August 16, 2004 
  MATLAB & Simulink 
 
While doing research on articles for ESL I came across several
references to MATLAB and Simulink from The MathWorks. These products
were identified as tool for algorithmic development and simulations
in the early stages of system design. I thought I would investigate
further.

The MathWorks was founded in 1984 and employs more than 1,000 people
worldwide, with headquarters in Natick, Mass. The MathWorks produces
technical computing and model-based design software for engineers,
scientists, mathematicians, and researchers across a board range of
industries. In 2003 The MathWorks had revenue of around $240
million. The firm claims more than 1,000,000 customers in over 100
countries, on all seven continents. The MathWorks products are in
use at 3,500 universities.

The two core packages are MATLAB, used for performing mathematical
calculations, analyzing and visualizing data, and writing new
software programs and Simulink, used for modeling and simulating
complex dynamic systems. The company also develops and markets an
extensive family of add-on products to meet the more specific needs
of vertical-markets.

MATLAB is a high-level technical computing language and interactive
environment for algorithm development, data visualization, data
analysis, and numerical computation. MATLAB is used in a wide range
of applications, including signal and image processing,
communications, control design, test and measurement, financial
modeling and analysis, and computational biology. Add-on toolboxes
(collections of special-purpose MATLAB functions, available
separately) extend the MATLAB environment to solve particular
classes of problems in these application areas.

The MATLAB language provides all the features of a traditional
programming language as well as supporting vector and matrix
operations while eliminating the need for memory management and
variable declaration. It also enables visualization of engineering
data via 2D and 3D plots and volume visualization. Functions are
provided for integrating C and C++ code, Fortran code, COM objects,
and Java code with application codes. It can read data from popular
file formats, such as Microsoft Excel; ASCII text or binary files;
image, sound, and video files; and scientific files. It can also
acquire data from hardware devices.

MATLAB tool boxes include Statistics and Data Analysis, Control
System Design and Analysis, Signal Processing and Communications,
Image Processing Test & Measurement, Financial Modeling & Analysis.

Simulink is a platform for multidomain simulation and model-based
design of dynamic systems. It provides an interactive graphical
environment and a customizable set of block libraries that let you
accurately design, simulate, implement, and test control, signal
processing, communications, and other time-varying systems. With
Simulink one can build complete end-to-end simulations, integrating
components such as analog/mixed signal, DSP, digital communications,
and control logic. Through simulation, one can ensure the system
performs to specifications, explore design trade-offs, and tune
parameters to optimize performance. Add-on products extend the
Simulink environment with tools for specific modeling and design
tasks and for code generation, algorithm implementation, test, and
verification.

Simulink enables one to build a detailed block diagram of a system
models by dragging and dropping blocks from the library browser onto
the graphical editor and connecting them with lines that establish
mathematical relationships between the blocks. One can define and
control the attributes of signals and parameters associated with the
model. Signals are time-varying quantities represented by the lines
connecting blocks. Parameters are coefficients that help define the
dynamics and behavior of the system. Solvers are provided to support
the simulation of a broad range of systems, including
continuous-time (analog), discrete-time (digital), hybrid
(mixed-signal), and multirate systems of any size. The model is
simulated and refined until it performs as expected. The refined
system model then becomes the executable specification that's shared
among design teams for system testing, verification, and
implementation.

Real-Time Workshop generates and executes stand-alone C code for
developing and testing algorithms modeled in Simulink. The resulting
code can be used for many real-time and non-real-time applications,
including simulation acceleration, rapid prototyping, and
hardware-in-the-loop testing. You can interactively tune and monitor
the generated code using Simulink blocks and built-in analysis
capabilities, or run and interact with the code outside the MATLAB
and Simulink environment.

On June 2nd MathWorks announced the availability on Windows,
UNIX/Linux, and Macintosh systems of the latest versions of its core
products, namely MATLAB 7 and Simulink 6. MATLAB pricing starts at
$1,900, while Simulink Pricing starts at $2,800.

In October 2003 MathWorks and Mentor Graphics announced the release
of Link for ModelSim, a cosimulation interface that integrates
MATLAB and Simulink into the hardware design flow for FPGA and ASIC
development. It provides a fast bidirectional link between MATLAB
and Simulink and Mentor Graphics' HDL simulator, ModelSim. It
enables direct cosimulation and lets you efficiently verify and
cosimulate ModelSim RTL-level models from within MATLAB and
Simulink. Link for ModelSim supports the developing software test
benches in MATLAB or Simulink for HDL entities, including HDL models
in larger-scale system models developed and simulated in Simulink ,
generating test vectors to test, debug, and verify HDL code against
its original MATLAB or Simulink specification and providing
behavioral modeling capabilities for HDL simulation in MATLAB and
Simulink and verifying, analyzing, and visualizing HDL
implementations in MATLAB and Simulink.

