Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: hmaly who wrote (75583)	3/26/2002 1:24:06 AM
From: Yousef	Read Replies (3) \| Respond to of 275872

Hmaly, Re: "Thanks for the post. I have a couple of questions. From previous discussions, I had assumed "defect density" meant the number of defects expected in a area of silicon before processing, which would of itself destroy a chip. Your formula for BE also includes a multiplier of n for the number of layers, but the MUR formula doesn't. Why? What determines defect density. The no. of bad chips, or the defects one can expect in silicon? If the defect density is compiled by totaling bad die/cm^2, are all processing errors called defect densities." Good questions ... I will try to answer them. Please realize that I don't work for INTC, so an engineer for INTC might not agree with my explanations. First, yield can be broken into at least two components -> Ytotal = YsimpleYspeed Ysimple - is the yield when the wafer is test at nominal voltage and frequency (these conditions are the easiest to meet). Thus parts failing Ysimple are usually due to a "hard" process defect (eg. a particle) Yspeed - is yield of a wafer tested at "smhoo" points (voltage +-10%, high and low frequency, ...) Parts failing these tests (that pass Ysimple test) are generally due to parametric issues in the process (eg. wide poly linewidth or Vt's off center, ...) Given this description of yield, defect density is based on the Ysimple failures. These are generally process defects that are randomly distributed on the wafer (many times though there are patterns to the defects - like the edge has more than the center of the wafer). So defect density (DD) is the number of "killer" defects per square cm of wafer area. As you correctly noticed, the Murphy calculation figures yield with no consideration for the number of "critical layers". Bose-Einstein calculations for yield do consider the number of "critical layers" in a process. So the BE method basically takes the yield for each layer and multiples it by the next layer --> Ylayer1Ylayer2Ylayer3...Ylayer20 -> for .13um process The number of critical layers is determined by the number of masking layers with a weight of 0->1 based on potential yield effect -> metal, poly, active, via layers all count 1 implant layers all count .5 .13um process has ~20 critical layers while the .18um process had ~18 layers. One advantage of using Bose-Einstein for Yield calculations is that this equation can be directly solved for DD (defect density). While the Murphy equation cannot. So defect density is calculated based on yield (Ysimple) and can be thought of as the number of hard defects per square centimeter of silicon. These defects come during the processing of the wafer. In fact, both INTC and AMD have in-Fab inspection equipment to find defects on the wafer and engineering groups whose sole job is to review data and determine root cause for these defects. The defect reduction methodologies of each company could be very different which could lead to very different results in reducing defects. Make It So, Yousef