Hardware/Software Design

+
of Embedded Systems
Sudeep Pasricha
Colorado State University
CS/ECE 561 Fall 2010

+Copyrighted Material adapted from Peter Marwedel, Rajesh Gupta, Frank Vahid and Tony Givargis

1
 Welcome to CS/ECE 561
 Instructor
 Sudeep Pasricha [email protected]
 Office Hours: Tu/Thu: 3:30-4:30 pm (ENGR C103A)

 Course Page
 http://www.engr.colostate.edu/~sudeep/teaching/cs_ece_561.htm
 Visit regularly for updates and announcements

 Class Meets
 Tu/Thu 5:30-6:45 pm in ENGR E202

2
 Course Requirements
 ECE 452 or equivalent Computer
Organization or Architecture course

 No graduate level courses necessary

 Assume knowledge of
 Computer Organization
 Programming in C/C++ or VHDL/Verilog
 Some familiarity with Data Structures, Algorithms

3
 Evaluation and Grading
 Homework Assignments – 25%
 Paper critiques, problems, programming, analysis
 Embedded System Project – 35%
FinalPresentation (10%)
Design project report (25%)
 Final Examination – 30%
Closed book
 Class Participation – 10%
Asking/answering questions, attendance, …
 Grading will be on a curve
4
 Reading Materials
 No Textbooks
 Materials will be from various books and journal articles

 References
 Lecture Slides
 Books
 Peter Marwedel – Embedded Systems Design, Kluwer 2003
 Vahid and Givargis – Embedded System Design, Prentice Hall 2002
 Embedded Systems Courses
 Sangiovanni-Vincentelli @ Berkeley, Gupta @ UCSD, …
 Conferences/Journals
 ACM TECS/TODAES, IEEE TCAD/TVLSI, Kluwer DAES,…
 Web Resources
 www.embedded.com, www.eetimes.com, …

5
 Past Course Projects (Spring‘09)
 Dynamically Adaptive Inverted Pendulum Platform
 NASA space grant symposium 2009 First Place Award
 Embedded RFIDs in Cybernetics
 Interactive Teaching Guitar
 Real Time Embedded Database Design
 Roving Wireless Sensor Network
 Novel Hybrid Photonic Network-on-Chip for Emerging
Chip Multiprocessors
 Accepted IEEE/ACM conference paper at Embedded Systems Week
(ESWEEK ‘2009)

6
 Past Course Projects (Fall ‘09)
 IPhone app – Digital Circuit Analyzer
 Hexapod Robot for Environmental Monitoring
 Home Security System
 Task Scheduling on FPGA
 Semi Automated Brewery Control System
 Fault Tolerant Routing for Networks on Chip
 Accepted
IEEE/ACM paper at Embedded Systems Week
(ESWEEK ‘2010)

7
 My Expectations from YOU

CS/ECE 561

 Get started on your project right away!

 Assemble a project team (2-3 members)
 Decide on topic, send 1 pg project summary by Sep 2

 Attend lectures, complete assignments on time

8
 Possible Class Project Topics
 Thermal aware and Low Power Design of Multiprocessors
 Using M5 or SESC simulators; Hotspot thermal analysis tool
 Design a system level floorplanner for 3D ICs
 Extend open source floorplanning tools such as Parquet, Hotfloorplan
 Network on Chips for Multiprocessor systems
 High level design/exploration/synthesis algorithms
 Fault tolerant design techniques (encoding, routing, redundancy schemes)
 RTL Synthesis/Gate level power and performance analysis

 Task allocation techniques for multiprocessors

 To optimize power, performance, thermal profile, reliability
 Processor logic synthesis and design exploration/optimization
 e.g., with open source RTL processor code from Opencores (OpenSPARC, OpenRISC)
 Design of contemporary and novel memory architectures
 e.g. techniques to improve lifetime of phase change RAM (PRAM), STTRAM
 e.g. exploring on-chip cache, memory hierarchy
 e.g. exploring Flash memories – performance, power, lifetime

9
 Possible Class Project Topics
 Wearable computing/Body sensor networks
 Sensors on a human body transmitting data wirelessly to smartphone/PC
 Hack an Android device
 e.g. #1 create wireless P2P communication between devices
 e.g. #2 attach a sensor and create an Android app to make it work

