A Sharper EDGE for GSM

Transcription

1 A Sharper EDGE for GSM Mark Pecen Vice President Advanced Technology Research In Motion Limited and

2 Industry roadmap major phases 3GPP wireless technologies - LTE expected to require more time to standardize and deploy radical shift in air interface, system architecture and management of user mobility - Evolved EDGE viewed by operators as alternate evolution path to LTE leverages existing infrastructure - GSM licenses in Europe have been extended to

3 GSM/EDGE Technology What is GSM/EDGE? Enhanced Data for GSM Evolution (EDGE) is a 3rd Generation superset of the General Packet Radio Service (GPRS) Three primary attributes High order modulation scheme (8-PSK) 3 bits per symbol instead of 1 Radio link adaptation Agility at radio layer to exploit local radio conditions Incremental redundancy Reduces re-transmissions of radio blocks by hybrid coding scheme Maximum practical data rate (4 timeslots downlink) kb/s Typical data rates approximately kb/s in North America

4 What is Evolved EDGE? High technology leverage applied to GSM Newest advances in technology applied to the largest deployed wireless system in the world 2 to 3 times current EDGE data rates 600 to 900+ kb/s downlink 200+ kb/s uplink 2 or more times current capacity both statistical and information theoretic gains Multiple carriers downlink fills more resources more of the time Information theoretic gain from 3 bits/symbol to 5 bits/symbol ½ of current EDGE radio latency From 18.5 ms/block to 9.2 ms/block All the benefits of the latest technology within the cost structure of existing technology

5 How will EDGE Evolve? Fairly simple leveraging of existing standards Approximately 10 features were standardized, but many were considered economically or technically infeasible by industry participants Realistic subset of Evolved EDGE features Multiple carrier Downlink Higher-order modulation, Uplink and Downlink Advanced algorithms for data latency reduction Mobile station diversity techniques Software upgrade for existing base stations

6 Downlink Multi-Carrier techniques 2.5 times the current throughput and trunking efficiency Utilization of more than one radio frequency carrier Multiple receivers and/or baseband in the downlink Multiple streams of information interchanged simultaneously Channel capacity greatly improved Not by increasing basic efficiencies of the air-interface, but because of statistical improvement in the ability to assign radio resources increases trunking efficiency Economies of scope Re-use and extend multiple receiver architecture for intelligent selection of MSRD or multi-carrier receive, depending on operational environment Single timeslots scavenged from 8 carriers on downlink to mobile TDMA Frame terminal having 2 receivers no need to schedule contiguous timeslots Timeslot Number RF Channel 1 RF Channel 2 RF Channel 3 RF Channel 4 RF Channel 5 RF Channel 6 RF Channel 7 0 R1 1 R2 2 R1 3 R2 4 R1 5 R2 6 T1 7 0 R1 1 R2 2 R1 3 R2 4 R1 5 R2 6 T1 7 0 R1 1 2 R1 3 R2 4 R1 5 R2 6 7 T1 RF Channel 8 R2 Fm1 Fm2 FmMAX M2 Method of timeslot scavenging using 2 receivers at mobile terminal M2 M2

7 High-order modulation More information per transmitted symbol E s 010 E s 000 E s E s Bi-phase modulation: 1 bit/symbol Quadrature phase shift keying (QPSK): 2 bits/symbol level phase shift keying (8PSK): 3 bits/symbol 16 Quadature amplitude modulation (16QAM) modulation: 4 bits/symbol 32 Quadature amplitude modulation (32QAM) modulation: 5 bits/symbol 64 Quadature amplitude modulation (64QAM) modulation: 6 bits/symbol

