Electronic Part Obsolescence Forecasting CALCE Electronic Products - - PDF document

1

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Electronic Part Obsolescence Forecasting

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

• Judiciously choose parts (part selection)
• Forecast part life cycles
• Life cycle planning

Understanding that electronic part obsolescence is going to be a significant contributor to system sustainment costs, what actions can be taken during the initial system design to minimize the impact of part obsolescence? While reactive obsolescence mitigation approaches can provide 3:1 paybacks, pro-active part obsolescence planning holds the promise for 20:1 or 100:1 paybacks*.

*R. Stogdill, “Dealing with Obsolete Parts,” IEEE Design& Test of Computers, pp. 17-25, April-June 1999

Pro-Active Approaches to Part Obsolescence Management

2

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Careful Part Selection

• Does the part manufacturer recommend usage in new designs?
• Is the manufacturer willing to support the part for the long term?
• Is the part technology likely to become obsolete (such as through-hole

versus surface mount)?

• Is the part single-sourced, or is there more than one supplier?
• Are upgrades/substitutes available?
• Can the part be emulated?
• Is the part "modular" in nature (e.g., ASIC)?
• Are the aftermarket suppliers likely to support this part?
• Is the part to be placed on (economically) discardable sub-assemblies?
• Has the design been partitioned to simplify solutions to expected part
• bsolescence problems?
• Does the design consider plans for future enhancements/upgrades?

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Objective: Track and archive obsolete parts and predict when existing parts will become obsolete. Approaches (tools and databases):

• i2 (Aspect, TACTech)

– TACTech – TACTRAC Health Model – Lifecycle Management (LCM)

• CALCE (University of Maryland)
• OPT (SHAI)
• MTI
• Total Parts Plus
• +many other commercial and government database and alert services

(GIDEP, IHS, …)

Forecasting Part Obsolescence

3

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

TACTech:

• Traditionally more focused on providing discontinuance information on parts

already obsolete rather than on lifecycle forecasting for existing parts

• Forecasts parts-specific obsolescence risk index
• Based on TACTech database
• May not be kept up to date with new parts in future (phasing out)

TACTRAC Health Model:

• Forecasts obsolescence risk associated with an entire BOM
• Understands physical hierarchy of system, i.e., can see common parts in

different parts of the system

• Provides alternative part assessment
• Post-processing (plots, reports, etc.) that TACTech service does not provide
• Based on TACTech database, but can be linked with Aspect database

Lifecycle Management (LCM):

• Lifecycle management shell around TACTRAC
• Lifecycle optimizer – recommend alternates for obsolete parts
• Workflow process manager
• Integrated with Aspect electronic parts database

V e r y E x p e n s i v e

i2/Aspect/TACTech Tools

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

TACTech Obsolescence Forecasting

Attributes 1 Emerging 2 Growth 3 Mature 4 Decline 5 Phase Out Speed <= 10 ns 200-500 MHz 11 to 24 ns 133-200 MHz 25 to 34 ns 60-133 MHz 35 to 45 ns 16-60 MHz 55 ns or slower <16 MHz Process BiCMOS DMOS CMOS Bipolar PMOS NMOS Sourcing depth Introduction of new technology, limited sourcing Secondary sourcing Mature sourcing, production peaking Production declines, discontinuance begins Not suitable for design, phase

• ut in progress

Usage Design in Increasing Leveling off Declining Phase out Substrate GaAs Silicon Germanium Complexity ULSI VLSI (100,000+) LSI (10,000+) MSI (1,000+) SSI (100+)

TACTech life cycle forecasting uses an approach in which the part life cycle stage is determined by averaging unquantifiable technological and market attributes.

4

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

TACTech Obsolescence Prediction

• Fundamentally TACTech outputs

– Data in the form of a number (index) that represents obsolescence rating of the component – The index can vary from 1 through 5 – Data is available only for active components; passive components are assumed to be available at all times in some form – TACTRAC post-processes the raw data above and allows it to be used at board and systems levels

• What does the index mean?

