Presented at the ECTC 2009 in San Diego, Dan Evan and Zeger Bok present material and process consideration, 2 high accuracy case studies, the systems required to enable these processes and techniques involved in the application. For complete download visit http://palomartechnologies.com/Applications/OptoelectronicPackaging.aspx
Micron Level Placement Accuracy for Optoelectronic Components
1.
2. • Terms
• Material & Process Considerations
• High Accuracy Case Studies
• Systems
• Case Study 1: Multi-channel
communication
• Case Study 2: Wafer-Scale Eutectic Die to
Wafer / P-Side Down Laser Attachment
• Palomar’s Complete Solution
3. The Solution: Model 6500
How Accurate is it?
75µ = Width of a Human Hair
SMT – 40um Die – 25um 5um
7µ 2µ 1.5µ
Red Blood Cell Bacteria Cell 6500 Placement
Accuracy
4. Term Definitions
• High Placement Accuracy Geometry
Z
Translation dX, dY
Rotation Tz
Levelness Tx, Ty
Y
X
Bondline dZ
• Interconnect Method (In-situ / Offline)
Eutectic
Adhesive
• Shift (Pre Cure – Post Cure)
5. Flatness
Cleanliness
Material Considerations
Symmetry
Image Quality
High Accuracy Attachment is run in production but
requires close attention to detail and control of
materials
7. Part Geometry for 1 um alignment
• 5um particle below one end with push edge
outside of 1um tolerance
Exaggerated
1.3 um
Drawing
400 um
5 um
100 um Human
Hair
8. • Part Geometry for 1 um alignment
5um particle with push edge outside of 1um
tolerance
Particle
between die
9. Process Considerations
1) Image Recognition
• Fiducial selection
Two Point Refs / Look Down Substrate
2) Pick Strategy
• Waffle, Gel, Double
3) Place Strategy
Two Point Refs / Look Up Chip
• Fixed Pattern vs. Pick die presentation format, collet, and
Relative process are critical. Lookup refs
remove pick error
Place die based on one substrate fiducial
4) Curing or based on previously placed die.
• Profile Pre vs. Post Cure Accuracy
11. 11
Model 6500 – Precision Eutectic Die Bonder
Combined speed, accuracy, and compact footprint of
the Model 6500 provide for high yielded throughput and
optimized cost of ownership in a high accuracy
assembly system.
o Post eutectic placement attachment accuracy
o ± 1.5 micron (3 sigma)
o ± 0.1 degree post-placement rotation
o Cycle time <7 seconds
o Programmable pulse-heating
o (500 C @ 100 C/sec ramp, +/- 2 C)
o Waffle Pack, Gel Pak
o <1 sq meter footprint
o Work area 300 mm x 150 mm
o 10-100 grams force, ± 1 gram (3 sigma)
o 6 tool turret – “On the fly tool change”
o Integrated data management and analysis
16. 16
High Accuracy Attach Cases
Application Attach Accuracy
Multi-channel communication modules (Active Epoxy ±3-5 um
Optical Cables)
P-Side Down Laser (Singulated & Wafer Scale) Eutectic ±1.5-3 um
LED Laser Print Head Epoxy ±3-5 um
Laser Marker Head (VCSEL Laser) Epoxy ±2-3 um
Lithography/Screen Interconnect Epoxy ±5 um
Thru Via Die Stacking Misc ±3-5 um
3D MEMS Stacking Epoxy ±5-10 um
17. Case 1:
Multi-Channel Communication, Active Optical Cables
Finisar
Luxtera
Zarlink
18. Luxtera Blazar LUX5010
Multirate 4x10G Optical Active Cable
Features Benefits
•No optical interoperability issues
•Easy maintenance - no need to clean
Optical Active Cable:
optics
Closed Optical System
•Eliminates costly fiber connectors
Multi-Rate (1-10Gbps) One solution for multiple applications
Single Laser for 4-Channels Better reliability
•Standards compliant electrical interface
(IB, FC, 10GbE)
SFP+ Electrical Interface
•Lower link costs: 8 CDRs eliminated
(4 per end)
•Lower link costs and better performance
