Guide to Reliability of Electrical/Electronic Equipment and Products--Robust Design Practices (part 1)

Home | Articles | Forum | Glossary | Books


1. INTRODUCTION

Technology has created an increasingly complex problem for delivering reliability. More systems are designed today faster than ever before under shrinking cost margins and using more electronics with ever more complex devices that no one has the time to test thoroughly. Concurrently, the tolerance for poor reliability is shrinking, even while expectations are rising for rapid technology changes and shorter engineering cycles.

Grueling project schedules, thirst for performance and cost competitiveness result in corners being cut in design, with as little verification and testing being done as possible. In all of this what is a company to do to deliver high-reliability products to its customers? Here are some hints. Use common platforms, product architectures, and mainstream software and keep new engineering content down as much as possible to have evolutionary rather than revolutionary improvements.

Use proven or preferred components. Perform margin analysis to ensure performance and reliability beyond the stated specifications. Conduct accelerated stress screening to expose and correct defects in the engineering phase before shipping any products. Conduct detailed design reviews throughout the design process composed of multidisciplinary participants.

This section discusses the importance of the design stage and specifically the elements of design that are vital and necessary to producing a reliable product.

Design has the most impact on the reliability outcome of a product. 80% of the reliability and production cost of a product is fixed during its design. Reliability must be designed in a product. This requires a lot of forethought and a conscious formal effort to provide just the required margin needed for customer application.

A given product's projected life in its intended application determines the amount of margin (robustness) that will be designed in and for which the customer is willing to pay. Figure 1 shows the product life cycles of various categories of computers. This figure shows that not all products require high reliability and that not all segments of a given product category have the same reliability requirements, i.e., a one size fits all mind set. For example, personal computers (PCs) are commodity items that are becoming disposable (much as calculators and cell phones are). Computers used in financial transaction applications (stock markets, banks, etc.), automobiles, telecommunication equipment, and satellites have much more stringent reliability requirements.

=========


FIGURE 1 Product life cycles for various categories of computers. (Courtesy of Andrew Kostic, IBM.)

============

TABLE 1 Detailed Design Tasks

Product design, including design for manufacture, design for test, and design for electromagnetic compatibility Logic circuit design Design verification, simulation, and emulation of application-specific ICs PWA design/layout and design for manufacture System software development Software reviews Enclosure design (mechanical and thermal) Reliability production and update (throughout the design phase) Part selection/finalization (throughout the design phase) BOM reviews (throughout the design phase) Failure modes and effects analyses Hardware design reviews (throughout the design phase) PWA test vectors and system test programs development Manufacturing qualification testing of new packaging and technology Diagnostic test development Signal integrity and timing analysis tests Thermal analysis and testing PWA outsource provider qualification Appropriate hardware design verification testing

==========

How product reliability has been accomplished has differed greatly be tween U.S. and Japanese companies. The Japanese companies generate many change notices during product design, continually fine tuning and improving the design in bite-sized pieces until the design is frozen upon release to production.

Companies in the United States, on the other hand, are quick to release a product to production even though it contains known "bugs" or deficiencies. Geoffrey Moore, in Living on the Fault Line, calls this "going ugly early" to capture market share. After being released to production a number of changes are made to correct these deficiencies. But in the meantime, plan on customer returns and complaints.

Reliability begins at the global system concept design phase; moves to the detailed product design phase that encompasses circuit design, application-specific integrated circuit (ASIC) design, mechanical interconnect design, printed wire assembly (PWA) design (including design for manufacturability), thermal design, and industrial design (case, cover, enclosure, etc.); and concludes with design for reliability, testability, and electromagnetic compatibility and design verification testing. Reliability is strongly dependent on the individual components (piece parts) used, suppliers selected, and the manufacturing and test processes developed.

Table 1 takes the detailed design block of Figure 1 and expands it to show the typical tasks involved in producing a high-end computer design. These tasks, some of which can occur in parallel, are applicable to the design of all products, but the choice as to which of these tasks and how many are used for a particular situation depends on the amount of robustness (design margin) needed for the intended customer application.

2. CONCURRENT ENGINEERING/DESIGN TEAMS

The quality of a product is highly dependent on the human organization that turns it out. Today's designs require close cooperation, alignment, and integration among all appropriate technical disciplines to fulfill the design objectives, beginning at the earliest place in the design cycle. This method of operation has been given several names, including concurrent engineering. Concurrent engineering can be defined as (IDA Report R-338, 1988) A systematic approach to the concurrent design of products and their related processes, including manufacturing and support. This approach is intended to cause developers, from the outset, to consider all elements of the product life cycle from conception through disposal, including quality, cost, schedule, and user requirements.

