Published on

Published in: Technology, Business
  • Be the first to comment

  • Be the first to like this

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. Making Wireless Infrastructure Building Reconfigurable Base Stations Alan Gatherer CTO Wireless Infrastructure Texas Instruments
  2. 2. Making Wireless Infrastructure Agenda Why we love reconfiguration Degrees of reconfiguration Where are we today? Software Defined Radio Reconfiguration architecture options Summary
  3. 3. Making Wireless Infrastructure Agenda Why we love reconfiguration Degrees of reconfiguration Where are we today? Software Defined Radio Reconfiguration architecture options Summary
  4. 4. Making Wireless Infrastructure GSM Refuses to Die Millions Worldwide GSM subscribers. 8th 1200 1117 November 2004 1100 1000 Source: “The 900 Numbers Add Up 800 with GSM and 700 WCDMA”, 600 500 400 300 213 105 200 15 18 60 100 0 Analog CDMA GSM + iDEN PDC US WCDMA TDMA GSM base stations showed over 16% and 13% growth respectively in 2003 and 2004 By some accounts, the GSM transceiver shipments in these two years grew over 50% Shipping into both new deployments as well as existing base stations
  5. 5. Making Wireless Infrastructure 2G is Diehard. Is 3G Diehard II? “We’ll use our GSM equipment 2G (GSM) ‘til it rusts….” 2 2.5 2.95 Frequency hopping GPRS SAIC EDGE ??? Channel reuse Vocoder improvements HARQ 3G (WCDMA) 3 3.2 3.5 HSDPA TxDiversity ??? HSUPA MIMO
  6. 6. Making Wireless Infrastructure Rush to Deploy Many competing technologies WCDMA, EVDO, 802.16, TD-SCDMA WCDMA deployed much earlier in its development cycle than GSM WCDMA / EV-DO continue to develop Competition with each other in Japan Competition with the promise of 802.16
  7. 7. Making Wireless WCDMA (R99) vs GPRS (R97) Infrastructure Change Requests / Qtr: June ’97 – Dec 04 Source: Agilent Technologies 1200 W-CDMA Total Early R99 WCDMA deployment 1000 GPRS Total W-CDMA Test Non-editorial CRs 800 GPRS Subscriber 600 Growth Early GPRS deployment 400 Global GPRS 200 0 99 00 01 02 04 02 00 01 97 04 97 Ju 8 03 98 99 03 9 ec ec ec n ec ec n n n n ec ec n n n ec Ju Ju Ju Ju Ju Ju Ju D D D D D D D D Source: ETSI CR Database 11 Jan 2005
  8. 8. Making Wireless Infrastructure No Convergence in 3G cdmaOne cdmaOne CDMA2000 CDMA2000 CDMA2000 IS-95-A IS-95-B 1X 1xEV-DO 1xEV-DV Voice Voice High Capacity Voice 2.4 Mbps Packet >2.4 Mbps Packet 14.4 kbps 64 kbps Packet 153 kbps Packet Higher Cap Voice CSD & PD Rel. A-307 kbps TD-SCDMA 802.16 TDMA ?? IS-136 EDGE Voice Up to 384 kbps 9.6 kbps CSD Packet W-CDMA GSM GSM GPRS W-CDMA HSDPA Voice 114 kbps Packet Higher Capacity >2 Mbps Packet 9.6 kbps CSD Voice Higher Cap Voice 2 Mbps Packet 1999 2000 2001 2002 2003 2004 2005 2006 GSM: Global System for Mobile Communications HSDPA: High Speed Downlink Packet Access CDMA: Code Division Multiple Access EDGE: Enhanced Data rates for GSM Evolution TDMA: Time Division Multiple Access EV: Evolution; DV: Data and Voice; DO: Data Optimized GPRS: General Packet Radio System W-CDMA: Wideband CDMA: General Packet Radio System
  9. 9. Making Wireless Infrastructure We Like to Make Money! Total Cost FPGA (Fixed and Variable Cost*# of Units Units) Total Cost DSP Total Cost Total Cost ASIC NB: Need to watch: Fixed Cost Power ASIC Form factor Fixed Cost DSP Fixed Cost FPGA Volume (Units)
  10. 10. Making Wireless Infrastructure Caveats For Cost Per Unit Power Continues to be critical to base station designs. Base stations are often in enclosed spaces and experience wide temperature variations. These systems need to operate with high reliability for years Means some reconfigurable devices. Even some DSPs fail the basic power limit for a hot spot Low power efficiency of power amplifiers has forced the industry to focus on Power Amplifier (PA) efficiency and not flexibility Cost There continues to be significant cost pressure on base station equipment. The cost of 3G systems is dropping rapidly. WCDMA has already fallen below GSM in cost per channel Complexity Development time
  11. 11. Making Wireless Infrastructure Cost 35% 30% PA is 25% consistently 20% the high cost item, so focus in 15% PA is cost 10% reduction, not reconfiguration 5% 0% 2G old 2G today 3G today DBB AFE PA Good reconfigurable target Not so good (yet)
  12. 12. Making Wireless Infrastructure Complexity: Data is Harder Than Voice Voice modem Data modem resource scheduling resource scheduling User User User User User User User User 1 2 3 4 N 1 User 3 4 2 resource resource Static for the life of the user Dynamic resource requirement Similar across users Widely varying resources No fragmentation Fragmentation possible Deterministic scheduling Dynamic scheduling Adaptive modulation, HARQ
