statistical monitoring of solder paste printing...*salo operations, nokia mobile phones, nokia...
TRANSCRIPT
「エ レク トロニ クス実 装学 会誌 」7〔3〕255~260(2004)●技術論文
Statistical Monitoring of Solder Paste Printing
Timo LIUKKONEN* and Aulis TUOMINEN**
はんだ印刷工程の品質管理方法
*Salo Operations, Nokia Mobile Phones, Nokia Corporation (PO. Box 86, 24101 Salo, Finland)
** Tampere University of Technology, Pori (P.O. Box 300, 28101 Pori, Finland)
概 要 現在 プ リント板生 産において,は ん だ印刷 の品質向上が必須 であ り,光 学 的手法 によるはん だ印刷状態の継続的 な
監視および管理が必 要 とされている。 ここでは,基 板上 にテス トパ ター ンを印刷す ることによるはんだ印刷品質の評価手法 に
ついて報 告する。 数種類 のテス トバ ター ンが,は んだ印刷工程 を通 じてテス トパ ッド上 に印刷 され る。通常,そ の印刷 され た
テス トパ ター ンの間隔とその面積 は,は んだ印刷 の状況 に応 じて体系 的に変化 する。 そのテス トパ ター ンを評価す ることによ
りはんだ印刷品質を事前 にク ラス分類 する ことができ,ま た同様 に多種 のプ リント板 にも適用で きる。例 えば複数 のライン間
ではんだ印刷の品質を比較 する こと も期待 で きる。 さ らに,は んだ印刷 に関連す る要因 のば らつきを減少 させ る手法 につ いて
も議論で き,は んだ印刷 の品質を体系的に評価 する ことができる。近年,テ ス トパ ター ンを用 いた ビデオカメラによる外観検
査か ら工程 全体を監視する手 法が報告 されているが,本 研究 では,は んだ印刷品質 を評価 する適切 なテス トパ ターンを開発す
ることによ り,製 品 ごとに最適 な印刷条件 を供給 することができた。 また作業性 を考慮 したカメラ システムを開発 し,そ れ ら
を生産 ラインにおいて適 用する ことができたので報告 する。 さ らに,こ の研究 にもとづ き生産 ライ ンの改善事例 につ いて報告
す る。
Key Words: Solder Paste Printing, Process Control, Paste Inspection, Process Monitoring, SPC
1. Introduction
According to estimations 50 to 70 percent of the total
defects in surface mount assembly lines are related to
the paste printing process, and approximately 30 to 50
percent of the total manufacturing cost is due to test
and rework expenses. The newest component package
types, such as 0.5 mm contact pitch CSP (Chip Scale
Package), are used more widely in today's electronics
industry. CSP assembly should not require specialized
equipment or processes beyond that currently being
used for other surface mount device attachments. Most
importantly, the stencil printing process must provide
uniform and consistent solder paste application. This,
together with strict process monitoring, can control
and maintain soldering quality1)•`3). The limited use of
feedback control in the printing process of existing
surface mount lines is most likely due to the complex-
ity and the high degree of natural variability included
in the process4). The stencil printing process is charac-
terized by having very significant amounts of noise
and by requiring a bounded and smooth output con-
tinuously5),6). Quality printing is crucial also for suc-
cessful component placement. Poor paste printing
quality limits the possibilities of e.g. placement opti-mization, because, instead of using PWB (Printed
Wiring Board) layout as the most determining factor,
placement sequences must be defined from inconsis-
tent paste output and thus from varying tackiness of
different component types.
In order to decrease the deviation in printing by con-
stant monitoring of printing result, a new method was
developed, based on at least one test pattern, that is
situated on the PWB area, i.e. on the product area or
on the panel frame. Because printing quality of the test
pattern is monitored instead of monitoring printing of any of the pads on the PWB itself, it could be stated,
that inspection is moved from inspecting of the prod-
uct towards inspecting of the printing process and its
parameters.
2. Development of the Test Pattern
The test pattern consists of a number of test ele-
ments, shaped like geometrical plane figures and of
varying degree of difficulty in view of the printing
process1. The example pattern in Figure 1 that is
printed on the surface of the test pad during normal
stencil printing procedure of the product, measures the
エ レク トロ ニ ク ス実 装 学 会 誌 Vol .7 No.3(2004) 255
Fig. 1 Test pad (size 3.0•~3.0 mm) before (left) and
after (right) printing of the test pattern. •gCross
•h on the left corner is a fiducial for a video cam-
era based monitoring method
printing quality both in horizontal and vertical direc-tion. This is very important, because the quality in
printing squeegee movement direction may be differ-ent from that in the transversal direction. However in
the case of e.g. a QFP (Quad Flat Pack) component
both directions have equal importance. Test pattern
consists of paste deposits of varying length and/or
width, and gaps of varying width between the de-
posits.The dimensions of the test pattern should be defined
from the difficulty of the PWB; it is recommended to
have the smallest gap and smallest deposit smaller
than any pad or pad spacing on the product itself.