8 Latency Reduction Send same amount of info in less time

9 Mobile Station Receive Diversity (MSRD) Smarter antennas increase network capacity and performance Use of two antennas and receive paths Increased tolerance to co-channel interference Improves service grade/data rates Greatly improved system capacity Addresses radio spectrum scarcity Why now? Smart antennas were restricted to base stations Learning curve effects offer cheap DSPs and other components As we evolve wireless, more intelligence is pushed into the radio layers simply because we can at reasonable cost Baseband signal in x(k) Starting point... RF front end and down conversion Channel estimate Received signal Gain factors in objective function min σ N i i= 1 r rˆ g Diversity combiner 2 i i Where the industry is going... RF front end and down conversion Pre-filter, p(k) y(k) TSC s(k) y(k) Advanced receiver Advanced receiver Compute cost function J(k) Slicer Feedback filter g(k) Coefficients z(k) Baseband signal ready for demodulation Decision logic Desired ^s(k) Fully demodulated data

10 Why does the industry need EDGE Evolution? Many new potential technology entrants/disruptions Excellent mobile talk/standby time Battery life is critical for mobile device Margins and battery life both better for mobile with one radio access technology Build additional global relevance for GSM Reduce probability of potential disrupters License fees Potential downward pressure on UMTS license fees Controlled technology switching costs for operators Higher present value of existing system deployments

11 Further benefits as industry evolves Service continuity extend value chains geographically Improve Global Mobile Data Fabric, GSM is 84% of wireless market in 2007 No WCDMA expected in many developing countries Service Continuity is key for future applications as users move between HSPA to EDGE coverage Less observable performance difference between EEDGE and LTE Give users less incentive to switch to disruptive technology Next evolution of GSM/EDGE before LTE deployment Certain problems must be solved no matter what the next generation of wireless brings not as simple as shutting off GSM one day and turning on LTE the next

12 Time frames Standards, deployment and contributors EDGE Evolution work item began in 2004 in 3GPP 3 rd Generation Partnership Project (3GPP) Technical Specification Group known as GPRS/EDGE Radio Access Network (GERAN) Standards nearing completion Release 7 for Technical Report (TR) (June 2006) 3GPP TR Release 7 for selected set of building blocks in core technical specifications (TS) (end of 2006) 3GPP TS , , , , , , , Stage 3 specifications and test specifications written during Deployment time frames Expect commercial mobiles in end of 2008, early 2009 Network upgrades possibly sooner for field trials of test mobiles Who are the players? The following companies have introduced contributions to the creation of the EDGE Evolution standards work item

13 Where are we now? Move toward commercialization Standards Core specifications almost complete Test specifications beginning Commercialization Companies now coming together Defining what is an Evolved EDGE product? Prototyping active Field trials planned

14 Industry Collaboration Mobile/network manufacturers, operators Careful selection of EEDGE features Greatest observable economic utility value Minimum technology risk Achieve fast turn-around Work with operators and vendors to Build prototypes Field test, refine and optimize deployment

15 Smart Phasing EEDGE Rollout Mobile terminal view of power consumption Phase 1 features UL/DL user data rate (kb/s) Chosen to get to market fast Minimum technical risk High utility value

16 Phase 1 Focus on quick success Minimize technology risk Easily deployed on base station as well as handsets Quickly deployed (2008 field trial 2009 commercial) Optimize observable utility value to users and operators: Data rates (600k+ downlink and 200k+ uplink) Battery consumption (similar to EDGE) System capacity (About 2x statistical capacity increase for data) Support proliferation of new mobile data applications Web 2.0 based mobile internet apps: social networking, user generated content, mash-ups, CRM, ERP, music/video downloads

17 Phase 2 Focus on improved speeds Minimize technology risk Taking advantage of a couple of Moore s law generations Deployment timeframe of Optimize observable utility value to users and operators: Data rates (944k+ downlink and 400k+ uplink) Battery consumption (similar to EDGE, little more) System capacity (Additional 1.5 to 2x information theoretic gain) Support proliferation of new mobile data applications Web 2.0 based mobile internet apps: social networking, user generated content, mash-ups, CRM, ERP, music/video downloads

18 Phase 3 May never happen in small handset form factor Take advantage of another couple of Moore s law generations Deployment timeframe of Issues of multiple antenna in small devices not solved yet High risk for handset form factor Expensive and risky but no observable benefit to end user, although higher capacity for operator Economic value re-capture doubtful for handset makers

