TechVista January 06 IIT Madras Wireless Broadband for India ashok Jhunjhunwala [email protected].

37
TechVista January 06 IIT Madras Wireless Broadband for India ashok Jhunjhunwala [email protected]

Transcript of TechVista January 06 IIT Madras Wireless Broadband for India ashok Jhunjhunwala [email protected].

TechVista January 06

IIT Madras

Wireless Broadband for India

ashok [email protected]

TechVista January 06

IIT Madras

• Motivation• Indian Requirements: Urban and Rural• Wireless Fundamentals: going beyond hypes• Current and Emerging Technologies

TechVista January 06

IIT Madras

Motivation

TechVista January 06

IIT Madras

Indian Telecom Market is booming

• Telephones:– 1994: 8 million

• Today: Over 110 million– 5 million lines added

last month– Fastest growing

telecom market in the world

• Broadband– Today : 1 million

• 2010 target: 50 million 0

100

200

300

400

500

2003 2004 2005 2006 2007 2008 2009 2010

Year

Pro

jec

ted

No

s o

f S

ub

sc

rib

ers

in

mill

ion

Number of Telephones in India

020

406080

100120

1948

1951

1961

1971

1981

1991

1997

1999

2000

2001

2002

2003

2004

Apr

il 20

05

Year

Nu

mb

er o

f T

elep

ho

nes

TechVista January 06

IIT Madras

Indian Market boomed• When Telecom became affordable

– Telecom Infrastructure Capex < $ 100 per line • with handset price of $30 onwards and tariff under 2 cents per minute• ARPU of about $ 8

• And easily Deployable: wireless• Competition drove the prices down

• Broadband will boom– At right price points– Competition and easy deployment

• Incumbents can use DSL on existing copper

• New Operators need Wireless Broadband

TechVista January 06

IIT Madras

India’s Requirements: Urban and Rural

TechVista January 06

IIT Madras

Dense Urban India• 200 sq meter dwelling area, 2 homes per dwelling area,

10000 homes per sq Km– All Cities and Towns have a highly dense part

• Market penetration by an operator: minimum 15% – 1500 homes per sq Km

• Broadband towers to serve at least 1 Km radius– Area Served: 3 sq Km– Number of Customers from each tower / cell: 4500

Chennai Madurai ErodeDense urban area 10km x10km 3km x 3km 1km x 1kmDense urban homes 1 Million 90,000 10,000Dense urban population 75% 35% 12%

TechVista January 06

IIT MadrasTo compete withlow end DSL Service

• Lowest end Service on DSL = 256 kbps Always ON

• 4500 homes x 256 kbps from each tower (cell-site) = 1152 Mbps– Impossible on Wireless– But Internet Bandwidth can be shared

• Low-end DSL Service offering in India: 1 G Byte download at $7 per month

256 x 60 x 60 x 100 /8 Kbytes per monthor 11.52 Gbytes

• To compete with this• Define Busy Hour: four hour 25 days in a month• Assume all 1 G byte downloaded by each customer spread over these Busy

Hours• At 256 kbps peak download, one can download 11.5 Gbytes in 100 hours• 11 customers can share a 256 kbps pipe

• Bit rate to be supported in each cell = 1152/11 = 100 Mbps– does not include any business connections: only possible on fibre / DSL

TechVista January 06

IIT Madras

Suburban Areas• Suburban Population Density: 2000 homes per Sq Km

• Let towers serve 5 Km radius or 75 Sq Km– 150,000 homes in a coverage area– 7.5% market penetration: 11200 customers from a tower (in a cell)– 260 Mbps required in a cell: impossible

(11200/11) x 256 kbpsor 260 Mbps

• Let Tower serve 3 Km radius or 27 Sq Km– 54000 homes in coverage area of a tower– 7.5 % market penetration : 4050 customers in a cell– 95 Mbps in a cell

• Suburban areas may need some business connections on wireless increasing the requirement

TechVista January 06

IIT Madras

Rural India

• Rural population density = 250 people per Sq Km

• 25 Km radius implies = 2000 Sq Km– 500,000 people– 1% penetration: 5000 connections– 116 Mbps from a tower (cell)– May be very difficult to provide such a

bit-rate at 25 Km range

• More doable– 15 – 20 Km radius– 1000 Sq Km: 2500 connections

• 8 to 10 connections per village including schools, business, Government

• Number of connections may be less, but higher usage as connections are shared

– 58 Mbps from a tower (cell)

15-2

0km

3 - 4 km

~5 km

Fiber PoP

village

Cellular coverage

• 250-300 villages per PoP

TechVista January 06

IIT Madras

Multi-operator Scenario• India has 4 to 5 operators competing: Also Desirable

• Spectrum is scarce– No operator should expect more than 10 MHz spectrum

• 100 Mbps in each cell site with total spectrum allocation of 10 MHz is a very tough task– Requires a Spectral Efficiency of 10 bps per Hz per Cell site

• And what happens if the DSL operator double the download to 2 Giga byte per month

• Spectrum requirement will double• Can be handled only by making cell size smaller

• And what about high end DSL?– 1.5 Mbps plus supporting Video broadcast / VoD– Impossible on Wireless in years to come

TechVista January 06

IIT Madras

Wireless Transmission Fundamentalsgoing beyond the hype

with apologies to experts for over-simplifications

TechVista January 06

IIT Madras

Single Channel TransmissionWhat does it take to transmit 10 bps per Hz?

