If you have worked in the semiconductor industry for more than a few years I am sure you have heard senior leadership speak about the need for your integrated circuit designs to be first pass successes and not the typical two to three spins or more to reach the targeted performance. The question is this: Is first pass success feasible and should be expected? I do not want to stir up a hornets nest with my response but the answer to the question is that it depends. Depends on what you say? Well, the answer depends on several different interwoven complexities that can determine if first pass success is possible. I would like to explore some possible ways to answer this question. Furthermore, the complexity of this question increases when developing complex mixed signal IC’s.
For IC development on the sub nanometer technologies it becomes very expensive to not get it right the first time or at least close to being right the first time. In the end, many project business decisions count on the design development as production ready by pass two or the business case may not work. As we move forward down the integration technology curve and more of the embedded system type solutions become standard practice, the idea of a first pass success in the nanometer technologies become even more important to meeting the needs of the business. In addition, the turn time from concept to final customer integration is shortening as well. Customers are expecting quicker turn times of the development of silicon – even if it is full custom- as their customers are demanding faster innovation to remain competitive. Therefore, the need to find a path to first pass success is important for the success of the project.
So what are some of the decision criteria if followed would help enable your designs to be first pass successes? Some of the conditions that will determine if first pass success is possible depends partly on the following:
Have you clearly defined first pass success with your management and the customer?
Producing a clearly defined criteria for first pass success is extremely important. If the criteria for first pass success is clearly communicated up front, meeting the goal is much more tractable. Furthermore, a clear definition will build confidence with your management and customers when you meet or even when you do not meet the goal. How is this? For instance, if the goal for first pass success is to meet a target specification for a new IP block and your design team clearly defines the risk level or better yet the confidence level of meeting the specification, but the new developed IP misses the target, is all lost? Here again the answer is that it depends. If the design misses the targeted specification by a percent value but the percent value aligns with the risk level there should be no surprises. Even though the leadership or the customer may not be happy at least there are no surprises and if communicated properly up front the customer may be able to plan properly for the potential shortfall in performance. This leads into the next question, does your team have some way to assess the proper risk to avoid design pitfalls?
Ensure your team has some way to assess the proper risk to avoid design pitfalls.
To answer this question, I propose that a proper risk criteria is established at the beginning of the project. The risk assessment should include the proposed design technology, the architecture risks, resource risks, and simulation model deficiencies to name a few. In figure 1, 2, 3 are snapshots of such a spreadsheet tool developed that provides a weighted score of the risks from the different contributors just mentioned. This “sniff test” spreadsheet includes various criteria for analog design to allow the design lead to understand where the weakest link exists in the proposed design and further allows management to understand the overall risk to the projects first pass success.
As part of any risk assessment, an important aspect is the idea of including a test vehicle to explore the new IP. This concept is very important for new analog IP and even older analog IP but on a new technology. Creating a test vehicle apart from the IC allows for the designers to check out critical aspects of the circuits that must be specially tested. As shown in figure 2 is a section titled “cycles of silicon learning” where such a test vehicle would be included as part of the risk. Obviously the lack of inclusion of a test vehicle would reduce the cycles of learning and add to the risk level.Figure 1Snapshots from a risk assessment tool
Figure 1: snapshots from a risk assessment toolFigure 2Snapshots from a risk assessment tool
Figure 2: snapshots from a risk assessment toolFigure 3Snapshots from a risk assessment tool
Figure 3: snapshots from a risk assessment tool
Define the part being created as a digital only IC, analog only IC, or a mixed signal IC
Feedback on this question is straight forward. However, the complexity of the design must be comprehended. This means if the design is a mixed signal IC then what are the various frequencies present on the die that can interact in a negative way. In addition, the various blocks on the IC must be understood how they would interact. Do the blocks have large dynamic range and how are the blocks placed on the IC? The design lead of complex embedded systems must understand how the signals will travel and what signals are critical. In addition, the design lead must understand what signals can be auto-routed and what signals must be routed by hand and properly isolated.
If the part being developed is part of a complex embedded system, ensure the system for which the part is being developed is well understood by your company’s team developing the product.
The worst case scenario is that the design team does not have enough knowledge of the system to understand potential problems for critical design targets. At a minimum, the design lead must understand the system well enough to properly interface with the customer to address concerns at the technical level. A lack of proper understanding often leads to over/under design and general inefficiencies that cost schedule, area, and reduce the probability of a first pass success. To avoid these type issues, it is important for the system teams that interface with the customer include the design lead on the meetings to enhance the required system understanding of the product being developed.
To improve the probability of success, it is important to establish a strong relationship with the customer to clearly understand the requirements for the part being developed.
This concept is very important as the design of complex systems often require tradeoffs with the various specifications targets. In addition, assuming that your team has the appropriate understanding of the system being developed, try to probe and find out if the customer understand their own system well enough to know the effects of the proposed specifications on their system- ask “what if” type questions to gauge effects of missing specifications. In addition, do you have the proper understanding such that you can fully vet that the proposed set of specifications will work in their system? This point must be emphasized. Without a clear understanding of the system, the specifications may be miss targeted and will not work in the system. Furthermore, even if you meet the proposed targets and the part does not work in the customers system you cannot call the part a success. Therefore, without a clear understanding of the system risks, the pathway to a first pass success is questionable.
In part 2 of the next article, I will present some additional questions that should be asked during the beginning phases of any project and add some additional comments.