•Supports reach up to 300 meters for
Single-Mode Optics
QDR (4x10Gbps)
•Eliminates need for EDC Components
Hot-Pluggable •Field replaceable
QSFP Form-factor •Provides more bandwidth density
•Lower cooling costs and
Low Power (0.5W/10G)
simplified thermal management
19. Active Optical Cable
Tyco PARALIGHT Active
Optical Cable
Assemblies provide E/O
Finisar brings serial active
and O/E conversion built
optical cable to 10G
into connector
applications
20. Active Optical Cable Channel
Electrical
Input
Light Channel
Coupler Coupler Detector
Source Fiber
Electrical
Output
Active Optical Cable has an Electrical Input and Output
All optical E-O, O-O, O-E elements are within cable
21. 21
AOC Specifications (um)
5 4 3 2 1 VCSEL Array N
VCSELs
Gap X Linearity Line
Die Gap Error in X ± 5
Die Linearity Error in Y ± 3
25. P-Side Laser Attach
P-Side Up example so stripe is visible.
Y Edge Alignment
+/- 3.0 um
+/- 0.1 Deg
X Stripe Alignment
+/- 3.0 um
W
L
InP Laser Diode Eutectically Attached (P-Side Down)
28. 28
Model 6500 WSP
Precision Eutectic Die Bonder
Combined speed, accuracy, and compact footprint of the Model
6500 provide for high yielded throughput and optimized cost of
ownership in a high accuracy assembly system.
o Post eutectic placement attachment
accuracy
o ± 3.0 micron (3 sigma)
o ± 0.1 degree post-placement rotation
o Cycle time 30 seconds
o Programmable pulse-heating Pick Tool
(500 C @ 65 C/sec ramp, +/- 2 C)
o Waffle Pack, Gel Pak
o Wafer Stage – Heated background temp
o <1 sq meter footprint
o Work area 300 mm x 150 mm
o 10-100 grams force, ± 1 gram (3 sigma)
o 1 Pulse heated tool optimized for Laser
o Integrated data management and
analysis
29. Wafer Scale Packaging Eutectic Die Bonder
Equipment Layout
Pulse Heated Pick Tool
with active cooling
Gel pack load of Lasers
Wafer Stage - Heated
Lookup camera
30. 30
Wafer Level Laser Diode Attach
Load Materials
Wafer Sub
Lasers
Pick Laser Lookup Ref
Pulse Reflow Place & Hold
Release Laser Repeat N..
Unload Materials
Wafer/Lasers
31. 31
Eutectic Process: Pulsed Heat Profiles
Pulsed heat
stage
• Computer
controlled
Programmable Temp up to
‘Point and click’
profiling 500 C
Parts at high
Temp
Temperature
accuracy +/-
for a limited
20C
time
Fast ramp
(up to 65
C/s). No
overshoot
32. XY PLACEMENT
(P-Side Measure)
3 3
2 2
1 1
0 d dX
0
X dY
1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15
-1 -1
-2 -2
-3 -3
P-Side accuracy measurements using a glass wafer
Time consuming so correlated to N-Side measurements
Instrument: Nikon VRM 3200
35. High Accuracy Attach Cases
Application Attach Accuracy
Multi-channel communication modules (Active Epoxy ±3-5 um
Optical Cables)
P-Side Down Laser (Singulated & Wafer Scale) Eutectic ±1.5-3 um
LED Laser Print Head Epoxy ±3-5 um
Laser Marker Head (VCSEL Laser) Epoxy ±2-3 um
Lithography/Screen Interconnect Epoxy ±5 um
Thru Via Die Stacking Misc ±3-5 um
3D MEMS Stacking Epoxy ±5-10 um
35
36. Palomar’s Complete Solution
Wire Bonders Die Bonders
Integrated Assembly Lines Custom Handling Systems Process Development
37. Contact & Resources
• Present your packaging challenge and we will propose a
solution at no cost
• Download “Automated Eutectic Die Attach”, by Palomar
Sr. Applications Engineer Zeger Bok
• Visit Palomar’s Blogs: “Interconnection” and “Wire
Bonders’ Speak”
• Download Palomar System Data Sheets
• Contact us directly
38. Micron Level Placement Accuracy Case Studies for Optoelectronic Products
Daniel D. Evans, Jr., Zeger Bok
Palomar Technologies, Inc.