– An index value of 1 means being in the beginning stages of life of the component and 5 being in the ending stages of the life of the component – The life span of the component is an average mean lifetime of the component category it belongs to, e.g., microprocessors, transistors, diodes etc.

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

TACTech: Disadvantages

The disadvantages of the TACTech approach are that it:

• does not capture market trends accurately, instead relies on

unquantifiable attributes such as technology complexity, usage, and sourcing depth.

• uses the number of sources (for procurement) in predictions, which

may give false or inaccurate results.

• does not forecast years to obsolescence of a device/technology group,

instead gives an overall life cycle stage for the part number. For example, a manufacturer discontinuing a part does not imply that its device/technology group is obsolete.

• makes the erroneous assumption that all microcircuits follow the

same life cycle curve, e.g., in reality, 16-bit RISC microcontrollers, 4- bit CISC microcontrollers, and 16M DRAMs all have different life cycle profiles.

5

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

TACTech Business Model

• Traditionally, TACTech’s success was based on an extensive

network of human contacts at IC manufacturers and distributors

• Potential competitors could not build or maintain a network

to complete with TACTech

• With IC manufactures adopting of the web as the preferred

delivery mechanism for PCNs and other data associated with electronic parts, the value of TACTech’s human network is decreasing

• Other web-based services (e.g., PCNAlert.com, Total Parts

Plus) are catching up quickly

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Mapping TACTech Data to Years-to- Obsolescence

Mean average life span of component categories

2.5 5 7.5 10

Discrete Devices

M a t u r i t y P h a s e

• u

t I n t r

• d

u c t i

• n

D e c l i n e G r

• w

t h

5 1 0 1 5 2 0

Microcircuits

M a t u r i t y P h a s e

• u

t I n t r

• d

u c t i

• n

D e c l i n e G r

• w

t h

( )

index ce

• bsolescen

TACTech i component the

• f

span Life L year Base B where, 4 1 i 1 L B date ce Obsolescen = = =       − − + =

• TACTech is a popular approach to obtaining obsolescence

data for electronic components, but the obsolescence index data must be mapped to obsolescence dates and uncertainties for use in life cycle cost modeling.

6

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Curve-fit sales data of primary attribute (for example, DRAM memory size) Evaluate years to obsolescence from trend equations Curve-fit sales data of secondary attribute (for example, package style) Modify life cycle stage and years to

• bsolescence of primary attribute by

that of secondary attribute if required Evaluate years to obsolescence from curve-fits of secondary attribute Plot curve fit characteristics vs. primary attributes to form trend eq.

Year

500 1000 1500 2000 2500 3000 92 93 94 95 96 97 98 99 00 01 02 03 04 05

Units shipped (in millions)

Life cycle profile parameters: µ = 1998.2 σ = 1.6 years 16M actual 16M forecast

Current Date (July 1999)

Life Cycle Curve for a 16M DRAM

Electronic Component Years-to- Obsolescence Prediction

• R. Solomon, P. Sandborn, and M. Pecht, “Electronic Part

Life Cycle Concepts and Obsolescence Forecasting,” IEEE Transactions on Components and Packaging Technologies, pp. 707-713, Dec. 2000. CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

1986 1988 1990 1992 1994 1996 1998 2000 2002 10 20 30 40 50 60 70

DRAM Size (Mb) Peak Sales Year, µ

(0.26M, May 1998) (1M, March 1991) (4M, February 1997) (16M, March 1998) (64M, May 2000)

µ actual µ forecast 1991.8M µ

0.0011

=

1 2 3 4 2 4 6 8 10 12 14 16 18

DRAM Size (Mb)

σ = 3.1M

• 0.23

Standard Deviation of Gaussian Fit to Sales Data, σ

Trend Equations for DRAMs

7

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Secondary Attribute Example

(Bias Voltage Trends)