  13. 13. Making Wireless Infrastructure Development Time Maximum Available Revenue Total Lost Revenue Due to Revenue Delayed Entry Maximum Revenue Due to Delayed Entry Market Rise Market Fall Development Ramp to Time Begins Time to Market Production Total Market Window
  14. 14. Making Wireless Infrastructure ASIC Industry Design Starts Declining 1,900 1,800 1,700 ASIC Design 1,600 Starts 1,500 1,400 1,300 1,200 2002 2003 2004 2005 2006 2007 Source: iSuppli 2Q04
  15. 15. Making Wireless Infrastructure FPGA Gaining in Wired/Industrial 7 PLD 15.8% ’03-’08 CAGR 11.1% 6 ASIC 11.0% ’03-’08 CAGR 5 2004 Sales 4 $B 4.7% 8.0% 3 4.0% 13.6% 2 15.6% 11.5% 1 2.2% 6.1% 5.8% 0 Computer Wireless Consumer Wired Auto Industrial Source: iSuppli 3Q04
  16. 16. Making Wireless Infrastructure Cost of ASIC Affecting Design Starts Example: Saving 25% of Unit Cost with ASIC 10,000KU Wireless Units Shipped to Payback Investment Computer Consumer Wired Access/ 1,000KU Basestation Wired Datacom Wired 100KU Telecom Investment $15M $10M $ 5M 10KU $10 $100 $1,000 Unit Cost of ASIC
  17. 17. Making Wireless Infrastructure Agenda Why we love reconfiguration Degrees of reconfiguration Where are we today? Software Defined Radio Reconfiguration architecture options Summary
  18. 18. Making Wireless Infrastructure What is Reconfigurable? Common base station MOST Common modem cards Common footprint Common chassis Common programming platform Common elements LEAST
  19. 19. Making Wireless Infrastructure Common Base Station On-the-fly reconfiguration of base stations Still very expensive Carriers generally don’t need this level of flexibility Some rural carriers see value when roaming as the dominant revenue source Hoteling? Vanu Uses off the shelf servers Cost reduction for small quantities as off-the-shelf DSP cards tend to cost more (but wait for ATCA), even though DSPs tend to be cheaper and lower power per function Lyrtech DSP/FPGA board “Universal Base Station” system, using flexcell ™ and SignalMaster software-design radio (SDR) platforms multi-band (AMPS, GSM, WCDMA, VHF and 802.11) ACT Produced a generic DSP / FPGA board for multi-standard
  20. 20. Making Wireless Infrastructure Common Base Station Issues Cost RF and Analog Cell planning for a 100% 100% reconfigurable network? Cell planning and site (As % of initial value) (As % of last value) 80% 75% acquisition is a major issue Revenue 60% Cell planning even occurs Cost 50% 40% for WLAN in the enterprise 25% Backhaul network 20% differences IP? Time (Years) What about mobility An idea whose time is yet to come Frequency agile standards such as 802.22 might have some reconfigurable elements to them
  21. 21. Making Wireless Infrastructure Common Modem Cards Much more likely in the mid term Apples to oranges 1 lb apples to 1 kilo apples 64 WCDMA may imply 72.6 GSM (I guessed) This isn’t a useful number Still, if you have a new standard that isn’t shipping in volume…… Inventory savings Reduced NRE
  22. 22. Making Wireless Infrastructure Common Footprint A single board for multiple standards with a few missing components that define frequency range and bandwidth
  23. 23. Making Wireless Infrastructure Common Chassis Use the same rack and backplane e.g., CPRI, OBSAI, ATCA Use the same controller cards Especially with a common network termination e.g., OBSAI Use the same power supply and clock distribution cards
  24. 24. Making Wireless Infrastructure OBSAI, CPRI, ATCA OBSAI (open base station architecture initiative) Defines the whole base station mechanicals and electric interfaces, including the control and transport cards. Some software messaging is also defined CPRI (common public radio interface) Focuses on the baseband to analog card interface ATCA (Advanced Telecom Computing Architecture) Optimized for telecoms transport Probably too expensive for radio baseband uATCA may be more suitable for radio A great prototyping platform at least, using RapidIO as the backplane for scalability
  25. 25. Making Wireless Infrastructure Common Platform Make the modem look as similar as possible at the software level across different standards Hardware abstraction Common APIs Reuse of common software blocks Same software tooling allows engineers to move between projects This is what SDR really is for wireless infrastructure!!
