureeffects
Effect of AGC unknown0 5 10 15 20 25 30 35
0
10
20
30
40
50
60
70
80
90
100
SNR (dB)
Pro
b. o
f S
yn
c E
rro
r (%
)
Vyas et al., “Characterization of an IEEE 802.11a Receiver using Measurements in an Indoor Environment”, Globecom 06
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 12 / 15
Accuracy Issue 6: PER Computation
PERR(k) = f(SNIR(k, t))
= 1 − Πl (1 − Pe(k, l))
SNIR(k, t) calculationPER Computation
Existing Viterbi models: AWGN or Rayleigh channels, upper-bounds,asymptotic performanceOFDM modulation not modeledValidity of computation by block
Lacage, Henderson, “Yet Another Network Simulator”, 2006
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 13 / 15
What About ns-3.3
./!012 /:!012 35!012 8/!012 7:!012 78!0;2 9!012 5!012
6
.
76
7.
86
8.
36
3.
/6
!"#$%&'(")+E!012F
!&J(C$C+>?,"$@?A$'+E!012F
Indoor LOS, 2 nodes, 15 m
UDP, saturated 0
5
10
15
20
25
30
35
40
0 50 100 150 200 250
Exp54mb48mb36mb24mb18mb12mb
9mb6mb
According to http://www.atheros.com/pt/whitepapers/Methodology_Testing_WLAN_Chariot.pdf, should be 30.5 Mbps
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 14 / 15
Outlook and Summary
Ongoing and Future work
Channel modeling and validation
Ray-tracing?Testbed in simulator
Packet detection and timing acquisition validation, AGC
PER calculation validation
Detailed, bit-level PHY for 802.11
Large scale validation with network traces
PHY: need modeling and validation
Exhibit clearly what the underlying assumptions are
What are the elements of importance for the overall simulation accuracy?
Ruben Merz — [email protected] — http://www.net.t-labs.tu-berlin.de/~ruben/ 15 / 15