March 06, 2006
Yield Analysis Automation - LogicVision
Please note that contributed articles, blog entries, and comments posted on EDACafe.com are the views and opinion of the author and do not necessarily represent the views and opinions of the management and staff of Internet Business Systems and its subsidiary web-sites.
Jack Horgan - Contributing Editor


by Jack Horgan - Contributing Editor
Posted anew every four weeks or so, the EDA WEEKLY delivers to its readers information concerning the latest happenings in the EDA industry, covering vendors, products, finances and new developments. Frequently, feature articles on selected public or private EDA companies are presented. Brought to you by EDACafe.com. If we miss a story or subject that you feel deserves to be included, or you just want to suggest a future topic, please contact us! Questions? Feedback? Click here. Thank you!


Are you familiar with RDC (Design Rule Checker) tools? Twenty to twenty-five years ago when you did DRC, you would plot out the chip and you would visually inspect whether all the rules were respected. Similar to that we built a system that automatically executes best practices for analysis techniques on the selected data and identifies problematic behavior. In addition a customer can take existing yield analysis procedures and use our product to automate them. What we have built is what we call a component based analysis model. It provides all of the atomic actions that the customer would use. It basically fetches a certain set of data and analyzes it a lot of different ways. It has
a lot of different analysis techniques and it help visualize by creating some specific plot or report. We have built these components which are pieces of software. The user is able to take these components, put them in a script and then automatically execute them. You are taking a work flow. You can automate a yield procedure with this component model.


So the product engineer is performing a Design of Experiment on the intended range of operational parameters and seeing how that relates to yield?

Right! That's the characterization phase. When they do these DOEs and build these test parts, they try to understand the behavior across all these ranges. The second thing that is done in volume production is that they still keep watching these parameters on every piece of silicon that is built to see if any problematic behavior occurs or a yield excursion that they can quickly fix. There is a characterization phase and a monitoring phase.


So the production phase would suggest to someone in manufacturing that the process is straying a bit?

That could be the issue but they are really looking at what the part is doing. So it could be the process or some other issue. But you are right. Most of the time the design doesn't change. It is the process that is straying. They look at the effect of that straying on the parameters. If in production there is something straying, they typically adjust the process.


If during characterization they find out that the behavior is not according to spec in a particular temperature range or in a particular voltage range, what does SiVision do beyond simply reporting that fact?

Spotting an excursion is the easy part. Helping them to figure out what caused that excursion is what the tool exists for. That's why we have for example correlation analysis. This enables one to correlate the parts that are failing for example where the limits that are being exceeded with the conditions that are causing it. To be able to identify that most of the parts that are failing are in this process corner. Correlate failures to an operating condition and then be able to help them analyze what is causing this problem and what they need to go and fix. Typically if the product engineers have an issue during characterization, they go to the design engineers and ask “Is this the behavior you were expecting in silicon?” The design engineers look at it and say “No, it isn't. Here's what you should look out for.” It's a team effort to figure out what the cause of the problem is. What we give them is all of this data in one area, all correlated. You basically know for this part here in the test, the conditions used and for the same part later on where they did wafer acceptance testing or etest which is even earlier what were the parameters. They are able to correlate much earlier in the manufacturing cycle and determine what they should be looking out for to ensure that the part works properly in the end. And if it is not
working properly, it helps to find out what they should go and adjust. Should I go back to the fab and tell them to stay away from this process corner or adjust this process parameter because when this process parameter is in this range, we get the yields we expect. Basically in helping them we also come up with actionable information to make a fix.


Let me give you a specific example where a customer derived value from our software. One of our customers that was in production saw a major yield hit. Their yield suddenly went down to 50%. They couldn't figure out what the problem was. The foundry couldn't figure it out either. There was a lot of finger pointing as to whether it was a design issue or not. By simply doing a correlation analysis with respect to the equipment, they were able to see that most of the failing parts were going through one particular piece of equipment. It was clear then when they went through that equipment that things were unlikely to work. When they didn't go through that piece of equipment, things
worked out much better. It was clearly an equipment problem which the foundry accepted. That's a simple example but from a dollar standpoint it provided great value because they were able to get their yield back on track.