  26. 26. Making Wireless Infrastructure Agenda Why we love reconfiguration Degrees of reconfiguration Where are we today? Software Defined Radio Reconfiguration architecture options Summary
  27. 27. Making Wireless Infrastructure Example: WCDMA Preamble Post Detect Process Messaging Resource Allocation Post & Control Search Process Control Channel MRC Despread est Symbol Rate Processing Data MRC Despread Needs some acceleration. Maybe software Software Often ASIC
  28. 28. Making Wireless Infrastructure Digital Baseband Trends WCDMA is the most complex CDMA chip rate generally needs some sort of acceleration EDGE / GSM, TDSCDMA, 802.16 can be done in software More data (and less voice) lead to more software- oriented solutions Less chip rate processing, more symbol rate processing HARQ and MAC functions now on the modem
  29. 29. Making Wireless Infrastructure Digital Up and Down Convert Signal of Signal of Signal of Signal of Internet Internet Internet Internet 2 GHz IF Q Q Specialized DUC/DDC chips on Analog RF to IF A/D Filtering & Decimation Gain/AGC & Formatting Chip/Symbol Processing Baseband Processing the market that are Analog Down parameterizable Conversion NCO (Tuning Freq) Digital Downconverter Channel Big OEMs may do their own ASIC GC5016 or GC4016 FPGAs also used IF 2 GHz Predistortion and crest factor Baseband Generation Formatting & Gain RRC Filter Interpolation D/A Analog IF to RF PA reduction are now Analog Up Conversion being used, but Q Q Digital Upconverter NCO (Tuning Freq) CFR & Optional focus is on cost – Channel DPD Crest Factor Reduction & not reconfiguration Q Q Digital PreDistortion Processing GC5016 or GC4116
  30. 30. Making Wireless Infrastructure Trends From Analog to Digital VCO are typically fixed SAWs are fixed Zero IF to remove SAWs Is used on Tx Rare in Rx for BTS due to stress on analog components Benefit: Carrier selection at BB PA and LNA are fixed in center frequency and bandwidth
  31. 31. Making Wireless Infrastructure Agenda Why we love reconfiguration Degrees of reconfiguration Where are we today? Software Defined Radio Reconfiguration architecture options Summary
  32. 32. Making Wireless Infrastructure SDR vs Reconfigurable Base Station Software defined doesn’t imply reconfigurable SDR is an object oriented philosophy of system design and may or may not imply reconfigurability: Reduces NRE. Allows smooth SDR migration to next platform and Here is lower time to market nice, too SDR gives hardware abstraction SDR allows software elements Want to be here to be plug and play SDR enables reconfigurability in many cases Reconfigurable Reconfigurable implies the design takes account of multiple radio standards
  33. 33. Making Wireless Infrastructure The CORBA Gatekeeper Applications Core Framework (CF) OE Commercial Off-the-Shelf (COTS) Non-CORBA Non-CORBA Non-CORBA Modem Security I/O Components Components Components Physical RF API Modem Modem Link, Network Security Security Security Link, Network I/O I/O Components Adapter Components Adapter Components Adapter Components Adapter Components MAC API LLC/Network API Security API LLC/Network API I/O API Core Framework IDL (“Logical Software Bus” via CORBA) CORBA ORB & CF Services & CORBA ORB & CF Services & Services Applications Services Applications (middleware) (middleware) Software Operating System Operating System “bus” Network Stacks & Serial Interface Services controls all Board Support Package (Bus Layer) Board Support Package (Bus Layer) transactions Black Hardware Bus Red Hardware Bus
  34. 34. Making Wireless Infrastructure How the ORB Works All components, whether software or hardware, have an Applications Programming Interface (API) that allows the “client” application to access them in a regular, object- oriented way via the “logical software bus” In this manner the objects can be added and removed easily At the heart of this easy addition and subtraction is the Common Object Request Broker Architecture (CORBA). Communication between objects is managed by the Object Request Broker (ORB) Database of object “I need ORB descriptions some O” C “O please” O
