Download Session III: Precursors for the Next Wave in Communications

SESSION III:
PRECURSORS FOR THE NEXT WAVE IN
COMMUNICATIONS
Session III Precursors for the Next Wave in Communications
Symposium Keynote: Raouf Y. Halim, Convergence Trends in
Communications: Implications for CPCC and Southern California
Ender Ayanoglu, Next Generation Wireless Local Area Networks: How to
Achieve 15 dB Improvement Over Today's Standards Proposals
Ahmed Eltawil, Wireless Broadband Systems: From Theory to Silicon
Payam Heydari, Novel Ultra-Broadband Communications Circuits
Syed A. Jafar, Generalized MIMO: Promises and Limitations
Hamid Jafarkhani, Recent Advances in Space-Time Coding and
Beamforming
Convergence Trends in Communications:
Implications for CPCC and Southern California
Raouf Halim
CEO, Mindspeed Technologies
Corporate Highlights
•Public since 06/2003
•Fabless communication semiconductor provider
•Grew revenues 46% to $119 million in fiscal 2004
•>550 Employees Worldwide, >400 Engineers
•Headquartered in Newport Beach, California
NASDAQ: MSPD
•Leading positions in high-growth enterprise and carrier infrastructure markets
•A broad product portfolio designed into top-tier customers worldwide
•Strategic suppliers: TSMC, Jazz, Amkor, ASE
4
Mindspeed Technologies, Inc.
Serving Top-Tier Customers Worldwide
5
Mindspeed Technologies, Inc.
Our Strategic Focus
Delivering highly optimized, software-rich solutions
Small/Branch Office
Environments
PSTN
Networks
Enterprise
Environments
Mindspeed Product Portfolio
Packet
Networks
Wireless
Networks
Enterprise (Private) Networks
6
Access/Metro (Public) Networks
Leader in VoIP, FTTx, SONET, . . .
Mindspeed Technologies, Inc.
Consumer Convergence Is Real !
Cell phone
Game console
Quad Play
Calculator
PDA
Walkman
Portable TV
7
VoIP, IP Video, Data, and Mobile = Quad Play
Mindspeed Technologies, Inc.
Carrier Convergence:
The Advent of VoIP/Quad Play
End Customer
Quad-Play Service Providers
Transport
(Voice, IP video, Data & Mobile)
Residential
RBOC
SOHO
MSO
IP Core
Wireless
VoIP
VoIPisisthe
thesingle
singlegreatest
greatest
enabler
enablertotothe
thecommunication
communication
convergence
convergencephenomenon
phenomenon
Enterprise
Mobile
8
Mixed
Mixedphysical
physicalmedia
media==Cable,
Cable,HFC,
HFC,
xPON,
FTTx,
Twisted
Pair,
Wireless
FTTx,
Twisted
Pair,
andand
Wireless
Mindspeed Technologies, Inc.
Enterprise Convergence:
The Advent of VoIP/Quad Play
End Customer
Quad-Play Service Providers
(Voice, IP video, Data & Mobile)
Residential
RBOC
TDM / LAN / WLAN Convergence
SOHO
MSO
Enterprise
Wireless
Mobile
9
Mindspeed Technologies, Inc.
. . . Creating A Plethora of New Opportunities
VoDSL
SOHO/ROBO
Gateway
IP TV
End Customer
Quad-Play Service Providers
(Voice, IP video, Data & Mobile)
Residential
Soft IP
Phone
POTs Card /
Acces Gateway
DSL
Home
Gateway
Packet
Cable
Voice Gateway
IP Phones
RBOC
SOHO
Trunking Gateway
C5 Switch /
Acces Gateway
MSO
IADs
Multil
Mode
Handsets
Enterprise
Trunking Gateway
BSC &
Media GW
Converged Switch / PBX
Wireless
Soft
IP
Phones
Mobile
IP Phones
10
Dual cellular/
wifi phone
A new generation of converged SD/HD video systems, wireless/wireline equipment, and consumer
devices
Mindspeed Technologies,
Inc.
Implications for Semiconductor Platforms
3rd generation Mindspeed VoIP SOC - over 300M transistors in 90nm
Flexible Interfaces:
1
TDM
UTOPIA
Ethernet
MII/GMII/RGMII
PCI
DDR
USB
2
3
1
2
4
5
Embedded Cores:
6x VLIW / 64bit DSP
2x 32bit RISC CPU
EMAC
Multi Mb SRAM
UART
USB Host
IPSEC
6
The Rise of Multi-Core Computing
11
Mindspeed Technologies, Inc.
Challenges for 90 nm SOCs
• Economics
