One Approach to IP Cores: Ten Years and Going Strong

by Hal Barbour and Bill Finch, CAST, Inc.
Woodcliff Lake, NJ  USA
   
Abstract :
The benefits of Intellectual Property (IP) based design have proven elusive for designers and especially for providers attempting to build a business around this promising approach. With ten years of success, CAST has shown that it is indeed possible to satisfy users and create a sustainable business selling and supporting IP cores. This article examines CAST’s philosophy and approach to the IP business, highlighting characteristics that may be useful for everyone involved with IP.

One Approach to IP Cores: Ten Years and Going Strong

The concept of developing systems with pre-designed blocks has held promise for several years.

To designers, reusing intellectual property (IP) is a way to overcome the conflicting goals inherent in building more complex systems with tighter time-to-market constraints. To entrepreneurs, the packaging of IP is the enticing foundation of a possible new business.

Yet both designers and entrepreneurs have come to feel that that the reality of IP-based design falls short of its great promise. Their experiences with poor IP quality and deficient product packaging lead some designers to feel they would have been better off just doing it themselves. IP companies struggle to find a common ground in which terms are acceptable to clients and pricing is acceptable to themselves. Often it is difficult to meet customer demands for diverse capabilities with the sort of standardized, off-the-shelf product that is essential for decent profits.

In considering this gulf between the promise and the reality of IP-based design, it is illustrative to look at one company that is perhaps the oldest surviving IP provider, CAST, Inc.

A Needs-Driven Evolution

In 1993, CAST’s founders combined their simulation background from HHB Systems and their systems development experience from Racal-Redac to serve a new market, providing cost-effective pre-packaged simulation models.

Customers appreciated this early form of design automation “intellectual property,” but soon started to ask if it were possible to buy synthesizable versions of the more complex simulation models. The company responded, and started shipping what may have been the industry’s first “IP cores.”

CAST was thus driven from the beginning by the growing needs of challenged designers, not by a clean-sheet business plan. The company also decided to control its growth carefully and to remain privately-funded. These two factors led to an early joint focus on value and customer satisfaction that still pervades the operation today.

Fitting Into the IP Marketplace with Standards-Based IP

As the IP marketplace evolved, differences among various types of products and companies became clearer.

Today the “semiconductor intellectual property” business covers four broad types of design IP:

• Star IP, cores and related products and services around which entire systems are based, mainly processors from providers like ARM.
• Standards-Based IP, cores for popular processors, bus and network interfaces, multimedia, encryption, and other necessary functions typically based on formal or de facto industry standards.
• Hard IP, post-synthesis physical blocks represented as IC mask layouts and used lower in the development process than the other types of “soft” (hardware description language source code) or “firm” (netlists) IP.
• Niche Solutions, IP and services from companies focused on vertical solutions for specific problems, such as high speed telecom protocol cores.

Of these, the Standards-Based IP business is probably the most diverse and arguably the most interesting market segment. This is where CAST participates.

Providers and sources of Standards-Based IP products include:

• EDA companies that offer libraries of IP cores as an adjunct to their main business of selling tools.
• Broad-range IP companies totally focused on providing and supporting IP products.
• FPGA and ASIC vendors offering IP to improve sales of their silicon production services and products.
• Small, often design service-oriented suppliers who try to sell cores they’ve developed under contract.
• Free cores contributed by open source volunteers, typically available on the web with no support.

CAST believes that of these sources of IP, the second — a broad-range, dedicated provider — offers designers the best combination of quality, value, and success. Building a competitive organization and developing desirable products for this market segment has been the firm’s focus the past few years.

The Secret of Success with Standards-Based IP

There is no secret, just a lot of hard work.

When a firm such as CAST is totally reliant on IP for all its revenue, all its energies must be focused first on developing the IP products designers need and, second, on making those designers successful and happy using those products.

Looking at a durable, standards-based IP provider like CAST offers a number of useful lessons in how best to do this.

Provide a broad range for “one stop shopping.”
Designers prefer to develop a relationship with a trusted provider from whom they can get several of the cores they require. This simplifies both business and legal arrangements. It also means the designer spends less time researching and choosing from the varying offerings of multiple vendors. CAST, for example, provides over a hundred different cores. Most are intended for new systems, including popular choices in several commonly-needed areas:

• 8- and 16-bit Processors and Microcontrollers — 8051 variations, 6805/11-compatibles, 165-family, and various digital signal processors (DSPs).
• Bus and Network Interfaces — Ethernet MACs, FireWire, USB, PCI, CAN, LIN, I2C, SPI, parallel ports and I2C and SPDIF audio interfaces.
• Multimedia — JPEG, JPEG 2000, MPEG-4, and various graphic processing “tool box” functions.
• Encryption — AES, DES, MD5, and other popular crypto functions.
• Peripherals — device controllers for NAND flash memory, smart card readers, and video displays; DMA controllers, interrupt controllers, and general purpose I/O interfaces.
• Communications — a family of universal asynchronous receiver/transmitters (UARTs) plus an SDLC global serial channel controller.