========

TABLE 2 Benefits of Concurrent Engineering

Case study Cost Schedule Quality

AT&T Circuit pack repair Total process time cut Defects down 30% cut 40% to 46% of baseline

Hewlett-Packard Manual costs down Development cycle Field failure rate 42% time cut 35% cut 60%

IBM Labor hours cut Design cycle cut 40% Fewer ECOs 45%

=========

Thus, concurrent engineering is the cooperation of multiple engineering disciplines early in and throughout the design cycle. It is important because a team organization can have the most immediate effect on quality and reliability.

Experience shows that concurrent engineering leads to:

Shorter product design and manufacturing

Lower field returns

Significant cost savings

Table 2 provides three examples of the benefits of the application of concurrent engineering.

A typical concurrent engineering product design team includes system de signers; logic designers; ASIC designers; PWA designers; mechanical, thermal, and structural designers; reliability engineers; test engineers; component engineers; manufacturing engineers; industrial designers; and regulatory (electromagnetic compatibility and safety) engineers. Reliability engineers can be considered the glue that keeps the design project together since they are concerned with Time degradation of materials

Physical measurements

Electrical and electronic measurements

Equipment design

Processes and controls

System performance simulation and analysis

Synthesis

Piece parts and suppliers selected for the design

However, a startling change is taking place. Design (concurrent engineering) teams are shrinking in size. This is due to the increased integration possible with integrated circuit technology and the available design tools.

3. PRODUCT DESIGN PROCESS

A development process needs to be both flexible and grounded on certain basic fundamentals and processes. One fundamental concept is that work proceeds sequentially through certain phases of a product's development. Although the pro cess is iterative, much of the work in each of the phases is performed concurrently, and earlier phases are often revisited as the work is modified, the sequence remains much the same throughout and phases complete sequentially. For example, though the product design may be modified as implementation proceeds, major design always precedes major implementation. Movement of a development project from one phase to the next represents an increased level of commitment to the work by a company. Exit from a phase is a result of interorganizational concurrence with the phase review deliverables required in that phase.

Concurrence implies commitment by each organization to the activities required in the next phase. Each phase has a specific objective. During each phase, major functions perform certain activities to achieve the objective. Typical product life cycle phases for a complex electrical equipment development project are Phase 0: Concept. Identification of a market need

Phase 1: Investigation and Requirements. Response to the need with a product description

Phase 2: Specification and Design. Design of the product(s) that constitute the development effort

Phase 3: Implementation and Verification. Implementation and testing of the product(s)

Phase 4: Preproduction and Introduction. Preparation of the product for general availability

Phase 5: Production and Support. Review of the performance of the product(s) in the field to determine future related work and to codify lessons learned for next generation development.

Let's look at each phase in greater detail.

3.1 Phase 0: Concept

Phase 0 answers the question is this the right product for the market. The objective of Phase 0 is to identify, as accurately as possible, the market need or opportunity for a development effort and to communicate the need or opportunity to Development (a.k.a. Engineering) so that they can perform a feasibility study.

Marketing identifies the need for the development effort/project and communicates the need to Product Management. If a project begins with a proposal of a market opportunity from Development, Marketing reviews the proposal, identifies the market for the program, and then communicates the information to Product Management, in much the same way as if the idea had originated in Marketing. Product Management produces a high-level product definition and market requirements document to define the market need or opportunity and communicate it to Development.

3.2 Phase 1: Investigation and Requirements

Phase 1 answers the question will we commit to develop and build this product.

The objective of Phase 1 review is to reach a decision to commit funds necessary to design and specify a development project (in Phase 2) and to enter it into a company's strategic plans.

Development (with the support of other organizations such as Manufacturing, Industrial Design, Product Assurance, and Reliability Engineering, for example) responds to the requirements document created in Phase 0 with a detailed Product Description. The Product Description addresses each of the requirement document's product objectives with a corresponding statement of planned product function, performance, documentation, and other product attributes.

All organizations (whether involved in the project at this phase or not) prepare a preliminary Operating Plan with estimated dates to support the Product Description. A Program Manager is assigned to lead the effort and forms the core team, with assistance from the heads of the major functions.

3.3 Phase 2: Specification and Design

Phase 2 answers the question can we approve this project for Marketing with committed Beta (external partner evaluation units) and first-customer-ship (FCS) dates. The objective of Phase 2 Review is to approve the completion of the product design and to commit both the ship dates and the funds necessary for its implementation.

Development completes the specification and design of the product, including creation of project level deliverables. Product Assurance works with Development to create a test plan for the products in the project (test libraries for software; test suites, facilities, and equipment for hardware test). Reliability Engineering reviews the piece part components and suppliers that have been selected, generates reliability prediction calculations, and compares the results with the product/ market goals. Support defines the field product support strategy.

All functions commit to their efforts by producing "final" operating plans.

Final, in this sense, means that a plan is complete, contains committed dates, and emphasizes the activities that each function will perform during Phase 3.

Operating plans are reviewed at each phase to ensure that they contain the information needed during the next phase of the development effort.

3.4 Phase 3: Implementation and Verification

Phase 3 answers the question can we authorize shipments to our Beta partners.

