Multi-point to Multi-point MIMO in WLANs

Fine-grained Spectrum
Adaptation in WiFi Networks
Sangki Yun, Daehyeok Kim and Lili Qiu
University of Texas at Austin
ACM MOBICOM 2013, Miami, USA
1
Current trend in WiFi
• Wireless applications increasing throughput
demand
• Channel width is increasing
802.11a/b/g
802.11n
802.11ac
20MHz
40MHz
Is wide
channel always better?
160MHz
• Benefit of wide channel: higher throughput
2
Disadvantage of wideband channel
• High framing overhead
•channel
High energy consumption channel
access preamble
access preamble SIFS
SIFS
• idle
Lower spectrum efficiency
wide due to frequency
idle
channel
wide
data
period
period
transmission
transmission
diversity
data
channel
ACK
ACK
25
20
20
SNR (dB)
SNR (dB)
25
15
15
10
10
5
5
0
0
20MHz channel
20MHz channel
3
Lessons
• Static spectrum access (wide or narrow
spectrum exclusively) is insufficient
• Need dynamic spectrum access to get the best
of both worlds
4
Ideal case: per-frame adaptation
• Clients select channel based on their preference
• AP needs per-frame spectrum adaptation to communicates with
different clients
• Preferred channel may change over time -> further increase the
need for per frame adaptation
5MHz
20MHz
10MHz
Spectrum efficiency
Energy efficiency
20MHz
time
5
Challenges
• Enable per-frame spectrum adaptation
• Sender and receiver agree on the spectrum
• Dynamically allocate spectrum efficiently
6
Related work
• Dynamic spectrum access (WiMAX, LTE, FICA)
– Requires tight synchronization among clients
– Significant signaling overhead
• Spectrum adaptation (SampleWidth, FLUID)
– Focus on spectrum allocation and ignore spectrum agreement
– Slow to adjust the channel width
• WiFi-NC
– Channel width is fixed to 5MHz
– Requires longer CP to reduce guard bandwidth
• IEEE 802.11ac
– RTS/CTS for dynamic bandwidth management
– Not fine grained (minimum channel width 20MHz)
7
FSA: Fine-grained spectrum adaptation
• Per-frame spectrum access
– Change spectrum per-frame
– Communicate with multiple nodes on different
subbands using one radio
• In-band spectrum detection using existing
preamble
• Efficient spectrum allocation
8
Transmitter design
20MHz bandwidth
OFDM signal
PHY
encoder
Reduces
bandwidth
Interpolation &
remove images
upsampler
LPF
Center frequency
shifting
CF
shift
. . .
. . .
. . .
RF
mixer
PHY
encoder
upsampler
LPF
CF
shift
9
Generating narrowband signals
• Convert 5 or 10MHz signal based on 20MHz signal
through upsampling and low pass filtering
LPF
upsampling
20MHz
frequency
20MHz signal
20MHz
frequency
Upsampling generates
images outside tx band
20MHz
frequency
Narrowband signal
10
Sending signals together
• Center frequency shifting is performed and the
signals in different spectrum are added
𝑓𝑠
𝑓𝑠
𝑠10 𝑛 = 𝑠10 [𝑛]𝑒 𝑗2𝜋∆
20Hz
Narrowband signal
20Hz
Center frequency
shifting
Shifted signal
𝑓𝑠
𝑠10 [𝑛]
𝑠10 𝑛
RF
20Hz
Mixed signal
𝑠[𝑛]
𝑓𝑠
𝑠 𝑛 = 𝑠10 𝑛 + 𝑠5 𝑛
adding another
narrowband signal
20Hz
Deliver to RF
11
Receiver design
CF
shift
PHY
decoder
...
LPF
downsampler
...
RF
Spectrum
detector
LPF
downsampler
...
CF
shift
PHY
decoder
12
Receiver design
CF
shift
Spectrum detector is key component
PHY
decoder
...
LPF
downsampler
...
RF
Spectrum
detector
LPF
downsampler
...
CF
shift
PHY
decoder
13 / 35
Spectrum detector
• Goal: Receiver identifies the spectrum used by the
transmitter
• Possible solutions
– Use control channel or frame
• Too much overhead
• Target for attack
• Control channel may not be always available 
further increase overhead
– Design special preamble [Eugene,12]
• Deployment issue
14
Spectrum detection using STF
• It is ideal to detect spectrum using existing 802.11
frame detection preamble (STF)
• One solution: Spectral and Temporal analysis of the
detection preamble (STD)
– Power spectral density to detect the total spectrum
width
– Temporal analysis to identify exact spectrum allocation
– Costly and inaccurate especially in noisy channel
• Our approach
– Exploit special characteristics of STF for spectrum
detection
15
Characteristic of 802.11 STF
• Time domain: 10 repetitions of 16 signals
t1 t2 t3 t4 t5 t6 t7 t8 t9
t10
• We
Frequency
domain:
12
spikes
out
of
64
exploit the subcarrier interval for the spectrum detection!
subcarriers with 4 subcarrier intervals
16 / 35
Spectrum detector design (Cont.)