– Escalating mask costs: >$500K for 0.15 mm; >$750K for 0.13 mm; and > $1M for 0.09 mm
– Longer development time: 9-18 months from feasibility to first sample
– Increasing total development cost: ~$10M -> $20M
• Design
– Power delivery: 2 amps to the core at less that 25 mV drop!
– Yield, leakage, redundancy, and Soft Error Rate
– 1.0V and below pose major analog design and power delivery challenges
• Packaging
– Signal integrity aware routing reaching practical limits
– Almost 3 orders of magnitude difference in minimum spacing on the die and the substrate of
the package
– Finer pitch peripheral pad arrangement increasing wire bond inductance and resistance
12
Mindspeed Technologies, Inc.
Block Diagram for VoIP / Data Routing
SoC Software Stack
Asterisk Open Source PBX
POTS
Signaling
Packet Signaling
(eg SIP, H.323)
TDM
Signaling
Open Source Router Code (e.g.Linksys)
PBX
Switching
Comcerto Channel Module
Networking and Routing Stacks
(IP,TCP,UDP, PPP, HTTP,ICMP,IPSec etc)
Comcerto Device Driver
Host Kernel (Linux) including packet filtering, crypto API
T.38
FOIP
V.27,
V.29,
V.17
RTP/RTCP or CPS
MPoA
MPoFR
G.711,729a/b/e
G.726,723a
AAL5
FRF.12
ATM
Driver
(WAN
Utopia)
HDLC
Driver
(WAN
HSSI)
G.168 Echo Cancel
TDM Driver
MSP Supplied Software
13
CSP Supplied Software
Over 2 million lines of code
Enet
Driver
LAN
Enet
Driver
WAN
PCI Driver
Eth, PPP Framing, IP, UDP Framing
Dual Port Serial Driver
Caller ID Gen & Det
DTMF Gen & Det
SPI Driver
Voice Packet classifier & switching/bridging
USB Driver
Shared Memory Interface driver
Hardware Crypto Modules
Virtual Ethernet driver (control, data)
CSP Customer Software
Mindspeed Technologies, Inc.
Implications for Southern California & CPCC
Convergence Brings a Plethora of Exciting New Opportunities to SoCAL & CPCC
• Unique with rich communications expertise
– Built primarily from defense legacy
• Unique with deep semiconductor and software expertise
• Unique with five world leading universities (e.g. UCI, Caltech, UCLA, UCSD,
and USC)
• However, we need :
- Tighter coupling between universities and industry
- An ecosystem for entrepreneurial culture
14
Mindspeed Technologies, Inc.
CPCC: Center for
Pervasive Communications and Computing
www.cpcc.uci.edu
Ender Ayanoglu, Next Generation Wireless Local Area Networks: How
to Achieve 15 dB Improvement Over Today's Standards Proposals
Ahmed Eltawil, Wireless Broadband Systems: From Theory to Silicon
Payam Heydari, Novel Ultra-Broadband Communications Circuits
Syed A. Jafar, Generalized MIMO: Promises and Limitations
Hamid Jafarkhani, Recent Advances in Space-Time Coding and
Beamforming
Next Generation Wireless Local Area Networks:
How to Achieve 15 dB Improvement Over Today's
Standards Proposals
Ender Ayanoglu
UC Irvine
The Henry Samueli School of Engineering
Research Symposium 2005
May 23, 2005
BICM-OFDM
• BICM-OFDM can achieve the full
frequency diversity order of L
over L-tap frequency selective
channels
• It has a simple Viterbi decoder
with modified metrics
• If an equalizer is used, then the
channel state information should
be included at the bit metric level
• Modified bit metrics are given as
min |  | | y  x |
2
2
0
10
Original High Complexity Decoding
Modified low complexity bit metrics using CSI Factor
Low complexity bit metrics without CSI factor
-1
10
-2
10
-3
10
-4
10
-5
10
> 18 dB
-6
10
x
 : Channel gain
y : Equalized signal
x : Constellation point
-7
10
25
30
35
40
45
BICM-STBC-OFDM
0
2 bits/sec/Hz per tone 2 Transmit 2 Receive Antennas over IEEE Channel Model D
10
4-state QPSK STTC
4-state 1/2 rate 16 QAM BICM-STBC-OFDM
64-state 1/2 rate 16 QAM BICM-STBC-OFDM
