Figure 1.1 Four Process Areas
Volume 2, Number 2, April 2002
How effective is your current test equipment? Is your first pass yield percentage a good measure of the production process quality? What is the relationship between effectiveness and yields?
Making sound investment decisions
for test equipment selection is not a simple process. Choosing alternative test
strategies based on process effectiveness is fundamental to cost management.
Ignoring a small detail when making production process yield analyses can be
very expensive. Understanding the basic process functions, terms, and calculations
will allow better choices to be made during test equipment selections or process
improvement investments.
The basic block diagram of a general process is comprised of four process areas.
Production—This is the source for items to be tested.
Testing—All of the production items proceed through the testing process for pass or fail determination.
Repair—If an item fails the testing process, it is sent for repair. Items may make several passes through the test/repair loop.
Final Assembly or Next Step—This step is representative of a number of process possibilities such as assembly, system test, final test, and packing and shipping.
Figure 1.1 Four Process Areas
The heart of the process is the determination
of the quality (pass or fail) of the production boards. We know the assemblies
coming from production (N) contain both good boards (Ng) and bad
boards (Nb). Therefore the production boards can be expressed with the
following formula: N = Ng + Nb.
As the boards are processed through
the tester, some of the good boards will be passed by the tester (Ngp)
and some of the good boards will be failed by the tester (Ngf). Likewise,
some of the bad boards will be passed by the tester (Nbp) and some of
the bad boards will be failed by the tester (Nbf). The fact that the
tester failed some of the good boards and passed some of the bad boards happens
because the tester is not perfect. This testing process can be shown in Figure
1.2.
Figure 1.2 Testing Process Equations
As you can see, the tester’s
determination of the “goodness” of the boards may be different than
the true quality level of the boards.
In other words, the production output of “good” boards is different
than the tester’s output of good boards.
The output of good boards from a
specific process step is called the yield of the process. The example above
reveals that yields may not be the same for each process step and is dependent
upon the ability to measure the goodness of the board. Two common yield classifications
are the production-related and the tester-related yields. A measure of the production
yield (Yp) can be stated as a percentage of the output of good boards
(Ng) relative to the total number of boards produced (N), as expressed
in the formula Yp = Ng/N.
For example, if the yield for a board
was 85 percent, then 85 of every 100 boards produced are good. The determination
or confirmation of the production yield is done using testing/inspection processes.
If the production yield (Yp) is equal to 85 percent, a perfect tester/inspector
would pass 85 boards on to the next step and the balance of 15 would be passed
on to the repair process step. Unfortunately, because testers are not perfect,
some of the bad boards will pass and some of the good boards will fail. If the
tester is unable to identify a specific fault, then this fault will not be detected
and will pass to the next process step as a good board.
A commonly used yield is the first
pass yield (Y1). This is a measure of the number of boards that test
good (Np) the first time they are tested relative to the total number
of boards tested (N). Expressed as a formula, Y1 = Np/N. The two
components that contribute to the determination of the first pass yield are
the production process and the tester’s ability to detect faults. In the
production environment, both of these elements are unknown; therefore, the first
pass yield percentage developed should only be used as an approximation of either
process or tester quality.
Other tester-related yields worth mentioning are the output yield to next step (Yf) and the output yield to repair (Yr). Again, these yields are calculated using the number of good boards that pass (Ngp) and the number of bad boards that fail (Nbf) test. They can be expressed in the following formulas: Yf = Ngp/Np and Yr = Nbf/Nf. Figure 1.3 below depicts these yield calculations.
Figure 1.3 Yield Calculations
Another way to look at yield percentage is to equate it to the probability of the board being good. So if the yield (Y) of a process step is 95 percent, the probability of producing a good board from that process is also 95 percent. Inversely, the probability of a board being bad is 1-Y, or 5 percent.
As we learned above, because testers
are not perfect, bad boards will pass and good boards will fail. We can calculate
the effectiveness of a tester for both of these shortcomings. Because these
two elements are independent functions, two different calculations are required
to characterize the tester’s effectiveness.
To determine bad-board test effectiveness
(Eb), we must describe the ability of the tester to perform the mission
of failing bad boards. Expressed as a percentage, Eb is the likelihood
of a bad board (Nbf) being detected as compared to the total number of
bad boards (Nb). Expressed as a formula, Eb = Nbf/Nb. For example,
if the Eb is 95 percent, 95 of every 100 bad boards tested would be detected
as being bad. Five percent of the undetected bad boards (100% – Eb)
would be considered good and would be passed on to the next process step.
To determine good-board test effectiveness
(Eg), we must describe the ability of the tester to perform the mission
of passing good boards. Also expressed as a percentage, Eg is the likelihood
of a good board (Ngp) being passed as compared to the total number of
good boards (Ng). Expressed as a formula, Eg = Ngp/Ng. For example,
if the Eg is 95 percent, 95 of every 100 good boards tested would be
passed as being good. Five percent of the failed good boards (100% - Eg)
would be considered bad and would be sent on for repair.
The effectiveness of any tester is usually less than 100 percent. This inefficiency causes us to waste resources to resolve these shortcomings. Usually the costs involved in this endeavor can be minimized and offer an opportunity to avoid these costs. In simple terms, if we can decrease the number of times we send bad boards to the next process step and send good boards back to the repair step, our operational costs can be reduced.
By selectively applying process improvements to solve problems causing our effectiveness to suffer, we can make strides in improving yields and tester effectiveness and in reducing costs in our processes.
Sources:
1. Hewlett Packard. "TEST EFFECTIVENESS, YIELDS, and CO$T$". February
1986.
2. Davis, Bredan. "The Economics of Automatic Testing". McGraw-Hill
Book Company (UK) Limited, 1982.
 |
 |
 |
 |
 |
| http://www.divsys.com | ||||
| http://www.rapidproto.com | ||||
| Voice:
(317) 299-9547 Fax: (317) 298-2055 |
||||
| sales@divsys.com |