This month issue An Independent Journal Dedicated to the Advancement of Chip - Scale Electronics January - February 2001 Email the editor

By Ron Iscoff, Editor

At the right price, you can find nearly every feature you want on the sophisticated die attach gear currently available. "Nearly," of course, is the operative word.

To find out what equipment makers might be missing, as well as what features users consider essential, we surveyed several packaging engineers.

Despite tremendous advances in throughput, yield and the ability to integrate die attach with wire bonding (and other EOL equipment), users want more, as the following comments show. (Four feature-packed die attach systems are shown in Figures 1-4.)

"Current and upcoming technologies in packaging are exerting undue pressure on assembly processes," observe Nohmar Acosta and T.M. Chung, both die attach process engineers; and K.H. Kwok, die attach technical support engineer, at AIT (formerly Hana Technologies), Hong Kong.

"At die attach, a few of the challenges that must be met head-on are large and thin die combinations, multi-chip packages, pin-point die placement accuracy and zero DPM," adds Acosta. "At AIT, we expect and demand more robust next-generation die attach machines that can help us keep abreast of the latest technology," the trio responded.

Wish List

Aside from the perennial higher UPH and less machine downtime requirements, the AIT wish list looks like this:

• More accurate die placement because new technologies leave little room for error. The ultimate target should be <1 mil for x and y axes, <1 degree for die rotation and 0.3 mil maximum die tilt

• No more jamming. This results in chip cracking and die scratches which can reduce yield.

• More powerful IQC (image quality check). The machine should be able to screen out-on a realtime basis-epoxy-related defects, such as epoxy on die or lead.

• Microcrack-free die ejector system for large and thin die, because wafers will become thinner.

• The machine should be capable of detecting wrong substrate orientation using the chamfer (pin 1) as the basis.

• In current models, there is a tendency for inconsistent dispensing as the syringe approaches the critical level. The machine should produce an alarm once the present level is reached.

• Reduced set-up time is also desirable, including quicker indexing set-up and flexible and user-friendly conversion kits. More lighting options for the PRS are also needed.

Figure 1. The Alphasem Swissline SL 9022, which will be introduced in May, can handle wafers up to 300 mm.	Figure 2. The DIAS Automation A-9020 system specifies a bonding cycle time of 1.5 seconds.

The new features wish list from the AIT team includes a built-in x-ray to scan epoxy voids under the die.

Another desirable feature is the capability to detect rejected (X-out) substrates. The machine will then not dispense epoxy on these marked rejects. This feature can save precious silver epoxy and may even improve UPH on low-yield (many X-out) matrix substrates.

The AIT engineers also like their die attach systems equipped with heated plates along the indexer length. This is to cure "snap-cure" pastes during the actual attachment process, which will boost to cycle time.

Finally, they would like the machine totally visible under a see-through cover, to prevent dust/fibers from settling on devices being processed.

Mani Govindasamy, senior manager-FOL process engineering at Amkor Technology Inc., Chandler, Ariz., offers this wish list:

• Automatic substrate sensing and die thickness should be a standard feature in all die bonders to provide consistent adhesive bondline thickness.

• Split-field optics should be employed for die-to-substrate alignment.

• The conversion time between waffle packs and wafers should be less than 10 minutes.

• All major die bonder models should offer 300 mm wafer handling by the end of this quarter.

Figure 3. The ESEC Micron 2, which accommodates wafers up to 200 mm, was introduced last May.	Figure 4. ESC's Apollo 3000 die bonder handles wafers up to 300 mm.

Major Goal: Process Integration

Govindasamy also looks for machines that deliver no loss of throughput from the die flipping mechanism. Furthermore, he says, "Process integration (substrate, wafer, etc.), with built-in heater block for curing, should be a major goal for reducing cost of ownership."

Stephan Chan, an assembly engineer at ASAT Ltd., Hong Kong, offers a common suggestion, that we received from most users: Improve die placement accuracy <± 1 mil and boost the UPH.

Other recommendations from ASAT's Chan include:

• The ability to handle multiple wafers, accommodate smaller die sizes <10 mils and two or more types of epoxy dispensers.

• Provide AutoCAD files to reduce the set-up time, with automatic adjustment for length, width, thickness and matrix of the leadframe.

• An inline system with automatic inspection capability to identify die attach quality and transfer the quality data to the wire bonder.

Like several competitors, Chan also wants an inline heater block capability for using snap-cure epoxies and bonding films.

Alex Wu, manager of the CSP Division at Meicer Semiconductor, Taiwan, wants a die bonder for µBGA packages with a high accuracy, integrated dispensing system.

"We can't use the original dispensing system supplied by the equipment vendor for µBGA packaging applications," Wu says.

"To handle today's tighter specs in die attach, we also need high resolution CCD cameras and a bonding arm based on an air-bearing system," he adds.

James Lai, supervisor of the R&D Group at Walsin Advanced Electronics, Taiwan, checks in to recommend these capabilities:

• The handling or input hardware should include a flexible design for various packages: leadframe-based, flex or rigid substrates.

• An auto-calibration function for x, y and z axes is needed.

• An auto-bonding force calculation method (or system) should allow easy set-up for specific bumps, bump widths and die sizes.

• A multi-function "all-in-one" machine, using multiple bonding methods should accommodate different raw material substrates.

Additionally, Lai adds, top/bottom heating stages are needed with a very high bonding accuracy. Lai calls for ± 5 microns. Add an accurate underfill dispensing system, as well as an integrated vacuum changer for die-attach curing.

Conclusion

Are users ever satisfied with their equipment? Fortunately not, as the comments above indicate.

If the equipment delivered perfection for today's package formats, it's likely that further advances would take much longer than they currently do. And at some point-which wouldn't take long-today's machine perfection would become tomorrow's obsolescence for advanced packages.

Next