Welcome to the jungle: early adoption in embedded development
Successful early-adoption trailblazers bring the right types of resources to bear on a problem, know how to recognize when they are heading into a dead end, and exercise flexibility to rapidly redirect or maintain the group's focus on the overall goal.
By Robert Cravotta, Technical Editor -- EDN, October 30, 2003
|
"First to market" is the characteristic marketers are often implying when they mention a product's "time to market." So, why do they shroud this coveted title in imprecision? There can be only one first. But there are several ways to define your development effort as the first: It can be the first to demonstrate a concept, the first to bring a viable product to market, or the first in profitability or market share. "Time to market" implies that your design will be complete before your competitor's because your development effort will be shorter and less risky than if you chose other options.
As a first-to-market developer, your design efforts in practice incur a unique set of risks by virtue of your trying to accomplish something in a way that no one has before you. If you are using precertified and preproduction components, you usually lack access to mature, targeted development tools; documentation; and training, because they are still being built and certified as you complete your design. Reference 1 is an EDN hands-on project that deals with some of the issues facing early adopters using preproduction technologies and products in pursuit of being first to market. Reader feedback from that article became the seed for investigating how you can improve your success when you are considering pre-production processors, components, and development-support resources for your next embedded project.
Researchers at Iowa State University in 1957 developed the technology-adoption life cycle model to track farmers' use of hybrid-seed corn. Others have since repeatedly applied and modified it to usefully describe adoption behavior for other application populations (references 2, 3, and 4). According to the model, early adopters are willing to embrace untried or incomplete technology offerings in exchange for potential competitive advantages; the early-majority adopters will wait for more mature and evolutionary offerings that provide specific benefits at less risk (see sidebar "Mapping the technology-adoption life cycle").
A significant revision to the original model to accommodate disruptive or discontinuous technologies is the introduction of gaps between the adoption groups in the time line. Each gap denotes an area that a technology provider must bridge for the adopters on either side. The "chasm," or "valley of death," between the early adopters and early majority is the most significant of these gaps. When mapping the revised adoption model to embedded-system designers, the chasm between the early adopter and the early majority coincides with the silicon and tool vendors' efforts to ramp up the breadth and depth of their development-support infrastructure, including the availability of accurate documentation, certified development tools, evaluation platforms, library components, application notes, and reference designs.
To the left of the chasm, early adopters lack access to a general pool of experts, because they are usually breaking new ground. To the right is a pool of experts that has practical, hands-on experience with the technology and products. Early adopters must often work with incomplete and precertified products and development tools, but, by the time the early majority begin using them, the products and tools support the required functions and have completed certification. The interaction between the vendor and the design team to the left of the chasm is more tightly coupled, because both sides mutually influence and learn from each other.
Mutual benefit
To mitigate the risks of using preproduction products and tools, early adopters work closely with processor, component, and tool providers, usually under a strategic alliance agreement or as a lead or beta partnership. In the case of preproduction processors, the early-adopter partnership includes the silicon vendor, possibly a separate tool vendor, and the lead customer. The parties lower their individual risks, more quickly ready their products for production, and improve their product quality when they work together to make the interdependent silicon and tools.
From the silicon provider's perspective, providing early access to the tool provider and the lead customer provides insight into how to tailor the processor to maximize development-tool effectiveness and best suit the customer's target application. The silicon vendor can incorporate suggestions from the other partners into a new iteration of the processor design, as well as advise both the tool vendor and the lead customer on how to better use the processor features. This effort allows the silicon vendor to train the sales channels to allow a gradual transition from a core-support team to a production-support team as the processor rolls out to more customers.
From the tool vendor's perspective, the partnership provides early insight into supporting the processor architecture and how the customer will use it before actually supporting a production environment. The tool vendor can incorporate suggestions from the other partners to better unleash the power of the processor, as well as advise both the silicon vendor and the lead customer how to modify the processor architecture and application code for the best results. This effort also allows the tool vendor to train the sales channels to allow a gradual transition from a core-support team to a production-support team as the processor and development tools roll out to more customers.
