DFM is simply enabling the ability of the design team to foresee and eliminate any barriers to wafer yield while they are in the midst of design. It could be tools, processes, procedures, reviews or checklists plus many others. Anything that is put in place to close the gap between designing and production yield will be a step towards design for manufacturability.

DFM has been around the IC design world for a lot longer than many would think. When design rule checking software came on scene in the early 80's that was certainly a step towards DFM. With the push towards smaller geometries DFM has been enjoying a renewed focus with concerns over interconnect variability and the ability of design teams to deal with them. The reality is that DFM must always be a consideration, regardless of the technology. The degree of implementation varies by technology, design philosophy or sample schedule requirements.

In many cases the desire to meet a sample date milestone is what drives the design team's focus, energy and decisions. This will invariably lead to cases where some of the activities that produce a high quality design fall by the wayside. The first picks that usually are cut are the DFM type tasks which usually adds a small risk to samples and a much larger risk to your production ramp.

Conversely, if you focus your efforts on getting the design into production you would want to do all the validation necessary to make sure you have a high margin design and your sample date would suffer. The relationship between a focus on sample dates and production dates can be seen at the beginning of this article.

What can you do about this? It would not seem prudent to sacrifice production ramp in today's market place. This leaves you with a sample date that may lose the business. Now is the time to look at the design flow that leads to samples and also review the requirements that drives the need for samples. Try giving these items some thought to help you brainstorm solutions to your samples issue:

Enhance your system trade-offs flow to close quicker on the required implementation.

Partition validation into two phases. Sample validation for tapeout and then follow up with DFM validation while 1st silicon in FAB.

Alternatives to samples such as hardware emulation or modeling so the customer can continue their system development without the need for samples

What is most important is that you do not arbitrarily eliminate DFM tasks in the interest of saving your sample date. That may very well cost you an entire silicon spin on your path to production ramp. This would delay your time to revenue by 10-12 weeks!

Everyone wants to create designs that will have the margin that is required to yield without any parametric fallout. How does one achieve this lofty goal? There are certainly tools out there that will help the cause. Which ones should you use and when should you use them? This is where your design process comes into play.

When you develop your design process with DFM in mind you are sure to have it covered to everyone's satisfaction. Without a process in place that dictates how you will be supporting DFM leaves your team open to making decisions on the fly, and if the pressure is on for making that tapeout date for samples, you will probably go to mask without the coverage you would want for manufacturability.

Some items in your design process that support DFM could be engineering spec templates, design travelers & checklists, design review procedures, design assumptions and best practices. Spend your time defining these items up front and you will reap the benefits on many projects down the road. It comes down to defining what you going to do, do it and then verify that you did what you said.

email: jeff@jorvigconsulting.com

phone: 480-895-0478

web: http://jorvigconsulting.com