85% 70% 60% 50% 40% 30% 27% 24% 21% 15% 20% 25% 40% 55% 60% 51% 40% 28% 0% 0% 3% 7% 9% 12% 20% 30% 35% 0% 0% 0% 2% 7% 8% 9% 12% 15%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 1993 1994 1995 1996 1997 1998 1999 2000 2001

Year Percentage IC Design Starts

5V 3.3V 2.5V >1.8V 85% 70% 60% 50% 40% 30% 27% 24% 21% 15% 20% 25% 40% 55% 60% 51% 40% 28% 0% 0% 3% 7% 9% 12% 20% 30% 35% 0% 0% 0% 2% 7% 8% 9% 12% 15%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 1993 1994 1995 1996 1997 1998 1999 2000 2001

Year Percentage IC Design Starts

5V 85% 70% 60% 50% 40% 30% 27% 24% 21% 15% 20% 25% 40% 55% 60% 51% 40% 28% 0% 0% 3% 7% 9% 12% 20% 30% 35% 0% 0% 0% 2% 7% 8% 9% 12% 15%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 1993 1994 1995 1996 1997 1998 1999 2000 2001

Year Percentage IC Design Starts

5V 3.3V 2.5V >1.8V

(Fairchild Semiconductor)

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Life Cycle Phases: Exceptions

• False starts: stall in part growth because of

− introduction of a superior competing product − identification of a problem associated with a product − failure to reach critical mass that allows economies of scale − failed emergence of a compelling application for the product − Example: GaAs, a niche technology might soon become obsolete and be replaced by CMOS, which requires less power and is cheaper to manufacture.

• Niche: Products serve niche markets only
• Revitalization: Product revitalization by defining new market segments and/or

product use

Revitalization phase Revitalization False start Niche Time Units shipped/ Market ($) Obsolescence Normal Life Cycle

8

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

DRAM Life Cycles From Trend Curves

(Skipped Generations)

1000 2000 3000 4000 5000 6000 7000 8000 9000 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Year Units (in millions) 16M 4M 8M 32M 64M 256M 2M 1M 0.004M 0.256M 0.512M 128M Skipped generations

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

ASICs differ from off-the-shelf parts in several ways, however, the critical differentiator from an obsolescence viewpoint is that for ASICs the customer has some level of ownership over the design information associated with the part. For an ASIC, the sales curve could represent any one of the following (or combinations of the following):

• Ability or willingness to fabricate the part.
• Ability or willingness to design the part. Loss of ability to design the part may

be due to loss of the required design tool support, loss of human expertise, unwillingness to commit design expertise (too busy designing for newer technologies), or lack of availability of the appropriate intellectual property (IP) – IP obsolescence.

• Lapsing of legal rights to use the necessary IP. The ASIC supplier has the

technical capability but no longer has the necessary legal rights to design or fabricate the part.

A Word About ASIC Obsolescence …

9

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Custom ASICs have been used by many companies, in an attempt to mitigate hardware obsolescence problems: Advantages:

• You have some ownership over the IC design (IP, mask set, etc.), and therefore,

have some level of control over its obsolescence, theoretically, a custom ASIC never needs to become obsolete

• Very effective way to avoid software obsolescence

Disadvantages:

• Owning the mask set doesn’t automatically mean that you can find someone to

fabricate the chip for you. You must keep up with: – Fabricator wafer size changes – Fabricator process changes

• Moving from one ASIC vendor to another may require re-qualification
• You are not immune to technology obsolescence
• Once you are on the ASIC highway, its hard to get off

Re-qualification?

Using Custom ASICs to Avoid Obsolescence Problems?

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Programmable Gate Arrays (PGAs), for some applications, may provide ASIC advantages without all the custom ASIC disadvantages: Advantages:

• You have some ownership over the custom programming of the gate array
• You need not worry as much about fabricator wafer size and process changes

Disadvantages:

• Programmable gate arrays may not be able fulfill your application needs or

performance requirements

• You are still not immune to technology obsolescence
• Once you are on the ASIC highway, its hard to get off (but its easier to get off

PGAs than to get off custom ASICs)

Using Programmable Gate Arrays to Avoid Obsolescence Problems?

