IPC9850_Surface Mount Equipment Characterization.pdf - 第20页
IPC-9850 Official Proposal May 2001 20 Table 4-3 Calculation Example Machine Number 1 (Note 1) 2 3 Run Series 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Run Series 2 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 1…
IPC-9850
Official Proposal
May 2001
19
Table 4-1 Component Types and Quantities for ‘Typical kit’
Quantity Component
Type
440 SOT-23
440 SOIC-8
880 1608C
880 1608R
880 1005C
880 1005R
The components are to be placed onto an adhesive laminated PWB. The land pads for the PWB shall meet IPC-SM-
782. (Specific board size and layout are not defined by this standard.) After the parts have been placed on the sticky
PWB, a manual visual inspection is made to determine how many placement defects have occurred. While manual
inspection is less than ideal due to its subjectivity, the investment in Automated Optical Inspection (AOI) equipment
for such a limited application cannot be justified.
In this standard, attribute defect rate information differs significantly from the performance metric defined by the
repeatability and accuracy parameters. Whereas the repeatability and accuracy performance parameters are defined
through continuous data -i.e. numerically how much is the component placement error-, the attribute defect rate
parameter is only a function of discrete defects i.e. components which are known to be placed such that a proper
solder joint will not form. Generally these discrete defects occur at such small rates that they are not measured in
percentages of the total number of placements but in defective parts per million (PPM) placements. These discrete
defects are referred to as Attribute Defects, and the anticipated frequency of their appearance is referred to as the
Attribute Defect Rate.
Attribute defects are defined as components placed upside down, tombstone, on side, missing, extra part, damaged
lead(s), damaged part, completely off land, and wrong polarity. An exception is that upside down components are
not to be counted as defects when using bulk feeders. Partially off-land and askew type placement defects are not
included, since the performance level for these types of defects is accounted for by the repeatability and accuracy
parameters. For this parameter evaluation, each component equals one opportunity for a defect and can have a
maximum of only one defect, regardless of the associated anomalies –such as the number of leads, etc. Things other
than the placement machine can cause attribute defects, but they shall be attributed to the machine in the absence of
any other clear cause, such as poor board fiducials, etc.
Estimating the average number of defects that will occur in each million placements is no small task. Estimating
defect rates less than 50 PPM requires very large sample sizes. Different sampling plans exist, each with it's own set
of characteristics and sample sizes.) This standard requires the placement of 88,000 parts. Due to the time and cost
associated with a machine supplier placing this many actual components with each machine manufactured, the
88,000 placements may be spread out over 20 consecutive machines in builds of 4,400 components. As a result, the
confidence level applies only to that group of machines rather than any specific member of that group of machines.
The data may be used as an estimate of the performance of the model type. When a new model machine is launched,
where only a single machine of its kind is available for testing, this single machine shall be utilized for the 88,000
placements.
Calculation Method The PPM level is the total number of observed attribute defects during 88,000 placements of
various component types, divided by 88,000, multiplied by 1,000,000. A minimum total of 88,000 components is to
be placed over 20 tests, where each test run places a component mix of at least 4400 components. The number of
defects is then calculated as a moving average of the last 20 tests.
Attribute Defect Rate = (1,000,000/88,000) *
∑
+
=
20i
itest
i
efectsAttributeD
The following example refers to Table 4-3 where two defects over the 88K parts yields Attribute Defect Rate of 23
ppm. For the first and second machines, the attribute defect rate is obtained by dividing the three attribute defects –
found during run series 2, 7, and 17-- divided by 88K parts, for a 34 PPM. Once the third machine is shipped, the
attribute defect level is reduced to 23 PPM, since the single attribute defect of run series 2 is replaced by the zero
attribute defect of machine 3, run series 22.
IPC-9850
Official Proposal
May 2001
20
Table 4-3 Calculation Example
Machine
Number
1 (Note 1) 2 3
Run
Series 1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Run
Series 2
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
Run
Series 3
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Number
of
Attribute
Defects
0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0
Note 1: The first machine built will have 20 runs of 4400 components.