The object of Phase 3 review is to release the product(s) for Beta testing.

Phase 3 typically includes several subphases. After hardware design and development, the preproduction product typically goes through both engineering and manufacturing detailed verification testing. In software, Development completes the design and coding, inspects the code, and performs unit and product testing; Product Assurance performs system testing of the integrated product(s) as a whole, and Program Management coordinates all Alpha (internal engineering and manufacturing) unit testing. Since many products combine both software and hardware, the subphases must overlap, i.e., a hardware unit must be available for software testing. So for life cycle purposes, a single Phase 3 is recognized.

Before completing Phase 3, Product Assurance verifies that Beta criteria have been met. The goal is to thoroughly validate the functionality and quality of the product so that the product that goes to Beta partners is the same product that goes to FCS.

3.5 Phase 4: Preproduction and Introduction

Phase 4 answers the question should we announce and authorize FCS of this product. The objective of Phase 4 review is to announce and authorize general availability of the product, i.e., transfer it to production.

Development produces all project-specific release-related material. Marketing produces data sheets, collateral support material, and announcement material for the product. Support manages Beta testing to ensure that the product is ready for general availability (the necessary support structure is in place). A postpartum review of Phase 4 should include a review and analysis of the entire development effort and how it could have been executed better.

3.6 Phase 5: Production and Support

Phase 5 answers the question is this product successful. The objective of Phase 5 review is to ensure that the product is and continues to be successful in the field by defining (if necessary) maintenance plans, product upgrade and enhancement plans, and new marketing strategies or retirement plans.

=====

TABLE 3 Component Engineering Product Design Phase Deliverables

Phase 0

Technology roadmap Concept BOM review and needs assessment Quality key technologies/suppliers before use (DRAM, SRAM, microprocessor)

---

Phase 1

Spicer model and library support

Limited special studies experiments

Circuit simulation performance model

BOM review and risk assessment

Component qualification plan and matrix

Electrical testing and support for failure analysis

---

Phase 2

Spicer model and library support

Special studies experiments

Circuit simulation performance model

BOM review support changes and new qualification plans

Potential suppliers list complete

Qualify components to plan

Electrical testing and support for failure analysis

----

Phase 3

Spicer model and library support

Special studies experiments

Manufacturing test support

BOM review support

Supplier qualification complete

Component qualification complete for first source

Electrical testing and support for failure analysis

Supplier defect and corrective action support

----

Phase 4

Spicer model and library support

Special studies experiments

Manufacturing test support

BOM review support

Component qualification complete for first source

Electrical testing and support for failure analysis

Supplier defect and correction action support

Manufacturing support

----

Phase 5

Spicer model and library support

Special studies experiments

Manufacturing test support

Emergency product support

Approved vendor list management, process and product changes, re-qualification as needed

Electrical testing and support for failure analysis

Supplier defect and corrective action support

Manufacturing support

=========

All involved functions meet periodically to review the performance of the product in the field. The first such Phase 5 review meeting should take place approximately 6 months after the product has become generally available (a shorter time for products with an extremely short lifetime, such as disk drives or consumer products), unless critical problems arise that require convening the meeting sooner. At the first Phase 5 review subsequent review dates are selected.

A Phase 5 review ends by recommending any of the following product actions:

Continue to manufacture and maintain the product.

Replace the product with the next generation.

Enhance the product.

Retire/make obsolete the product.

Table 3 is an example of the typical phase deliverables for a component engineering organization at a high-end computer server manufacturer.

4. TRANSLATE CUSTOMER REQUIREMENTS TO DESIGN REQUIREMENTS

Marketing typically describes, develops, and documents the market need or opportunity for a product based on customer requests and inputs. This document contains such general information as product function, weight, color, performance (sensitivity, selectivity, power output, and frequency response, for example), fit with earlier versions/generations of the product, product migration, price, availability date (market window of opportunity), some measure of reliability or warranty allowance, and so on (this is the Phase 0 input). From this, a product description (Phase 1) is developed that addresses specific objectives and translates the general requirements to a high-level engineering requirements document that is used as the basis for the technology assessment and detailed design activities that follow.

=========

TABLE 4 Component and Technology Questionnaire

1. What interconnect schemes do you plan to use? Do they involve using only existing qualified hardware, or are there new interconnect requirements? (If new, we would like to work with you on requirements definition, application verification, supplier selection, and supplier and part qualification.)

