Subject: Student project at Philips Research

corydon

Screening for Improved IC Quality and Reliability

The manufacturing process of Integrated Circuits (ICs) is complex and hence defect-sensitive. The raw manufacturing yield is typically anywhere between 30% and 99% depending on the maturity of the process technology used, its feature size, and the silicon area of the IC in question. All ICs produced undergo a series of (automated) tests that determine which ICs can be sold to a customer and which ones should be discarded. These test sets are improved on a continuous basis, to enhance their defect detection qualities and/or to reduce their associated costs.

Two test-related issues affect the customer’s impression of this process.
•        Test escapes. Test sets cannot be guaranteed to be perfect, and hence defective ICs might incorrectly pass the test. These ICs might end up causing malfunction of the customer’s product.
•        Reliability issues. Some defects do not exist yet when the IC is tested in the factory, but are only formed after several hours of usage of the IC, e.g., through circuit deterioration. Also these defects might end up causing malfunction of the customer’s product.
Although these issues are important for all customers, especially the automotive industry puts a lot of emphasis on so-called “zero defect” delivery.

Screening is the process of rejecting chips that pass all their tests, but are nevertheless suspect of being test escapes or latent reliability problems. Two common screening methods are neighborhood cluster detection and parametric outlier detection. In neighborhood cluster detection, a die with too many failing dies as its neighbors on the wafer is suspect. In parametric outlier detection, a die with measurement values within specification, but sufficiently different from other passing dies, is suspect. Also, combination methods exist. Screening methods vary both in their effectiveness for finding true test escapes and latent reliability problems, but also in how many ‘innocent’ good dies they reject.

Philips Semiconductors already applies screening methods to a limited extend in its production. The topic of this student project is to assess the effectiveness of these screening algorithms and to refine them, by means of test data analysis for an industrial IC. The target
-2




- 2 -

IC is a mixed analog/digital circuit in 180nm technology, with approximately 900 parametric measurements for each die. The student will be analyzing data from around 1 million pieces produced over the last months.

The project will include the followings tasks.
1.        Set-up of a software environment to analyze the production test data.
2.        Implement existing screening algorithms for baseline comparison.
3.        Invent and implement new and improved screening algorithms.
4.        Filter for known effects.
5.        Compare effectiveness of various screening algorithms.
Oral and written presentations on the work performed are of course part of the project. Depending on the results of the project, one or more joint scientific publications may be possible.

The project will take six to nine months. It will be carried out on-site at the High Tech Campus in Eindhoven, The Netherlands, due to the required high-intensity knowledge transfer and supervision, as well as the availability of specific tools and circuit designs. A multi-disciplinary team of experts from Philips Research and Philips Semiconductors will supervise the student through regular meetings.

Working in this project at Philips Research in Eindhoven implies working in a stimulating, multi-disciplinary environment at the forefront of technology, with knowledgeable colleagues, and an excellent (computer-) infrastructure. For more information: see
•        http://www.research.philips.com
•        http://www.hightechcampus.nl/.

Candidate students for this project need to send an application letter to the contact person below, accompanied by a resume and course curriculum plus grade list. Candidates should have a background in electrical engineering, computer science, or physics. Experience with computer programming is required. Experience with databases, data manipulation, and data mining, as well as knowledge in statistics is an advantage.

For more information or application:
  Erik Jan Marinissen
  Philips Research Laboratories – IC Design / Digital Design & Test
  Prof. Holstlaan 4 – WAY-41, 5656 AA  Eindhoven, The  Netherlands
  E-mail: Erik.Jan.Marinissen@philips.com
  Tel. +31 40 274-3227 / Fax +31 40 274-4113