2728 Loker Avenue West
Carlsbad, CA 92010
Phone: (800) 854-3467 E-mail: info@bonders.com
Abstract
Applications requiring ultra high placement accuracies of
1µm to 3µm are resurfacing in several optoelectronic
applications such as Arrayed Laser Print Head assemblies, P-
Side Down Laser Attachment applications, and Multi-
Channel Optical Communication products. An overview of
the technologies, placement accuracies, and attachment
methods is presented for two cases. With placement
accuracies for surface mount machines typically around
40µm, 10µm for die attach machines, and 1µm for ultra high
accuracy placement machines, this paper will cover the
differences in measurement, material, and process controls Figure 1 - Geometric Six Degrees of Freedom
that are required to successfully achieve ultra high placement
accuracies of 3µm. Interconnects are comprised primarily of two methods for
optoelectronic assemblies: adhesive (epoxy) or metallurgical
(eutectic solder).
High Accuracy Die Attach Requirements and Challenges
The application breakout given in Table 1 is useful to These attachment options can be either in-situ (serial) or
explore general application, attachment, and accuracy batch (parallel). In-situ attachment is completed during the
requirements for ultra high accuracy die attach. The main placement operation for each component and will have lower
breakout is by application product or technology. For the throughput since the attachment time for each component is
purposes of this paper, each of the applications explored is for added to the pick and place time. Batch attachment is
specific attachment technologies (epoxy or eutectic) and completed after all components are placed so the actual
general ranges of required placement accuracy. attachment can be completed as a parallel process. Batch
attachment methods typically have higher throughput
Table 1 - High Accuracy Application Overview compared to in-situ.
CASE Attach Accurac Another consideration for the attachment method is its
effect on placement accuracy. To understand the effect on
VCSEL Array Epoxy ±3-5µm batch attachment methods, it is important to measure the pre-
cure and post-cure accuracy of the components. In-situ
P-Side Down Laser Eutectic ±3-5µm attachment methods can have higher placement accuracy,
especially if the design does not include self-centering.
LED Laser Print Head Epoxy ±2-5µm
Material and Process Considerations
Lithography/Screen Interconnect Epoxy ±3-5µm When approaching micron level placement accuracies,
there are several important factors that need careful
Thru Via Die Stacking Misc. ±3-5µm management and control:
3D MEMS Stacking Epoxy ±5-10µm • Substrate flatness, cleanliness, and fiducial clarity (in
particular in case of edge alignment)
Terms and Definitions • Die flatness, cleanliness, and fiducial clarity
Pick and place is composed of geometric accuracy and • Attach material uniformity, shrinkage, symmetric
interconnect method. To completely define placement application, and curing stability
accuracy would require specifying all six (6) degrees of Measurement Considerations
freedom, as shown in Figure 1. Most pick and place accuracy
applications specify Z as a bond line and placement accuracy Measuring the actual placement accuracy down to 1-3µm
as X error, Y error, and Theta-Z error. For the purposes of requires both well characterized equipment and measurement
this paper, the specific requirements will be listed for each of methods [1]. Even if the equipment can support the
the cases studied. resolution, the parts usually have imperfections beyond the
39. targeted accuracy. Measurement methods must include ways One construction of a complete cable has an array of
to handle these imperfections. VCSEL transmitters which require high accuracy placement
The remainder of this paper will explore two cases for to allow better fiber optical coupling to the VCSEL lasers.
high accuracy placement and attach. These assemblies generally require photodiodes in the same
package as well but these are easier to optically couple and
require less placement accuracy.