• Depending on the transmitter spectrum width, the
received STF has various subcarrier intervals
20MHz
Subcarrier interval: 4
10MHz
Subcarrier interval: 2
5MHz
Subcarrier interval: 1
17
Spectrum detection using STF
• 20MHz transmitter to 20MHz receiver
20MHz
20MHz
transmitter
20MHz
receiver
STF in the frequency domain at the 20MHz receiver
18
Spectrum detection using STF
• 10MHz transmitter to 20MHz receiver
Two subcarriers of
10MHz transmitter
is merged into one
10MHz
subcarrier of
transmitter
20MHz receiver
20MHz
20MHz
receiver
STF in the frequency domain at the 20MHz receiver
19
Spectrum detection using STF
• 5MHz transmitter to 20MHz receiver
20MHz
5MHz
transmitter
20MHz
receiver
STF in the frequency domain at the 20MHz receiver
20
Spectrum detection using STF
• The subcarrier interval difference let us easily
identify the spectrum
20MHz
20MHz
transmitter
20MHz
receiver
STF in the frequency domain at the 20MHz receiver
21
Spectrum detector design (Cont.)
10MHz
Transform spectrum detection
5MHz into pattern matching.
10MHz
10MHz
10MHz
5MHz
5MHz
22
Spectrum detector design
Cross-correlation
check
RFfrontend
Maximum likelihood
pattern matching
802.11
preamble
detection
FFT64
Received signal
sampled in
20MHz rate
spectrum
detection
Magnitude of
64 subcarriers
• Optimal Euclidean distance based spectrum detection
𝐗 = arg min
𝑖
64
𝑘=1
𝑦𝑘
𝑘 2
− 𝑥𝑖 .
• Binary detection
𝐗 = arg min 𝐗 𝑖 ⊕ 𝐘
𝑖
23
Spectrum Allocation
Controller
buffer
AP
client
AP
client
AP
client
client
24
Spectrum Allocation (Cont.)
• Input
– Destinations of buffered frames
– CSI between APs and clients
– Conflict graph
• Goal: Minimize finish time
– Avoid interference
– Harness frequency diversity
• Knobs
– Spectrum
– Schedule
– AP used for transmission
25
Spectrum allocation (Cont.)
• Break a frame into mini-frames
• Break the entire spectrum into mini-channels
• Greedily assign a mini-frame to a mini-channel that
minimizes the overall finish time while avoiding
interference
• Find a swapping with an assigned mini-frame that
leads to the largest improvement, go to step 3
26
Evaluation methodology
• Implemented testbed in Sora
– 2.4GHz
– 20MHz maximum bandwidth
• Evaluates detection accuracy and latency,
spectrum allocation performance in testbed
• Trace based simulation for spectrum
allocation in large-scale network
27
Spectrum detection accuracy
1.0
Delivery rate
Probability
0.8
Detection rate - STD
0.6
Detection rate - FSA
(binary)
Detection rate - FSA
(ED)
0.4
0.2
0.0
20 - 15
15 - 10
10 - 5
SNR range (dB)
5-0
28
Spectrum detection delay
1
CDF
0.8
STD
0.6
FSA
0.4
0.2
0
0
100
200
Detection delay (µs)
Median detection delay 4.2 us < detection delay budget
29
Throughput evaluation – no interference
Throughput (Mbps)
20
FSA
Fixed
15
10
5
0
1
2
3
4
5
6
7
8
9
10
FSA improves throughput by exploiting frequency diversity
30
Throughput evaluation – interference
Throughput (Mbps)
15
FSA
Fixed
10
5
0
1
2
3
4
5
6
7
8
9
10
With narrowband interference, the gain grows larger
31
Summary
• FSA – a step towards enabling dynamic
spectrum access
– Flexible baseband design
– Fast and accurate channel detection method
– Spectrum adaptation
32
Q&A
Thank you!
33
Comparison with WiFi-NC
Throughput (Mbps)
60
WiFi-NC
FSA
40
20
0
10
15
20
SNR
25
30
Simulation in fading channel width RMS of delay spread = 100 ns
WiFi NC incurs lower SNR due to sharp filtering
34
Discussion
• Detection accuracy
• Antenna gain control
• Bi-directional traffic
35