From the lead customer's perspective, the partnership provides early insight into using the processor features and development tools. The silicon and tool vendors provide extensive development support so that the lead customer can successfully complete the early-adopter project. The lead customer can also suggest and influence the next product iteration from both the silicon and the tool vendors for a more tailored device and tool set. As part of the partnership, the lead customer often receives advanced samples and preferential pricing.
Feedback and design iterations can require multiple cycles. Ideally, each member successfully balances incorporating the suggestions from the other partners and modifying how they implemented their part. The partnership effort more quickly wrings out the technology and gets everyone to market and provides everyone with better insight into the application-usage model. This effort helps better suit the new processor, tools, documentation, application notes, evaluation boards, and reference designs to the target application space and allows silicon and tools vendors to support enough sales volume to recover their engineering costs. The lead customer, in turn, gets to lead the market with a new product.
Candidate customers for these partnerships are usually large key customers with expected large-volume applications. Smaller customers can be early-adopter candidates if they represent a new market into which the silicon vendor is trying break. The list of candidate customers is limited, because the "hand-holding" required to support early silicon does not scale beyond a few customers. The silicon vendor may create and provide, in speculation, proof-of-concept demonstrations for candidate lead customers. The silicon vendor may offer technical schematics and designs to show its customer how to use the new technology component. In some cases, silicon vendors have even financed their customers' efforts to prove out the value of the new product.
Large early-adopter partnerships are temporary teams, and, in some cases, the vendor commits to a level of resources, including on-site support, to help the customer's project succeed. Each member of the partnership shares a mutual obligation in managing the risks. In these cases, the partnership is a technical commitment. Any contract is to secure the business but with an understanding about the robustness of the silicon and tools. Partnership contracts for smaller early adopters may specify expectations of capabilities and performance and may include funding from the customer to guarantee that vendors provide support within a specified time schedule. Contracts can include damages to protect for the downside; however, those damages may be worth little if your product rollout fails and costs you valuable customers and your reputation.
Outfitting your team
Your ability to su cceed with preproduction technology depends on your team's ability to cope with and accommodate the unknown, gain access to missing expertise and skills, and persevere through the various challenges that crop up during your design, building, and verification effort. It is critical that you determine and document your requirements and your value-added emphasis so that you can choose the right parts, recruit the right people, and accept the right level of risk for your project. Using preproduction products and services means that you are accepting schedule and budget risks, so make sure that you have a good reason to do so. Some vendors claim that some early adopters use preproduction offerings because they are new and cool, even though they could probably work with processors or components that are one generation older and incur less risk from mature production-support processes. Before accepting the risk of a preproduction technology, determine whether it offers you an opportunity for real feature differentiation, a measurable increase in development productivity, or a justifiable decrease in your bill of materials within your needed time window; road maps are not deliverables and do not allow you to complete your project on time.
In addition to understanding your project requirements, you need to build a committed team within and outside your company to help execute the plan. You must honestly evaluate your management's risk tolerance and your design team's knowledge and skills so that you can recruit the appropriate people and develop productive relationships with your vendors to compensate for your lack of experience and skill. When evaluating a vendor's development-support services, ensure that the vendor provides a good technical fit for your team, brings complementary business skill to the partnership, has adequate resources to commit to your project, and has the capabilities to meet your production requirements. Market fluidity means there is no guarantee that silicon will be introduced, especially if the silicon vendor can't find a large enough base to support the product. Make sure what the vendor says about your project makes sense to you; if it doesn't, you could be facing a serious mismatch. Be cautious and follow up if you feel that the vendor is faking its answer; you cannot afford that type of risk. If your project application is revolutionary, consider bringing in experienced consultants to provide an independent, objective perspective that can help you with the interview and selection process.
