|
PCB Inspection
Outlook for 2005
by Tom Lecklider, Senior Technical Editor
Regardless of their labor rate, inspectors alone can’t determine area-array
solder-joint quality.
The use of smaller components, lead-free soldering, and greater PCB component density requires higher performance
from automated PCB inspection systems. In separate ways, each of these factors
has resulted in the need for higher resolution and accuracy from the inspection
process.
Lead-free soldering affects X-ray inspection because of the altered mass
absorption of the new solder alloys. Measurement of solder-paste volume by 3-D
techniques now must account for slightly different fillet curvatures and a
rougher surface finish. Higher temperatures are used, which means that board
warping and component materials may become issues. In addition, new pad shapes
have been proposed to assist automated determination of soldered-joint quality.
Smaller components such as 0201-size resistors really are minute. You need a
10-µm pixel size to get the required magnification for accurate inspection. At
least, this is the highest magnification one manufacturer offers via a zoom
lens. Today, there’s nothing smaller than a 0201, although one website mentioned
the next step, 01005, which would be 4× smaller than a 0201.
Optical zoom or a higher pixel-count target must be used to provide high
resolution. Both approaches are being pursued.
One way to achieve very high resolution, as described by George Blackwell,
director of product marketing at Cognex, is by using line-scan cameras that also
can reduce the peak power level needed to position the PCBs. Full-frame cameras
require the board-positioning system to stop the PCB motion and allow time for
settling before the board’s image can be recorded. Stopping and starting takes
much more power than a continuously moving conveyor such as used in a line-scan
application.
And, both circuit and component densities are increasing, affecting throughput.
There are many optical or X-ray images to capture and a huge number of
computations to be performed on large, densely populated boards. The speed with
which the actions can be accomplished largely determines the throughput.
Increased circuit density results from the trend to use more area-array packages
in each assembly. Boris Mathiszik, vice president of sales at phoenix|x-ray
Systems + Services, commented, “In the past, typical assemblies may have
contained one to five area-array packages, but now designs frequently have 30 or
40. In addition, assemblies are becoming more complex. Double-sided reflow is
more prevalent, often with mirrored BGAs.”
For boards populated on both sides, Mr. Mathiszik said that conventional 2-D
X-ray inspection with top-down viewing might not be the best approach. Instead,
at high magnification, oblique images of components are less likely to be
affected by parts on the other side of the board.
Smaller components, lead-free soldering, and greater component density are the
three main themes affecting PCB inspection, but they’re not the only ones. Cost
always restricts the technologies and performance available for an application.
And, manufacturers continue to innovate, resulting in a wide range of inspection
products with varying capabilities and benefits. It’s not surprising then that
few inspection processes have become completely standardized.
Implications for the Industry
With these trends forming the background to the inspection industry, we asked
managers to describe their outlooks for 2005.
X-Ray
“We see nothing but increased opportunities for X-ray inspection, both in the
product assembly and the nondestructive testing (NDT) markets,” said Lance
Scott, president of FEINFOCUS/COMET North America. “All of the industry drivers
are pointing to increased inspection needs. For instance, in electronics
assembly,” he continued, “the move to smaller, denser assemblies has led to a
greater use of area-array devices, which in turn, is dictating extensive use of
X-ray inspection to identify faulty interconnects.
“Software controls have eliminated the need for an X-ray physicist to operate
the tool and put high-resolution inspection in the hands of the manufacturing
technician. For example, in the case of wafer-bump inspection,” he explained,
“void detection requires analysis of multiple bumps. This means searching for
slight variations in density, which often is outside the scope of manual
inspection.”
According to Mr. Scott, wafer bumping is recognized as the largest challenge in
semiconductor packaging because of the voids that often form beneath the bump
surface. High-resolution X-ray inspection is the only technology capable of
detecting and measuring these voids.
John Spofford, the vice president and general manager of Agilent Technologies’
Manufacturing Test Business, added that millions of units potentially could be
affected by undetected failures in hidden joints. “As liability and risk move
further back in the value chain, X-ray imaging will become a more important
manufacturing tool. It’s fixtureless and ready for lead-free processes,” he
said.
Optical
Optical inspection deals with visible connections, and as components and their
joints become smaller, generating the best possible image is more difficult.
According to Alan L. Paris, the industrial microscopy marketing director at
Leica Microsystems, “Advances in optical engineering software, manufacturing
tools, and coating technology have improved optical performance in even the most
basic inspection microscope. The price/performance ratio of today’s optical
microscopes is exceptional compared with levels only a decade ago.”
|
It used to be that the A
in AOI was enough to justify the investment.