19 Summary Way forward for today Coordination Among Parties Recommend confining phase 1 scope to 1. Dual Downlink Carrier, 2. Latency Reduction and 3. 4 TS uplink Reasonable restriction on complexity and technical risk Good probability of success Highly observable benefits Coordinate prototyping effort Hardware & software changes on mobile Software changes only on network Coordinate lab and field trials Most fundamental measurements/verification first in lab Study behaviour in operator environment Create actionable rollout plan

20 Appendix A The building blocks of Evolved EDGE

21 Dual Carrier Downlink >2x Downlink Throughput Slot N Slot N + 1 (Idle Frame) Slot N + 2 Slot N + 3 Rx1 Tx NCM Neighbor Cell Measurements Uplink Timeslot Downlink Timeslot Slot N Slot N + 1 (Idle Frame) Slot N + 2 Slot N + 3 Rx1 Rx2 Tx (1) Neighbor Cell Measurements Uplink Timeslot Downlink Timeslot

22 Dual Carrier Mobiles in Single Carrier Networks Increased multislot capability! Use the second receiver for neighbor cell measurements when a dual carrier mobile is operating in single carrier mode. The first receiver doesn t have to retune to make measurements, therefore the mobile can support a higher multislot class than its hardware dictates one more slot per TDMA frame!

23 Dual Carrier More than just increased throughput 90% Multiple Timeslot Assignment Probability Prob. of a 5 timeslot assignment 80% 70% 60% 50% 40% 30% and the probability of getting an assignment for several contiguous timeslots on the downlink decreases accordingly. 20% 10% 0% 100% 90% 80% 70% 60% 50% 40% Prob. of being assigned 1 timeslot As the radio system fills up with users, the probability of finding a single timeslot on an adjacent cell decreases

24 How dual carrier can increase system capacity In the case of congestion, the probability of finding contiguous timeslot assignments on one frequency drops significantly. Multislot assignments could still be possible if the timeslots can be scavenged from more than one frequency. This alleviates or removes the need for the BTS to schedule contiguous timeslots. The availability of a second independently tunable receiver in the mobile allows it to receive back to back timeslots on different frequencies while not violating the reaction times of its declared multislot class. Using timeslot scavenging can increase data rates and trunking efficiency due to an increased statistical likelihood of utilizing a greater number of radio resources over a given period of time. Example shown in next figure single timeslots are scavenged from 5 carriers on the downlink to a dual carrier capable mobile.

25 Dual Carrier Downlink Timeslot Scavenging

26 Multi-Carrier - timeslots easily scavenged Statistical gain using existing constraints of TS Timeslot Number RF Channel 1 RF Channel 2 RF Channel 3 RF Channel 4 RF Channel 5 RF Channel 6 RF Channel 7 RF Channel 8 Fm1 Fm2 FmMAX MAX St: Single timeslots scavenged from 8 carriers on downlink to mobile TDMA Frame terminal having 2 receivers no need to schedule contiguous timeslots R1 R2 R1 R2 R1 R2 T1 M2 R1 R2 R1 R2 R1 Where: timeslot number t R Nt Nr S tr Method of timeslot scavenging using 2 receivers at mobile terminal u Nt Nr Maximize the tr number of slots : StRutR; u { 0,1} over all RF carriers t= 1 R= 1 Subject to the a a = 1 tr tr+ 1 ut 1R = 0:1 restrictions u = already in the Else:0 EDGE specifications R2 T1 M2 R1 R2 R1 R2 R1 R2 T1 M2 = = = = = = = RF deck number (both transmit and receive) Maximum number of timeslots in allocation Maximum number of RF decks available for simultaneous reception in mobile terminal Radio resource of timeslot t on RF deck R for mobile receiver Utilization of timeslot t on RF deck R permitted according to the reaction time constraints (Boolean) Timeslot availability for allocation to mobile downlink (Boolean)

27 Data Latency Reduction New Advanced Algorithms Latency reduction will help to enable new services and increases statistical system capacity Support is added for : RTTI (Reduced Transmit Time Interval) FANR (Fast Ack/Nak Reporting) Methods can be combined Updated assignment messages for new configuration information New channel coding required for Fast Ack/Nack Reporting