Since these cores cover a range of system requirements, designers may find several of the cores they need in the CST inventory. A single System-on-Chip (SoC) product for secure image viewing, for example, might utilize a USB interface, an 8-bit controller, an encryption function, and JPEG image processing.

CAST also serves the surprisingly-large replacement parts market, offering software-compatible cores for previously-discrete chips such as 2901 and 59016 bit-slices, 80186 and other processors, DMA controllers, CRT and keyboard peripheral controllers, timer/counters, and UARTs.

Use and retain application area experts.
No one company can have sufficient experience with the broad range of applications encompassed under standards-based IP.

Some providers hire contractors or purchase and resell IP to cover areas beyond their in-house abilities. This too-often leaves holes in the firm’s ability to support users. A better approach, proven by CAST, is to form ongoing partnerships with technology experts in each application area. This means, for example, that CAST customers can almost always speak with the engineer who actually developed the core they intend to use, a claim many larger companies can’t make.
Customers buy from, seek support from, and generally interface directly with CAST. When detailed technical issues arise, however, CAST becomes an efficient communications conduit connecting users with the core developers. The primary development partners are:

• Evatronix S.A in Poland, focused on processors and interfaces,
• Alma Technologies S.A. in Greece, focused on multimedia, and
• Ocean Logic PTY. Ltd. in Australia, focused on encryption.

Additional partners include divisions of the Fraunhofer Institute focused on buses and controllers, and First Silicon Solutions in the US, providers of on-chip debugging for processor cores.

The unity among these partners is critical: CAST has worked hard to select and nurture partnerships so that the entire “extended family” is integrated and aligned towards producing quality products and ensuring customer success. Communication among them is fast and frequent, using technologies such as instant messaging (IM) and voice over IP  telephony (VOIP).

Offer competitive prices and simple licensing.
IP must be affordable to win in the designer’s “make-versus-buy” calculation. Achieving this while ensuring quality means running a lean operation focused on building and supporting IP. For example, CAST has just a dozen key employees, but fields a virtual team of 60 – 70 experts in ASIC and FPGA engineering, VHDL and Verilog, IP development, marketing, and other critical disciplines.

Besides being cost-effective, the licensing and usage terms for the IP should be easy to understand and comfortable to both the designer and his or her legal department. Straightforward contracts with up-front fees and discounts for reuse (or occasionally royalties when they better suit the customer) have proven the most effective for CAST.

The pricing of IP cores typically reflects some combination of the difficulty in developing the core’s function and the desirability of the core in the marketplace. Competition for sales in some areas can be fierce, and CAST has learned to focus on customers who appreciate the extra levels of support and developer contact that the company provides. (Often designers burned by poor IP in the past are the most appreciative of these significant benefits.)

Deliver complete, high-quality products.
The industry is littered with horror stories about cores not working or not being worth the trouble they took to use them. The lessons are clear: for designers to be successful, the cores they purchase must work correctly, be well documented, and include everything needed to integrate and verify the core.

An independent IP provider like CAST doesn’t survive if it doesn’t learn how to do these things well. This must start with a rigorous core development process, including uniform standards and coding practices shared across all participating groups. Verification must be extensive and include code coverage testing, simulation, and those methods suitable for each particular core. Every core should ship complete with testbenches and suitable scripts and aids ready for the designer to use. Documentation should be comprehensive and well-written, covering function, design, and techniques for integrating the core.

A recent tool for improving core quality is the Quality IP (QIP) initiative from the VSIA. CAST is an early supporter of the QIP, and is modifying core documentation and packaging accordingly.

Provide quick, effective customer support.
Because using IP is rarely a “drop in” process, superlative support is a requirement for customer satisfaction. This support must be fast so as not to interfere with the designer’s timeline, and it must be knowledgeable so as to provide real solutions.

By developing a company culture committed to delivering this level of support, CAST is able to respond to most issues worldwide within 24 hours, and can typically connect the designer with the original developer of a core when necessary.

Summary

Pundits and analysts repeatedly declare that there is no viable business model for IP. Starting with simulation models in 1993 and expanding to cores as the market evolved, CAST has uniquely proven them wrong with over ten years of successful operation as an independent IP provider. By building an efficient, effective organization, focusing on real designer needs, and building and supporting great products, CAST has satisfied hundreds of demanding customers. These customers, in turn, have experienced the competitive benefits of using standards-based IP cores.