George Blackwell
Cognex |
Software has been key to many recently introduced Leica tools such as 3-D
analysis that can accurately measure solid-surface topography, surface areas,
and volumes. “Only because of today’s powerful PCs and digital imaging
technology can the processing capacity and speed required for 3-D analysis move
from research to the production floor,” Mr. Paris added.
Yet other innovations have extended what were physical limits to performance in
the past. Here software has addressed such fundamental problems as small depth
of focus through the application of multifocus algorithms for optical image
stacking.
Also operating in the optical imaging space is JAI PULNiX. The company develops
and manufactures smart cameras that include neural networks. Without requiring
complex programming, the cameras take only microseconds to make image-related
decisions that mimic human intelligence.
“Machine vision has a wide variety of applications spread across a range of
fields,” explained Toshi Hori, the company’s CTO. “In general, the industry is
moving to higher resolutions and speeds, lower costs, and miniaturization of
components. Applications that stand to substantially benefit from these trends
include semiconductors, electronics, assembly and packaging, medicine,
pharmaceuticals, military, and aerospace.
“More and more smart cameras will likely enter the market. New communications
technologies such as gigabit Ethernet will reduce the need for large computers
on factory floors,” he continued, “and make possible more network-based
automation. We believe that most current trends—such as faster processors,
smaller component size, and higher resolution—will continue in the foreseeable
future.”
Image Processing
Smart cameras are well suited for many applications and offer a very compact
solution. On the other hand, several vendors agree that faster PCs have been key
to today’s improved pattern-recognition performance. And, for either
conventional or smart cameras, lighting plays a critical role. Unless the item
to be inspected is correctly lit, image-processing algorithms will not provide
consistent performance.
National Instruments’ Kyle Voosen, a vision/motion product engineer, explained
how PCs have become faster. “The new PCI Express technology supports data
transfer speeds up to 680 MB/s and higher speeds in the future. This provides a
low-cost alternative to traditional approaches to high-speed inspection systems
that generally were expensive.”
Modern pattern-matching algorithms require fast PCs because the algorithms are
complex. However, he elaborated, in addition to matching objects, newer software
also classifies them. This is very useful in large-mix manufacturing
applications where you can quickly sort out parts regardless of the many
possible combinations that may exist.
According to Francois Bertrand, vice president of sales and marketing at Matrox
Imaging, “The fundamental science behind a number of techniques, such as
geometric pattern recognition (GPR) and feature extraction, has been known for
quite a while. However, the recent increase in PC processing power has made it
practical to use compute-intensive algorithms in industrial inspection
applications. The Matrox Geometric Model Finder tool,” he continued, “locates a
model or multiple models by using geometric features instead of pixel-to-pixel
correlation.”
Originally, normalized grayscale correlation (NGC) was the dominant
pattern-matching method. As more complex requirements developed, companies
turned to GPR. For example, the appearance of chemically and mechanically
polished (CMP) wafers exhibits severe contrast changes, making NGC-based tools
unsuitable. GPR tools can have a greater tolerance to lighting variations
including specular reflection as well as module occlusion and variations in
scale and angle.
Truly uniform lighting, noted Mr. Bertrand, was very difficult to provide, and
advances in analysis algorithms have helped to relax previously stringent
lighting requirements. For example, in addition to coping with shadows and
reflections from objects in the vicinity of the inspection camera, GPR
algorithms can identify features such as solder paste that become partially
transparent under strong lighting conditions.
To give a good idea of the capabilities of modern feature-location software, Mr.
Blackwell explained the different uses addressed by a few of the Cognex
algorithms. PatMax® is the company’s basic geometric pattern recognition tool.
PatMax scene alignment (SA) provides partial-match alignment that eliminates the
need for the entire pattern to be present in an image. PatFlexTM extends PatMax
image processing to include the severe perspective changes and spatial
distortion found in curved or warped surfaces such as flexible circuits.
“When an acquired image is unavailable or unsuitable for training,” he
continued, “Synthetic PatMax can be used to generate reference models. This tool
leverages the increased availability of electronic data and parts databases to
automatically generate alignment targets from CAD data. Users can import CAD
data, extract edges, or sketch freehand to create reference models without
relying on image data.”
With regard to threats and opportunities for inspection companies in 2005, Mr.
Blackwell cited the continuing shift of electronics manufacturing to China as a
major concern. Because of the country’s very low cost of labor, the decision to
use AOI vs. manual inspection is based on criteria other than elimination of
labor. In addition to throughput, these factors include the reliability,
accuracy, robustness, and consistency of the inspection result.