-1
10
-2
10
-3
10
Bit Error Rate
• Multiple antennas can be added
at the transmitter and receiver
using Space Time Block Codes
to BICM-OFDM
• BICM-STBC-OFDM with N
transmit and M receive antennas
achieves the maximum diversity
order of NML in space and
frequency over L-tap frequency
selective channels
-4
10
6 dB
-5
10
-6
10
8.5 dB
-7
10
-8
10
20
22
24
26
28
SNR in dB
30
32
34
36
Single Beamforming
BBO vs BICM-STBC-OFDM over 50 ns channel
0
10
STCB 16 QAM
BBO 4x4 QPSK
BICM-STBC-OFDM 2x2
BICM-STBC-OFDM 4x4 1/2 rate STBC 16 QAM
-1
10
-2
10
Bit Error Rate
• The channel is known at the
transmitter
• Only one symbol is transmitted at a
time over all the transmit antennas
• Single Beamforming achieves the
full spatial diversity of NM over flat
fading channels when N transmit
and M receive antennas are used
• When used with BICM-OFDM, BBO
(BICM-Beamforming-OFDM)
achieves the full spatial and
frequency diversity order of NML
over L-tap frequency selective
channels
• Beamforming provides coding gain
compared to STBC based systems
-3
10
-4
10
-5
10
6.5 dB
-6
10
-7
10
10
12
14
16
18
20
SNR in dB
22
24
26
28
BICM-Multiple Beamforming
(BICM-MB)
•
•
More than one symbol is
transmitted over N transmit
antennas
The diversity order of uncoded
multiple beamforming decreases
with increasing number of
symbols transmitted. The diversity
order of N transmit M receive
antennas multiple beamforming
system is (N-S+1)(M-S+1) when S
symbols are transmitted
BICM-MB achieves full spatial
diversity order of NM while
achieving full spatial multiplexing
of S=min(N,M). We designed an
interleaver/code design criterion
which satisfies full spatial
diversity while maintaning full
spatial multiplexing
BICM-MB vs Zero Forcing with BICM over a flat fading channel
0
10
ZF w BICM 2x2
BICM-MB w 11a int 2x2
ZF w BICM 2x3
BICM-MB w 11a int 2x3
BICM-MB w new Int 2x2
BICM-MB w new Int 2x3
BICM-MB w new Int 3x3-2subs
-1
10
-2
10
Bit Error Rate
•
-3
10
-4
10
14 dB
-5
10
> 20 dB
18 dB
-6
10
10
15
20
25
SNR in dB
30
35
40
BICM-MB-OFDM
BICM-MB
OFDM vs ZF with BICM-OFDM using new Interleaver Design, 16 QAM 1/2 rate, IEEE Channel Model D
0
10
BICM-MB-OFDM 2x2
BICM-MB-OFDM 2x3
BICM-MB-OFDM 3x3 2 streams
-1
10
ZF w BICM-OFDM 2x2
ZF w BICM-OFDM 2x3
-2
10
-3
10
Bit Error Rate
• If the channel is
frequency selective,
then OFDM is used to
combat ISI
• BICM-MB-OFDM can
achieve full spatial and
frequency diversity of
NML while maintaining
full spatial multiplexing
of S=min(N,M), by using
an appropriate
convolutional code.
-4
10
-5
10
15 dB
-6
10
9.5 dB
-7
10
-8
10
20
25
30
35
SNR in dB
40
45
California: Prosperity Through Technology
Symposium
May 2005
Ahmed M. Eltawil
State of the Wireless Industry
1800
Mobile
1600
Mobile
subscribers
Fixed
1400
Mobile Internet
1200
Fixed Internet
1000
Mobile
Internet
subscribers
800
•
In 2004 mobile
subscribers exceeded
fixed subscribers
600
400
200
0
1995
2000
2005
2010
Subscriptions worldwide (millions)
• Can we do even better ?
• If so, why isn’t the potential fully realized yet ?
Sources: Wireless-world-research organization
State of the Wireless Industry
•
There are
numerous reasons
for this delay.
•
A prominent
reason is the gap
between expected
theoretical
performance and
practical issues.
• 3G Services are an example of a technology that
“slipped” by more than two years !!
Sources: Siemens Organization
Experimental Approach to
Wireless Communications
Source
Coding
Channel
Coding
Mod.
Channel
Decoding
Demod.
Device
Level?
Circuit
Architectures
Communication
System Design
SoC
Architectures
 Study
Map functionality
toof
optimal
circuit
topology