 Hack a Wii/Xbox/PS3
 e.g., hack motion controller/balance board to help train stroke victims
 Hack a Roomba robotic vacuum
 Tack on a camera, hack the firmware …
 Vision Algorithms for motion estimation, object/gesture recognition
 Use a webcam or higher res camera, implement image processing techniques
 Energy Harvesting Platform Demonstration
 Use solar/vibration/sound/thermo/piezoelectric energy to supplement battery in sensors
 Comprehensive literature review (individuals only)
 e.g. RFIDs, cyber-physical systems, embedded cellphone architectures
 Profile and summarize trends for a conference (e.g. ESWEEK 2009)

10
 Course Outline
Hardware Components
Hardware

si gn )
Concept HW/SW De ut, …
Specification Partitioning Layo
,
e s is
yn th
( S

De
sig
Estimation - (C
om n
pil
Exploration a ti
on
, …)

Software Components Software

Evaluation (area, power, performance, reliability, security, …) 11

 Embedded Systems
Automobiles

Medical Entertainment

Handheld

Airplanes
Military

ICS212 FQ05 (Dutt) Introduction 12

 What’s an Embedded System?
 Embedded systems =
 information processing systems
embedded into a larger product
 Two types of computing
 Desktop
– produced millions/year
 Embedded – billions/year

 Non-Embedded Systems
 PCs, servers, and notebooks
 The future of computing!
 Automobiles, entertainment,
communication, aviation,
handheld devices, military and
medical equipments. 13
An Example Embedded System
Digital Camera Block Diagram

14
 Components of Embedded Systems
Memory Controllers Interface

Software
(Application Programs)

Processor
Coprocessors

ASIC

Converters

Analog Digital Analog15

Components of Embedded Systems
 Analog Components
 Sensors, Actuators, …
 Digital Components
 Processor, Coprocessors, Memories, Buses
 Controllers, Application Specific Hardwares
 Converters
 Analog-to-Digital (A2D), D2A, …
 Software
 Operating systems
 Middleware
 Applications (MPEG-x, GSM-kernel, …)
16
 Automotive Electronics Platform Example

CS/ECE 561 FQ08 (Pasricha) Introduction 17

 Automotive Electronics Market Size
Cost of Electronics / Car ($)
1400
1200
1000
800
600
400
200
0
1998 1999 2000 2001 2002 2003 2004 2005

Market
8.9 10.5 13.1 14.1 15.8 17.4 19.3 21.0
($billions)

90% of future innovations in vehicles:

based on electronic embedded systems

18
 What can go wrong: BMW 745i
 2, 000, 000 LOC
 Windows CE OS
 53 8-bit P
 11 32-bit P
 7 16-bit P
 Multiple Networks
 Buggy!
 improper synchronization between two digital electronic control
units
 would stall the engine and lead to a catastrophic accident in
some cases
 Recent Toyota break/acceleration fiasco
19
 Digital Convergence – Mobile Example

Entertainment
Communication

Broadcasting
Computing

Imaging Telematics

 One device, multiple functions

 Center of ubiquitous media network
 Smart mobile device: next drive for semicon.
Industry 20
 Mobile graphics/games
Resolution today ~176x208 – 480x320
In Japan, QVGA (320x240) is the norm
Nokia series 90 is 640x320; Nokia N93 is 320x200
Sony Ericsson S700i, M600i, K800i: 240x320
IPhone, IPod Touch is 480x320
900
Millions of units
Jon Peddie Research
800 Handheld Multimedia Devices report

700 MM phones
Non 3D MM phones
600 Native API based phones

500

400

300
Latest IPhone 4 has 960x640
200

To 1024x768 and 100

beyond in the future 0

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

21
 Growing Importance of ES
.. embedded chips form the backbone of the electronics driven world in
which we live ... they are part of almost everything that runs on electricity
[Mary Ryan, EEDesign, 1995]

22
 Characteristics
 Application Specific
 Applications are known a priori
 Optimize for cost, area, power, and performance

 Digital Signal Processing

 Signals are represented digitally
 Reactive
 Reacts to changes in the system’s environment
 Real-time
 Compute certain tasks before deadline
 Distributed, Networked, …
23
 Characteristics
 Reliability
 Probability of system working correctly provided
that is was working at t=0
 Maintainability
 Probability of system working correctly d time
units after error occurred.
 Safety
 Not harmful for user
 Security
 Confidential and authentic communication
24
Embedded
ICS212 – Applications
Lecture 2

25
 Aerospace
 Flight control
 Stability: real-time differential
feedback loops
 Positioning & navigation
 GPS, INS
 Instrumentation
 Data acquisition, display,
processing, and archive
 Radar
 Communication

26
 Automobiles
 Engine management
 Fuel, ignition, timing
 Emission control
 Instrumentation
 Data acquisition, display,
processing, and archive
 Safety & stability
 Airbags, active control
 Entertainment & comfort
 Radio, A/C, …