28 Reduced Transmit Time Interval (RTTI) More data in shorter frame period 10ms TTI achieved by interleaving 4 bursts in 2 frames over two slots. USF is encoded in 20 ms to be compatible with legacy MS when there is mixed BTTI (Basic TTI) and RTTI on PDCHs USF is encoded in 10 ms if PDCH-pair is exclusively RTTI Method will be signaled to MS during TBF establishment PDCH-pair need not be allocated on adjacent slots UL PDCH-pair need not be allocated on the same slots as DL PDCHpair PDCH-pair must be allocated on the same carrier in case of a dual carrier allocation MS can have more than one PDCH-pair assigned to it. Dynamic and Extended Dynamic Allocation will be supported

29 Latency Reduction Fast ACK/NACK Reporting (FANR) Piggybacks ACK/NAK information in radio data block New PANI/PAN fields added to RLC header (Piggyback ACK/NAK Indicator) - PAN field is bit map of 20 bits New Channel Coding has been added to support PANI/PAN fields PAN is encoded using tail-biting rate 1/3 convolutional code with puncturing to 48 bits and is protected by a 6 bit CRC. New puncturing maps are defined to allow the addition of the extra 48 coded PAN bits to the radio block MS can send PAN field based on DL data rather than waiting to be polled Network can request ACK/NAK report and provide MS with ACK/NAK information in one data block, or can poll for ashort ACK/NAK bitmap Network can continue to use legacy ACK/NAK reporting methods PAN field can change (to reflect current ACK/NAK conditions) between retransmissions of a RLC block

30 Higher Order Modulation and Higher Symbol Rate More bits per symbol more throughput, higher capacity Higher order modulation is a spectral efficiency enhancement. Continuing to improve spectral efficiency is key for any wireless technology as spectrum a scarce resource that drives this industry Uplink and downlink enhancement is being developed simultaneously to add symmetry and reuse technology where possible Includes the addition of QPSK, 16QAM and 32 QAM as well as increased symbol rate (20% increase) and turbo coding in some cases (algorithms reused from RAN)

31 But we cannot add bits/symbol indefinitely No magic want faster data rates? Need more link margin! States that the maximum number of bits/symbol we can support is limited by i) bandwidth and ii) signal quality Pse( ηs ) < erfc π ηs sin M Channel capacity, i.e. max number of bits/symbol 10-1 η s = E s E s N 0 S C = B log N 10-2 How much bandwidth do we occupy? What is the quality of our signal? C = capacity in bits/symbol B = Bandwidth occupied by signal (Hz) S = Signal strength (W) N = Noise level (W) = ktb = thermal noise floor Bit-Error Rate QAM 16QAM 8PSK E 0 /N 0 (db)

32 Higher-order modulation schemes A fundamental information theoretic improvement More bits per symbol allows more data interchange per unit time Fundamental technological improvement Faster data rates Exploits local coverage maxima - takes advantage of the geographical locations associated with probabilities of high C/I ratio and enables very high data transfer rates whenever possible Enhances EDGE network capacity with little capital investment by extending scope of existing platform Statistically significant gains and avoids high technology switching costs of replacing system Why did we wait until now? Were impractical for wireless only a few years ago due to factors such as processing power and variability of interference and signal level Equalizers very complex RIM s new equalizer technology plus very low cost/high density silicon enables hardware-implementation of calculations that were impossible only 5 years ago Modulation and coding scheme (MCS) Existing EDGE througput (bits/sec) Enhanced EDGE - 16-QAM extension (bits/sec) S C = B log N Enhanced EDGE - 32-QAM extension (bits/sec) Enhanced EDGE - 64-QAM extension (bits/sec) MCS-1 35,200 MCS-2 44,800 MCS-3 59,200 MCS-4 70,400 MCS-5 89, , , ,200 MCS-6 118, , , ,200 MCS-7 179, , , ,200 MCS-8 217, , , ,000 MCS-9 236, , , ,400 Existing EDGE System: 8PSK 3 bits/sym Data rates for 4 timeslots 4 bits/sym 6 bits/sym 5 bits/sym Channel coding for MCS-5 9 reused on 16, 32 and 64QAM