How far can one transmit?

• Price increases dramatically beyond a certain Transmit power (Tx device / Battery)

– 1 Watt is optimum today

• Assuming Signal Bandwidth of 1 MHz, Noise at the receiver = k T Δf = 4 x 10-15 Watts

• Signal needs to be five times larger than noise (signal to noise ratio of 5) for digital wireless

– Required rec signal = 2 x 10-14 Watts

• Maximum propagation loss allowed (system Gain)= 5 x 1013 or 137 db

– Higher loss allowed if Bandwith less than 1MHz

0.001 0.01 0.1 1 10

Power Output (W)

Pric

e

1.38x10-23 x 300 x 106

or 4 x 10-15

10-14 W

Noise =10-15 W

1W

Propagation Loss in Rural Areas

• Rural Planes: 40 m tower can cover 15 – 20 km radius

– 10 -12 m pole at village for LOS• Losses are only around 130 dB or 1013

– 6 m pole for Non-LOS (foliage)• Losses around 160 dB or 1016

10-20m

40m10-12m

1W

10-13 W

Noise =10-15 W

10-16 W

TechVista January 06

IIT Madras

Urban Propagation Loss

• Urban Range of 2 - 3 Kms, Built-up Area– Has around 160 dB or 1016 propagation losses for inside building

coverage

• To Sum up: Propagation Losses – 160 dB for NLoS (mobile)– and 130 dB for LoS (Fixed) coverage – will enable up to 2 Mbps (using 4 level modulation) transmission in 1

MHz with no interference

• But System Gain with 1W Transmitter is only 137 dB– Good enough for LoS, but requires more for NLoS communications

1W

10-16 W

TechVista January 06

IIT MadrasHow can we Enhance System Gain?

• Base Station can transmit 3 to 5 W (4 to 7 db gain)– Base Station to Subscriber Communication can have higher bit-rate

• Antenna: can focus beams and enhance power in certain directions– Base Station Antenna gain: 12 to 16 dB– Subscriber Antenna

• ~1 dB for NLoS• 10 to 15 dB for LoS

• Turbo Coding (Error Correction) and hybrid ARQ : 3 to 5 dB

• Signal Processing, multi-antenna diversity techniques can give another 3 to 5 dB

TechVista January 06

IIT Madras

If we had higher system gain?

• Multi-level Modulation

– If Signal is 5 times larger than noise• Can have four distinct levels and

noise will not result in error (bit-error)• 2 bit can be transmitted per symbol (QPSK)

– in 1 MHz can transmit 1 M Symbols per sec or up to 2 Mbps

– If SNR is another 3 times (5 db) higher, one can transmit 3 bits per symbol (8-QAM) or 3 Mbps in 1 MHz

– For every additional 5 db SNR, bit ratel can be increased to 4 Mbps, 5 Mbps, 6 Mbps …

• 16-QAM, 32-QAM, 64-QAM … 256-QAM

• But where can this extra System Gain come from?– And are there penalties?

11

10

10

00

TechVista January 06

IIT Madras

Penalties due to multiple paths

• Time Dispersion– Can be corrected by equalization– If bit duration is too small,

correction not possible• TDMA fails beyond 1-2 Msps

• Frequency Dispersion– different frequency components of the signal fade differently due to

multi-path scattering– CDMA takes advantage of it

• But too many RAKE fingers in CDMA if bit-rate beyond 4 to 5 Msps

• Neither TDMA nor CDMA can have data rates > 5 M Symbols per sec

TechVista January 06

IIT Madras

OFDM for higher data rates

• Divide the transmission band into multiplefrequency bands & place bits on each carrier

– Different frequency components fadedifferently and has different amount of interference

– Signal to Noise (plus interference) ratio will vary from band to band• where SNR is low: Assign less number of bits per symbol (even zero)• where SNR is high: Assign more number of bits per symbol• Distribute the power available to maximise the bit rate

• Channel condition (fade / interference) will vary from instant to instant– Assign more or less number of bits per symbol on dynamic basis– Combats frequency dependent fading and interference

• Technology of future: IEEE802.11, 802.16, Flash-OFDMA, others

frequency

TechVista January 06

IIT Madras

• OFDM and Signal Processing techniques like coding, equalisation, diversity can enable us to have a single point to point link with– Close to 10 bit per second per Hz

– But can we do this for multiple users?

TechVista January 06

IIT Madras

Multi-Channel Communications

Divide area into cells toenable reuse

• Assume 10 MHz spectrum assigned– Say it is assigned to Central cell

• Can a neighboring cell use the same spectrum?• Spectrum reuse 1:1 in every cell

• Use of same spectrum in neighboring Cell can interfere with each other– Can not be reused unless Signal to Interference (SIR) is high– Same channel can be reused only in a far away cell– An Example Reuse 1:7: but then only 10/7 or 1.4 MHz can be

reused in every cell– Only 1:1 reuse will enable all 10 MHz to be reused in every cell

TechVista January 06

IIT Madras

Will Sectorising a Cell help?