A vendor's commitment to frequent and honest two-way communication is critical to the success of early-adopter projects. It can help align each side's expectations with each side's needs and potential accomplishments during the joint-development effort. Periodic technical interchanges can help each side avoid relearning lessons. You should document agreements between you and the vendor to avoid mismatched expectations. Time is critical in an early-adopter project, so you need to understand how to receive engineering-support. Do you rely on a single point of contact for support, or do you need multiple points of contact? How do you escalate an engineering support request to gain access to the vendor's design resources?
It is important to establish a method of determining how and when to expect updates as well as how to disclose the status of outstanding defects in each release, even if it means using a simple log of defects and an accounting for how each defect is resolved. You want to obtain information about issues as early as possible so you do not waste effort in a known problem area. Speaking of issues, it is essential that you understand how to determine whether a problem is a defect or a feature request. Defects require immediate resolution; feature requests need a contract addendum and require negotiation.
Because you will likely be using "work-in-progress" components and tools, understanding the verification and coverage process is essential. Complete and accurate documentation is critical for risk mitigation and problem resolution. Otherwise, you are at greater risk for undetected problems that you need to recognize and escalate as quickly as possible. Never assume that another part of the team is resolving an issue; document issues and follow up on the resolution to gain closure on each item. This type of discipline can help you more quickly identify when you are in trouble. Your development partners should also be able to broaden the team perspective to more quickly identify when you are in trouble. Redirecting your efforts and recovering from a potentially fatal issue is easier when you can identify when you are about to stumble into a hole than when you are already at the bottom of that hole.
Value the qualities that demonstrate your future partners' stability, such as their experience supporting early-adopter projects, their successful support of projects similar to yours, and their track record in managing promises and expectations while delivering products, support, and documentation in a reasonable time. Be wary if you do not feel that the vendor views early-adoption projects as symbiotic activities in which both parties commit resources and time and from which each party benefits. You might also consider your vendors' support partnerships. Are the partners that are relevant to your project reliable and reputable? A list that includes unproven or disreputable partners should raise a red flag that the vendor requires closer scrutiny before you make a commitment. Often, when a problem arises, it is unclear why or where the defect originated. Therefore, it is important for your internal team and partner teams to be able to work together to troubleshoot the new silicon, design hardware, and design tools to best categorize the problem for resolution.
Partitioning your design effort into separable modules based on the technological risk makes it easier for you to accommodate failed risks. As much as possible, focus your efforts on the new technology and tools at the beginning of your project so that you can work out the risk and test planning before tackling the mature technology. You'll avoid major problems later in the project and be able to develop backup plans for hardware shortages, key components' missing schedule, or new technologies that fail. Your backup plan should identify out-of-the-box fallback options available today and highlight your criteria for evaluating how and when your technology vendor/partner reaches a steady state with the technology product sufficient to meet your production needs.
Consider limiting the number of new technologies you will accept as part of your early-adopter project and properly accommodate these technologies in your planning. Each new technology you adopt, whether a preproduction product or a product with which you have no experience, represents a learning curve and another source of risk that can combine to kill your project. Don't overlook how adopting "only" one new technology can increase your learning-curve risk because of d ependencies. As an example, just switching your processor vendor may also require you to adopt a different set of development tools, rely on a different operating system, port your legacy code to the new architecture, and develop new working relationships with each of the vendors supporting your effort.
Take into account how well you will be able to reuse your software. To keep your development effort moving forward, you may use noncertified, preproduction simulators, silicon, boards, and software tools. Depending on the early information you received from your vendor partners and how you built your code, you may experience a few days' to a few weeks' impact on your project when you convert to the certified, production versions. This area is one in which separating your design based on technological risk can limit your exposure. Be sure to build more contingency or pad into your schedule and budget than you would for a similar early-majority technology adoption for unplanned engineering effort for troubleshooting and porting.
Take advantage of your vendor's engineering resources to review your designs. You are learning about its product even as the vendor is learning about it. These types of review checkpoints can help you more quickly spot impact from changes or misunderstood documentation and avoid costly fixes that could require several months. This process should also address your use of any undocumented features, which are features that do not make it to the final documentation for any number of reasons—perhaps because they do not work as intended under all operating conditions. You know about these features because you planned for them during the preproduction process and you speculated that they would be viable. You might continue to use these features because they offer significant advantage and work well for the context in which you use them. You should be aware that those features could disappear in future revisions of the device, forcing you into a redesign and verification effort or a life-end buy.