33 Downlink Higher Order Modulation Data download to mobile Two levels of MS support defined Support Level A 8 PSK, 16 QAM, 32 QAM with turbo coding (reused from RAN, defined in 3GPP TS ) Support Level B QPSK, 16QAM, 32 QAM with turbo coding and higher symbol rate (1.2 x legacy symbol rate) Support of Level B includes support of Level A Involves some challenges in hardware design (linearity requirements) but measurement data has shown that there is sufficient link margin in most cases to support higher order modulation Offers 25% - 50% increase in data rates at all C/I Significant complexity - all new modulation and coding schemes, training sequences, etc. are being defined

34 Uplink Higher Order Modulation Data upload from mobile Three levels of MS support defined Support Level A adds 16 QAM with legacy symbol rate Support Level B adds QPSK (instead of 8 PSK), 16QAM and 32 QAM with higher symbol rate (1.2 x legacy symbol rate) Support Level C adds 32 QAM and possibly turbo coding MS support of Level B implies support of Level A. MS support of Level C implies support of Levels A and B. Involves some challenges in hardware design (linearity requirements) but measurement data has shown that there is sufficient link margin in most cases to support higher order modulation Offers 25% - 50% increase in data rates at all C/I. There will be an impact on power consumption (in uplink this will depend largely on PA efficiency optimization) and maximum output power due to back off required by higher PAPR Significant complexity - all new modulation and coding schemes, training sequences, etc. are being defined New MCS are not necessarily going to be aligned with the downlink

35 HOM with Higher Symbol Rate (HSR) The Pulse Shaping Question However, RRC filters have poorer spectral properties compared to LGMSK filters, which violates the GSM spectral mask. In the downlink this causes increased ACI for legacy mobiles, which is a problem. Gains for HSR are improved with widening of the transmit pulse shaping filter. This also lowers the PAPR of the signal for higher order modulations, which permits higher mean output power. Root Raised Cosine (RRC) filters with different bandwidth have been considered.

36 Implementation Details Timing advance for HSR is still measured in normal symbol rate units Blind detection of modulation uses the symbol rotation similar to GMSK/8PSK GMSK uses π/2 8PSK uses 3π/8 QPSK uses 3π/4 16 QAM uses π/4 32 QAM uses -π/4 USF encoding uses 12 symbols for LSR and 16 symbols for HSR USF encoding when multiplexing LSR with HSR on same PDCH is an issue Legacy mobiles may false detect USF Operators don t like trunking inefficiency due to reserving PDCHs for legacy mobiles Desire to be able to multiplex REDHOT A and REDHOT B. Proposals involve resampling I/Q input data to detect USF of REDHOT B on a REDHOT A MS.

37 Choosing an MCS Example: Downlink HOM with HSR There are many options for new MCSs. Part of the challenge is choosing as many new MCS give roughly the same performance. Too many options for the MS to support will limit implementation. Source: 3GPP Tdoc GP , Marvell Semiconductor Inc.

38 MSRD Performance summary Standardization is completed Network only needs to recognize mobile capabilities, however mobile changes are significant Significantly higher processing required Two receive chains, second antenna in the handset greatly complicates implementation Builds on DARP Phase I (SAIC), but unlike SAIC it can be used for 8PSK (and other HOM) AWGN sensitivity gains of ~3 db (coverage improvement) Co-channel gains of ~7 db (network capacity improvement) Changes to core specifications are essentially complete ( and ) Work is proceeding on new test cases for

39 MSRD Radio Channel Model Each path (transmitter to receive antenna) is independently faded Independent white noise is added to each receive path ρ is used to specify correlation between antenna inputs as antenna separation typically < λ/4. G is used to specify Antenna Gain Imbalance φ is used to add random phase between receivers (as LOs may not be phase coherent between bursts) 2 1 ρ ρ jϕ e 2 1 ρi ρ i 2 1 ρi ρ i

