TDEB Top Bar

Trending
Semiconductor Industry News — The Digital Electronics Blog
Follow Us

TDEB Masthead

The Semiconductor Hub

Semiconductors.  Chips and Systems.  Innovation.

Indexed by leading Search Engines, AI and referenced by Engineers worldwide 500+ Articles 10M+ Pageviews 30+ Reports 50+ Citations

The Economics of Test, Part 1: What Drives the Cost of Test

The Editor's Desk
by
0
Share Facebook X LinkedIn WhatsApp Email

What are the factors that influence the cost of test? Quality and test costs are related, but they are not the inverse of one another. As we shall see, an investment in a higher-quality test often pays dividends during the test cycle.

Test-related costs for ICs and PCBs include both time and resources. As pointed out in previous sections, for some products the failure to reach a market window early in the life cycle of the product can cause significant loss of revenue and may in fact be fatal to the future of the product.

About the Author:
Name: Joachim Bauer, Test Engineer
Experience: 13+ Yrs
Location: Nice, France

This is Part 1 of the series. Part 2 breaks down test costs, Part 3 covers testing, yield and feature size, and Part 4 covers the cost of locating board faults.

Test in plain words

Every chip and every circuit board is tested before it ships. Testing costs money: test equipment, engineers to write test programs, extra circuitry on the chip to make it testable, and above all time on the tester and time before the product reaches customers. Skipping or shortening tests saves some of that money, but defective parts then reach customers, and fixing problems in the field is far more expensive than catching them at the factory.

Why quality and cost are not simple opposites

It is tempting to think of test as a dial: turn it up for quality and pay more, or turn it down to save money. The author's point is that the relationship is not that simple. A better test, one designed with care and supported by design-for-test (DFT) features, can find more defects in less tester time. It costs more to prepare but less to run, and it lets fewer bad parts escape. Semiconductor Engineering's 2024 article Hidden Costs And Tradeoffs In IC Quality makes a similar point, describing the link between quality and test cost as cyclical rather than linear.

The cost of being late

The second idea in the original text is time to market. Test development sits on the critical path: a product cannot ship until its test program is ready and its yield is understood. If a slow or troublesome test program delays a launch, the company can miss the window when prices and demand are highest, and for a short-lived product that lost revenue can outweigh every other test cost.

What drives the cost of test

  • Preparation: writing and debugging test programs, designing DFT logic.
  • Execution: tester time, handlers, probe cards, operators.
  • Silicon overhead: the area added to every die by DFT circuits.
  • Imperfect test quality: bad parts that pass (escapes) and good parts that fail (yield loss).
  • Time: the revenue lost if test delays the product.

Part 2 of this series looks at each of these in more detail.

Deep dive: the numbers and details

This part holds the published cost ratios and two worked examples. Skip it if the overview is all you need.

The rule of ten

A defect costs more to deal with the later it is found. NXP's article on test economics and the cost of late detection gives approximate relative costs:

Where the defect is foundRelative cost (approximate)
Wafer test1
Package test10
Customer incoming inspection100
System assembly1,000
Field failure10,000 or more

NXP calls this the rule of ten. The same article cites an estimate that for every 100 killer defects found at test there are roughly one to two latent defects that can cause reliability failures in the field.

What the tester costs

In the Semiconductor Engineering article, an industry expert describes production testers as typically costing hundreds of thousands or millions of dollars. The article also cites test time reduction work that found potential savings of 15% to 24% of test time without affecting quality, and quotes an estimate that marginal defects that escape can cost a customer 10 to 100 times more than the scrap cost of the part.

Worked example: test time and the rule of ten

The figures below are illustrative only. Suppose a tester costs 0.05 dollars per second to run, a part needs 2 seconds of test, and you ship 10 million parts a year.

  • Execution cost: 2 s x 0.05 $/s = 0.10 dollars per part, or 1 million dollars a year.
  • Cutting test time by 20% saves 200,000 dollars a year.

Now suppose the shorter test lets an extra 50 defective parts per million escape. That is 500 extra escapes a year. If each escape costs 100 dollars to handle at the customer, the escapes cost 50,000 dollars: the shorter test still wins. If each escape becomes a field failure costing 1,000 dollars, the escapes cost 500,000 dollars and the "saving" becomes a 300,000-dollar loss. The same change in test time can be a good or a bad decision depending on where the escapes are caught, which is the author's point that quality and cost are not simple opposites.

Worked example: the cost of a late launch

Again with illustrative numbers: a product is expected to earn 20 million dollars over an 18-month life, with demand and prices highest in the first months. If test program problems delay the launch by three months and the product's end of life does not move, the company loses its best three months of sales. Even a rough estimate of one sixth of lifetime revenue is more than 3 million dollars, far more than most test program budgets. This is why test engineers are brought in early and why DFT is planned with the design rather than added at the end.

Takeaways for engineers

  • Count escapes, not only tester seconds, when you judge a test change.
  • Put test development on the project schedule from the start.
  • Spend on DFT when volumes are high or when field failures are expensive.
The Editor's Desk
About the Author
The Editor's Desk
Share Facebook X LinkedIn WhatsApp Email

Post a Comment

0 Comments

Your comments will be moderated before it appears here.

Post a Comment (0)

#buttons=(Ok, Go it!) #days=(20)

Our website uses cookies to enhance your experience. Learn more
Ok, Go it!