Is this the first release of SiVision by LogicVision?

This is the first release after the acquisition. We did have a release before the acquisition which was pretty much the first version of the product. We have focused over the last year on building the next generation component based automated analysis system. This is the first release of that system.


How many customers were there for the prior release?

We have a small group of customers. The ones we can talk about include a customer in Texas called Microtune who builds digital tuners. Another customer is Agere who is also a major LogicVision customer.


What is the pricing and packaging of SiVision?

The pricing of SiVision starts at about $70K on a per design basis. It's a TBL, one year license. If you have more than one design, we give you discounts on multiple designs.


Per design or per designer basis?

Per design! One of the points that I didn't make is that the key capability of the product is that it is a browser or web based system. When a set of product engineers are working on it, we want to enable them to work with their foundry, their test house and with their designers more efficiently. We have built it basically like a web tool. Anyone anywhere in the world who you would like to be on the system and to be able to see the behaviors that you are seeing can do so. We try not to limit ho0w many people use it. That's why we sell it on a per design basis, not on a user basis.


For a typical design, how many people might be using this tool?

Obviously it depends on the design. On a single design there might be 2 to 5 product engineers. A lot of times these product engineers are working on multiple designs. It could be a larger team but it is typically 2 or 3 PEs working on a design. It really depends upon what stage of the post silicon you are working on. If it is characterization, then there is intense activity for a few weeks or months. If it is production monitoring, there may be a single or a couple of PEs watching multiple designs. They are not actively doing stuff to it. They are just looking out for excursions.


How is the information presented to the users: text, tables, graphs?

The information that comes in is first cleansed. That's a big part because the data that comes in raw needs to be formatted. As I mentioned our data base takes all of this data and correlates to which means that you can cross reference this is the part this tracks from the very beginning, when it is part of a wafer to when it is packages. We cross correlate. Different views at different stages of test are available. We have a host of different plotting and mapping techniques. Product engineers use box plots quite a bit, histograms and some plots. All of these ways of analyzing data are there.


« Previous Page 1 | 2 | 3 | 4 | 5  Next Page »


You can find the full EDACafe event calendar here.


To read more news, click here.



-- Jack Horgan, EDACafe.com Contributing Editor.


Rating:


Review Article Be the first to review this article

EMA:

Featured Video
Editorial
Peggy AycinenaWhat Would Joe Do?
by Peggy Aycinena
Retail Therapy: Jump starting Black Friday
Peggy AycinenaIP Showcase
by Peggy Aycinena
REUSE 2016: Addressing the Four Freedoms
More Editorial  
Jobs
Development Engineer-WEB SKILLS +++ for EDA Careers at North Valley, CA
FAE FIELD APPLICATIONS SAN DIEGO for EDA Careers at San Diego, CA
Manager, Field Applications Engineering for Real Intent at Sunnyvale, CA
ACCOUNT MANAGER MUNICH GERMANY EU for EDA Careers at MUNICH, Germany
Principal Circuit Design Engineer for Rambus at Sunnyvale, CA
AE-APPS SUPPORT/TMM for EDA Careers at San Jose-SOCAL-AZ, CA
Upcoming Events
Zuken Innovation World 2017, April 24 - 26, 2017, Hilton Head Marriott Resort & Spa in Hilton Head Island, SC at Hilton Head Marriott Resort & Spa Hilton Head Island NC - Apr 24 - 26, 2017
CST Webinar Series



Internet Business Systems © 2016 Internet Business Systems, Inc.
595 Millich Dr., Suite 216, Campbell, CA 95008
+1 (408)-337-6870 — Contact Us, or visit our other sites:
AECCafe - Architectural Design and Engineering TechJobsCafe - Technical Jobs and Resumes GISCafe - Geographical Information Services  MCADCafe - Mechanical Design and Engineering ShareCG - Share Computer Graphic (CG) Animation, 3D Art and 3D Models
  Privacy Policy