The MathWorks Connections Program directory is a broad collection of
commercially-available products and services based on the MATLAB
technical computing environment. Relevant listings in the directory
are offerings from Altera, Xilinx, Accel, Synplicity and startup
Catalytic.

The Xilinx System Generator for DSP is a plug-in to Simulink that
enables designers to develop high-performance DSP systems for Xilinx
FPGAs. Designers can design and simulate a system using MATLAB,
Simulink, and Xilinx library of bit/cycle-true models. The tool will
then automatically generate synthesizable Hardware Description
Language (HDL) code mapped to Xilinx pre-optimized algorithms. This
HDL design can then be synthesized for implementation in Virtex-II
Pro Platform FPGAs and Spartan-IIE FPGAs. As a result, designers can
define an abstract representation of a system-level design and
easily transform this single source code into a gate-level
representation. Additionally, it provides automatic generation of a
HDL testbench, which enables design verification upon
implementation.

The Xilinx System Generator for DSP bridges the gap between the
high-level abstract version of a design and its actual
implementation in a Xilinx FPGA. New capabilities include HDL
Co-Simulation, Hardware in the loop (HIL), and the ability to
estimate FPGA resources from within Simulink.

AccelChip's high-level DSP synthesis tool provides a direct path
from the MATLAB technical design language to industry-standard FPGA
and ASIC design flows for DSP designers. With AccelChip, designers
can easily load DSP algorithms as script M-files and the tool
automatically determines the design's structure and provides an
intuitive debugging environment. AccelChip automatically performs
floating- to fixed-point conversion and then produces synthesizable
RTL (VHDL, Verilog) models and simulation testbenches to eliminate
the time-consuming and error-prone manual creation process.
AccelChip uses the Resource Description Language (RDL) to optimize
the creation of RTL models based on the resources available within
the targeted FPGA device.

Altera's DSP Builder links MATLAB/Simulink environment to its own
Quartus II environment. Designers can use the blocks in the DSP
Builder library to create a hardware implementation of a Simulink
system modeled in sampled time. The DSP Builder library contains bit
and cycle accurate Simulink fixed-point blocks, which cover basic
operations such as arithmetic or storage functions, as well as
complex functions such as forward error correction or filtering. The
DSP Builder library SignalCompiler block reads Simulink model files
(.mdl) and writes out VHDL files and Tcl scripts for hardware
implementation and simulation. This HDL design can then be
synthesized for implementation in Altera APEX II, APEX E, FLEX 10K,
FLEX 6000, and Mercury device families.

Starting from a floating-point algorithm in MATLAB, Catalytic users
will rapidly convert to fixedpoint versions, analyze them with
accelerated simulation and explore them with MATLAB visualization
features, saving development time and eliminating difficult-to-find
errors. Catalytic's products augment MATLAB with the ability to
define, simulate and test true fixed-point math and include a
convenient user interface to manage the conversion process.
Catalytic's simulation acceleration expands the user's ability to
verify and test his floating-point as well as fixed-point
algorithms. Catalytic is now in an extensive Beta testing period
prior to launching products later in 2004.

The Synplify DSP product automates the implementation of RTL from a
Simulink system-level specification. No hand-coding of any RTL is
required. Included in the Synplify DSP solution is a set of
functional blocks commonly used in DSP design such as filtering
(FIR, IIR), transforms, math functions, CORDIC, signal operations,
memories, and control logic. These functions were designed to be
technology-independent and are tightly integrated into The MathWorks
environment, allowing the algorithm designer to utilize Simulink
features such as discrete time simulation, multi-rate management,
fixed point quantization, scope debugging, and more.

Also on August 11th Cadence announced (see link below) that it and
Cybernet Systems in Japan had co-developed the PSpice SLPS interface
between OrCAD's PSpice technology and MATLAB and Simulink from The
MathWorks. This will allow co-simulation of electrical and
mechanical systems.

MathWorks biggest competitor is home grown systems within target
prospects. Mathematica from Wolfram Research could be considered a
competitor to MATLAB. Mathematica seamlessly integrates a numeric
and symbolic computational engine, graphics system, programming
language, documentation system, and advanced connectivity to other
applications. Since Mathematica was first released in 1988, its user
base has grown to over a million. Mathematica is used today in all
of the Fortune 50 companies, all of the 15 major departments of the
U.S. government, and all of the 50 largest universities in the
world.

Letter to the Editors

Thanks for a good review of the status of data exchange and
interoperability, as it affects the EDA market.

One of my frustrations in this area is that so many promises of
these efforts seem to go unfulfilled.

