Let's Get Physical

 

How do FinFETs impact Physical Verification (DRC/LVS)?

I am often asked the question, “Does IC Validator support FinFETs?”.  I wrote this post to talk in general about what it means for a DRC/LVS tool to support FinFETs.  Many EDA tools are impacted by the use of FinFETs – most notably SPICE simulators and parasitic extraction engines.  So how is physical verification impacted?  Let’s start by reviewing how a FinFET compares to a standard MOSFET.

FinFET and Planar MOSFET Layout and Model

The FinFET structure has been around for some time, but has only been introduced commercially in recent years.  Most semiconductor foundries are introducing FinFETs as they build technologies below the 20nm process node.  A FinFET offers many electrical advantages over traditional planar MOSFETs that I won’t discuss here.  With a FinFET, the channel is ‘raised’ into a fin, with the gate wrapped around it.  From a 3-dimensional perspective, this is quite complex, but for mask manufacturing purposes, it requires a standard set of 2-D masks.

 The standard device model of a FinFET is a 3-terminal device with optional bulk nodes.  The FinFET netlist topology is identical to that of the planar MOSFET.  In addition to the length and width parameters, it will include additional device properties describing FIN configurations.

FinFETs and DRC

Boiling it down to basics, a DRC tool’s job is to look at layers of 2-D geometries and report errors due to interactions and measurements:

  • on a single design layer
  • between multiple layers
  • between layers in context of a complex net-connection structure (ie same-net checks, antenna checks, net-voltage checks)

 If we consider this in context of a FinFET – you can see that a FinFET is built from a simple collection of 2-D masks.  While it may have many additional rules (compared to a planar MOSFET) – the rules will all fit into one of the 3 categories above.  What we are finding in practice is that to effectively manufacture these devices at extremely tight pitches, the design rules for FinFET layers are much more restrictive, and in some ways, much simpler to manage as a designer than planar MOSFET rules.  For example, while MOSFETs can be drawn without consideration to pitch, FinFETs must often be very carefully placed with respect to their neighbors so that the entire array of 3-D structures can be manufactured correctly.

FinFETs and LVS

An LVS tool’s job is to extract and compare MOSFET devices (plus any other device types which are not relevant to this discussion).  At a minimum, it must extract a MOSFET with gate, source, and drain terminals.  For most technologies, the LVS tool must also extract at least one bulk terminal.  In addition to topology, the LVS tool must extract and compare certain standard properties, for planar MOSFETs they are generally width and length.

 As you can see from the image above, the 2-D layout of a FinFET is nearly identical to a planar MOSFET.  Therefore, the task of identifying and extracting the FinFET device is trivial for an LVS tool that can extract a 4-terminal MOSFET.  Calculation of the FinFET properties requires slightly different measurement techniques than for a planar MOSFET.  Most notably, a FinFET’s “width” is actually discritized by the number of fin/poly crossings.  This is captured by a number-of-fins property instead of width.  Fortunately, most modern LVS tools have had features to support this for many technology nodes.

One requirement for LVS is emerging, driven by new simulation models.  Since the manufactured silicon is very dependent on neighborhood, the simulation models need to count the number fins in a diffisuion, even those not directly connected to the device being extracted.  This a new requirement on LVS not seen in previous technologies, and can require new software features to properly implement.

Summary

DRC and LVS tools that support recent process nodes should have no difficulty supporting the additional demands of FinFETs.  A foundry writing DRC and LVS runsets to support FinFETs will face additional complexities of adding new DRC rules, and supporting FinFET device definition/properties for LVS.  As a designer, you may experience different types of rules, and different failure modes in LVS.  However, the software itself should support FinFETs without any new features or modification.

 Keep in mind that I refer here only to the impact of FinFET – there are many other features of sub-20nm technologies that will impact Physical Verification tools that I will discuss in future blog posts.

 I invite any feedback and comments.

Share and Enjoy:
  • del.icio.us
  • Digg
  • Facebook
  • Google Bookmarks
  • Print
  • Twitter
  • StumbleUpon
  • LinkedIn
  • RSS