Thus good printing quality of the test pattern, being
the most critical object, reliably indicates good printing
quality of the rest of the board7). If PWB has a lot of CSP components, it could be useful to have the pattern
also consisting of circular shaped figures of varying di-
ameter and varying spacing in between, some of them
being preferably smaller than CSP pads on the prod-
uct.
3. Setting the Monitoring Targets
Changes in the printing process can be seen as
changes in the test pattern. Certain kinds of process
changes affect the area of the paste deposits, some
changes affect the gaps (i.e. area free of paste) between
the deposits. If e.g. viscosity of paste gets too low,
there is still enough paste on the deposits, probably
there is too much, but the gaps get smaller, because the
paste spreads also in the areas that should otherwise be free of paste. Thus it is important to monitor both
the deposits and the gaps separately to notice any un-
desired change that may happen in the printing
process. Some paste printing machines have an em-bedded camera based inspection system, but especially
monitoring the areas between the deposits is very dif-
ficult for these systems, because of poor contrast be-
tween the dark paste and dark solder mask of the
PWB. The test pattern has clear contrast between the
deposits, because also this area of the test pad is cov-
ered with e.g. gold, tin-lead or copper.
4. Defining Printing Quality
The paste deposits and the gaps can be examined ei-
ther visually (by microscope etc.) or by means of a
monitoring equipment e.g. CCD (Charge Coupled De-
vice) camera. If the widths of the test elements on the
test pattern have been chosen appropriately, the results
of the monitoring can be classified according to how
well the paste printing process has succeeded for each
test element. If the smallest paste deposit element has
an area of paste of just the required size or amount and
the smallest gap has acceptable area free of paste, the
printing process can be regarded as successful in all re-
spects; the printing process belongs to the first or best
printing class. Of course the printing of the bigger de-
posits and gaps must also be successful. If the smallest
deposit or gap has failed, but the rest of the deposits
and gaps are acceptable, the printing process can be
classified as belonging to the second best class. Simi-
larly the whole process can be classified into sub-
classes that reflect the printing quality1.
Monitoring can be automated by suitable CCD cam-
era based equipment, which takes a picture of the test
pad from every board that passes it. Image processing
software then calculates the area of each paste deposit
in the test pattern and the area free of paste between
the deposits. Average or preferably weighed average
of all the test elements separately for deposits and
gaps are then calculated to present the quality of the
printing. SPC (Statistical Process Control) gives addi-
tional benefit to this monitoring method. Variation in
any process is caused by common causes and special
causes. The common causes are •ginbuilt•h in the
process, affecting all products, and can be identified
and eliminated with continuous process improvement
tools. Special causes are not present in the process con-
tinuously, but are present only in certain circum-
stances, and can be identified with SPC tools8)•`10) SPC
can help to avoid taking actions too sensitively in out-
of-control situations, because SPC does not give an
alarm until there is an expected certainty that e.g. an
undesired trend is noticed in the printing process. SPC
can be used in both manual (visual) and automated so-
256 エ レク トロ ニ ク ス実 装 学 会誌 Vol,7 No.3(2004)
lution of this monitoring method.
5. Case Study
Before introduction of the printing test pattern to the
production line a printed PWB was weighed once per hour to get indication about the amount of paste on
the PWB for process control purposes. Based on the
weighing the printing parameters were adjusted to in-
crease or decrease the amount of paste. Unfortunately
this procedure increased the total variability of the
printing process because the process was overadjusted due to too small sample size and too small sample fre-
quency. Figure 2 shows the principle of the negative effect of process overadjustment on distribution10).