“It used to be that the A in AOI was enough to justify the investment,” he
continued. “As electronics manufacturing continues to move to China, that’s no
longer the case. It could take a while to saturate China’s labor market.”
Continuing the optical-inspection and image-analysis themes, George Ayoub,
president and CEO of Machine Vision Products, said, “AOI is mainly affected by
advances in camera/electro-optical technologies and the computing power
available to run robust algorithms in real time. Recent advances include faster
large-format cameras, color-stable cameras, brighter and longer-life LEDs, and
faster computers.” The company has introduced real-time color-processing
algorithms and innovations that support concurrent 2-D/3-D solder-paste
inspection.
“Although the future of AOI includes assembled PCB inspection,” Mr. Ayoub
explained, “process control will be derived from AOI information to form a
closed-loop feedback manufacturing solution. Doing so allows the entire process
to be monitored and maximum efficiency achieved quickly.”
Process Inspection
View Engineering produces off-line optical measuring equipment used to calibrate
screen printers, component placement systems, and glue-dispensing machines.
Determining the deviations from nominal positions for all these operations helps
eliminate a major source of surface-mount manufacturing errors.
Steven Rosso, the company’s president, explained the benefits of a recently
developed software product called Elements: “The software is dedicated to SMT
electronics assembly applications such as stencils, PCBs, component-placement
verification, solder-paste registration, and volume measurements. Using the CAD
file of each application, an operator can identify the measurements of interest
and assign tolerances for acceptable results before beginning the measurements.
“The program determines the optimal sequence of inspection for reduced cycle
time and generates a comprehensive report including a histogram and scatter-plot
analysis that clearly identifies the process trends. Because Elements eliminates
programming,” he continued, “it allows quick changeover of parts on an SMT line,
which typically is being used for high-mix and low-volume manufacturing.”
Mr. Rosso also discussed the relationship among inspection, outsourcing, and
contract manufacture (EMS). As manufacturing continues to become concentrated in
EMS companies, he believes the standardization of inspection should increase.
Regardless of which EMS organization performs the work, the OEMs contracting for
it will influence how their parts are inspected and measured.
Outsourcing to overseas companies involves a different competitive dynamic than
what American inspection equipment vendors are used to, he explained. This could
mean modification of equipment to meet local safety regulations, for example, or
the transfer of technology to the local office to ensure adequate local support.
Summary
For most electronics manufacturing inspection applications, there are a number
of proven solutions. Generally, that’s true, but as little as a year ago, it
wasn’t necessarily the case. Timothy Ludy, product marketing manager at Data
Translation, provided background details:
“A 25% increase in PC CPU speed per year has allowed our Vision Foundry product
to take huge advances in the type of algorithms it uses. A 3.4-GHz Pentium 4
just about keeps pace with the faster applications,” he said, “and clearly one
year ago, projects such as a high-speed, RGB color-print inspection system
probably would have required DSP-assisted hardware.”
Of course, fast PCs are only one part, although an important part, of an
inspection solution. Obtaining an optical or X-ray image and optimally applying
the most relevant algorithms involve many other disciplines from optical design
and material physics to complex mathematical modeling. And, in addition to
technology, as View Engineering’s Mr. Rosso pointed out, dealing with overseas
contract manufacturers requires carefully tuned business skills.
Because his company provides both optical and X-ray imaging solutions, YESTech’s
CEO Don Miller has a slightly different view of the inspection industry. “The
goal is to go beyond just improving defect detection to provide meaningful
process-control information that supports yield improvement in a proactive
manner. The key to this goal,” he concluded, “is to offer an efficient system at
a price easily justifiable by our customers.”
FOR MORE INFORMATION
on Agilent Technologies
X-Ray inspection systems
www.rsleads.com/412ee-291
on phoenix|x-ray inspection systems
www.rsleads.com/412ee-292
on FEINFOCUS x-ray
inspection systems
www.rsleads.com/412ee-293
on Leica Microsystems
optical stereomicroscopes
www.rsleads.com/412ee-294
on JAI PULNiX optical
imaging products
www.rsleads.com/412ee-295
on National Instruments
machine vision products
www.rsleads.com/412ee-296
on Matrox imaging products
www.rsleads.com/412ee-297
on Cognex optical imaging products
www.rsleads.com/412ee-298
on Machine Vision Products equipment
www.rsleads.com/412ee-299
on View Engineering products
www.rsleads.com/412ee-300
on Data Translation Vision Foundry
www.rsleads.com/412ee-301
on YESTech inspection equipment
www.rsleads.com/412ee-302
|