40 MSRD Performance (Continued) Comparison with DARP Phase I and Conventional (DTS-2) MCS-1,tux km-1950MHz,DTS-2,GMSK,a=2,o=2,r=0,g=0dB,rank-2c,n= MCS-5,tux km-1950MHz,DTS-2,8PSK,a=2,o=2,r=0,g=0dB,rank-4c,n= BLER 10-2 BLER MSRD DARP phase I Conventional C/I [db] GMSK/MCS MSRD Conventional C/I [db] 8PSK/MCS5

41 How does MSRD increase system capacity? BCCH carrier repeated over 12 cells 2.4 MHz for GSM Example: A fractionally loaded system may repeat f11 f15 on each of the cells 1/1 re-use pattern high co-channel interference higher system utilization With 1/1 frequency re-use, you can only load these cells up to around 25% capacity each But with 4 of them you get 100% of a 4/3/12 normal re-use cell, so you have doubled your capacity by adding 800 KHz (33%) more spectrum and MSRD mobiles can give you at least 400% of a typical system with only 33% additional spectrum

42 Mobile Station Receiver Diversity Adoption? User experience Slightly improved throughput Slightly better voice quality in low signal areas Larger more costly handset Less battery life Conclusion Hard for manufacturers to make an appealing product at a reasonable price point Difficult to convince consumers to switch to a device that is more expensive, potentially larger and with poorer battery life when the benefits seen by the consumer are slight Statistically significant capacity gains for operators are difficult to achieve without some adjustment to frequency planning Network gains Mobiles can tolerate more interference Significant penetration of MSRD mobiles would allow increased cell loading Reduced need for improved infrastructure in areas of low signal coverage

43 Other Small technical improvements 4 slot up at reduced power for EDGE (part of EDGE anyway we are working on this for the mobile) Battery and Efficiency improvements to slot allocation and scheduling to allow adaptive use of 1,2,3...8 slots down and 2,3,4 slots up as available TBF close bit from either side when the application is done

44 Appendix B Comparison of today s EDGE to UMTS

45 Appendix B.1 Comparison UMTS to EDGE: 64Kbyte file download User data rate (kb/s) Y downloads of 64K file was repeated on a heavilyloaded EDGE network Row s EDGE throughput significantly higher than UMTS for 64K transfer (typical web page) Even under heavy network loading, EDGE still outperformed UMTS for this use case Summary: for data transfers less than 1Mbyte, EDGE is more appropriate than UMTS

46 Appendix B.2 Comparison UMTS to EDGE: 1Mbyte file download User data rate (kb/s) Y downloads of 1Mbyte file was repeated on a heavilyloaded EDGE network Row s UMTS performs better for 1M file than 64K file download but not statistically significantly better then EDGE EDGE has similar throughput, but lower under heavy system loading Summary: for most mobile applications, EDGE is the appropriate technology

47 Appendix C Network performance considerations: RIM s Global Monitoring and Testing System

48 RIM Global Monitoring and Testing System (GMTS) Remote monitoring and data acquisition Specially equipped BlackBerries Carried by certain RIM employees Automatic gathering of information from applications and network Information is sent periodically or on event-driven basis to central server over BlackBerry network Searchable database Locates data by user PIN, IMSI, network or other identifier Sophisticated queries possible based on geographic location, parameter settings or other complex combinations

49 What kind of data can we gather? Really almost anything available to the mobile Application information Application-specific data and parameters Application states, exceptions GPS location data if user permits Samples of voice data and codec parameters Radio stack information States, exceptions of all layers Mobile parameter settings Network settings such as timers, counters, re-try limits, information sent by network in System Information messages Just about anything else that may be useful GMTS is easily customizable If useful data are identified, it is likely that we can capture them and send them back to the database