Case 1: Multi-channel Communication Products The remainder of this case study will share geometry,
(VCSELs) accuracy, and attachment requirements of the VCSEL arrays.
Several new products related to optical communications
such as Active Optical Cables use multiple channels of
The VCSELs used in this study are 300µm square by
transmit and receive pairs to produce the high bandwidth
200µm thick and are presented in 2x2 Gelpacks. The
communication required for high performance computer
substrate can be any material but more even materials will
connections or digital audio/video connections. The basic
produce more consistent results. The material in this study
concept of an active optical cable link as shown in figure 2 is
was a flexible circuit mounted to a PWB backing and
an optical cable that contains electrical input and output. The
presented in a common carrier in groups of 10 per carrier.
input conversion from electrical to optical (E-O) and the
The VCSELs are placed into 84-1LMIT1 epoxy which is
output conversion from optical back to electrical (O-E) is
deposited just prior to the pick and place process.
included as an integral part of the cable so that the end user
The arrangement of the VCSELs is shown in figure 4.
sees only an electrical cable without the problems associated
VCSELs 5 through 1 are placed left to right according to a
to optical cable connections [2]. The various optical to
specified pitch. VCSEL number 5 is the master VCSEL and
optical (O-O) interfaces as well as the actual fiber are also
all others are placed relative to it. The required accuracy for
included in the cable. A complete cable as shown in figure 3
each VCSEL is ±5µm in X and ±3µm in Y from the target
would include multiple channels for transmit and multiple
location.
channels for receive.
Figure 4 – VCSEL Array, Gap X ±5µm, Gap Y ±3µm
Placement results are evaluated pre-cured (wet) and post-
cured to verify that shifting of the epoxy during cure does not
adversely affect the results.
Pre-cure results are shown in figures 5 and 6. Statistical
Figure 2 – Active Optical Cable Schematic of one summaries of pre-cure results are given in Table 2. Pre-cure
Transmission Path (Cables have Multiple Tx/Rx Pairs) results are within specification for both X and Y.
Figure 5 – Pre-cured (Wet) X Error Results
Figure 3 – Active Optical Cable Example looks like a USB
Cable with Electrical Input and Electrical Output (Courtesy of Figure 6 – Pre-cured (Wet) Y Error Results
Finisar)
40. Table 2 – Pre-cure (Wet) Statistics case study will focus on pick and place and attachment
X Y capability. Substrate time at temperature and its effect on
Average 0.1 0.0 solder reflow and solder aging are explored as well.
Maximum 3.9 2.5 Measurement methods for P-Side down laser attachment
Minimum -3.6 -2.7 provide some challenges during process development and will
Range 7.5 5.2 be covered as well.
3StDev 4.5 2.6
The placement requirements are based on positioning the
Post-cured results are shown in figures 7 and 8. Statistical laser for optimal coupling of the laser optical output to
summaries of post-cure results are given in Table 3. Post- coupling optics as shown in figure 9. The gray section
cure results are also within specification for both X and Y. represents the wafer substrate area upon which the laser will
be solder attached. The wafer substrate has fiducials which
determine alignment targets and gold pads plated with 80/20
AuSn eutectic solder. The laser is 300µm long by 250µm
wide by 125µm thick with backside (P-Side) plated gold pads
for solder attach. The P-Side of the laser provides alignment
features that are a combination of fiducials and the alignment
point edge. Both are referenced during the alignment phase
using an upward looking camera.
