Not Your Father’s Base Station: Meeting The Data Challenge That Shannon Can’t Help

Not Your Father’s Base Station Meeting the data challenge that Shannon can’t help

Alan Gatherer TI Fellow and CTO of High-Performance & Multicore Processing Texas Instruments

Agenda • Modulation meets the end of the road • The network has gotten very messy • Self-optimization to the rescue • Femto cells: part of the solution or part of the problem? • Some math to show how smart we are • Conclusions

Sorry, but we’re out of ideas…. TDMA! FDMA! CDMA! OFDMA!

I think I’ll go play with my mice

SDMA/ MIMO!

• Point to point capacity we have nearly got covered • Future of capacity is at the network layer

And we don’t need more buttons to press • WCDMA followed by HSPA introduced data to cellular • We have LOTS of options – Data rates, FEC, scheduling, QoS, latency

• We have made for a very hard system to manage “The belief among operators is that 3G represents a missed opportunity to automate network processes, and that much of the ongoing cost to configure and manage Node Bs, radio network controllers, and core network elements is accounted for by the need to allocate expensive technicians to mundane, yet cumbersome, tasks.” – Heavy reading

Plan B: more cells, smaller cells ``Wireless capacity has doubled every 30 months over the last 104 years’’ – Martin Cooper, ArrayComm

• Million-fold increase since 1957 – 25 x from wider spectrum – 25 x from slicing spectrum, better modulation and coding 01001  1010110101

–1600 x from reduced cell sizes and transmit distance

But that’s cheating!

5

So the answer is….it’s really, really messy Relays, Distributed Antennas

Big Cells Macro BTS

Small Cells Pico BTS

Various Backhaul

Overlay

Residential Femto

and…even more messy to manage!

Self-optimization to the rescue •

SON functions: “Aims to configure & optimize the network automatically to minimize operational effort and improve network performance” - 3GPP TS 36.902 SON Use Cases and Solutions

•

Why SON? – Number & structure of network parameters is complex/large – Parallel operation of multiple technologies such as 2G, 3G, LTE, etc. – Rapid expansion of base stations (& Home NB’s/eNB’s) require minimum human effort

•

Aspects of SON – Self Configuration: Automatic installation & basic configuration for system operation (for NB/eNB) – Self Optimization: UE & eNB measurements are used to auto-tune the network operations – Self Healing: Detection & automatic failure recovery mechanisms

We all know it’s a good idea • Industry momentum – Operators: T-Mobile, Orange, KPN – Manufacturers: Nokia, Motorola, Ericsson, Nortel, NEC, Huawei, NTT DoCoMo

• Standards momentum – 3GPP • NSN, Huawei, NEC, Ericsson, Samsung, T-Mobile, Orange • RAN2, RAN3 – use cases and procedures • SA5 WG - requirements

– Next Generation Mobile Networks (NGMN) • T-Mobile, NTT DoCoMo, Nortel, KPN

– Femto Forum • Huawei, NEC, ALU, Ubiquisys

– SOCRATES FP7 [EU] • Vodafone, Ericsson, NSN, University research

Self organizing network (SON) SON framework for eNB eNB Power on

Basic Setup

Configuration of IP address Association with a GW

Self Configuration (Pre -operational state )

Authentication Software & configuration data

Initial Radio Configuration

Neighbor list configuration Coverage parameters configuration Others (Physical cell identity configuration )

Self Optimization (Operational state )

Optimization

Neighbor list optimization Capacity & coverage optimization Others (energy savings , interference reduction , mobility optimizations )

Self Healing (Operational state

)

Self -healing

Failure detection

& localization

Healing schemes Others

Femto without SON? • SON might give you – – – –

Idiot install Dynamic reaction to traffic change Long term “cognitive” optimization of the network Fallback in the event of failure

• Without these can you really add femto to your network? – Well, probably yes (because people have done it) – But…

Halloween costume idea

Dress up as a femto cell. They’re pretty sc

Femtocells: standardization efforts • LTE-A is the first standard to explicitly address femtocells and relays – Treats femtocells as a relay through the network

• 3GPP pursuing “femto aware” – Heterogeneous networks – In-band backhauling for relays – Inter-cell Interference Coordination (ICIC) – Lots of room for research!