27
 Robotics
 Implies autonomous
operation
 N physical degree of
freedom
 Artificial intelligence
 Control heavy
 Mission oriented
 Repair, search, rescue,
investigate, and perform
physically difficult tasks

28
 Process Control

 Industrial automation
 Plant monitoring and production control
 Similar to control systems but with emphasis on
management

29
 Sensor Nets
 Many sensor nodes each
capable of sensing,
computation/storage, and
communication
 Structure safety
 Search and rescue
 Military use
 Energy Efficient
 Distributed Modern Sensor Nodes

UC Berkeley: COTS Dust

UC Berkeley: Smart Dust

UC Berkeley: COTS Dust

UCLA: WINS Rockw ell: W INS JPL: Sensor W ebs

30
 Multimedia
 An exercise in
 Signals, analog to digital conversion,
quantization, sampling, processing, and
digital to analogue conversion
 Information theory, entropy, Huffman
codes, compression, lossless
compression
 Images, audio, video
 Virtual Reality
 Presentation
 Quality of service
 Think lots of data (formats and
standards too)!

31
 Consumer Electronics
 Home appliances
 Yesterday’s appliances: add
computation
 Tomorrow’s appliances: add
networking (Internet)
 Office electronics
 Integration
 Electronic paper (filing, printing,
sending, and receiving)
 Home/office automation
 Common fantasy about the
automated home or office of the
future with lights and appliances
that operate by themselves or
with minimal effort

32
 Network Components
 Stitching LANs
 Bridge
 Connects two parts of the same network
 Router
 Link networks using different network
identities

 Extending ports
 Switch
 Transmit to recipient only
 Hub
 Transmit to all

 Handle large volume of highly

structured data with little
transformation

33
 Medical Instruments
 Perform diagnosis
(screening/evaluation)
 Data collection
 Appraisal of that data
 Developing a plan of action

 Observation or monitoring

 Sensing and instrumentation

 Accuracy and precision

 Other applications:
 Radiation therapy
 Artificial hearts, arms, legs, …

34
 E-Business
 Information processing
systems
 ATM
 Cash registers
 Scanners
 Credit-card readers

 Often the interface behind

a database

 Automation and
convenience

35
Summary of Application Domains
 Looked at a number of application domains

 Large amount of overlap between these domains

 Often each domain has associated standards,

design methodologies, and certification
programs

 The future appears to suggest a fusion of design

differences into a single methodology

36
 Real-Time Systems
 A systems where correctness depends on
logical results and the time the results are
produced
Safety-critical: incorrect operation leads to human
loss
Mission-critical: incorrect operation leads to failed
mission

 J.A. Stankovic et al. “Strategic Directions in

Real-Time and Embedded Systems”

37
Embedded System Design Challenges

 Low cost  Mixed digital/analog requirements

 Light weight  Shrinking time-to-market

 Reliability  Short product lifetime

 Low power  Real-time processing

 Portable  Inherent concurrency

 Complexity  HW/SW co-design

 Ease of use  Secure

38
 Recent Design Challenges
 Design Complexity
 Multiple use-case applications
 Ultra low power and thermal aware
 Highly adaptive
 Active power management (voltage scaling, etc.)
 Alternative energy source (scavenge, solar, etc.)
 Internet aware
 IncorporateRF technologies
 Networking capabilities
 LargerOS, middleware, etc.
 Understanding of many/changing protocols
 Co-operative operation (trashcan & refrigerator)
 Fault Tolerance
 Security
39
 Complexity Driver: Application Pull
1TOPS/W 3D TV
3D gaming

3D ambient
Structured
interaction
decoding
Ubiquitous
3D projectednavigation
Autonomous
display
driving
HMI by motion
StructuredGesture detection
encoding
100GOPS/W Expression
recognition
Gbit radio
Collision
H264 Adaptive avoidance
encoding route
Gesture Language
Emotion
dictation recognition recognition

10GOPS/W
UWB A/V Sign
5 GOPS/W Image
recognition 802.11n
streaming recognition
Mobile Si Xray
Base-band
H264
decoding Auto
Fully recognition
personalization
(security)

2005 2007 2009 2011 2013 2015

[IMEC] Year of Introduction

40
Design Complexity: Moore’s Law
Transistor capacity doubles every 18 months

Source: Intel

Driving force behind

 Electronic systems research and industry
 Proliferation of embedded computing systems
41
 High-end Processors: Intel only?