2. What basic technologies are you planning to use? What will the basic voltages for your circuits be (e.g., 1.8, 2.5, 3.3, 5, 12, or 15 V? Other?) Please identify the component types that will be used. How many total new components do you anticipate using that will need to be qualified? How many component part numbers do you anticipate using in your BOM (all will need source checks and may need approved vendor list/source control drawing generation/updating)? Specify the following:

DRAM (speed/type/size/package)

SIMM style SRAM (speed/type/size/package)

SIMM style PLD/FPGA Tools Prototype only or production too? ASICs (speed/size/package) Tools Microprocessor Support tools Bus structures Microperipherals SCSI Special communication circuits Special functions Special modules/custom circuits DC/DC converters Delay lines Oscillators Special functions (i.e., line interface modules) Analog ICs Digital logic ICs Delay lines (standard or custom/active or passive/single edge or both) Fiber optic/optic interface Passive components Terminators Filters Other Discrete Semiconductors LED FET Other PON circuit (please identify which one) Nonvolatile memory (type/size/package/speed) EPROM EEPROM Flash SEEPROM Fuses Switches Voltage regulators Mechanical fasteners Connectors Cabling Clock-driving circuits and phase-locked loops

Of these technologies, which do you anticipate will be custom or will need to be procured to nonstandard requirements? Which are new to your group? Which do you believe will require special studies (for ground bounce, SSO/SSI, edge rate, jitter, timing, load sensitivity, symmetry, power)? Which do you believe need special characterization or qualification analysis?

3. What kind of PWA or like assemblies will you be using? What kind of material?

4. What manufacturing processes do you anticipate using (MCM, TAB, BGA, CSP, flip chip, SMT, two-sided SMT, SMT-PTH, PTH)?

5. Are you planning to use any of the following categories of devices:

ECL TTL/bipolar (AS/ALS/LS/S/F)

_16-MB or smaller DRAMs?

_4-MB or smaller SRAMs? Bipolar PLDs? 8- or 16-bit microprocessors? EEPROMs (PLDs slower than 10 nsec?)

If so, you will be facing severe sourcing risks and may face end-of-life issues.

6. Which devices will need SPICE model support? How will you verify timing? Signal integrity? Loading? Component derating? What derating guidelines will be used? Identify the following:

Clock design margin? Analog margin, signal integrity? Thermal/junction temperatures of components? Power plane noise/decoupling? Power of IPS/UPS?

7. Regarding testing:

Do you plan to test each supplier's part in your application? Do you plan to measure all signals for timing, noise margin, and signal integrity? Do you plan to test power fail throughout the four corners of your product specification? Do you plan to perform HALT evaluation of the product (test to destruction-analyze-fix-repeat)? If so, which assemblies? Will all suppliers' parts be used in the HALT evaluation? Will you be testing worst/best case parts (which ones)? Will you HALT these?

8. What kind of manufacturability verification/process verification do you plan to do?

9. What special equipment/software will be needed to program the programmable logic?

10. For new suppliers, what discussions and design work and supplier qualification/component qualification has been done to date? Who has been the principal contact or interface between the circuit designer and the supplier?

Are these discussions/efforts documented?

11. What are the key project assumptions?

Major design assemblies Design reviews planned (which/when/who) List of key designers and managers (and their responsibilities today) Schedule:

Design specifications Prototypes Alpha Beta FCS Number of alpha tests (by whom?) Number of prototypes (built by whom?) Product support life Product sales life Total product quantity during product life (per year if estimated) Design and phase review dates (Phases 1,2,3,4,5)

=========


FIGURE 2 Matching user technology requirements and supplier offerings.

5. TECHNOLOGY ASSESSMENT

Early on in the conceptual design stage it is important to understand what technology needs are required to fulfill the documented product description as listed in the engineering requirements document. For example, for a new computer server development effort, such issues as system speed, microprocessor type, bus architecture, memory requirements (DRAM, SRAM, size, speed, etc.), application specific integrated circuit requirements, on-board power requirements, and interconnect needs (how? mechanical? electrical? size? bus interface?) must be addressed. The process starts with a questionnaire, similar to that shown in Table 4, sent from Component Engineering to the design organization. Component

Engineering works with Design Engineering at the concept phase to determine the type of components that will be needed in the circuit design, outlining the scope of work regarding components for both parties. Typically, the component engineer sits down with the design team members and works through the myriad technical issues listed in the questionnaire, starting at a general level and then digging down to more specific component issues. This helps determine if new technologies are needed to meet the product requirements. If new technologies are required, then the component engineer needs to find answers to the following questions:

Does the technology exist?

Is the technology currently in development? When will it be available?

Will the technology development need to be funded?

Will the technology need to be acquired?

To use this product/technology do I need to have a contingency plan developed to ensure a continuous source of supply for manufacturing? There is also the issue of old products and technologies, those products and technologies that the designers have used before in previous designs and are comfortable with. In some cases the designers simply copy a portion of a circuit that contains these soon-to-be obsolete components. How long will the technologies continue to be manufactured (1.0-µm CMOS in a 0.25-µm CMOS world, for example)? How long will the specific product be available? Where is it in its life cycle?

This process helps to direct and focus the design to the use of approved and acceptable components and suppliers. It also identifies those components that are used for design leverage/market advantage that need further investigations by the component engineer; those suppliers and components that need to be investigated for acceptability for use and availability for production; and those components and suppliers requiring qualification.