  35. 35. Making Wireless Infrastructure How the ORB Works The objects do not have to be aware of each other, only the ORB In order to be recognized by the system, each object has a domain profile The ORB, along with the core framework (CF) that defines the interfaces is a “middleware” Middleware is common in base station modems today But not usually this complex
  36. 36. Making Wireless Infrastructure Agenda Why we love reconfiguration Degrees of reconfiguration Where are we today? Software Defined Radio Reconfiguration architecture options Summary
  37. 37. Making Wireless The Dependable Programmable Infrastructure Processor DSP or GPP Differences are somewhat subtle these days DSPs tend to be . . . GPPs tend to be . . . Focused on embedded market Focused on computing market Lower power per function Focused on decision making Lower cost per function Focused on arithmetic Good for housekeeping functions that also and control require control
  38. 38. Making Wireless The Dependable Infrastructure Programmable Processor A really good platform for the type of OO C, C++, SDR programming that everybody does “Mild” multi-core is appearing e.g., TNET3010 Less than 10 cores Usually aimed at DSP farm applications e.g., Transcoding, RNC, GGSN Can be light on MIPs due to limited parallelism
  39. 39. Making Wireless Infrastructure Good Ol’ FPGA Large amounts of parallelism Programming style like RTL Needs support to work in a SDR environment Traditionally used for glue logic Embedded multipliers and cores are moving FPGAs into DSP Power and Cost have been traditional impediments
  40. 40. Making Wireless Infrastructure “Reconfigurable” Processors A catch all for anyone who isn't GPP/DSP or FPGA e.g., Morpho, Freescale, Picochip, PACT Usually have parallelism between DSP and FPGA Claim to be easy to program but this is far from clear Usually involve a mixture of C code and RTL like structured coding Debug can be difficult Not suitable for control code
  41. 41. Making Wireless Infrastructure “Reconfigurable” Processors Fall into 2 categories Parallel array of programmable elements Parallel array of ALUs Issues Local connectivity Utilization of resources in real applications Programmable Elements Parallel Elements
  42. 42. Making Wireless Infrastructure TI’s View: Why DSPs Real time algorithms need a device which effectively handles both math intensive functions and control code functionality !! 1. Multiplication and 2. Control and Real-time Arithmetic Operations Decision Making Massively GPP Parallel RISC FPGA/ASIC Reconfigurable DSP is best solution for cost-effective, high-performance signal processing problems
  43. 43. Making Wireless Infrastructure DSP Block Diagram: TMS320TCI6482 C64x+ Core WI Chip rate INST. inst L1P L1D (32k) (32K) L2 (2MBytes) EDMA 3.0 (Switch Fabric) GPIO I 2C PLLs Boot VCP2 TCP2 Timers Others Gigabit HPI Serial DDR2 VLYNQ McBSP UTOPIA EMIF 64 Ethernet PCI-66 RapidIO EMIF
  44. 44. Making Wireless Infrastructure Agenda Why we love reconfiguration Degrees of reconfiguration Where are we today? Software Defined Radio Reconfiguration architecture options Summary
  45. 45. Making Wireless Infrastructure Why/What is Reconfigurable Reconfiguration being pushed by: Standards evolution Complexity Cost NRE and time to market Not by multimode modems Reconfiguration comes in many forms − from software to common mechanicals
  46. 46. Making Wireless Infrastructure Progress SDR is an OO approach to complex system development It enables reconfiguration rather than being driven by it Most architecture progress is in DBB: More DSP / GPP MHz More DSP focused FPGA New reconfigurable architectures
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.