Figure 7 – Post-cured X Error Results
Figure 8 – Post-cured Y Error Results Figure 9 – P-Side Down Laser Alignment to Substrate
Requirements
Table 3 – Post-cured Statistics
The alignment specifications are defined in figure 9 and
X Y
revolve around the P-Side intersection of the laser stripe
Average 0.5 0.0
location in the X direction and the laser output edge to
Maximum 4.7 2.4
substrate datum in the Y direction. Both X and Y directions
Minimum -2.2 -2.6
have a ±3µm alignment tolerance. The angular alignment of
Range 6.9 5.0 the laser edge is also defined as a ±0.1degree tolerance in
3StDev 4.2 2.5 Theta-Z.
Case 2: Wafer Scale – Eutectic Die to Wafer P-Side Down
Laser Attachment
P-Side down laser attachment using eutectic solder is a
mature process and has been used extensively for long haul
and metro distance laser transmit modules [3]. The P-Side
laser attach to heat spreader has traditionally been done using
singulated heat spreaders with a pulsed heat in-situ reflow
stage. The heat profile recipe is critical for proper solder
phase in the final bond.
Recent work in P-Side down laser attachment has
expanded the technology to Wafer Scale – P-Side Down
Laser Die to Wafer attachment. This technology shifts away
from singulated submounts to bonding directly on a wafer or
on substrates. Design considerations for component heat
flow, power, and reliability are significantly affected by the
specific geometry, materials, and interconnect process. This
41. Figure 10 – Steady State Heat Wafer Stage and Pulsed Heat used for initial process development by measuring placement
Pick Tool Optimized for Laser Die Eutectic Solder Attach accuracy through the backside of the glass. The P-Side
Figure 10 shows a close up view of the Palomar measurement results in figure 12 show that all samples were
Technologies Model 6500 with wafer stage, pulsed heat pick below ±3µm in the X and Y directions.
tool, and lookup camera alignment algorithm that was used
for execution of the tests.
The pulsed heat tool is controlled through a temperature
ramp profile as shown in figure 11. This is not the profile
used for the study. While holding the laser in place on the
wafer, the tool ramps from a background temperature to the
reflow temperature, holds that temperature during the reflow
time, and is then cooled down to the background temperature.
Figure 12 – P-Side Measurement Results for two Glass
Substrate Runs
P-Side measurements are time consuming due to
preparation of the sample materials and are nearly impossible
to perform on actual wafer level substrates. To mitigate this
problem, a correlation between P-Side and N-Side
measurements was established which resulted in a more
effective N-Side measurement specification of ±4.5µm in X
and ±3µm in Y. The correlation was based on known
misalignment errors between the lithography masks of both P-
Side and N-Side of the wafer. All measurements could then
be performed on a Nikon VMR3200 Automated Optical
Inspection Microscope.
Figure 11 – Generic Example Pulsed Heat Tool Profile The wafer level substrate was then built and measured
using N-Side measurement techniques as shown in the chart
The pulsed heat tool used to pick the laser die includes the of figure 13. Most samples were within specification for both
following capabilities and benefits: X and Y dimensions. Device level testing is normally
• Pulsed heat controller required to determine actual yield.
• Temperature up to 600C
• Temperature accuracy ± 20C
• Fast ramp (up to 65C/s) - no overshoot
• Parts at high temperature for a limited time
• Programmable ‘point and click’ profiling
The wafer assembly process steps include:
• Load wafer onto steady state heated stage
• Load laser die (pre-flipped with P-Side down) onto
machine
Figure 13 – N-Side Measurement on Wafer with Correlated
• Repeat the following process for all bond sites: Tolerance of X ±4.5µm and Y ±3µm
o Vision find wafer at next available site
o Vision find N-Side of next available laser die Solder aging at temperature is typically not considered for
and pick it singulated substrates since the solder is exposed to
o Vision find P-Side of laser die on pick tool temperature for a short duration of time only. When moving
using lookup camera to wafer level substrates however, solder can be exposed to
o Align and place laser die to substrate site background temperatures over 200C for tens of hours
o Apply bond force and initiate pulsed heat profile depending on the number of bond sites and the equipment
o Release laser die upon completion of pulsed heat throughput.