Changing landscape
Embedded-development products and tools are evolving toward higher integration and complexity. Silicon designers must be aware of and account for how their customers will develop applications using the new silicon. Development tools and early-adopter application designers are influencing silicon designs to include instruction-set enhancements and better on-c hip debugging. Simulation tools are supporting more awareness and interoperability of the tools, including the compilers, operating systems, debugging tools, and analysis tools that allow you to identify and resolve many design issues before committing the silicon design to fabrication.
Changing market conditions and fluctuating risk tolerance shape how early-adopter projects progress. According to many vendors, early-adoption projects during the last few years differ in tone from five years ago. Five years ago, venture cash focused more on the potential payoff than the risk of an early-adopter project, and it was three to four months from first contact with the customer to the close of a sale. Today, the focus is more risk-averse. Vendors are seeing more scorekeeping, and their past records on on-time delivery, preservation of value, and consistency of risk are affecting their opportunities for early-adopter partnerships. The time from first contact to close of sale has grown to six to nine months as embedded-system designers shift more of the analysis work before closure to prove that the vendor's product will meet the project needs.
As the market and the available technology continues to change, where the risk manifests itself also changes. The risk always exists that the perceived market for your early-adopter project may not materialize, because the project may be postponed or canceled. Silicon is becoming a smaller area of risk, but software and tools continue to grow in risk. The risk that a processor or component will even make it to the market is still real, but the success of such devices is relying more on the availability and integration of the development tools, operating systems, drivers, middleware, and focused application-development services to support these devices. The source of risk continues to change, especially as innovation continues to introduce disruptive technologies and processes and design cycles continue to shrink.
The trend toward shorter development cycles leads to less room in the schedule to integrate new technologies, which is a limiting or a blocking point for early adopters. In response to this pressure, some early adopters are doing generic platform-based designs, using new processors to preserve their ability to use the new technology for competitive advantage. These endeavors start as speculative engineering projects with no committed end customer, but they can provid
e a basis for a quick customized design when you find a committed customer, minimizing the engineering risk at the expense of investment risk.
The challenges facing embedded-system pathfinders are changing as the markets continue to change, and disruptive technology sometimes makes the previously impossible possible. Support for all but the largest early adopters is moving toward smaller system integrators, opening up opportunities for early adopters in smaller companies or projects, but moving the support from a "free" partnership model to a charged service. Small adopters can also often obtain access to models and boards in advance of production silicon. What remain similar are the qualities you need to find in a vendor partnership and your ability to identify and redirect around dead ends and pitfalls while pursuing your goal.
Special thanks to Tom Helfrich for sharing his experience and lessons learned as an early-adopter developer.
| References |
|
Author Information
 You can reach Technical Editor Robert Cravotta at 1-661-296-5096, fax 1-661-296-1087, e-mail rcravotta@edn.com. |
For More Information...
For more information on products such as those discussed in this article, contact any of the following manufacturers directly, and please let them know you read about their products in EDN.
Analog Devices
1-800-262-5643
www.analog.com
ARC
1-408-437-3400
www.arc.com
Green Hills Software
1-805-965-6044
www.ghs.com
IDT
1-408-727-6116
www.idt.com
Infineon
+49-911-654-4262
www.infineon.com
Intel
1-800-628-8686
www.intel.com
Metrowerks
1-800-377-5416
www.metrowerks.com
Microsoft
1-425-882-8080
www.microsoft.com
NEC
1-408-588-6000
www.necelam.com
Telairity
1-408-764-0270
www.telairity.com
Texas Instruments
1-800-336-5236
www.ti.com
Vytek
1-888-554-2024
www.vytek.com
Wind River
1-800-545-9463
www.windriver.com
Zilog
1-408-558-8500
www.zilog.com