feasibility
dynamic
power
control

understand
and
evaluate
comm.
algorithms
Wireless
 Develop,
Identify VLSI
architecture
Channel
 Accurately
For example,
DDFS,
Cordic’s
etc.

model
transceiver
impairments

Voltage
and
frequency
scaling

Identify
memory
hierarchy

between performance
andarea
constraints
 Tradeoffs
Study performance
vs. power vs.

Effects
of
leakage
(e.g.
Multi-V
th )
 Study Hw/Sw Partitioning
Source
Decoding
DAC
ADC
RF
RF
Communication
System
Design
Device
Level
SoC
Architecture
Prototype Testbeds
Circuit
Architecture
Diversity Gains for WCDMA
• Intention:
– Study the impact of space
diversity on WCDMA mobile
terminals.
• Issues:
– Robustness of
communication algorithms
under stress conditions.
– Tradeoff between time and
space diversity.
– Power consumption (2 RX
chains)
Measured Gains
Speed: 120 Km/h
• Ior/Ioc=9 dB
• 384 Kbps DCH
• 3 Multipath
Speed: 3 Km/h
• Ior/Ioc=9 dB
• 384 Kbps DCH
• Flat Fading
•
•
1
1
1 Rx
2 Rx
1 Rx
2 Rx
0.1
BLER
BLER
0.1
0.01
0.001
-20
0.00
1 -13
-18
-16
-14
-12
-10
DPCH_Ec/Ior (dB)
2..5 dB
0.01
10 dB
-8
-6
-4
-12
-11
-10
-9
-8
DPCH_EC/Ior (dB)
-7
-6
Current and Future Projects
•
Opportunistic communication
– Spectrum is highly congested within shared bands and there is a need to
study radios that can optimally utilize the available spectrum.
•
Co-operative radios within Ad-hoc networks
– Within an ad-hoc network, different radios experience different fading
conditions to a base station.
– In a co-operative scheme a multi-hop network can be utilized to improve
aggregate throughput.
•
Wideband channel modeling and emulation
– Increasingly important in advanced wireless standards especially those
that depend on MIMO.
•
Yield issues ?
– Wireless integrated circuits are becoming dominated with memory,
specially in standards that utilize OFDM.
– Yield issues should be revisited in an effort to improve “effective” yield
based on knowledge of desired application, namely wireless.
Novel Ultra-Broadband Communications
Circuits
Payam Heydari
Broadband IC Lab
UC Irvine EECS/CPCC/Cal-(IT)2
5/23/2005
UCI Research Symposium
Challenges in Ultra-Broadband IC Design
• Circuit level:
 Design of silicon-based RF circuits with
BW >500MHz: challenging
 The conventional design techniques (e.g.,
matching for the optimum power gain and
NF) must be revisited
• Transistor/device level:
 Parasitics not negligible
 Lumped models not verified at multi-GHz
frequencies
 Highly layout dependent
 Greater accuracy required
 Passive elements’ losses not negligible
 Scalable models desired
 Technology scaling: stacking not possible
Digital
p- Epitaxial Layer
p+ Substrate
D/A
A/D
Analog
Non-Uniform Downsized Distributed Amplifiers
(2005 IEEE ISSCC)
Rd
VDD
Ld /2
Ld /2
Ld 1
.
2 K
…
Ld /2KN
Ld /2KN
+
Vout
…
CD
CD
(W/L)
+
VBIAS
…
Vin
Lg /2
Lg /2
Lg 1
.
2 K
CD/KN
(W/L)/K
C
c

CD/KN
ISS /K
ISS

Rg
…
…
K=down sizing factor; N = number of stages = 3; VDD=1.8V
Ld = Lg = 363pH; Rd = Rg = 50W; K = 1.5 ; W/L = 180mm/0.2mm
Measurement Results
The die photo
Area: 1.025x1.29 mm2
Drain-Line Inductors
Gate-Line Inductors
A Novel FF-Based Frequency Divider
(2004 Transactions on VLSI )
VDD
RD
RD
Vout1
X
Vin1
Y
M2
M1
VCLK+
M5
VBIAS1
M3
VREF
M8
M7
M9
VBIAS1
M6
M10
M4
Vin2
Latch 2
Vout2
VCLKVBIAS2
M11
Latch 1
 A novel FF-based FD fabricated in a 0.18mm CMOS process for a targeting
frequency of 40GHz
 The latch and the tracking circuits employ two distinct tail currents
 Makes it possible for simultaneous optimization of delay and gain
Measurement Results
The die photo
Area: 650X715mm2
Input signal at 40GHz and output signal at 20GHz
Measured input sensitivity vs. frequency
A UWB Mixer Circuit
(2005 Trans. VLSI  2005 RFIC Symp.)
VDD
ZIF
LIF /2
LIF
LIF /2
Cc
LIF /2
LIF
LIF /2
VLO
LLO /2
LLO
LLO /2
M13
LLO
M23
LLO /2