In my early years of EDA work, the CAD Framework Initiative (CFI)
was supposed to solve all of our intra-vendor problems. I listened
to speakers describe a bright future where design work would be
unencumbered by vendor-specific flows; we'd be free to choose the
best tool for any job, knowing that CFI-developed standards would
allow smooth data transfer.

Of course, it didn't work out that way. I''m sure some standards and
interfaces came out of CFI, but I'm still living in an EDA world of
uneven standards and inconsistent efforts.

Sadly, your column seems to confirm this status. Most efforts are
directed by vendors who rightly have their own interests at heart;
other sincere, independent work is being done - but mostly in
pursuit of a single, particular problem.

Still, I hope for a world where designs can be easily exchanged,
viewed, edited and manipulated as needed.

Regards,

Phil Lindberg
Design Automation, Technical Services Department
The Johns Hopkins University Applied Physics Laboratory



Letter

I enjoy reading your column every week. However, after reading this
week's column on my company (Intel), I feel obliged to comment.

You presented a wealth of facts and data. What I find lacking is
YOUR opinion on how the company is doing and how the company is
projected to perform in the future. An opinion from an expert like
yourself will be far more interesting and valuable to our readers
than a collection of facts, most of which can be mined from various
sources. This is actually what I enjoyed from the previous editor
(Peggy).

Thanks for writing the weekly column and I hope for more interesting
articles in the future.

Dr. Joon Kim from Intel, Portland, OR

Response 

After selecting a topic I do a considerable amount research
devouring as much information as I can find, digesting it, filtering
it, organizing it and then writing and editing a column. I also try
to speak to people at companies related to the weekly topic. A
reader with sufficient interest, time and inclination could
presumably acquire the same information that can be quickly and
easily read in my column. My opinion is reflected in the topics I
choose, the information I included and exclude and so forth. When
you say it would be valuable to express my opinion on how a company
is doing and how a company is projected to perform in the future,
are you speaking from a technological, business or stock price
perspective? I co-author quarterly columns on this website and its
sister website MCADCafe concerning the financial performance of IP,
EDA and MCAD vendors. However, we do not speculate on the future
business results or stock prices. So called experts in these matters
have a difficult enough time explaining the past much less the
future. In the article on Intel I pointed out that AMD had been able
to capitalize on their decision to support 32-bit compatibility in
their 64-bit chip and Intel was playing catch up in this arena. I
also noted recent Intel product delays and reported in a subsequent
column the response from CEO. Incidentally, Intel just announced
that its Sonoma platform launch would be delayed until Q1 due to
some quality issues with the Alviso chipset, a component of the
Sonoma platform. The Alviso chipset promises to boost performance of
Intel Centrino notebooks, which use its Pentium M processor.

Note: After each weekly commentary it is not unusual to get a letter
saying I omitted a particular firm or suggesting that I should
include a particular firm in a later article. My time and the
editorial space simply do not permit me to investigate and report on
every vendor in a given field. I apologize to those who have been
and in advance to those who will feel slighted. Two examples follow.

Letter

I recently joined your newsletter audience. 

I know that the PLM topic is not the most current, but I thought you
should know that you left out another key PLM company - Arena
Solutions. Arena is the fastest growing PLM company in the industry
and the world's only on-demand PLM company.

I would like to discuss further if you consider the topic still
relevant to you audience.

Please give me a call or email to discuss further. 

Regards, 

Mike Nelson
Director, Business Development
Arena Solutions

Response

I am aware of Arena Solutions. The company began in February 2000 as
bom.com, a clever play on Bill Of Material. Unfortunately, this name
became a negative when the dot.com implosion was frequently referred
to as dot.bom. The company was renamed Arena Solutions in January
2003.

Product Lifecycle Management is a combination of CAX, Product Data
Management and Collaboration. The PLM market is composed of firms
that offer CAD products (Dassault Systemes, PTC, UGS, and AutoDesk)
and those that do not (MatrixOne, AgileSystems). The former have
much larger revenue due principally to their CAD business. MatrixOne
and AgileSystems have annual revenues of approximately $100 million.

Arena is much smaller, although the firm reports that it has
achieved nine consecutive quarters of at least 30 percent growth and
annualized revenue growth of more than 300 percent during the
12-month period ending March 31, 2004. Arena numbers more than 150
customers.

Arena Solutions is unique among PLM vendors in that its business
model is exclusively on-demand solutions, aka hosted solutions or
ASP (Application Service Provider). Of course there is nothing to
prevent a System Integrator like IBM or EDS from providing PLM as a
hosted service to a client.

Arena PLM contains three application suites-Product Arena, Sourcing
& Costing Arena and Integration Arena. The Product Arena suite
includes the core Product Definition module and optional modules for
Manufacturer Management and Change Management. Arena offers these
products as a bundle for a one-time cost of $9,995. User licenses
are available for $995 per year with no set-up or installation fees.
The Product Definition module, the foundation for Arena PLM,
provides a centralized environment for managing product data around
a bill-of-materials framework.