• But.. today femtocells are just little base stations • $100 BOM leads to only $20-30 of DSP physical layer • And don’t even think about MIMO!! • Will the volumes justify this price point?

From concept to reality….

Concept

Reality

3GPP RAN4. “Downlink interference from a closed-access femtocell will result in coverage holes in the macro network….considerably more significant than when the femtocell is deployed on a dedicated carrier.” 13

On the other hand…. • Reduced TX-RX distance – Spatial reuse – Service poor coverage areas

IP

IP IP

• Femtocell differences – Connected via DSL or cable modem – Customer pays for their own backhaul – Customer pays for air conditioning and power – Customer may pay for equipment too?

14

[ChaAndGat08] V. Chandrasekhar, J.G Andrews and A. Gatherer, “Femtocell networks: a survey,’’ IEEE Communications Magazine, Sep. 2008

So there is some room for research… • How to do power control? – Assist users with poor channels – Limit interference to neighboring cells

• Overlay macro-femto deployment – Interference “couples” target Signal-toInterference-Plus-Noise Ratio (SINRs) at macrocell and femtocells

• Femtocells – Placed by end-consumers in self-interest – Femto-users seek higher data rates Key Questions 1) Given N arbitrary FC SINR targets, what is the highest cellular SINR target? 2) How does relative placement of femtocells affect per-tier SINR targets? 15

System model • Central Macrocell (B0) • N femtocells Bi, • One user/slot/cell • All in the same frequency band

Min. SINR target

Current SINR

Transmit power Channel power, user j to Bi

Background What was the question again? – Given a desired SNR for the femtos, what is the largest macro SNR?

Max-Min SINR Assignment [Zander92, Grandhi94]

Max-min SINR

Normalized channel powers For convenience, define the normalized cross-tier channel powers given as below: Normalized interference of macro-user onto femtos

Normalized interference of femto-users onto macro

Shared spectrum two-tier network Consider a two-tier network with target SINRs γc and γf at macrocell and femtocell hotspots. Rewrite

Interference from femtocell users to macrocell Normalized channel power between femtocells Interference from macro-user at femtocell

Highest macro-femto SINR targets

Macro SINR

Normalized intra & cross-tier channel power

Femto SINRs

Tower and maze approximation • Femto sees 2 walls to neighboring femto • Macro sees one wall to each femto • So drop the femto to femto matrix term

Lower bound with common femto SNR • If there is a common Femto SNR then:

Link Budget

• Gives an interpretation of the link budget available in terms of the normalized interference

Decibel Link Budget

SINR Contours

Square grid with 64 femtocell 60

. Isolated configuration

40

Nearly 40 dB Link Budget Loss

A

Гc (dB )

20

0

B

-20

. Co-located configuration

-40 Blue solid : Exact (Theorem 1) Black dotted: Link Budget Apprx. -60

0

5

10

15

20

Гf (dB )

23

[ChaAndSheMuhGat08] 1. Link budget analysis is simple and accurate 2. Motivates interference-aware femtocells

25

Strange effect: correlation of interference patterns • Maximize link budget by “de-correlating” the two interference vectors – You don’t want the normalized interference to peak in the same place for macro to femto and femto to macro

• When the macro user is close to its tower it sees a similar pattern of interference as its tower creates – Because all cell phones are close to their towers – Hence the vectors are correlated

• When the macro user is away from its tower it sees a different pattern than the tower is creating – Hence vectors are uncorrelated

• Result improves as the macro cell user moves away from its tower!!

Conclusions • The winners in next-generation base stations will solve the messy networking problem • SON is essential to the solution -- and a hot topic • Femtocells are just part of this picture • Without managing interference, femtocell deployments are likely self-defeating • There is still a lot to learn in this area

Acknwledgements V. Chandrasekhar, J.G. Andrews, Ramanuja Vedantham, Sandeep Bhadra 25

Thank You!

Alan Gatherer TI Fellow, CTO, High Performance & Multicore Processing email: [email protected]