Email: canzian@dei.unipd.it, zanella@dei.unipd.it, zorzi@dei.unipd.it

Overlapped NACKs: Improving Multicast Performance in Multi-access

Wireless Networks

IEEE Globecom 2010 Workshop on Pervasive Group Communications (PerGroup 2010)

Authors : Luca  Canzian, Andrea  Zanella, Michele  Zorzi

• Basic problem• Our approach• Case study: Bluetooth• Performance comparison• Conclusion

2

Outline

• Basic problem• Our approach• Case study: Bluetooth• Performance comparison• Conclusion

3

Outline

Multicast: only 1 transmission!!!-> resource saving (time/frequency, power) Increasing rate More bandwidth for other devices/applications

4

Multicast communicationWireless network

5

ACK based reliability

ACK

ACK

ACKACK

Devices send ACKs if the packet is correctly received Reliability Scalability Coordination mechanism

6

NACK based reliability

NACK

Devices send NACKs if the packet is NOT correctly received Scalability Reliability Coordination mechanism

NACK

• Basic problem• Our approach• Case study: Bluetooth• Performance comparison• Conclusion

7

Outline

8

Overlapping NACKs

NACK

NACKs sent in the same resource!! (-> collisions) Scalability No coordination mechanism? Reliability

NACK

1 1 1 1 1 1 1 1 1 1 1 1 1 1

1 1 1 1 1 1 1

Nack slave 1 Nack slave 2

Resulting Nack

9

DPSK exampleQ

I

Q

I

After 1 symbol period

Q

I

-120°

120°120°

240°

0

10

Double-NACK structure

1 1 1 1 1 1 1 0 0 0 1 1 1 1 1 1 1

Q

I

Q

I

11 1 1 1 1 1 1 0 1 0 1 1 1 1 1 1 1

I

Q

I

Q

00

10

01

11

DQPSK phase-shifts

First NACK Second NACKPhase displacement

SLA

VE 1

SLA

VE 2

• Basic problem• Our approach• Case study: Bluetooth• Performance comparison• Conclusion

11

Outline

Bluetooth technology

• Ad-hoc Personal Area Network (PAN)

• Direct communication only between master and slaves

• Time-Division Duplexing (TDD) communication

• Packets last 1, 3 or 5 slots

• GFSK (1 Mbps), π/4-DPSK (2 Mbps), 8DPSK (3 Mbps)

• No multicast (multi-unicast) !!!-> SIG proposal

12

13

ACK and NACK based protocols

SIG (Special Interest Group) proposal

Our proposalOverlapped NACKs

• Basic problem• Our approach• Case study: Bluetooth• Performance comparison• Conclusion

14

Outline

15

Throughput vs slaves

16

Throughput vs BER& 3-DH5

17

Reliability

BER=1E-5 BER=1E-4

DH1 6.9E-11 / 2.1E-10 6.3E-09 / 1.9E-08

DH3 3.0E-09 / 9.0E-09 1.7E-07 / 4.9E-07

DH5 9.7E-09 / 2.9E-08 3.5E-07 / 1.0E-06

2-DH1 2.7E-10 / 8.2E-10 2.3E-08 / 6.8E-08

2-DH3 1.1E-08 / 3.4E-08 3.8E-07 / 1.1E-06

2-DH5 3.4E-08 / 1.0E-07 6.6E-07 / 1.9E-06

3-DH1 6.4E-10 / 1.9E-09 4.9E-08 / 1.4E-07

3-DH3 2.4E-08 / 7.1E-08 5.5E-07 / 1.6E-06

3-DH5 6.9E-08 / 2.0E-07 9.9E-07 / 2.8E-06

On average 1 packet every 625000 is lost !!!

Average PER / Worst case PER

• Basic problem• Our approach• Case study: Bluetooth• Performance comparison• Conclusion

18

Outline

19

Conclusion

1. New simple NACK based protocol for reliable multicast communications

2. NACK structure to be robust to collisions

3. Comparison with an ACK based protocol for Bluetooth

4. Results show:

• Little reliability loss

• High throughput gain

Email: canzian@dei.unipd.it, zanella@dei.unipd.it, zorzi@dei.unipd.it

Overlapped NACKs: Improving Multicast Performance in Multi-access

Wireless Networks

IEEE Globecom 2010 Workshop on Pervasive Group Communications (PerGroup 2010)

Authors : Luca  Canzian, Andrea  Zanella, Michele  Zorzi

21

Time synchronization

1 1 1 1 1 1 1 11 1

Q

I

Q

II

Q

11 1 1

1 1 1 11 1 1 1 1 111 1 1

OVERLAPPING INTERVAL

22

Time synchronization

1 1 1 1 1 1 1 11 1

Q

I

Q

II

Q

11 1 1

1 1 1 11 1 1 1 1 111 1 1

OVERLAPPING INTERVAL

0

0

1

0

23

Time synchronization

1 1 1 1 1 1 10 10 1

OVERLAPPING INTERVAL 1

1 1 1 1

1 1 1 1 1 1 10 11 1 1 1 1 1

1

Q

I

Q

I

Q

I

1

OVERLAPPING INTERVAL 2I

Q

00

10

01

11

DQPSK phase-shifts

24

Frequency offsetEqual amplitude case

Red turns of 0°Blue turns of 60°

Green turns of 30°

Green turns of 210°!!!-> discontinuity of 180°

Conclusion:• Most time high SNR and constant phase drift

-> it can be compensated • Some time SNR drastically decreases and phase discontinuity

-> this must be taken into account to set correlator threshold

25

Frequency offset1 signal much stronger than the other

Red turns of 0°Blue turns of 60°

Conclusion:SNR of the received signal decreases a little

Green ~ Red + noise Max phase error

Max amplitudeerror

26

b: average number of bits transmitted in a cycle d: average duration of a cycle

N N-1 1 0. . .

Pe N

N (1-P )e

N-1 (1-P )e

1-P e

PePe N-1

1

(N-1) (1-P ) P

N (1-P ) P

e e

N-1

N-2

ee

N (1-P ) P N-1ee

Markov model Nbite BERP 11

Renewal Process:db

ttuntildtransmittebitsTh

t

lim

Nji

otherwise

ijPPji

Pji

eje

ji ,...,1,0,0

1,

Markov model

27

Probability that the master does not recognize the overlapping of i double NACKs (missed detection probabilities)

N N-1 1

0

. . .

PeN

PePe N-1

1

(N-1) (1-P ) P

N (1-P ) P

e e

N-1

N-2

ee

N (1-P ) PN-1

ee

L L L

[1-P (N-1)]

[1-P (1)]

[1-P (1)]

(N-1) (1-P ) Pe e

N-2P (1)

[1-P (1)][1-P (N-1)]

N N-1 1

(1-P )eN

(1-P )eN-1

1-P e

[1-P (N)]

PeN

P (N)

N (1-P ) PN-1

ee P (N-1)

Pe N-1

P (N-1)

N (1-P ) PN-1

ee P (1)

nr

nr

nr

nr

nr

nr

nrnr

nr

nr

nr

Pe P (1)nr

1 1 1

. . .

)(iPnr

N

iLiP

NiPER

1

N

iLiPB

1

Probability to be absorbed in Li

iLP