A project was started to introduce an automated
new tool for paste printing process control. The
process control method to be developed should be able to detect changes in the following paste printing qual-
ity characteristics:
1) Area of the paste deposits
2) Area of the gaps between the deposits
3) The locations of the paste deposits (possible off-
set)
The printing test pattern, which was described in the
previous chapters of this paper, was developed sys-tematically through evolution of various versions:
1) The first pattern was used only for paste cover-
age horizontally, individual pads were used for
different test elements in size
2) Instead of having individual pads for each paste
deposit, a large pad was designed common to all
deposits; this finally enabled detection of gaps
between the deposits (excellent contrast was
achieved)
3) Vertical row of test elements was introduced at
the end, when the dimensions of the test elements
were defined completely
A patent was applied (and granted later in 2002) to
the principle of using a test pattern for paste printing
process control1). A CCD camera based device was de-
veloped to automate the control procedure of the de-
signed test pattern. The device is placed above the pro-
duction line conveyor where the boards are measured
while they are moving beneath the camera unit. The
system consists of the following parts (see Figure 3):
•\ Video camera
•\ Proximity sensor with optical fibre
•\ Light tower for SPC alarms
•\ Diffuse On-Axis light tip for the strobe light unit
Instrument rack includes the following parts:
•\ Personal Computer (PC) and keyboard
•\ Vision processor located in PC bus
•\ Video monitor
•\ Trackball
•\ Strobe light unit with optical fibre
Image processing software (to calculate area for de-
posits and gaps) and SPC software (to analyze the re-
sults) were installed inside the PC unit. The proximity
sensor is used to synchronize the camera scan start.
The image processor reads the input trigger, fires the
strobe light and starts the camera scan. The video
monitor is also controlled by the vision processor and
all system functions are displayed on it. The user inter-
face commands are executed by using the trackball.
The image-taking sequence is as follows:
1) PWB arrives below the camera
2) Sensor commands the vision processor to take a
picture
3) Strobe light flashes and video camera takes the
picture of the moving PWB
4) Image processor analyzes the picture, converts
the pixels to percentages of coverage and sends
the result to SPC
5) The control charts are updated on the video
monitor
6) Light tower status is updated
The pilot system was installed between paste print-
ing machine and the first placement machine. Because
the most critical package on the board was a QFP176
with 0.4 mm lead-pitch, the dimensions of the printing
test pattern were selected especially for it as shown in
Figure 4. Image processing software calculates the
Fig. 2 Process overadjustment may increase the
process variability since the process average is shifted each time an adjustment is made. Even-tually the number of products being out-of-specification (LSL/USL=Lower/Upper Specifi-cation Limit) is increased
エ レ ク トロニ クス 実 装 学 会 誌 Vol .7 No.3(2004) 257
Fig. 3 Mechanical construction of the system
Fig. 4 Width of the test elements in the test pattern
(all element lengths 1.0 mm; the outer dimen-
sions of the test pad 3.0•~3.0 mm)
amount of black pixels (i.e. paste) in the area of the de-
posit and converts it into percentage (in relation to the
maximum •gideal•h paste coverage; stencil opening rep-
resenting the maximum paste area) separately for dif-
ferent paste deposits. The same calculations and con-
versions into percentages are made for the gaps using
white pixels (area free of paste); see Figure 5 for a prin-
cipal example. The two final values to be plotted on
the SPC-charts after each board are (weighed) average
of paste deposits' percentages and (weighed) average
of gaps' percentages. Users can apply different weight
values for different test elements if desired.
Because individual values from successive PWBs do
not form a normal distribution (Gaussian curve), val-
ues from successive PWBs are calculated into sub-
groups and averaged in order to get a normally dis-
tributed process, to which SPC criterias can then be
applied (The Central Limit Theorem, see8),9)). Both the
quality of the deposits and quality of the gaps has
their own X-bar/R-control chart on the video display
of the system. X-bar-chart will show the averages of
the successive subgroups and R-chart will show the
range inside each subgroup8)•`10). See Figure 6 for an
Fig. 5 Example of area of paste deposits (black pixels inside deposits) and area of gaps (white pixels
between deposits) converted into percentages
per test element (gaps' percentage values shown in brackets)
258 エ レ ク トロニ ク ス実 装 学 会 誌 Vol.7 No,3(2004)
Fig. 6 X-bar/R control chart including average values
(top chart) and range values (bottom chart) of subgrouped printing results of paste deposits.
Out-of-control situation is present
example of an X-bar/R-chart.
In Fig. 6 a typical out-of-control situation is pre-
sented, showing towards the end of the chart the print-
ing quality dropping very fast when a new solder
paste lot is taken into use in printing, and viscosity of
that is far too low. Gray scale camera picture from the
test pattern before and after the lot change to poor vis-
cosity can be seen in Figure 7.