The answers allow the component engineer and the designers to develop a technology readiness and risk assessment and answer the question is it possible to design the product per the stated requirements. A flow diagram showing user technology needs and supplier technology availability and the matching of these is shown in Figure 2.

As the keeper of the technology, the component engineer Is responsible for keeping abreast of the technology road maps for assigned circuit functions (connectors, memory, microprocessor, for example)

Conducts technology competitiveness analyses

Works with the design organization to develop a project technology sizing

Understands the technology, data, and specifications

Ensures that the design guidelines and computer-aided design (CAD) methods and libraries are available and meet the designer's requirements All of this activity results in developing a work plan, helps focus the design effort, and facilitates in developing a "first cut" bill of materials (BOM).

6. CIRCUIT DESIGN

The design of today's electronic products and equipment is a complex task. This is primarily due to the steady increase in integrated circuit density, functional capability, and performance, for both digital and analog ICs. Concomitant improvements have been made in interconnect devices (printed circuit boards, connectors, and back-planes) and the materials used to make these devices; discrete semiconductors (e.g., power transistors); power supplies; disk drives; and the like.

A second reason for increased design complexity is that the electronics industry is evolving into a new horizontal structure around the four levels of packaging: chip-level integration (ICs), board-level miniaturization surface mount technology, productization (PCs, peripherals, communication sets and instrumentation), and system integration. Each of these requires distinctly different design, manufacturing, and management skills and techniques.

6.1 Circuit Types and Characteristics

A given electronic circuit often contains both analog and digital sections. There are significant technical differences between these sections and careful attention must thus be paid as to how they are electrically interconnected and physically located on the printed circuit board (PCB). As digital IC processes migrate to below 0.25-µm (the deep submicron realm), the resultant ICs become much noisier and much more noise sensitive. This is because in deep submicron technology interconnect wires (on-chip metallization) are jammed close together, threshold voltages drop in the quest for higher speed and lower operating power, and more aggressive and more noise-sensitive circuit topologies are used to achieve even greater IC performance. Severe chip-level noise can affect timing (both delay and skew) and can cause functional design failures.

The analog circuitry needs to be electrically isolated from the digital circuitry. This may require the addition of more logic circuitry, taking up board space and increasing cost. Noise can be easily coupled to analog circuitry, resulting in signal degradation and reduced equipment performance. Also, interfacing analog and digital ICs (logic, memory, microprocessors, phase-locked loops, digital-to-analog and analog-to-digital converters, voltage regulators, etc.) is not well defined. Improper interfacing and termination can cause unwanted interactions and crosstalk to occur. Then, too, the test philosophy (i.e., design for test, which will be discussed in a later section) must be decided early on. For example, bringing out analog signals to make a design testable can degrade product performance. Some of the pertinent differences between analog and digital ICS are listed in Table 5.

========

TABLE 5 Analog and Digital IC Technology Comparison

---

Analog

Transistors full on Large feature size ICs High quiescent current Sensitive to noise Design tools not as refined/sophisticated Tool incompatibility/variability Simulation lags that of digital ICs

----

Digital

Transistors either on or off

Cutting edge feature size ICs

Low quiescent current

Sensitive to signal edge rates

Sophisticated design tools

Standard tool sets

Sophisticated simulation process

=========

6.2 Design Disciplines and Interactions

Design of electronic products involves dealing with myriad issues. These include component (part) tolerances, selection/restriction of part suppliers, component/ part selection and qualification, design for test (in-circuit test, self-test), under standing failure modes and mechanisms, understanding the relationship in component volume and price cycles, and understanding component life cycles. Many assemblies have design elements that are not optimized, resulting in increased labor time, increased production and rework costs, as well as yield and quality issues.

Knowledgeable engineers with a breadth of experience in designing different types of products/systems/platforms and who are crosstrained in other disciplines form the invaluable backbone of the design team. All of the various design disciplines (each with their own experts)-circuit design (analog and digital), printed circuit board design and layout (i.e., design for manufacturability), system design (thermal, mechanical and enclosure), design for test, design for electro magnetic compatibility, design for diagnosability (to facilitate troubleshooting), design for reliability, and now design for environment (DFE)-are intertwined and are all required to develop a working and producible design. Each one impacts all of the others and the cost of the product. Thus, a high level of interaction is required among these disciplines. The various design tasks cannot be separated or conducted in a vacuum.

The design team needs to look at the design from several levels: component, module, printed wiring assembly (PWA)-a PCB populated with all the components and soldered-and system. As a result, the design team is faced with a multitude of conflicting issues that require numerous tradeoffs and compromises to be made to effect a working and manufacturable design, leading to an iterative design process. The availability and use of sophisticated computer-aided design tools facilitate the design process.

