Department of Information Engineering University of Padova ITALY

Department of Information Engineering University of Padova, ITALY Special Interest Group on NEtworking & Telecommunications Mathematical Analysis of Bluetooth Energy Efficiency Andrea Zanella, Daniele Miorandi, Silvano Pupolin {andrea. zanella, daniele. miorandi, silvano. pupolin}@dei. unipd. it WPMC 2003, 21 -22 October 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Outline of the contents q Motivations & Purposes q Bluetooth reception mechanism q System Model q Results q Conclusions WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

What & Why… Motivations & Purposes WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Motivations q Bluetooth was designed to be integrated in portable battery driven electronic devices Energy Saving is a key issue! q Bluetooth Baseband aims to achieve high energy efficiency: Ø Units periodically scan radio channel for valid packets Ø Scanning takes just the time for a valid packet to be recognized Ø Units that are not addressed by any valid packet are active for less than 10% of the time WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Aims of the work q Although reception mechanism is well defined, many aspects still need to be investigated: q Ø What’s the energy efficiency achieved by multi-slot packets? Ø What’s the role plaid by the receiver-correlator margin parameter? Ø What’s the amount of energy drained by Master and Slave units? Our aim is to provide answers to such questions! How? Ø Capture system dynamic by means of a FSMC Ø Define appropriate reward functions (Data, Energy, Time) Ø Resort to renewal reward analysis to compute system performance WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