profile including forced cooling cycle to below Tables 4 and 5 provide a summary of the results from a
reflow temperature study on the effects of exposure to time at temperature. A
combination of time intervals at a background temperature of
Since it is not possible to directly measure the P-Side 230C was used to evaluate both non-reflowed solder and
interface of the completed assembly, glass substrates were reflowed solder in completed assemblies. The study involved
42. waiting the pre-bond time intervals and then attaching a set of
4 laser die each. These laser die sets where then sheared after
Case 1: Multi-channel Communication Products
waiting the post-bond time intervals. This particular test
(VCSELs)
sequence allowed testing of solder exposure to temperature
before laser die attach (0 to 72 hours), solder exposure to The VCSEL based application placed an array of 5
temperature after laser die attach, and a combination of both VCSELs into epoxy. Post-cure results better than ±5µm in X
(0 to 96 hours = Pre-bond Attach Time + Post-bond Shear and ±3µm in Y were achieved.
Time).
Case 2: Wafer Scale – Eutectic Die to Wafer P-Side Down
Table 4 – Solder Attach Average Shear Strength vs. Time Laser Attachment
Shear The P-Side-Down 80/20 AuSn eutectic laser attach onto
Post-bond Shear Time
Shear Average Time wafer produced results of ±3µm in X and ±3µm in Y for P-
[Hrs]
[grams] [Hrs] Side measurements on glass substrate. N-Side measurements
0 24 48 72 0 on the wafer included shifts between N-Side and P-Side but
0 270 275 310 253 24 results still indicated ±3µm in X and ±3µm in Y when N-Side
Pre-bond 24 217 210 241 193 48 measurements were correlated back to the P-Side.
Attach The effect on solder quality of time at temperature was
48 239 206 202 72
Time studied and found to have some effect on average shear
[Hrs] 72 192 189 96
strength. Additional studies could be completed to verify
Grid 1 2
repeated results. Follow- on studies could be conducted on
the reliability of laser devices as well.
Table 5 – Solder Attach Range Shear Strength vs. Time
Shear
Post-bond Shear Time
Shear Range [Hrs] Time
[grams] [Hrs] Acknowledgments
0 24 48 72 0 Palomar Team members who contributed to equipment
0 36 128 128 87 24 and process development include Mike Artimez, Don Beck,
Pre-bond 24 53 21 90 38 48
Tom Boggs, Bill Hill, Dan Martinez, Ricardo Saldana, and
Attach Tim Hughes.
48 6 7 119 72
Time
[Hrs] 72 24 31 96 References
Grid 1 2 1. Bok, Z. et al, “Micron Level Placement Accuracy Case
Studies for Optoelectronic Components”, The 41st
Table 4 shows a reduction in average shear strength from International Symposium on Microelectronics,
270 grams to a minimum of 189 grams over all aging Providence, RI, November 2008
conditions compared to zero pre-bond attach time and zero 2. Dick Otte, Promex Industries, Chair of iNEMI
post-bond shear time. Visual inspection of the sheared Optoelectronic Technical Working Group (OE TWG)
samples showed no observable differences in solder joint Roadmap draft dated September 26, 2008, p. 16
quality at 60X magnification. 3. Weiss, S. et all, “Fluxless die bonding of high power
Table 5 shows the effect on the range of shear strength for laser bars using theAuSn-metallurgy”, Electronic
each data set. The data did not show continuing degradation Components and Technology Conference, 1997.
of shear strength range for pre-bond attach time or post-bond Proceedings., 47th Volume , Issue , 18-21 May 1997
shear time exposure. Page(s):780 – 787
Summary and Conclusions
Two separate optoelectronic application case studies have
been reviewed. Both cases required ±5µm placement
accuracy or better using adhesive or metallurgical attachment
as shown in Table 6.
Table 6 - High Accuracy Application Cases
CASE Attach Accuracy
VCSEL Array Epoxy ±3-5µm
P-Side Down Laser Eutectic ±3-5µm