SPC control limits were calculated from the •gnor-
mal•h process during a long-term monitoring period,
representing stable process and having natural print-
ing variability in it. 3-a (sigma) limits were used as
control limits, 2-a and 1-ƒÐ limits as warning lim-
its8)•`10). Classic out-of-control criterias used in SPC
were chosen to the SPC software to give an alarm in a
situation of special cause existing in the process. For
more information on the criterias, see e.g.8)•`10)
The X-bar/R-chart presented together with the
shape of the distribution in Figure 8 shows promising
results during several weeks' usage of the printing
process control device. It shows an increase for aver-
age paste coverage of ca. 5%.
6. Conclusions
Continuous monitoring of printing of the test pat-
tern offers tool to monitor the paste printing process.
Key target is a stable printing process, where any un-
desired changes in the process itself are noticed rap-
idly. Many factors affect the quality of the stencil print-
ing process, including squeegee pressure, printing
Fig. 7 Paste printing pattern on the test pad before
(left) and after (right) the paste lot change
Fig. 8 X-bar/R-control chart of paste coverage from several weeks' production period after the installation of paste print-ing monitoring device showing improvement: average paste coverage (upper chart) has increased ca. 5% and its average range (lower chart) has decreased. However, a slight backlash is seen at the end of the period on both
charts
エ レ ク トロニ クス 実装 学 会誌 Vol .7 No.3(2004) 259
speed, solder paste viscosity and rheology, and air
temperature and humidity within the production envi-
ronment4). Changes in these factors affect the overall
printing quality of the PWB. If the test pattern can be
situated to a place common to all products produced
in the production line, e.g. on the panel frame, it is not
necessary to do any changes to the setup of micro-
scope, CCD camera or any other monitoring device,
when product is changed to another. This makes it
easy to compare printing capability between different
products and production lines giving also possibility
to better standardize production methods.
If we could get also volume or height values of the
paste deposits of the test pattern, it would give us
more possibilities to classify the printing process. This
would however require more sophisticated monitoring
device e.g. laser based and remarkably increase the
cost of the system.
In our case study using the developed test pattern,
the developed video camera based control device and
embedded SPC software we were able to improve the
average paste coverage during a several weeks' evalu-
ation period. The preliminary results are very promis-
ing and encourage us to further study the effect of the
method especially on variation of the paste printing
process on a longer term. (2003.10.22-Žó—•)
References
1) A. Forsten and T. Liukkonen:•gMethod for Monitoring
Solder Paste Printing Process•h, United States Patent no:
US 6,342,266 B1,Jan. 29, 2002
2) J. Pan, G. Tonkay, R. Storer, R. Sallade and D. Leandri:
•g Critical Variables of Solder Paste Stencil Printing for
Micro-BGA and Fine Pitch QFP•h, Proceedings of the
IEEE/CPMT International Electronics Manufacturing
Technology Symposium, October 1999, pp. 94-101, 1999
3) V. Solberg:•gDeveloping Solder Stencils for CSP•h, EP&P
Electronic Packaging & Production, Vol. 41, No. 7, pp.
22-23, 2001
4) L. Barajas, E. Kamen and A. Goldstein:•gOn-Line En-
hancement of the Stencil Printing Process•h, Circuits As-
sembly, Vol. 12, No. 3, pp. 32-34, 36, 2001
5) L. Barajas, E. Kamen, A. Goldstein, M. Egerstedt and B.
Small:•gA Closed-Loop Control Algorithm for Stencil
Printing•h, Proceedings of the Surface Mount Technology
Association International Conference SMTA02, Boston,
MA, June 2002, pp. 51-58, 2002
6) L. Gopalakrishnan and K. Shirari:•gSolder Paste Deposi-
tion through High Speed Stencil Printing for a Contract
Assembly Environment•h, Journal of Electronics Manu-
facturing, Vol. 8, No. 2, pp. 89-101,1998
7) R. J. K. Wassink and M. M. F. Verguld:•gManufacturing
Techniques for Surface Mounted Assemblies•h, Electro-
chemical Publications Ltd., Bristol, England, p. 367, 1995
8) F. Breyfogle:•gImplementing Six Sigma: Smarter Solu-
tions Using Statistical Methods•h, John Wiley & Sons Inc.,
New York, USA, pp. 35-36, 160-167, 288, 1999
9) D. Montgomery:•gIntroduction to Statistical Quality Con-
trol•h, 4th ed., John Wiley & Sons Inc., New York, USA,
pp. 68,153-177, 2001
10) Chrysler Corporation, Ford motor company, and General
motors company:•gStatistical Process Control (SPC)•h,
Reference Manual, Carwin Continuous Ltd., Essex, Eng-
land, pp. 9, 21-65, 137, 1995
260 エ レ ク トロニ ク ス実 装 学 会 誌 Vol.7 No.3(2004)