Successive Interference Cancellation: A Back of the Envelope Perspective
Souvik Sen, Naveen Santhapuri, Romit Roy Choudhury, Srihari Nelakuditi
2
Simple Case of Wireless Transmission
Decoding successful if:
AP
Signal Noise
SNR =
T1
> Threshold=
3
Interferer
What if parallel transmissions?
T1AP
T2
Decoding successful only if:
Signal Interference + Noise
SINR = > Threshold=
4
Collision
Collision Interferer
T1AP
T2
Decoding fails when:
Signal Interference + Noise
SINR = < Threshold=
5
Successive Interference Cancellation
Interferer
T1AP
T2
1. Decode strongest signal first
6
Successive Interference Cancellation
Interferer
T1AP
T2
1. Decode strongest signal first
2. Model and subtract
7
Successive Interference Cancellation
Interferer
T1AP
T2
3. NormalDecode
1. Decode strongest signal first
2. Model and subtract
It is as if SIC can “uncollide” signals, resulting in two successful transmissions
8
Capacity with SIC
SNR =
Rblue = Sblue
noiselog 1 +
SINR =
R*green = Sgreen
Sblue + noiselog 1+
T1T2
Interferer
AP
RSIC = Sblue + Sgreen
noiselog 1+
Green bit rate has to be far less Blue bit rate remains same
Strong signal penalized, weak signal gets all the benefits
9
Channel Capacity w/o SIC
SNR =
Rblue = Sblue
noiselog 1 +
T1T2
Interferer
AP
SNR =
SgreenRgreen =
noiselog 1 +
RSIC = Sblue + Sgreen
noiselog 1+
RwoSIC = max( Rblue, Rgreen )
Gainsic =
10
SIC Capacity Gain
11
SIC PHY Capacity Gain
Max SIC capacity gain when equal signal strengths
We were tempted to schedule packet transmissions of similar signal strengths ...
As MAC protocol designers ...
Our interpretation was that ...
maximizing SIC capacity will maximize throughput
13
SIC: A Packet Perspective
MAC Layer throughput can actually suffer
T1T2
Interferer
AP
HOLE
Stronger green packet has to be at low rate
Weaker blue packet can be at a high rate
Packet Transmission Time
Rate
14
Mathematically ...
T1T2
Interferer
AP
TimeSIC = LRblue
LR*green
max ,=
Transmission Time
TimewoSIC = L
Rblue
LRgreen
+=
Transmission Time
15
Mathematically ...
T1T2
Interferer
AP
GainSIC =
TimeSIC = LRblue
LR*green
max ,=
Transmission Time
TimewoSIC = L
Rblue
LRgreen
+=
Transmission Time
16
SIC Throughput Gain
17
SIC Throughput Gain
Max throughput gain when signal strengths are 2:1
18
Capacity Vs. Throughput
We expected: Maximizing SIC capacity will immediately maximize throughput
Reality: Equal signal strengths maximize capacity Disparate signal strengths (2:1) maximize throughput
Capacity
by reducing size of the hole?
Can’t we improve MAC layer throughput with SIC
Certainly possible:
1. Power control2. Scheduling3. Multirate packetization4. Packet packing
by reducing size of the hole?
Can’t we improve MAC layer throughput with SIC
Certainly possible:
1. Power control2. Scheduling3. Multirate packetization4. Packet packing
But at what cost?
We study SIC enabled throughput in two scenarios
1. Common receiver
2. Distinct receivers
We begin with
1. Common receiver
23
(1) Power Control
Reduce power of blue Tx such that
SINR*green = Rgreen
Rblue
= 2 *
Reduce
24
(2) Client Pairing
T2 T3
T4
T1
T1, T2 T3, T4
25
(2) Client Pairing
T2 T3
T4
T1
T1, T3 T2, T4
26
(3) MultiRate Packetization
Multirate Packetization Send the strong packet at high rate after weak packet has finished
R*green
Rblue RgreenRblue
27
(4) Packet Packing
Packet Packing Send multiple packets to fill up the hole Hard because stronger signal modeling becomes difficult
R*green
Rblue
28
Monte Carlo Simulations
SIC
RatePowerControl
Packing
29
Considerable Improvement with Adaptation
Monte Carlo Simulations
SIC
RatePowerControl
Packing
2. Distinct receivers
2. Distinct receivers
Main Concern:
• Bit Rate of T1R1 is optimal
• R2 has to decode T1’s signal at this bit rate• Despite the presence of T2’s signal
T1
R2T2
R1
Gains available when several topological constraints hold:
T1
R2T2
R1
How often do these SIC permissible topologies occur?
33
Gain with SIC in less than 10% of the cases
Monte Carlo Simulations
(AP Transmit Range)
34
Not many topologies support SIC …thus limited scope for protocols
Does MAC Adaptation Help?
Implication on Network Architectures?
T1
R2
T2
R1
Enterprise WLANs:• Clients likely to associate with stronger AP• Such scenarios unlikely
Residential WLANs:• Neighbors AP may be stronger• Some SIC scenarios possible
37
Conclusion
Successive Interference Cancellation A PHY layer capability to “uncollide” transmissions
Throughput gain not immediate from SIC Permissible bit rates impact the length of packet transmission times Creates under-utilization of the channel
Protocol adaptations possible to cope with problem Some gains available for common receiver scenarios However, limited gains for networks with distinct receivers
38
Take Away Message:
SIC aware protocol design fraught with pitfalls …
Consider doing a back-of-the-envelope calculation before plunging into system design
Questions, comments?
Thank you
Duke SyNRG Research Grouphttp://synrg.ee.duke.edu
Thank You