4.2 Reliability Parameters
In contrast to the machine placement performance parameters that specify worst case performance, this standard
uses reliability parameters to describe typical performance. Reliability metrics are difficult for a supplier to specify
because there is no reasonable way to verify these parameters for an individual machine prior to shipment, without
rigorously exercising the machine for thousands of hours. Not only would this be costly, it would result in a new
machine becoming a used machine. Since it is not reasonable to expect all suppliers to dedicate machines for
reliability tests, in this standard the suppliers shall use reliability information collected by users of their equipment.
The problem of specifying reliability metrics is now compounded by the fact that users are collecting the data with
which reliability estimates are made by the suppliers. Each user has its own set of understanding and ability to track
reliability metrics amidst the pressures of daily production. The application of the equipment and skill of its
operators and technicians are believed to be significant factors that contribute to the final observed reliability and
maintainability. In this standard, terminology and metrics have been developed that help suppliers and users collect
and exchange reliability data.
The high yield and reliability of modern SMT placement systems requires very large amounts of data in order to
estimate these metrics. The desire for large data sets having fairly precise estimates of reliability performance must
be balanced with the desire to report reliability as early as possible. Confidence intervals for many machine
parameters can be made on the basis of what is observed in a sample because the underlying statistical distributions
associated with those parameters are fairly well understood. However, at the initiation of this standard, some
suppliers indicate that they do not have a complete enough understanding of the statistical distributions associated
with one or more of the reliability metrics in this standard.
Some types of placement machines produce failure data with χ
2
distributions. At first glance, somewhere around
half the machines would have somewhat better than the reliability levels experienced during the field studies, while
the rest of the machines would experience somewhat worse reliability levels. In fact, if the machines have constant
failure rates, 63% of the machines will have assists and failures more frequently than the observed MPBF and
MPBA. This is due to the fact that most reliability distributions (such as χ
2
) are skewed to the right, causing their
median to be smaller than their mean.
Performance similar to the typical MPBF and MPBA performance of the equipment should be enjoyed when the
recommended preventive maintenance (PM) activities are completed in a regular and timely manner. When
preventive and corrective maintenance is not performed in accordance with the maintenance manual, reliability may
be degraded.
To address confidence interval concerns, equipment suppliers shall only complete the reliability parameters with
data based on at least 3 times as many placements as the reported mean placements between failures (MPBF)
parameter (see 4.2.3). For example, it is not proper to claim MPBF of 1 million placements, without observing that
IPC-9850
Official Proposal
May 2001
21
level of performance over 3 million placements. The same is also true for the mean placements between assists
(MPBA) parameter (see 4.2.1), were at least 3 times as many placements as is reported for MPBA need to be made.
Since several machines shall be utilized for the collection of data for the reliability metric on multiple machines, the
following method shall be utilized for merging the data sets. For example, assume a supplier has data for two
machines of a particular model type. The first machine was observed for 300,000 placements and experienced 2
assists while the second machine was observed for 500,000 placements and experienced 3 assists. The calculation
for MPBA for the model type would be the total number of cycles divided by the total number of assists. In this
case, MPBA would be
000,160
3
2
000,500000,300
=
+
+
cycles.
Other metrics shall also be calculated this way when multiple machines have been observed.
An assist is defined as an unplanned interruption that occurs during an equipment cycle where all three of the
following conditions apply:
Ø The interrupted equipment cycle is resumed through external intervention (e.g., by an operator or user, either
human or host computer).
Ø There is no replacement of a machine part (defined specifically to distinguish from component parts being
placed by the machine), other than vendor specified machine consumable parts.
Ø There is no further variation from specifications of equipment operations.
This definition was clarified not to include replenishment of components since it is tied to the type of feeders and
tape capacity choices made by the user. It was agreed that replenishment is to be specified as a scheduled downtime.
A failure is defined as any unplanned interruption or variance from the specifications of equipment operation other
than assists.
Preventive Maintenance (PM) is a machine stop required by the supplier's published PM schedule.
See table for 4-4, below, for examples of the terms ‘Assist’, ‘Failure’, and ‘Preventive Maintenance” as applicable to
this standard: