Tag: SMT，Surface Mount Technology，AI

X, Y, θ Space

for square Components With Bumps or Columns

Points that lie inside the box represent combinations of x, y, and θ that meet our specs.  

· Points inside the “football” represent combinations that result in 75% feature to pad coverage or better.

· Points that lie inside the box but outside the “football” represent combinations that meet our x, y, and θ specs but yield less than 75% coverage.

· Points that lie inside the “football” but outside the box represent where we are not allowed to operate due to the x, y, and θ specs but would actually give acceptable coverages ( 75%).

· The regions described above (i.e. the portion of the “football” that lies outside the box, etc.) vary in volume depending on the component geometry (span, feature and pad diameters) and the spec. 

Fiducials and Pad Sites

 

A
fiducial is a board feature used for global and local error correction
to determine the difference between programmed coordinates and actual
locations on the board. This ensures that parts are not placed before
their locations are verified.

 

A pad site is a pad pattern on the production board that can be used in the same manner as a fiducial.

 

The most
typical types of fiducial failures are caused by improper color, size of
fiducial, and lighting values. Other factors such as the confidence
level and search area can also be trouble spots but as the programmer’s
experience level increases, these will be less likely to cause problems.

 

How many
fiducials to use on a board or circuit will depend on board quality and
the amount of time the manufacturing process can allocate to finding
fiducials. The following is a general guide as to the number of
fiducials used and the benefits of accuracy.

 

Number of Fiducials Found	Correction Possibilities
1	X and Y
2	X, Y, and Theta
3	X, Y, Theta, and Uniform Stretch
4-5	X, Y, Theta, and Independent X & Y Stretches
6-10 (max)	X, Y, Theta, Independent X & Y Stretches, and Corners not equal to 90 °

 

The total number of fiducials and pad sites that can be used for a global correction cannot exceed ten.

 

To use a combination of fiducials and pad sites for global error correction, you must assign them in the Circuit List window.

 

The total number of fiducials and pad sites that can be used for a local correction cannot exceed five.

 

To
use a combination of fiducials and pad sites for local error
correction, you must assign them in the Local Fiducials dialog box in
the Placement List window.

 

When
creating a fiducial or pad site, use the Tab key to move between the
data fields. If you use the Enter key, the fiducial placement is
attempted and error checking is performed.

 

To successfully create valid fiducial placements:

 –  Fiducials must be placed within the borders of the board.

 –  Fiducials cannot be placed directly on offsets. (Fiducials placed on circuits

   are automatically duplicated on all the offsets associated with that circuit.)

 –  Fiducials cannot be partially on a board or circuit.

 

If
a fiducial is on an offset and that offset is rotated, the fiducial
location is rotated but the fiducial is not. Only fiducials with
rotational symmetry are supported in this manner. All others will not be
found.

 

If
multiple fiducial or pad site definitions are selected when using the
Fiducial or Pad Site Copy function, all new fiducials and pad sites are
distanced from the originals by the same X and Y Offset values.

 

If
fiducials or pad sites are consistently not found by the vision system,
lower the confidence level. If the vision system finds objects other
than the fiducials or pad sites, increase the confidence level.

 

When
defining a search area, keep in mind that it should be large enough to
allow some tolerance in board handling, but not so large that additional
board features are found instead of the fiducial or pad.

 

Some recommended lighting levels for fiducials and pad sites.

 

Fiducial Type / Pad Site	Inner Ring	Outer Ring
Tinned / Tinned	80 / 80	20 / 20
Solder Mask over Bare Copper (not   recommended) / Gold	0 / 0	50 / 35
Bare Copper with Copper Bright / Bare Copper	0 / 0	35 / 35

 

 

The pad site functionality is not available for the Odd Form system at this time.

 

In
most cases, standard lighting cannot be used to image a pad site since
solder paste or flux may not allow a good contrast between the pad site
and the circuit board. Special lighting settings may need to be
installed in order to image the pad site. If Pad Site Find is the only
way to get component corrections, and lighting is the only issue,
consult your UIC Application Engineer.

 

Use
the Fiducial Lighting procedure located in the Operation Features
Module within the User’s Guide, to determine whether a pad site can be
imaged with the PEC camera. Verify contrast and the lighting level
required.

 

When to use Pad Site Find

1) When fiducials do not exist on the circuit board

2) When the pad site accurately represents a component type

3) When fiducials do not give an accurate enough correction

4) When accuracy is more important than speed

 

If
any errors occur finding pad sites, you will be taken to the Fiducial
Repair screen. In the case of failed pad site finds, manual alignment is
not recommended. For GSM1 systems, select the Reject Board button to
remove the board. For GSM2 systems, palm down the machine to manually
remove the board.

 

The need for a pad site correction is more typical of fine pitch placements such as C4 placements or fine pitch BGA’s.

 

Pad
sites are based on component definitions. To associate a pad site
definition with a component, the component must be defined in the
database. Refer to the New Component module for information on adding a
component to the database.

 

 

PEC Lighting

 

On
the GSM machine, a Pattern Error Correction (PEC) camera passes an
image to the vision system which attempts to recognize a programmed
fiducial or pad site based on parameters in the Fiducial or Pad Site
List. These parameters consist of type and size, center of fiducial
identified by its “X,Y coordinates”, and the search area identified by
“Search Area X,Y”.

 

After
the PEC camera moves to the programmed location of the fiducial, it
illuminates the Search Area using the programmed “IN/OUT” (inner
ring/outer ring) light levels. Within the search area of the image,
light intensity differences between the fiducial and the board help the
vision system detect the fiducial’s edges.

 

The
vision system is able to detect the North, South, East, and West edges
of the fiducials by relying on the differences in contrast between the
board and the fiducial color. Called vector points, triangles of red,
blue, green, and yellow are displayed in the Vision Window.

The
vision system uses six vector points per edge (N, S, E,W). In order for
the vision system to obtain 100% confidence, 24 out of 24 of the vector
points must be detected on an edge of a fiducial. The default
confidence level is 80% (19.2 rounded up to 20 vector points).

 

Since
the success of fiducial finds depends on the vision system’s ability to
discern the contrast between the board and the fiducial, some
combinations of fiducials (or object(s) to find) and their backgrounds
may call for different types of PEC cameras. Currently 2-sided and
4-sided lighting is being used and FlexLight, a new feature, will soon
be available. The 2-sided PEC camera was non-symmetrical in its lighting
pattern. It illuminated in one direction, from the North and South. The
4-sided PEC camera improved on this by illuminating in four directions,
from the North, South, East, and West. Originally both cameras used red
LED’s. When looking at solder-mask covered fiducials, the red light
would be absorbed by the solder mask (green). To overcome this problem,
green LED’s were added. The 4-sided scheme expanded the capability to
illuminate gold fiducials on white ceramic as well as fiducials on
flexible circuits.

 

FlexLight (trademark) is an enhanced PEC lighting module. It was originally
developed to address the imaging challenges associated with advanced
substrates such as ceramics and flexible circuits. Although FlexLight
was initially targeted at these markets, it can effectively image a wide
variety of substrate materials ranging from FR-4 to more exotic
materials. The chief advantages of FlexLight are: 1) Symmetric
illumination, 2) Polarization flexibilty,

3) Wavelength flexibility, 4) Ease of reconfiguration, and 5) Monolithic design.

 

A
mechanical support structure holds eight LED petals and an inner LED
ring. Each petal is a small printed circuit board containing 10 LED’s.
The petals can contain light sources of various wavelengths ranging from
blue to red. The petals and the inner ring can be exchanged in a
“plug-and-play” fashion. This allows the illumination wavelengths of
the module to be quickly and easily changed. It also facilitates ease
of service in the field. The supporting electronics allow the petals to
be configured in various series and parallel combinations to support a
wide variety of LED’s.

 

The structure supports an optional polarizing film that covers four of the eight petals as shown in the following diagram.

 

Corner Feature Enhancement for Multipattern Components

 

Multipattern
components consist of components or objects (RF shields, connectors
etc.) which cannot be described adequately as either leaded or leadless
components, but rather are defined in terms of an arrangement of
geometric features. The multipattern object is located by locating each
of the features of which it is comprised, using a single or multiple
fields of view. One such feature, which is commonly used to locate
rectangular or pseudo-rectangular objects, is the corner feature. At
present, this feature is defined simply by entering the length of each
of the two line segments, which make up the 90 degree corner (the
horizontal corner edge length and the vertical corner edge length).
With this special software, this feature definition has been extended to
allow for two more optional parameters. These parameters define
“ignore zones” at the apex of the corner, and allow the image processing
to ignore these regions of the edges when locating the corner. By this
means corners which are rounded, chamfered or poorly defined at the
apex can still be located by using segments of the corner away from the
apex, which subtend 90 degrees to each other.

 

The diagram below indicates the meaning of each of the parameters.

 

 

X2, Y2 should not exceed 25% of X1, Y1

If X2 or Y2 = 0, the standard corner find is employed

 

 

Enhanced Product Setup

 

 

A very helpful feature when programming components is Enhanced Product Setup. It consists of two parts, Enhanced Component Setup
and Enhanced Board Setup. Each process involves a live image, of the
object being taught, to be manipulated while the programmer sees the
changes as they are being made.

 

When
defining a new component, fill in as many data fields as possible while
paying special attention to the following; Component Height, PreOrient,
Number of Leads, Lighting Type, Camera Type, Default Feeder, Default
Orientation, and Reject Station.

 

Enhanced Component Setup supports, Four Spindle, C4, OFA (Oddform Assembly) and High Accuracy (UFP) Heads.

 

If
anything goes wrong with the Platform machine during this entire
process (reject station not mounted, feeder not mounted, exclusion zone,
drop bin not defined, centering fails due to invalid parameter, etc…)
recover by palming the machine down, and up again. Then push the Start
button and proceed to pick the part again.

 

If
the Platform machine was not calibrated correctly prior to using EPS,
the scale of the drawing may be incorrect and the Draw Component
function cannot be used.

 

All
changes made are immediately written back to the database scroll list
where the part was defined. Exit the Inspection screen at any time to
view the results of the changes there. Nothing is saved permanently
until the part is saved.

 

 

Common ECS Hinderances and Solutions

 

Before
the part can be picked, all the values associated with component
definition must be entered. This is necessary because these values are
all needed to inspect a component.

 

All
changes to the drawing are immediately applied to the definition
database of the component. If a mistake is made, rectify the error by
using the Undo function. No change is permanent until the component is
saved.

 

To
switch from editing the body of the drawing to any of the
leads/bumps/features, click on the leads/bumps/features. To switch back
to editing the body, click where there is no lead/bump/feature.

 

Due
to the method used for programming leads, it can be difficult to line
up all the leads over their displayed counterparts. This is because
pitches are measured from the center of the side of the component, and
when they are adjusted, leads move symmetrically out or in from/to the
center. To help the adjustment, if there is an odd number of leads,
position the single lead in the center of a side over its corresponding
displayed counterpart. If there is an even number of leads, position the
two center leads over their displayed counterparts before adjusting the
pitch.

 

To
define a C4 component it is sometimes convenient to define only one
bump initially, and add bumps when the image is displayed, wherever
necessary until the part is found. This is a good procedure because it
may be difficult to determine how bumps will image before seeing an
image of the part.

 

When
dealing with a large number of leads/bumps at once (over 50), the
drawing function will automatically move only the single lead selected,
instead of all the leads. This is done to increase the performance of
the drawing operations. If less than 50 leads/bumps are selected, they
will all be repositioned at once to give a better indication of their
final positions.

 

One
of the more difficult things to deal with is when the displayed part’s
rotation is slightly off. Make sure that the feeder pick position is
optimal to present the part accurately. Use the pick/inspect/drop-off
sequence more than once if necessary until the part is basically square
on the screen.

 

Lead
groups can cause additional problems. The drawing always assumes that
all leads are present on a side, but does not draw some of them if they
were deselected in the leadgroup screen. This can make it difficult for
pitches to be adjusted.

 

If
the component is too large to fit into a single field of view, the
vision system will take more than one image and stop at the first image
where it could find all leads/bumps/features. This might be the first
image seen, or the last. If the part is found successfully, it will be
the last. This makes editing of the components, by using the Draw
Component function, difficult. Sometimes it is more convenient in this
case to go back and forth between the Database Component Definition
screen and the Inspection screen.

 

When
viewing a component on the monitor, the image detail may require
enhancement. With the use of Vision Level Diagnostics, the operator can
increase or decrease the detail of the viewed image by raising or
lowering the current vision level. By increasing the Vision Level
Diagnostics to a level 5 setting, the operator can view the image with
the maximum amount of detail. Using a lower vision level results in a
decrease in display detail.

 

 

Specific Component Programming

 

If
a change is necessary while adding a new component to the database, do
not change the component type, exit and begin the procedure again.

 

The
Accuracy field applies only to a GSM2 (Dual Beam) machine. When the
value is set at high, this means stop the opposite beam while I place
this particular part with the other beam. Our accuracy studies indicate
there is no need to ever run the machine with this value set to high. It
adversely effects throughput and does not contribute to the accuracy of
the machine when placing standard SM devices. Ignore this field for any
other machine configuration.

 

For
parts that do require a more accurate placement it may be advantageous
to turn on preorient. This indicates to the machine that the part will
be rotated to it’s place rotation prior to being scanned through the
upward looking camera.This allows the machine to minimize the amount of
correction required after being centered and inherently contributes to a
more accurate and repeatable placement. It does however adversely
affect throughtput. Therefore, if you find you the placement accuracy
does not meet your expectation with preorient turned off, turn it on and
reevaluate the accuracy/repeatability of your placements.

 

When choosing a lighting level for BGA, C4, or C4-Pattern components, a level of +7 should only be used with side-lighting.

 

 

C4 Types

 

The
following restriction applies to programming C4 components on a machine
equipped with an AISI 3500 vision system: A maximum of 16 unique C4
components, with 20 programmed features per component, can be contained
in a product. This restriction is based on the number and type of
programmed C4 features.

 

Placement
pressure values above 350 grams are typically used for C4 applications.
If the placement head is not C4 capable, these pressures will not be
possible.

 

The
current bump process is ‘A’, selected as the default. Bump processes
B-E are reserved for future UIC vision inspection algorithms.

 

The X or Y Vector value will be ignored if the X or Y Number value equals 1.

 

The % Bumps Required for a C4 component is the percentage of bumps required to return an accurate image.

 

If
C4-Pattern is not available from the Component Types list box, you must
create a new database. This is done by using the New option under the
Database menu bar heading. If desired, existing component definitions
can then be brought into the new database using the Merge option.

 

For C4-Pattern, the value for Critical should be chosen as Yes.

 

There
should be no entry in the Min Precise Patterns, Pattern Inspection,
Location Tolerance X, Location Tolerance Y, or Relative Distance fields.

 

BGA Types (Requirements and Limitations)

 

A special version of software is needed, developed after an RFQ, for use with UPS 2.x

 

The component can only be processed in a single field of view

 

The appropriate magnification, circular lit camera (circular lit cameras take up 2 additional feeder slots

 

The vision system must be an AIS630 Lantern vision system only.

 

The % Bumps Required for a BGA component is the percentage of bumps required simply to display an image.

 

MISSING BALL DETECTION FOR BGA COMPONENTS

 

Centering
– the vision system identifies the defined features (bumps) and
determines the x, y, and theta corrections required for an accurate
placement. Bump Process A should be chosen in the component definition.

 

Inspection
– after the centering process is complete, an additional algorithm is
applied to determine if any bumps are missing. When centering and
inspection are is desired, Bump Process E should be chosen in the
component definition.

 

This
software inspects BGAs for missing balls using a two step approach.
First the regular ball find algorithm is executed and five candidates
are selected as potential missing ball sites. The selection is based on
either the failure to locate a ball at an expected site, or a low
correlation, or ball recognition score. Then an intelligent pattern
recognition algorithm is trained on sites which are known to contain
good ball images, and the trained algorithm is used to classify the
suspect sites and verify the presence/absence of a solder ball. Various
graphic overlays are used during the execution of the algorithm:

 

·   It
will be necessary to use circular lighting for bump imaging in order to
realize optimum reliability. This is because the image quality of balls
with the standard lighting is poor.