Tools

Circuit design has changed dramatically over the past decade. For the most part, gone are the days of paper and pencil design followed by many prototype bread board iterations. The sophistication of today's computer-aided design tools for digital integrated circuit designs allows simulations to be run on virtual bread board designs, compared with the desired (calculated) results, debugged, and corrected-and then the process repeated, resulting in a robust design. This is all done before committing to an actual prototype hardware build. But there are two big disconnects here. First, digital IC design tools and methods are extremely effective, refined, and available, whereas analog IC and mixed-signal IC tools are not as well defined or available. The analog tools need to be improved and refined to bring them on a par with their digital equivalents. Second, there tends to be a preoccupation with reliance on simulation rather than actually testing a product. There needs to be a balance between simulation before prototyping and testing after hardware has been produced.

===========

TABLE 6 Examples of CAD and EDA Design Tools Concept and entry Verification Design reuse

HDL Simulation Intellectual property Schematic capture Verilog simulation Libraries Behavioral synthesis VHDL simulation Memories/macros RTL synthesis Cycle-based simulation Acceleration and emulation Fault simulation SPICE simulation HW/SW coverification Design for test Virtual prototype DFT tools Hardware/software Automatic test pattern generation Silicon PLD design Special purpose tools Miscellaneous FPGA place and route Design for manufacture Analog design FPGA tool set Mechanical design DSP tool set Wire harness design Mixed signal design IC design Acceleration/emulation Extractors Physical verification CBIC layout Floor planning Process migration Custom layout Gate array layout Reliability analysis Delay calculator Metal migration analysis Signal integrity analysis EMI analysis Power analysis Thermal analysis SPICE Timing analysis PCB design EMI analysis Physical verification Thermal analysis MCM/hybrid design Power analysis Timing analysis PCB design Signal integrity analysis Virtual prototype evaluation Autorouter

===========

There are many CAD and electronic design automation (EDA) tools avail able for the circuit designer's toolbox. A reasonably comprehensive list is pro vided in Table 6. Electronic systems have become so large and complex that simulation alone is not always sufficient. Companies that develop and manufacture large digital systems use both simulation and hardware emulation. The reasons are twofold and both deal with time to market: (1) Simulation cycle time is several orders of magnitude slower than emulation. Faster simulations result in quicker design verification. (2) Since today's systems are software intensive and software is often the show stopper in releasing a new product to market, system designers cannot wait for the availability of complete hardware platforms (i.e., ASICs) to begin software bring-up.

A methodology called the electronic test bench aids the design verification task. The "intelligent test bench" or "intelligent verification environment" has been developed to further ease the designer's task. The intelligent test bench is a seamless, coherent, and integrated EDA linkage of the myriad different kinds of verification tools available and is controlled by a single, high-level test bench.

The smart test bench is driven by the increasing variety of point tools needed for complex IC logic verification. It includes simulation, hardware/software co verification, emulation, formal model checking, formal equivalency checking, and static analysis tools, including lint checkers, code coverage tools and "white box" verification tools that generate monitors or checkers. It makes choices about which tool to use on which portion of the design and derives a test plan. The value of the intelligent test bench is that it eliminates the need for engineers to spend time writing test benches, and it is transparent in that it utilizes familiar software languages.

Prior to use, the various models and libraries need to be qualified. In the case of the SPICE models for the various electronic components used in the design, several questions need to be asked, including:

Is there a viable model for a given functional component? What is the accuracy of the model? Is it usable in our system? The qualification of component SPICE models can be segmented into four levels:

Level 0: model not provided by component supplier; substitute model.

Level 1: validate model to functional specifications.

Level 2: validate DC paramters and limited AC/timing parameters.

Level 3: compare to driving transmission line on PCB; compare quantitatively.

to unspecified parameters (simulation to measurement) with limited timing.

6.3 Understanding the Components

Here's the situation: a given circuit design by itself is okay, and the component by itself is okay, but the design/component implementation doesn't work. The following example illustrates the point. The characteristic of a circuit design is that it generates low frequency noise, which by itself is not a problem. The design uses phase-locked loop ICs (PLLS), which are sensitive to low frequency noise to achieve the performance required. The PLL by itself is also okay, but the design doesn't work. A designer must understand the circuit design and the characteristics of the components used.

Experience is showing that the circuit designer needs to understand both the circuit design application and the characteristics of the components. Does the component function match the application need? In many instances the designer is not exactly sure what components or component characteristics are needed or even what a given component does (i.e., how it performs). Examples of questions facing the designer include the following: How does one use a simple buffer? With or without pull-up resistors? With or without pull-down resistors? No connection? How is the value of the resistors chosen (assuming resistors are used)? It has been found that oftentimes the wrong resistor values are chosen for the pull-up/pull-down resistors. Or how do you interface devices that operate at different voltage levels with characterization data at different voltages (these ICs have different noise margins)? This also raises the question what is the definition of a logic 1 and a logic 0. This is both a mixed voltage and mixed technology issue.

