Clustering for data compression

Computer-aided design and analysis of circuits and semiconductor – Nanotechnology related integrated circuit design

Reexamination Certificate

Rate now

  [ 0.00 ] – not rated yet Voters 0   Comments 0

Details

C716S030000

Reexamination Certificate

active

06591405

ABSTRACT:

FIELD OF INVENTION
Invention relates to profileometry in the field of non-destructive measurement of wafer layers in the basic wafer-fabrication operations of layering, patterning, doping and heat treatments.
BACKGROUND OF INVENTION
A relationship exists between the physical and optical properties of an IC device and the information from light shined on the IC device by instruments such as an ellipsometer and a scatterometer. The physical and optical properties of a wafer are highly sought after by chip manufacturers, since microchip fabrication machinery can be tuned and optimized once these parameters are known, preferably through non-destructive techniques that operate in real time.
Unfortunately, during wafer fabrication tremendous quantities of data are produced that cannot readily be used by a chip manufacturer to control processes in real-time.
The present invention addresses the concern of quickly turning raw data into information in microchip fabrication, using cluster analysis, a technique of multivariate analysis that creates degrees of association between items and the groups assigned to the items.
SUMMARY OF INVENTION
The invention resides in providing a profileometry system with software that provides for cluster analysis to match incoming real-time data signals collected from profileometry instruments, such as an ellipsometer or scatterometer, with a library of data stored in computer memory. The library data is first represented by clusters, having cluster representative data that is stored in primary memory while the clusters are stored in secondary memory. The incoming real-time data is initially matched with the cluster representatives in primary memory to find the closest matching cluster representative. When a match is found, the cluster associated with the cluster representative having the closest match is loaded from secondary memory to primary memory. This technique avoids the excessive disk thrashing found in prior techniques. The invention makes further use of the observation that in chip fabrication by and large data from measuring instruments is often mathematically continuous.
In a preferred embodiment, the heuristic min-max algorithm of T. Gonzalez is used as the preferred clustering and partitioning method.
Furthermore, various refinements are introduced, such as to account for special cases of borderline matches.
In sum, the advantages of the present invention are, in one preferred embodiment, manifold. First, the number of comparisons that have to be made for each incoming signal significantly reduces from 400K comparisons to less than 10K comparisons. Consequently, there is achieved an improvement in speed by at least 40 times. This can be achieved by selectively comparing only those signals associated with clusters of the best matched cluster representatives. Second, in the absence of enough RAM to load the entire library—which is often the case—the present invention exploits an important aspect of process control systems: the variability of incoming signals, under similar conditions, is small during a certain period of time. This aspect suggests that during a certain period of time the best matches for all the incoming signals can be found in only a subset of clusters. Thus an incoming signal is most likely to find its best match in the data that is already residing in RAM. This significantly reduces the number of swappings that take place during any given period of time. Furthermore, since the amount of data that needs to be swapped is a small fraction of the entire size of the data of the library, the time elapsed due to swapping from hard drive to RAM is negligible, and hence the time elapsed due to matching incoming signals to library signals is not that affected.


REFERENCES:
patent: 5386558 (1995-01-01), Maudlin et al.
patent: 5486998 (1996-01-01), Corso
patent: 5991699 (1999-11-01), Kulkarni et al.
patent: 5995644 (1999-11-01), Lai et al.
patent: 6337654 (2002-01-01), Richardson et al.
T. Feder et al., “Optimal Algorithms for Approximate Clustering”, Proceedings of the 20th ACM Symposium on Theory of Computing, pp. 434-444, 1988.
T. Gonzalez, “Clustering to Minimize the Maximum Intercluster Distance”, Theoretical Computer Science, vol. 38, pp. 293-306, 1985.

LandOfFree

Say what you really think

Search LandOfFree.com for the USA inventors and patents. Rate them and share your experience with other people.

Rating

Clustering for data compression does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Clustering for data compression, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Clustering for data compression will most certainly appreciate the feedback.

Rate now

     

Profile ID: LFUS-PAI-O-3087707

  Search
All data on this website is collected from public sources. Our data reflects the most accurate information available at the time of publication.