• Divide a Cell in three 120 degree sector– Use Directional Antenna to tx and rec in a sector

• Can same channel be reused in different sectors?– Depends on SIR: can be reused if SIR > 10

• Reusable in first case, not in second case– Can not be reused in neighboring sectors unless

orthogonal code (CDMA) is used

• May be reusable in some sectors of neighboring cells– Improves Reuse factor

• Reuse in every cell and each of three sector will imply a Reuse factor of 3:1– 30 MHz available in a cell with 10 MHz spectrum

TechVista January 06

IIT MadrasNew techniques to enhance Reuse

• Opportunistic Scheduling– Schedule order of transmission based on

• who has the best channel conditions– can maximize data-rate to such users

• Optimizing over several sectors and multiple cells

• Multiple Input Multiple Output (Space Time diversity) – Essentially like beam-forming to improve re-use

• Like virtual sectors or pipes

TechVista January 06

IIT Madras

Wireless Fundamentals summary

• Not only important to transmit maximum bit rate in each Hz of available spectrum– Signal Processing, Error Correction, Multi-level

Modulation, OFDM

• Equally important to reuse spectrum as often as one can– Opportunistic Scheduling, MIMO

• Still in infancy, a lot of work needed over the next few years

• To get close to 10 Bits per second per Hz per Cell

TechVista January 06

IIT Madras

Wireless Systems today and tomorrow

TechVista January 06

IIT Madras

Beware of Marketing hypes

– 2.5 G Mobile: 3G-1x/GPRS/Edge• 2 bps/Hz downstream per sector in the middle of a sector• drops to about a tenth as one leaves the center of the sector

as interference from neighboring cells catch up• Effective 0.7 bps per Hz per cell

Computerworld: http://www.corante.com/unwired/

Will 3.5G(HSDPA/HDR) domuch better?

• Sector throughput at best1.1 bits/sector/Hz

The Evolution of UMTS:3GPP Rel.5 and Beyond, June 2004, 3G Americas

HDR

TechVista January 06

IIT MadrasNew Wireless Internet AccessTechnologies• HDR/ HSDPA : today

– 1.8 bps per Hz per cell: mobile– Cost medium

• WiMax : mid 2007– 4 bps per Hz per cell: mobile– 802.11D is only 1.8 bps per Hz per cell: not good– Cost: have to wait

• Flash-OFDM & iBurst: – 4 bps per Hz per cell: mobile– Cost: medium to high

• So do we give up?

TechVista January 06

IIT Madras

LoS Systems: Inherent Strengths

• Disadvantage: LoS enables only Fixed Connection – not mobile– Requires installation Effort for each subscriber

• subscriber patch antenna sees interference only from BTSs behind serving BTS spectrum re-use in every cell – 8 sectors or even 12

• Multi-level modulation

• H and V polarization for each carrier doubles capacity

LoS Systems

• Have much lower path Loss– Can use higher gain subscriber antenna, eight or even

twelve sectors and horizontal and Vertical Polarisation

• Result– But gives 10-20 times longer range as compared to NLoS– 5-10 times higher spectrum efficiency

Line-of Sight LOS

Low interference from surrounding cells

Broadband corDECT• Designed for Indian situation to compete with DSL

– 256/512 kbps dedicated: today• 10 bps per Hz per cell: fixed LoS• Cost very low

– Optimised as a LoS System• Can compete with low-end DSL today

US$ 200 per line deployed Exchange and tower in town Works at 55 C Power requirement: 1 KW

TechVista January 06

IIT Madras

Decreasing costs fromProgressive Volume growth economics

Comparison of AlternativesB

road

ban

dN

arro

wb

and

FixedLocal Area Wide Area

Typ

ical

Use

r D

ata

Rat

e

802.11g family

Constant or high costsFrom Niche applicabilityOr saturation growth

India User

Requirements

1G

2.5G

3G

3G+

WiMAX

Emerging Technologies

Portability

LMDS

Satellite

Dial-up

Cable/DSL/cable-wireless

BB corDECT

corDECT

802.11bfamily

TechVista January 06

IIT Madras

Internet

Modem

Cable Headend

TVCable

Radio tower

And to compete with high end DSL

Cable WirelessDown Stream on Cable Upstream on Wireless

2 Mbps DL and 256/512 kbps UL for each sub

Video on demand possible

TechVista January 06

IIT Madras

To Sum Up

• Broadband is waiting to boom in India

– Requires competition• Broadband Wireless technology competing with DSL can enable

this• Difficult in near future, except with Fixed LoS Systems

– Innovations required towards this

– Requires the right price point

• Broadband will also require a appropriate home device

• The Right Services

TechVista January 06

IIT Madras

Novatium NetPC enabling Broadband

• Connected to a Server on LAN or Broadband– No virus, no back-up required, no upgradadtion every four years,

negligible maintenance– Management at server

– Target price: US$ 80 plus monitor• Works with Windows, Solaris, Unix and Linux Servers