Additionally, the characteristics of specific components are not understood.

Many of the important use parameters are not specified on the supplier's data sheets. One needs to ask several questions: What parameters are important (specified and unspecified)? How is the component used in my design/application? How does it interface with other components that are used in the design? Examples of important parameters, specified or not specified, include the following:

1. Input/output characteristics for digital circuit design.

Bus hold maximum current is rarely specified. The maximum bus hold current defines the highest pull-up/pull-down resistors for a design.

It is important to understand transition thresholds, especially when interfacing with different voltage devices. Designers assume 1.5 V transition levels, but the actual range (i.e., 1.3 V to 1.7 V) is useful for signal quality analysis.

Simultaneous switching effect characterization data with 1, 8, 16, 32, and more outputs switching at the same time allows a designer to manage signal quality, timing edges, edge rates, and timing delay as well as current surges in the design. Pin-to-pin skew defines the variance in simultaneously launched output signals from package extremes.

Group launch delay is the additional delay associated with simultaneous switching of multiple outputs.

2. Functional digital design characteristic.

Determinism is the characteristic of being predictable. A complex component such as a microprocessor should provide the same output in the same cycle for the same instructions, consistently.

3. Required package considerations.

Data on the thermal characteristics (thermal resistance and conductance), with and without the use of a heat sink, in still air and with various air flow rates are needed by the design team.

Package capacitance and inductance values for each pin must be provided. Oftentimes a single value of each parameter is shown for the entire package. Package differences should be specified.

Power dissipation when a device is not at maximum ratings is needed. Power curves would be helpful. What does the power and cur rent look like when a component switches from and to the quiescent state? The maximum junction temperature and the conditions under which it occurs must be specified. When the IC is powered to its nominal ambient operating condition, what is its junction temperature? Are there any restrictions on the component's use when power supply cur rent, frequency of operation, power dissipation, and ambient tempera ture are considered simultaneously?

Oftentimes the designer is not sure which supplier's component will work in the design. This is too much information for a designer to know about components. What is required is the support and collaboration of a component engineer at the beginning and throughout the design cycle who is an expert knowledge source of the functional components/suppliers the designer is using or contemplating using. Conducting regularly scheduled bill-of-material reviews is a good way to ensure that the right component for the application, a currently produced component (one that is not being made obsolete), and the right supplier are chosen. A later section discusses the concept of BOM reviews in greater detail.

7. POWER SUPPLY CONSIDERATIONS

Power supply voltages and requirements have changed over the past 10 years, being driven by the Semiconductor Industry Association (SIA) IC technology road map. The industry has migrated from 5-V requirements to 3.3 V, then to 2.5 V, and now to 1.8 V and below. This presents some significant system design issues such as dealing with ICs that have mixed voltage (VCC) levels, noise margins, crosstalk, and the like.

This change in supply voltage has driven the PWA and system designers to consider the issue of using centralized versus decentralized (on PWA) or distributed power supplies. Supply voltages less than 3.3 V have caused problems when trying to use centralized power architectures. Lower voltages drive current higher, causing resistive drops in the back-planes of large systems. This makes it difficult to distribute the required power with efficiency and safety. This also raises the questions: Do all PWAs get the same voltage and do all components on a given PWA get the same and correct voltage? To distribute the higher cur rents without significant voltage drops in these systems requires the use of large and expensive conductors. Other problems with centralized power include greater inductance and noise issues. As voltage and current pass through the wire or PCB trace, there is a greater loss of voltage.

Historically, power distribution in communications and large computer systems (base stations, switches, routers, and servers) has been accomplished using back-plane mounted DC/DC converters to convert a 24-48 V distribution bus to usable voltage rails for the analog and digital processing functions in a system.

Higher power density requirements, standard packaging footprints, increasing re liability requirements (in MTBF), and cost drive these designs.

More and more PWA and system designers are turning to the use of DC/ DC conversion at the point of use (on the PWA) to help remedy the situation.

The DC/DC converters for distributed power applications come in a standard footprint called a brick. Brick, half-brick, and quarter-brick form factors cover power levels from 500 to 5 W, respectively. Actual power density depends on such system factors as heat sinking, ambient temperature, and air flow as well as the efficiency and packaging detail of the converter. The use of distributed or on-board DC/DC converters is not without its issues. These converters must deal with unique load characteristics such as high di/dt (rate of change of current with respect to time), precision voltage tolerance, or multiple-output sequencing. Their use is driving thermal design and packaging innovations, PCB materials, and layout issues to achieve proper heat sinking/cooling.

Portable applications require minimal DC/DC converter power dissipation, size, and weight. Converters for these applications use special control techniques and higher frequency operation. Analog IC suppliers have developed specialized product families that power supply and product/system designers use to meet the needs of these specific converter applications. However, this can present issues for the systems engineer such as noise, voltage, and current spikes and other electromagnetic interference (EMI) circuit interfering issues. Another issue in portable designs is power management under wide load conditions (switching from the quiescent, or sleep, operating mode to a fully active mode). To prolong battery life, DC/DC conversion must be efficient at both heavy and light loads.