·   This
algorithm uses a training method based on balls which are found. If
the image quality is such that noise can be incorrectly labeled as a
ball, it is possible to mis-train the algorithm and fail to correctly
identify missing balls.

·   Only components which fit into a single field of view can be processed.

·   In
order to switch on missing ball inspection the customer must select
“processing type E” in the product editor (the default is A). This
processing type flag is provided to allow for customer defined image
processing and in general is not used. It is expected that using this
flag will have no impact on the overall functionality of the machine,
since processing types B-D are still available for customer specific
tuning.

·   This will be a special vision release to support the missing ball inspection.

· The five missing ball candidates are labeled by blue crosses with blue boxes.

· The trained existing balls around the missing ball candidates are labeled by blue crosses only

· The recognized missing ball is labeled by a small red cross on the center of the candidate label

 

If
the colored graphics are an annoyance, you can change the Vision
Diagnostic Level. The value is probably set at 4 or 5. The range is
between 0-5. The lower the value the faster the machine.

 

BGA Type	1.4x UPS Pick and Place Capable	2.x UPS Pick and Place Capable	Special Camera Requirements for inspection	Missing Ball Inspection Capability
CBGA (ceramic)	Yes	Yes	None	Need Analysis
CCGA, White (ceramic-column)	Yes	Yes	None	No
CCGA, Dark (ceramic-column)	Yes	Yes	None	No
uBGA	Yes	Yes	2.6-3.0 Mil/Pixel Camera	Need Analysis
PBGA (plastic)	Yes	Yes	None	Yes
TBGA (taped)	Yes	Yes	Circular Lighting	No

 

Camera	Maximum Single Field of View Size	Minimum Pitch	Minimum Ball Diameter
Super High Mag (0.5 mil/pixel)	4mm (0.160”)	0.125mm (0.005”)	0.075mm (0.003”)
High Mag (1.0 mil/pixel)	10mm (0.39”)	0.25mm (0.010”)	0.125mm (0.005”)
Medium Mag (2.6 mil/pixel)	20.8mm (0.8”)	0.5mm (0.20”)	0.25mm (0.010”)
Medium Mag (3.0 mil/pixel)	24mm (0.8”)	0.5mm (0.20”)	0.25mm (0.010”)
Standard Mag (4.0 mil/pixel)	32mm (1.25”)	0.8mm (0.031”)	0.4mm (0.016”)

 

Leaded Components

 

Lead
information must be programmed symmetrically. Information entered for
Sides 1 and 2 of the component is input to Sides 3 and 4, respectively.
The data can then be edited. To accommodate nonsymmetrical components or
components with different lead lengths or pitches, the Lead Groups option may be used.

 

Lead
groups can cause additional problems. The drawing always assumes that
all leads are present on a side, but does not draw some of them if they
were deselected in the leadgroup screen. This can make it difficult for
pitches to be adjusted.

 

If
0.0 (zero) is entered in any of the following Tolerance data fields,
that inspection is bypassed; Lead Tolerance From Body, Lead Tolerance
Across Body, Lead Spacing Tolerance, Lead Length Positive Tolerance,
Lead Length Negative Tolerance, Coplanarity Tolerance, and Colinearity
Tolerance.

 

If
an excessive number of components are rejected, check the component
definition relative to vendor specification sheet for the component.
Also, use ECS (Enhanced Component Setup) to adjust inspection parameters
(geometry, lighting, etc…).

 

Lead Groups

 

The
Lead Groups window is not used to toggle leads off for the purpose of
increasing the speed of vision inspection (SMC components only). This
will only result in a rejected component. All components must be defined
as they physically exist. Non-symmetrical leads can be accommodated by
defining the component as a Special-Leaded Component.

 

Lead
1 in the component database is not necessarily the component’s
electrical pin 1. It is only the first lead in the lower left corner of
the component when the component is in the 0 ° orientation. We define/assign leads as beginning with lead one in the
lower left hand corner and count up as we define the part in a
counter-clockwise fashion.

