SI CHINA     SI JAPAN
Login  |  Register          Free Newsletter Subscription
Subscribe
Email
Print
Reprint
Learn RSS

Vision System Adds New Life to Die Bonding Process

Paul Rogers ,
IBM Microelectronics, Burlington, Vt. -- Semiconductor International, 7/1/1999

IBM Microelectronics (Burlington, Vt.) had a number of 20-year-old ESEC 2005 die bonders that were prone to errors due to the complexity and number of mechanical steps involved in the bonding process. For example, if the indexing track on which lead frames travel malfunctioned, it could prevent lead frames from reaching a station where adhesive tape is applied to the leads. Similarly, if a die is not secured onto the machine's die heating block due to insufficient vacuum suction, upon making contact with the lead frame, the die temperature may not be hot enough to melt the bonding adhesive.

To monitor the accuracy of the die bonding process, IBM had traditionally relied upon manual inspection. Typically, an operator would pull a lead frame from the die bonder and examine each bonded device under a medium-power microscope to determine if the die-to-lead frame alignment was in or out of specification.

However, the inspection was only performed on a random sample basis in which only three out of 32 lead frames in a cartridge were examined. With 10 devices per lead frame, this could result in a loss of up to 100 devices a day. The only solution to eliminating the losses would be 100% inspection. But this was cost-prohibitive if done manually. As was purchasing a new die bonder that would be too expensive to have modified specifically to their die bonding process since various aspects of our bonding process were not industry-standard.

A cost-effective solution to eliminating yield losses was found in machine vision technology. With the reduced cost of vision processors, it was possible to retrofit the old machine with 'vision.' IBM's team chose to equip the ESEC 2005 with a Cognex MVS-8100 machine vision system running PatMax object location software. The MVS-8100 is a PCI frame grabber-based system that plugs into a 233MHz Pentium Industrial PC. The frame grabber, vision software and PC are contained in an industrial cabinet approximately 6 feet away from the die bonder.

Fig 1 The machine vision system performs the diel-to-leadframe alignment verification and reports the x,y and theta coordinates of the die. (Source: IBM)
The process is straightforward. As a newly bonded device passes under the vision camera, which is mounted a few inches above the die surface, a PLC triggers the camera to acquire an image (see Figure). Once acquired, the image is sent to the vision system and analyzed by the PatMax software that first looks at the distinct geometric features of the image and then compares it to a trained reference image stored in PC memory. If the die is aligned to the lead fingers within the X-Y position tolerance set (typically ~100 microns maximum), the device 'passes' and is then moved along to the wire bonding stage. If the device falls outside the tolerance set, a signal is automatically sent to the die bonder to stop the machine. At that time, a message appears on the PC monitor alerting the operator that a device has been improperly bonded and indicates the number of microns the die-to-lead frame is misaligned. The device is then manually retrieved and placed into another machine that prints out correction values that are then used to make the necessary adjustments.

In addition to shutting down the die bonder when an out-of-spec device is encountered, the vision system gives a 'warning stop' upon sensing that a die is shifting out of position. If the device is nearly three quarters of the way out of tolerance, the system can temporarily stop the machine and tell the operator that something is wrong and that some machine component needs to be cleaned or adjusted. At that point, an operator can take action. Another key benefit of the vision system is that it is able to tolerate significant changes in die appearance that occur during the process. Depending on how light hits the surface of a die, the image contrast may vary depending on the thickness and uniformity of overcoat used. Reflections off the die can also affect the appearance of the bond pads, which the vision system uses as reference points to measure from. In addition to tolerating lighting variations, the system can adapt to variations in the optical scale of the devices due to die breaks and debris. This same insensitivity to optical scale enables IBM engineers to train a reference image on one system and download the image to other machines throughout the fab.

Since implementing the vision system, IBM has significantly reduced yield losses while realizing greater control over the die bonding process, noted Richard Charlton, staff engineer at IBM. A greater than 95% improvement in yield is lost. Instead of losing ~100 devices a day, only two or three are lost. Image quality is also much less of a concern, so precise lighting and optical settings are less of an issue, thus reducing setup times. Because of these results, plans are in the works to use vision technology to pre-inspect the tape apply mechanism on the die bonding tools and verify its proper application of tape to the lead frames

 

Paul Rogers, advisory engineer, product equipment design, IBM Microelectronics.
Phone: 802-769-2528
Fax: 802-769-2880

E-mail: progers@ibmus.com

Email
Print
Reprint
Learn RSS

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS



 
Advertisement
SPONSORED LINKS

More Content

  • Blogs
  • Podcasts
  • Videos

Blogs


Sorry, no blogs are active for this topic.

» VIEW ALL BLOGS RSS

Videos

Advertisements





NEWSLETTERS
Plug in and get the latest SI news, trends and industry updates delivered free, directly to your inbox!

SI NewsBreak and Special Reports (Weekdays)
Wafer Processing Report (Monthly)
Lithography Report (Monthly)
Metrology Report (Monthly)
Clean Processing Report (Monthly)
Packaging Report (Twice Monthly)
©2008 Reed Business Information, a division of Reed Elsevier Inc. All rights reserved.
Use of this Web site is subject to its Terms of Use | Privacy Policy
Please visit these other Reed Business sites