50 Example: EDGE Random Access Channel timer Radio network congestion and access time are affected The mobile starts T3168 after sending the PACKET RESOURCE REQUEST message on RACH to request uplink resources or additional resources for an uplink TBF The T3168 is stopped after receiving PACKET UPLINK ASSIGNMENT message or PACKET ACCESS REJECT message The mobiles initiates the packet access procedure after expiry of T3168 and restarts T3168 unless it has already been repeated 4 times This continues until a PACKET UPLINK ASSIGNMENT message is received with a valid starting time Count of Networks Timer values vary over fairly wide range: zero to 4000 ms 0 0 ms 500 ms 1000 ms 1500 ms 2000 ms 2500 ms 4000 ms T3168

51 Example: EDGE Downlink data transfer timer This setting varies widely within and among networks When mobile has received all the data blocks on the downlink transfer, it sends a PACKET DOWNLINK ACK/NACK with FBI=1 and starts T3192 If the network has new data to transmit for the MS, a PACKET DOWNLINK ASSIGNMENT or PACKET TIMESLOT RECONFIGURE message with the Control Ack bit set to '1' is sent to MS while T3192 is running. Then MS stops timer T3192, consider the previous downlink TBF released and continue to monitor assigned downlink PDCHs When timer T3192 expires, MS shall release the downlink TBF; stop monitoring the PDCHs and begin to monitor the paging channel Count of Networks Value has severe impact on downlink data transfer rates due to cascading effects on Transport and Application layers 0 0 ms 200 ms 500 ms 1000 ms 1500 ms T3192

52 Many other parameters may be easily collected GMTS is easily customizable to collect what s needed 70 BS_CV_MAX (0,0,0) (1,1,8) (1,1,16) (1,2,20) (1,2,32) PAN_DEC, PAN_INC, PAN_MAX Count of Networks BS_CV_MAX Count of Networks RESELECT_HYSTERESIS Count of networks db -109 db -107 db -105 db -103 db -101 db -99 db RXLEV_ACCESS_MIN -97 db -95 db -92 db -90 db -87 db -85 db -83 db -79 db -61 db Count of Networks

53 Other factors such as backhaul have performance impact Our data analysis efforts identify areas for collaboration Channel capacity per timeslot, (kb/s) Total data channel capacity per carrier, (kb/s) Total data backhaul capacity required, (Mb/s) Comparison to voice backhaul requirement Number Traffic type of carriers Enhanced full-rate speech GPRS data, CS EDGE data, MCS Evolved EDGE, 32QAM

54 Conclusion: Why do we want to gather these data? RIM is not a typical equipment manufacturer RIM is a manufacturer and service provider Network operators are our partners and many times our distribution channel Any information we can gather to help improve our performance provides economic benefit across the value chain Network operator User RIM

55 Mark Pecen Mark Pecen serves as Vice President, Advanced Technology for Research in Motion Limited (RIM), makers of the BlackBerry wireless devices, systems and services. He is responsible for long-term corporate strategy for advanced wireless technology investments, commercialization of applied research and customer collaboration on future technology deployment. Pecen is the founder of the RIM Wireless and Networking Advanced Research Centre and founder of RIM CTO Board. Current priorities include development of technologies for the evolution of existing and creation of Next Generation wireless systems. His labs are active in applied information theory, smart antennas, crosslayer wireless network design and protocols, mobility management, radio link control, statistical analysis and simulation and modeling of wireless systems. Prior to joining RIM in 2005, Pecen held the title Distinguished Innovator and Science Advisory Board Associate (SABA) member, representing the top 1% of technology leaders at Motorola, Inc. Since 1988, Pecen has invented a number of technologies adopted in global standards for the Global System for Mobile telecommunication (GSM), General Packet Radio Service (GPRS), Enhanced Data for GSM Evolution (EDGE), Universal Mobile Telecommunication System (UMTS) and various Wireless Local Area Networks (WLAN) standards. He serves on the boards of several technology-focused industrial, academic and governmental associations in North America and Europe and is the board chairman of Quantum Works, a consortium that funds and manages applied research into quantum computing across a network of universities in Canada. Pecen holds more than 70 patents in the areas of mobile communication, networking and computing, and is a graduate of the University of Pennsylvania, Wharton School of Business and the School of Engineering and Applied Sciences.