What standard says… Bluetooth reception mechanism WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Access Code field 72 54 AC HEAD 0 -2745 PAYL access code packet header q CRC payload Access Code (AC) Ø AC field is used for synchronization and piconet identification Ø All packet exchanged in a piconet have same AC Ø Bluetooth receiver correlates the incoming bit stream against the expected synchronization word: ü AC is recognized if correlator output exceeds a given threshold Ø AC does check HEAD is received Ø AC does NOT check reception stops and pck is immediately discarded WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Receiver-Correlator Margin q S: Receiver–correlator margin q Determines the selectivity of the receiver with respect to packets containing errors ØLow ü S strong selectivity risk of dropping packets that could be successfully recovered ØHigh ü S weak selectivity risk of receiving an entire packet that contains unrecoverable errors WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Packet HEADer field 72 54 AC HEAD access code packet header q 0 -2745 PAYL CRC payload Packet Header (HEAD) Ø Contains: ü Destination address ü Packet type ü ARQN flags: used for piggy-backing ACK information ü Header checksum field (HEC): used to check HEAD integrity Ø HEC does check PAYL is received Ø HEC does NOT check reception stops and pck is immediately discarded WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Packet PAYLoad field 72 54 AC HEAD access code packet header q 0 -2745 PAYL CRC payload Payload (PAYL) Ø DH: High capacity unprotected packet types Ø DM: Medium capacity FEC protected packet types ü Ø WPMC 2003 (15, 10) Hamming code CRC field is used to check PAYL integrity: ü CRC does check positive acknowledged is return (piggy-back) ü CRC does NOT check negative acknowledged is return (piggy-back) Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Conditioned probabilities Receiver. Correlator Margin (S) AC 72 bits HEAD 54 bits 2 -time bit rep. (1/3 FEC) DHn: Unprotected DMn: (15, 10) Hamming FEC PAYLOAD CRC h=220 2745 bits 0: BER WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Retransmissions A B B B NAK MASTER ACK G SLAVE A q Automatic F X H H B X DPCK Retransmission Query (ARQ): Ø Each data packet is transmitted and retransmitted until positive acknowledge is returned by the destination Ø Negative acknowledgement is implicitly assumed! ü Errors on return packet determine transmission of duplicate packets (DUPCK) ü Slave filters out duplicate packets by checking their sequence number Ø Slave does never transmit DUPCKs! ü Slave can transmit when it receives a Master packet ü Master packet piggy-backs the ACK/NACK for previous Slave transmission ü Slave retransmits only when needed! WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Mathematical Analysis System Model WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Hypothesis q Single slave piconet q Saturated links Ø q Unlimited retransmission attempts Ø q Packets are transmitted over and over again until positive acknowledgement Static Segmentation & Reassembly policy Ø q Master and slave have always packets waiting for transmission Unique packet type per connection Sensing capability Ø Nodes can to sense the channel to identify the end of ongoing transmissions Ø Nodes always wait for idle channel before attempting new transmissions WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Packet error probabilities q Let us define the following basic packet reception events Ø ACer: AC does not check ü Ø HECer: AC does check & HEAD does not ü Ø Packet is recognized but PAYL contains unrecoverable errors PRok: AC & HEAD & PAYL do check ü q Packet is not recognized CRCer: AC & HEAD do check, PAYL does not ü Ø Packet is not recognized Packet is successfully received Packets experiment independent error events WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Reception events Reception Event Index Downlink pck reception events q q q Uplink pck reception events 0: both downlink and uplink packet are correctly received 1: downlink packet is correctly received, uplink packet is received but with errors in the PAYL field 2 U 3: downlink packet is correctly received but uplink packet is not recognized by the master unit Ø q 4 9: downlink and uplink packets are not correctly received Ø WPMC 2003 Master will transmit DUPCKs Master will retransmit useful packets Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Mathematical Model q Normal State (N) Ø q Duplicate State (D) Ø q q Master transmits packets that have never been correctly received by the slave Master transmits duplicate packets (DUPCKs) Since error events are disjoint, the state transition probabilities are given by The steady-state probabilities are, then, WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Reward Functions q q For each state j we define the following reward functions Ø Tj= Average amount of time spent in state j Ø Dj(x)= Average amount of data delivered by unit x {M, S} Ø Wj(x)= Average amount of energy consumed by unit x {M, S} Ø The average amount of reward earned in state j is given by Performance indexes Ø Energy Efficiency: Ø Goodput: G WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Notations q Let us introduce some notation: Ø Dxn (Dym) downlink (uplink) packet type, n=1, 3, 5 Ø L(Dxn) = PAYL length (bit) for Dxn packet type Ø w. TX(X) / w. RX(X)/ wss(X)= amount of power consumed by transmitting/ receiving/ sensing the packet field X Ø WPMC 2003 pj = Pr( j) Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Time reward ( T ) Transmission Reception/Sensing MASTER SLAVE n+m MASTER SLAVE n+1 WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Data reward ( D ) q Master gains Data reward when Ø Ø q System is in state N Slave perfectly receives the master packet Slave gains Data reward when Ø Ø Slave recognizes the master polling Master perfectly receives the slave packet WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Master energy reward ( W ) Receives entire uplink packet Receives only AC field Receives till the first uncorrected field and senses till the end of the packet Always transmits a downlink packet WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Slave energy reward ( W ) q Slave’ energy reward resembles mater’ one except that, in D state, Slave does not listen for the PAYL field of recognized downlink packet since it has been already correctly received! WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Performance Analysis Results WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Energy Efficiency q Downlink traffic only (M>S) and S=0 q Energy efficiency gets worse in Rayleigh channels q DH 5 outperform other packet formats for almost every SNR value q For SNRd. B=14 18, DMn outperforms DHn WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Master Slave swapping q Swapping Master and Slave role: Ø Ø WPMC 2003 DM 5 & DM 3 energy efficiency increases up to 15 % for SNR 20 d. B Unprotected pck types show slightly reduced performance gain Performance gain drastically reduces for increasing values of the Rice factor K For AWGN channels, master slave swapping does not lead to any significant performance improvement Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Master Slave swapping q WPMC 2003 Swapping Master and Slave role: Ø DM 5 & DM 3 energy efficiency increases up to 15 % for SNR 20 d. B Ø Unprotected pck types show slightly reduced performance gain Ø Performance gain drastically reduces for increasing values of the Rice factor K Ø For AWGN channels, master slave swapping does not lead to any significant performance improvement Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Impact of parameter S AWGN q Rayleigh The receiver correlator margin S has strong impact on system performance Ø AWGN: improves with S, in particular for low SNR values Ø Rayleigh: gets worse with S, except for low SNR values q Relaxing AC selectivity is convenient, since G gain is much higher than loss q Impact of S, however, rapidly reduces for SNRd. B>15 WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003

Conclusions q Main Contribution Ø q mathematical framework for performance evaluation of Bluetooth piconets Results Ø In case of asymmetric connections, Slave to Master configuration yields better performance in terms of both Goodput and Energy Efficiency ü Ø q Slave never transmits DUPCK Parameter S may significantly impact on performance ü Short and Protected packet types improve performance with S ü Long and Unprotected packet types show less dependence on this parameter Results may be exploited to design energy–efficient scheduling algorithms for Bluetooth piconets WPMC 2003 Yokosuka, Kanagawa (Japan) 21 -22 October 2003