1,000,000,000 Sony Graphic Synthesizer

IBM Power4 NVIDIA NV40
NVIDIA NV30 GPU
NVIDIA NV25 GPU
100,000,000 Sony Graphic Synthesizer
Intel Pentium 4
IntelPentium III

10,000,000 Intel Pentium II

Number of Transistors

Intel Pentium

Intel 486
1,000,000 NVIDIA NV20 GPU
Motorola G4
Intel 386
AMD Athlon XP
Intel286 Intel Centrino
100,000 NVIDIA NV35 GPU
Intel8086 ATI Radeon X800

10,000
Intel8080
Intel
4004 Intel8008
1000
1970 1975 1980 1985 1990 1995 2000 2005
42
 High-end: Embedded Processors!

1,000,000,000 Sony Graphic Synthesizer

IBM Power4 NVIDIA NV40
NVIDIA NV30 GPU
NVIDIA NV25 GPU
100,000,000 Sony Graphic Synthesizer
Intel Pentium 4
IntelPentium III

10,000,000 Intel Pentium II

Number of Transistors

Intel Pentium

Intel 486
1,000,000 NVIDIA NV20 GPU
Motorola G4
Intel 386
AMD Athlon XP
Intel286 Intel Centrino
100,000 NVIDIA NV35 GPU
Intel8086 ATI Radeon X800

10,000
Intel8080
Intel
4004 Intel8008
1000
1970 1975 1980 1985 1990 1995 2000 2005
43
 Who wants to be a Millionaire

 You double your investment everyday

 Starting investment - one cent.
 How long it takes to become a millionaire
a) 20 days
b) 27 days
c) 37 days
d) 43 days
e) 360 days
f) Lifetime ++
44
 Who wants to be a Millionaire
 You double your investment everyday
 Starting investment - one cent.
 How long it takes to become a millionaire
a) 20 days One million cents
b) 27 days Millionaire
c) 37 days Billionaire
 Believe it or not
 Each of us have more than a million ancestors in
last 20 generations.
 Doubling transistors every 18 months
 This growth rate is hard to imagine, most people
underestimate
45
 Time-to-Market
 Time required to develop
a product to the point it
can be sold to
customers
 Market window
Revenues ($)

 Periodduring which the

product would have highest
sales
Time (months)
 Average time-to-market
constraint is about 8
months
 Delays can be costly
46
Losses due to Delayed Market Entry

 Simplified revenue model

Peak revenue  Product life = 2W, peak at W
Peak revenue from  Time of market entry defines
Revenues ($)

delayed entry
On-time a triangle, representing
Market rise Market fall market penetration
Delayed  Triangle area equals revenue

D W 2W
Time  Loss
On-time
Delayed  The difference between the
entry entry on-time and delayed triangle
areas (shaded region)
47
 Delayed Market Entry (cont.)

 Area = 1/2 * base * height

 On-time = 1/2 * 2W * W
Peak revenue  Delayed = 1/2 * (W-D+W)*(W-D)
Peak revenue from
 Percentage revenue loss
Revenues ($)

delayed entry

Market rise
On-time
Market fall
= (D(3W-D)/2W2)*100%
Delayed  Try some examples
1. Lifetime 2W=52 wks, delay D=4 wks
D W 2W Loss = (4*(3*26 –4)/2*262) = 22%
Time
On-time
Delayed 2. Lifetime 2W=52 wks, delay D=10 wks
entry entry Loss = (10*(3*26 –10)/2*262) = 50%

 Delays are costly!

48
 Cost
 Unit cost
 the monetary cost of manufacturing each copy of the system,
excluding NRE cost
 NRE cost (Non-Recurring Engineering cost)
 The one-time monetary cost of designing the system
 Total cost
 NRE cost + unit cost * # of unit
 Per-product cost
 total cost / # of units = (NRE cost / # of units) + unit cost

• Example • e.g. Iphone 4

– NRE=$2000, unit=$100 • Component costs: $188
– For 10 units • A4 µP: $10.50
– total cost = $2000 + 10*$100 = $3000 • Retina Display: $28.50
– per-product cost = $2000/10 + $100 = $300 • Gyroscope: $2.60
• Accelerometer: $0.65
Amortizing NRE cost over the units
• Selling price: $600+
results in an additional $200 per unit 49
 Course Outline
Hardware Components
Hardware

si gn )
Concept HW/SW De ut, …
Specification Partitioning Layo
,
e s is
yn th
( S

De
sig
Estimation - (C
om n
pil
Exploration a ti
on
, …)

Software Components Software

Evaluation (area, power, performance, reliability, security, …) 50