 

If you select the Remove All Leads option, all component leads are toggled off and considered to be phantom leads. If a lead was already toggled off when the Remove All Leads option was selected, it would remain off.

 

If
you select the Enable All Leads option, all component leads are toggled
on and are inspected by the vision system. If a lead was already
toggled on when Enable All Leads option was selected, it would remain
on.

 

Special Leaded Components

 

Program the component as if all leads on the same side are identical and symmetrical with each other.

When defining a component with different pitches, find the greatest common denominator and enter that as the pitch.

 

The machine memory supports a maximum of 15 lead groups per component.

 

When
all lead information is entered, select the Lead Groups option. Select
the leads you want to be ignored by the vision system. The leads are now phantomed with just a broken line to indicate their existence.

 

Example:

Let’s
use the 23pin SMT connector as an example… There are physically 12
leads on one side of the device and 11 on the opposite side. It would be
a reasonable approach to define both sides as having 23 leads with a
pitch of 1mm, and turning off every other lead in a manner where the
database matches the physical description of the part. However, by
turning off every other lead this creates 23 lead groups, and this is
why the machine hangs up!

 

We
define/assign leads as beginning with lead one in the lower left hand
corner and count up as we define the part in a counter-clockwise
fashion. For example, for a 14 pin SOIC, lead # 1 is in the lower left
corner and lead # 14 is in the upper left corner (assuming the part is
defined with the leads facing north and south). There are two lead
groups when we define a 14 pin SOIC. Lead group 1 is defined as leads
1-7 and lead group 2 is defined as leads 8-14. However, if you turn off
lead 4 there are now 3 lead groups (lead group 1 = leads 1-3, lead
group 2 = leads 5-7, and lead group 3 = leads 8-14). Notice lead 4 is not included.

 

By
turning off every other lead you are creating 23 lead groups. We only
have enough RAM on the machine controller to support a maximum of 15
lead groups. However, the number of lead groups is dynamic and can be
limited (reduced) by the number of components, component placements, and
process complexity. Therefore, the number of supported lead groups can
be £ 15, depending on the product complexity.

 

Program
the part as it is… Assuming the part is coming in tape and the12
leads are facing 6 O’clock and the 11 leads are facing 12 O’clock, let’s
define the part as having 12 leads on side 1 at a pitch of 2mm and side
3 as having 11 leads at a pitch of 2mm.

 

 

Component Terminology

 

Acronym    Name

 

BGA      – Ball Grid Array

uBGA      – micro Ball Grid Array

CBGA      – Column Ball Grid Array

C4 or Flip Chip  – Controlled Collapse Chip Connection

COB      – Chip On Board

CSP      – Chip Scale Package

DCA      – Direct Chip Attach

FPT      – Fine Pitch Technology (20 to 40 mil pitch)

ILB      – Inner Lead Bonding

MCM      – Multi Chip Module

MELF      – Metalized ELectrode Face bonded

MSP      – Mini Square Pack

OLB      – Outer Lead Bonding

OMPAC    – Over Molded Plastic pad Array Carrier

PBGA      – Plastic Ball Grid Array

PLCC      – Plastic Leaded Chip Carrier

PQFP      – Plastic Quad Flat Package

QFP      – Quad Flat Package

SOD      – Small Outline Device

SOIC      – Small Outline Integrated Circuit

SOJ      – Small Outline J lead

SOT      – Small Outline Transistor

SQFP      – Shrink Quad Flat Package; QFP with a lead pitch of .016” or less

TAB      – Tape Automated Bonding

TSOP      – Thin Small Outline Package

UFPT      – Ulta Fine Pitch Technology (<20 mil pitch)

V-QFP      – Very Small Quad Flat Package

V-SOP      – Very Small Outline Package

 

 

Industry Terms

 

CER-QUAD    – Digital Equipment Component

C-QUAD    – Northern Telecom Package

Tape Pak    – Trade Mark/National Semiconductor

V-PAK    – Vertical Package (Texas Instruments – memory package)

1