When regulation is performed on a circuit card, a small switching regulator may provide single or multiple output voltages. Low dropout (LDO) voltage regulators or charge pumps are used to service additional loads. Power control ICs for portable applications at lower current levels often include integrated power switches for optimal size and cost, while at higher current levels pulse width modulated (PWM) control is provided to external FET switches. Robust control and protection features are needed not only to protect the regulator/converter itself, but to protect the loads as well. To maintain efficiency during extended, lightly loaded conditions portable power ICs must be able to change to a low frequency mode of operation to reduce gate charge loss in the power switches (transistors).

Thus, the design and selection of the power distribution system is a critical systems design issue that permeates circuit design; system design; component selection; supplier selection; PCB design and layout; thermal, mechanical, and enclosure design (including cooling); reliability; and Electromagnetic Compatibility (EMC). See also later sections on Thermal Management, Signal Integrity, and Design for EMC for further power supply consideration details.

8. REDUNDANCY

Redundancy is often employed when a design must be fail safe, or when the consequences of failure are unacceptable, resulting in designs of extremely high reliability. Redundancy provides more than one functional path or operating element where it is critical to maintain system operability (The word element is used interchangeable with component, subassembly, and circuit path). Redundancy can be accomplished by means of hardware or software or a combination of the two. I will focus here on the hardware aspects of redundancy. The use of redundancy is not a panacea to solve all reliability problems, nor is it a substitute for a good initial design. By its very nature, redundancy implies increased complexity and cost, increased weight and space, increased power consumption, and usually a more complicated system checkout and monitoring procedure. On the other hand, redundancy may be the only solution to the constraints confronting the designer of a complex electronic system. The designer must evaluate both the advantages and disadvantages of redundancy prior to its incorporation in a design.

Depending on the specific application, numerous different approaches are available to improve reliability with a redundant design. These approaches are normally classified on the basis of how the redundant elements are introduced into the circuit to provide an alternative signal path. In general, there are two major classes of redundancy:

1. Active (or fully on) redundancy, where external components are not required to perform a detection, decision, or switching function when an element or path in the structure fails

2. Standby redundancy, were external components are required to detect, make a decision, and then to switch to another element or path as a replacement for the failed element or path Redundancy can consist of simple parallel redundancy (the most commonly used form of redundancy), where the system will function if one or both of the subsystems is functional, or more complex methods-such as N-out-of-K arrangements, where only N of a total of K subsystems must function for system operation-and can include multiple parallel redundancies, series parallel redundancies, voting logic, and the like.


FIGURE 3 CPU with parallel redundancy.

For simple parallel redundancy, the greatest gain is achieved through the addition of the first redundant element; it is equivalent to a 50% increase in the system life. In general, the reliability gain for additional redundant elements decreases rapidly for additions beyond a few parallel elements. Figure 3 shows an example of a parallel redundant circuit. This is a block diagram of a computer central processing unit with parallel secondary cache memory, microprocessors, and ASICs.

Adding redundant elements (additional circuitry) may have the effect of reducing rather than improving reliability. This is due to the serial reliability of the switching or other peripheral devices needed to implement the particular redundancy configuration. Care must also be exercised in applying redundancy to insure that reliability gains are not offset by increased failure rates due to switching devices, error detectors, and other peripheral devices needed to implement the redundancy configurations. An example of this is the use of standby or switching redundancy. This occurs when redundant elements are energized (i.e., in their quiescent states) but do not become part of the circuit until they are switched in and only become part of the circuit after the primary element fails.

Thus, the redundancy gain is limited by the failure mode or modes of the switching device, and the complexity increases due to switching. In many applications redundancy provides reliability improvement with cost reduction. However, simple backup redundancy is not necessarily the most cost effective way to compensate for inadequate reliability. The circuit designer has the responsibility to deter mine what balance of redundancy alternatives is most effective, if any. This is a significant factor in total life cycle cost considerations. Redundancy may be easy and cost effective to incorporate if a circuit block or assembly is available off the shelf in comparison to starting a design from scratch or conducting a redesign. Redundancy may be too expensive if the item is costly or too heavy if the weight limitations are exceeded and so on.

These are some of the factors which the electronic circuit designer must consider. In any event, the designer should consider redundancy for reliability improvement of critical items (of low reliability) for which a single failure could cause loss of system or of one of its major functions, loss of control, unintentional actuation of a function, or a safety hazard. Redundancy is commonly used in the aerospace industry. Take two examples. The Apollo spacecraft had redundant on-board computers (more than two, and it often landed with only one computer operational), and launch vehicles and deep space probes have built-in redundancy to prevent inadvertent firing of pyrotechnic devices.

Top of Page PREV.   NEXT Article Index HOME