ZLO
VBIAS,LO
LLO /2
ZLO
VBIAS, LO
Cc LRF /2
VRF
M22
M21
M11
Cc
VLO+
M12
+
LRF
LRF /2
ZRF
VBIAS, RF




Provides a wideband matching for up to 8.72GHz
A two-stage distributed mixer was fabricated in a 0.18μm CMOS
Experiments showed a conversion gain of more than 2.5dB
The DC power consumption was 10.4mW
Measurement Results
The die photo
Measured two-tone test of the mixer at
RF=5.016GHz and LO=4.488GHz
Measured s11
Generalized MIMO: Promises and Limitations
Syed A. Jafar
UC Irvine
The Henry Samueli School of Engineering
Research Symposium 2005
May 23, 2005
Research Interests
•
•
•
•
•
•
Generalized MIMO
Next Generation Technologies
High Mobility Communications
Multi-user Capacity Analysis
Optimality of Simple Transceivers
Low complexity algorithms for optimal resource
allocation.
Generalized MIMO
•
?
Throughput grows as min(M,N)
M
M
?
M
N
M
N
M
N
N
N
Channel Uncertainty, Cooperation and
Usable Degrees of Freedom
–
–
Multiple users, multiple antennas provide additional degrees of freedom.
If these degrees of freedom are usable, tremendous throughput gains
are possible.
The additional degrees of freedom depend on the channel uncertainty at
the transmitter and receiver and the ability to jointly process signals.
With increasing channel uncertainty and without cooperation, the
throughput gains quickly disappear.
Perfect channel estimation, feedback and perfect cooperation are
unrealistic, especially in increasingly mobile scenarios.
The success of future wireless systems requires:
–
–
–
–
•
•
•
Shaping the channel uncertainty.
operating at the best point on channel uncertainty-throughput curve.
near-optimal, joint adaptation, resource allocation and scheduling.
Next Gen. Tech. (CDMA, OFDM)
• Users vary their rates by choosing the
spreading factor, number of codes,
modulation scheme etc.
• Optimal adaptation to maximize
throughput ?
• Power loading used to maximize
throughput.
h2
(0,1)
(0,0)
(1,1)
• Power loading used to control
PAPR.
(1,0)
• Optimal throughput subject to
PAPR constraints.
h1
High Mobility Communications
•
•
•
•
Rapidly varying channel
Mobility and channel knowledge
Low processing complexity
Comparative analysis of
– Coherent schemes
– Non-coherent schemes
– Partially coherent schemes
• Impact on cooperative schemes
• Optimal transceiver design
Recent Advances in
Space-Time Coding and Beamforming
Presenter: Li Liu
Advisor: Hamid Jafarkhani
UC Irvine
The Henry Samueli School of Engineering
Research Symposium
May 23, 2005
Research Focus
• High efficiency coding & modulation schemes for
wireless communications.
• New algorithms for MIMO systems.
• Solutions for both open loop and close loop system.
• Simple/low-cost implementation.
• Strategy: Space-Time Coding
Open Loop Wireless System
Space-Time
Encoder
Space-Time
Decoder
 Coherent S-T Coding: High cost + High performance
 Differential S-T Coding: Simple + lower performance
Differential space-time trellis codes based on extended super-orthogonal codes.
Differential space-time trellis codes based on super-pseudo-orthogonal codes.
Features of the Novel Differential
Space-Time Trellis Codes
 Two classes of S-T trellis codes with high rate and full diversity.
 Superior performance, 1dB gain over previous STTCs.
 Simple decoding.
 Outperform the differential SOSTTCs as well as TC-DUSTM.
Overall, the state of art on differential space-time modulation.
Closed Loop MIMO System
Feedback Channel
Space-Time
Modulator
Beamforming
Algorithm
Channel
Estimation
ML Decoder
 Superior performance + High complexity.
 Traditional Beamforming: Requires accurate channel info.
 STC + Beamforming: Robust with partial channel info.
Novel Space-Time Trellis Codes Using
Channel Phase Feedback
• Flexible design strategy for any constellation, any rate, and
any number of feedback bits.
• Simple feedback, no need for full search on VQ codebook
• Simple ML decoding.
• Low PAPR.
• Good performance, 1.5 dB performance gain over existing
schemes.
The state of art on close loop space-time modulations.