Software Defined Modems for “The Internet of

Things”

Dr. John Haine, IP Operations Manager

www.cognovo.comwww.cognovo.com

What things?

� “20 billion connected devices”

� Manufactured for global markets

� Low cost

� Lifetimes from months to decades

� Different...

Networks

Frequencies

Protocols

� Evolving networks & spectrum allocations

� Versatility & programmability required

2

Multiple systems and bands

� Cellular

� GSM, HSPA, LTE(A),

EVDO, WiMax,...

� Short range

� WiFi, Zigbee, UWB,

� Conventional cellular

bands, 400 – 2600 MHz

� License-exempt – 2400 &

5800 MHz

� “Digital Dividend” � WiFi, Zigbee, UWB,

Bluetooth,...

� Emerging techniques

� “WhiteSpace”

� Other systems

� Tetra & derivatives

� “Digital Dividend”

spectrum

� “Shared” bands

3

Focus on the radio modem

“Apps”IP packets

4

Modem – HSPA, LTE etc

OS – Android etc.

“Photons”Radio

Digital

processing

I/Q stream

Relative computational load for LTE Video

decoder:

~500 MO/s

DSP

arithmetic:

~50 GO/s

5

LTE Modem

Baseband Complexity

200

250

300

350

400

450

500

Bitr

ate,

Mbp

s

LTE 50GOp/s

LTE-A 1Gbps 300GOp/s

60

...Arithmetic

6

0

50

100

150

1990 1995 2000 2005 2010 2015

GSM

GPRS 200MOp/s

HSPA 5GOp/s

HSPA+ 20GOp/s

0

10

20

30

40

50

60

1990 1995 2000 2005 2010 2015

Cum

ulat

ive,

Mod

ulat

ion

& C

odin

g S

chem

es

GSMGPRS

HSPA

HSPA+

LTE

LTE-A

WCDMA

Bit shuffling...

Conventional Modem Architecture

Applications

Protocol

Stack

Layer 1

7

Software

HardwareProtocol

Processor

System

RF

H/w Drivers

DSP

Dedicated

arithmetic

hardware

Conventional Modem Architecture

� Hardware is inflexible

� Different / upgraded / new protocols

� Multi-mode needs separate hardware engines – increases chip

area & cost

� Distributed memory less efficient, increases area / cost

� Difficult development process� Difficult development process

� FPGA prototyping cannot fully model final device

� Software development depends on chip availability

� Hardware bugs need software work-arounds which may not

be ideal

� Cost of large chip developments rising exponentially

� Need to amortise over maximum number of applications

8

Processor SystemApplication software

Product development – the software paradigm

My PC is a...

� Typewriter

� Calculator

� Phone

� TV receiver

� Music player

...often simultaneously!

Software development...

� Is well understood

� Allows continuous improvement

� “Infinite flexibility”

� A platform for unpredictable

products

9

Drivers

“Real-world” interfaces

Example - universal radio/TV receiver for the PC

http://www.crestatech.com/products/ctc200.html

10

http://www.crestatech.com/products/ctc200.html

http://www.mirics.com/

Software Defined Modem (a/k/a SDR)

Processor SystemApplication software

� Hardware architecture becomes standards – independent

� Simplify production logistics – single hardware platform for

multiple standards & standards combinations – the “world

modem”

� Update firmware in the field to fix problems and/or change air

interface

� Extend product lifetime11

Drivers

Multi-band radio module

Barriers to SDM

Computation load – Moore’s Law is not enough!

� As geometry gets smaller, voltage gets lower, clock speeds

static or decreasing

� Overall single processor Mips/MHz/mW stops improving

� Large processor systems have to become multi-core

Standards complexityStandards complexity

� Ever more complex modulation & coding schemes

� Wider RF bandwidth

� Need for legacy support

� Management of task scheduling within and amongst protocols

RF technology

� Wideband operation, multiple frequencies

� Time and frequency duplex modes

12

Lifting the barriers

Vectorisation

� Operate on large blocks of samples (vectors) (e.g. FFT)

� SIMD computing architectures apply same operation on multiple

data values in the same processor cycle - vector arithmetic

Multi-processing

� VLIW computing architectures allow multiple operations on � VLIW computing architectures allow multiple operations on

independent data in the same processor cycle

� Multiple processors on same die to increase MiPs/MHz/mW

� Novel hardware architectures to enable deterministic

performance

Programming

� Automatic compilers to support multiprocessing

� SDM OS to support task scheduling across multiple cores

� System tools for multiprocessing design, development and debug

13

Lifting the barriers

Integrated RF technologies

� New methods such as MEMs for RF switching

� Monolithic filtering technologies such as FBARs

� DPD / ET for efficient wideband power amplifiers

� Multi-mode & multi-band integrated radios

14

Software Defined Modem (SDM)

Design

Environment

Applications

System

SDK

System

integration &

Modem specific

operation

Protocol Plane Control Plane

Layers 2, 3, 4

Protocol

Stack

Layer 1

Data Plane

Physical

Layer

PHY Kernel PHY Kernel

15

PHY coding &

optimisation

SDK

Kernel

SDK

Software

Hardware

integration &

modelling

Generic s/w

execution

environment

Modem Compute Engine

Stack

SDM-OSPHY Kernel

Library

Layer 1

Processor

PHY

Processor

Protocol

Processor

System

RADIO

DigRF

Ideal M2M module architecture

� Flexible broadband RF

� Software defined modem

� All signal processing in programmable

processor

� Air interface defined by software

download, at any stage in product life-

cycle

Modem

Microcontroller

system

Interfaces

cycle

yields...

� One hardware platform for all standards

� Upgrade / change air interface during

end-product manufacture or after sale

16

RF transceiver

DSP

Standardisation & regulatory barriers to SDM

� Mode incompatibility

� Device may need to be multiple-network aware

� Protocols need to permit “listening” for “foreign” air-

interfaces – difficult for “legacy” modes

� Approvals

� Approvals given for defined hardware platform and software � Approvals given for defined hardware platform and software

version – SW version may change frequently as bugfixes &

improvements are made

� Tests may need to recognise operation in “foreign” modes

� Separate approval regimes may apply to modes in the same

device

17

SDM as an enabler for the “Internet of Things”

� Advances in silicon integration & computer

architectures now allow advanced radio modems to be

implemented in software (the “software defined

radio”)

� RF technology allows multi-band / multimode

operationoperation

� Future products will be able to have a single radio

platform to provide radio connectivity across many

diverse networks and bands - massive scale economies

can result

� Standards and approvals regimes need to adapt to the

new possibilities that SDM creates

18