The first Sourcing & Costing Arena module is the core Vendor
Management module, a framework for capturing full sourcing
information - suppliers, costs, lead times and more for managing the
AVL, item sourcing, purchasing BOMs, purchase histories and quotes.

Arena Integration adapter links include MFG/PRO by QAD and SolidWorks.

Customers include companies such as Accordian Networks, Accumux
Technologies, Strida Bicycle, IronPort Systems, Deepsea Power &
Light, Metallic Power, Atlas Snowshoe, Color Kinetics and Alien
Technology.

Arena has an eye-catching website at arenasolutions.com.

Letter

I was surprised to see you overlook Runtime Design Automation in
your July 26 column on Grid computing. They were covered favorably
in John Cooley''s DAC 2003 trip report and profiled in depth in Tom
Moxon''s 2002 EET series on design flows.

I was glad to see you covering grid computing as it clearly has
become the default platform for most electronics design: a direction
that was less clear six years ago. Even four years ago a 500 machine
farm was considered large but now numbers 10-20X that size are the
norm for major semi and system houses.

The interesting thing about Runtime is that they own the short run
job niche (e.g. flows that combine jobs that are each less than 15
cpu minutes, such as you would encounter in library
characterization) due to their extremely low startup delay for new
jobs (one to two seconds vs. 30 seconds to a minute for LSF). If EDA
flow designers adopt a "web services" model of knitting together
many short run jobs in a grid the Runtime scheduler and flow manager
may become the default.

I am not an employee or consultant to Runtime and you should
probably get a formal statement from either Frank Bailey their VP of
marketing or Andrea Casotto their CEO.

S. K. Murphy

Response

I also receive a letter from and talked with Andrea Casotto, founder
and CTO of Runtime Design Automation. His company was founded in
May, 1995. This technology was initially developed at the University
of California at Berkeley between 1989 and 1991; it was extended and
tested at Siemens from 1992 to 1995. The company is relatively
small. Its best year revenue wise was $1M. The firm offers two
products Flowtacer/EDA and Flowtracer/NC (NC as in network
computing).

Flowtracer/EDA is a dynamic, automatic, selfcorrecting, multi-user,
multi-platform dependency management system. Flowtracer/EDA manages
any EDA flow such as those used to design vast libraries of cells,
IP cores, multi-million gate ASIC's and SOC's. It discovers
parallelism; dispatches jobs to the local grid and queues limited
resources, such as CPU's, RAM, and licenses. Andrea claims that
their event driven scheduler offers advantages in responsiveness
particularly when scheduling a large number of small jobs.

Flowtracer/NC optimizes the use of hardware and software resources
by automatic resource matching, dispatching, queuing, and load
balancing. Each job is sent to the fastest machine that offers all
the necessary resources. Flowtracer/NC implements all the features
expected of a modern network computing system, including: fair-share
allocation of CPU cycles; job priorities and preemption; and
matching of jobs to computers based on RAM, swap, load, etc.
Flowtracer/NC enables users to identify jobs that have completed,
failed and are still waiting. It provides complete and accurate
metrics that include: unmet demand, wait time, and no demand or idle
time. The largest known single flow manages more than 500,000 files
and 100,000 jobs. The product sells for $600 per machine rather than
per cpu.

Runtime Tracing is the key element of Flowtracer technology. All
input and output dependencies for each tool are computed by the
tools themselves at run-time. The technique used to activate runtime
tracing is a combination of interception of operating system calls;
encapsulation by means of a script; or instrumentation of the source
code.

RTDA customers include Altera Corporation, Ammocore, ATI
Technologies, Azul MicroSystems, Cadence, Faraday, Flextronics,
Microsoft, ON Semiconductor, ST-Microelectronics, SUN Microsystems,
Tera Systems, Virtual-Silicon, and Virage Logic.


Weekly Highlights

Cadence Announces New OrCAD Technology to Help Shorten PCB Design
Cycles; PSpice Enhancements Provide Highest Level of Integration
Between Analog Simulator and MathWorks' MATLAB

Carbon Design Systems Joins 0-In's Check-In Partner Program

Sequence Co-sponsors International Symposium on Low Power
electronics and design

Pericom Semiconductor Delivers Industry Dominant Performance DDR2
Buffer and PLL for Registered DIMMs

LSI Logic DVD Recorder Processor Enables JVC's First '3-In-1'
Digital Video Recorders

Atmel's Offer for Electronic Identity Strengthened by Its New 72
KByte EEPROM Dual Interface Secure Microcontroller

LSI Logic Broadens Processor Offering for RapidChip(TM) Platform
ASIC Designs

NVIDIA and BOXX Introduce First Gelato-Ready Server Node