转帖] Economic Justification of Halt Tests

版权属于作者


转帖] Economic Justification of Halt Tests
Economic Justification of Halt Tests:
The relationship between operating margin,
test costs, and the cost of field returns
_Edmond L. Kyser and Nahum Meadowsong Cisco Systems, Inc.
Email: ekyser#cisco.com
nameadow#cisco.com_
_Introduction



Increasing pressures for cost reduction

Prototype build is a leading cost item in product development

Halt test requires destruction of (at least) one prototype at a critical stage (earliest stable hardware and software)

How is this cost best justified?

The benefit of a Halt test is AVOIDED COST –


Improved reliability (reduced RMA rate)

Reduced test cost (eliminate/reduce ORT, RDT)


What is the relationship between Halt results (operating margin) and RMA rate?

Under what conditions is the cost of Halt justified?



**Operating Margin vs. Reliability
Can Halt tests be used to predict field performance?**


No !
A Halt test is NOT a highly accelerated life test – it’s not a life test at all. There is no appropriate acceleration algorithm, no acceleration factor. The deliverables of a Halt test are operating margin and failure modes.
Yes !
Operating margin is an indicator of field performance. Low margins indicate poor performance (short life), and high margins indicate good performance (long life). Halt tests determine operating margin, and failure modes show where margins may be improved.





The issue of relating Halt results (operating margin) to field reliability is NOT a yes/no issue, but rather how to express the relationship correctly.


It will be an Empirical relationship

It will not be independent (There will be other variables)

It will be probabilistic in nature - confidence interval

It will be product dependent


The following plot shows an Empirical relationship between Operating margin and Reliability for similar products (high performance line cards for Cisco routers)


!(http://www.reliantlabs.com/ima ... e1.gif)




Other Independent variables influence RMA rate:

Board complexity, measured by active component count
-Includes ASIC, IC, FPGA, Transistors, Crystals, Diodes

The following plot shows Normalized RMA rate vs. parts count, FOR A LARGER RANGE OF PRODUCT.
The correlation is about 1/10 that of Normalized RMA vs Operating Margin for Line Cards



!(http://www.reliantlabs.com/ima ... e2.gif)


Cost justification 1: Eliminate RDT


The following 2 slides show traditional RDT and Halt RDT
Traditional RDT (80% confidence in MTBF > 75,000 hours) The dashed blue line shows RDT requires 40 boards tested for 10 weeks. (assuming 1 fail and Arrhenius acceleration due to 50C temp)
Halt RDT requires 1 board for 1 week. The dashed blue line shows 80% confidence for Operating margin of 30C indicating Normalized RMA rate below 0.55


Standard RDT

!(http://www.reliantlabs.com/ima ... e3.gif)

80% CONFIDENCE LEVEL

!(http://www.reliantlabs.com/ima ... e4.gif)

Cost Justification 2: Improve Reliability


From slide 6, if the operating margin is increased N °C , the normalized RMA rate is reduced 0.0192N
Each RMA costs approximately WPC$, where WPC is the cost of producing the board, the Whole Product Cost.
Thus the Benefit of a Halt test that increases Operating Margin N °C is
The benefit of a Halt test
Benefit = (# of RMAs prevented) (cost of an RMA)

= N(0.0192)(RMA intercept)(Pvol)(WPC)$

Where
Pvol is the (annual) production volume

The cost of a Halt test
Cost = WPC + Esalary + DEP + CON + CA
Where
Esalary = fully burdened weekly salary
DEP = Weekly Depreciation of equipment
CON = Consumable costs
CA = Corrective action costs

The break-even point is where costs = benefits.
For a Halt test to be cost effective, it must, on average, increase the operating Margin N °C, where



N =



WPC + Esalary + DEP + CON + CA

_!(http://www.reliantlabs.com/images/testing/line.gif)_
(0.0192)(RMA intercept)(Pvol)(WPC)$
_