10

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Obsolescence Prediction Tool (SHAI)

• Predicts when electronic components are likely to become outmoded and their

production discontinued.

• Obsolescence of a part is predicted by first querying a parts database to identify

key attributes of the part or family of parts, such as its device technology, capacity, and type of packaging. These attributes are then used to query a technology trends forecast database to retrieve generic obsolescence predictors of the part or part family, based on expert opinions. These generic predictors are then fed into a prediction algorithm to derive an obsolescence prediction classification.

• The predictor’s ability is derived from the use of a range of artificial intelligence

techniques including case-based reasoning, object-oriented programming, and knowledge engineering.

• It will operate in either an Internet environment via a web-based client server

architecture, or in a stand alone environment on a single user’s desktop computer.

• The technology used in this project also be used for any non-electronic

components subject to discontinuance due to rapid technical change.

(http://www.shai.com)

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

MTI Obsolescence Forecasting

• Manufacturing Technology Incorporated (MTI) provides obsolescence

prediction information to customers.

• MTI uses proprietary market (α) and technology (β) factors to predict

demise of families and technology/market groups (devices with like market and technology factors).

• The proprietary (α, β) factors are pre-determined from “industry

knowledge”.

• Using (α, β) factors, MTI calculates a "safe" usage window for a device

family.

• The MTI algorithm does not use life cycle curves, and hence is unable to

determine life cycle stage for the technology/market family. (http://www.mtifwb.com/static/obsolescence.shtml)

11

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Total Parts Plus

• Provides PCN and product discontinuance tracking and

notification

• Substitute part information
• Aftermarket supplier information
• Die manufacturer information
• Projects production status of your part, projected availability

in years, and End of Life (EOL) status

(http://www.totalpartsplus.com/products.htm)

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Databases and Alert Services

There are a growing number of companies and organizations providing databases and databases connected to alert services that archive information on parts that have been discontinued. Examples include:

• PCNAlert.com (Cogent) – delivers manufacturer-specific PCN and

PDN information via email

• GIDEP (Government-Industry Data Exchange Program) –

government-wide central system for sharing fact-based technical information (US and Canada)

• COMET (Collaborative Obsolescence Management and Evaluation

Tool) – provides a hierarchical summary of current obsolescence or DMS status of electronic systems

• Arrow Risk Manager (www.arrow.com)
• Total Parts Plus (http://www.totalpartsplus.com/products.htm)

12

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

Data Sources: Market Research Organizations

• Instat, www.instat.com
• Dataquest, www.dataquest.com
• Semico, www.semico.com
• IC Insights, www.icinsights.com
• Integrated Circuit Engineering Corporation

(ICE), www.ice-corp.com

• World Semiconductor Trade Society (WSTS)
• Electronic Trend Publications

CALCE Electronic Products and Systems Center University of Maryland Obsolescence/Technology Insertion

• In the kinetic theory of gases, each atom or molecule in a gas moves

randomly so that the position or velocity of any one of them cannot be known. Nevertheless, using statistics, the rules governing their overall behavior can be worked out with great precision.

• The forecasted window of obsolescence for an individual part in a specific

application is not an accurate predictive exercise on which it makes sense to consider confidence levels or risks. The value of the obsolescence forecast is in the long-term aggregate picture.

• In other words, forecasting the obsolescence of many parts in many products

and using those forecasts to adjust bills of materials, and optimize design and sustainment is advantageous. The value of a single obsolescence forecast on a single product is hit-or-miss and is most likely much smaller than the aggregate value of forecasting for every product.

• Individual part obsolescence forecasts are noise, value is in forecasting for

many parts over many products.

A Comment on Obsolescence Forecasting