BlackHawk Analytics

    Semiconductor product design for the 21st Century:
    Faster, Multiple Goal Optimization, Robust To Process.
BlackHawk Analytics
Our Approach:
Analytic Surrogate Models

Surrogate models are created in such a way that they
locally replace simulation models, such as SPICE
simulations, for the circuit being designed.    

These surrogate models are typically accurate to the
original SPICE runs with approximately 0.1% error or less.

Surrogate models can be produced for:


  • Optimizing a product design with competing  
    performance goals (such as noise, speed, power,
    frequency) at multiple process corners.

  • Examining process variation impact on a specific
    design throughout the expected process window.

  • Examining the sensitivity of a specific design to
    voltage variation across the expected usage          
    conditions.

  • Examining the sensitivity of a specific design to  
    temperature variation across the expected usage
    conditions.

The surrogate models can usually be rapidly evaluated
across the equivalent of millions of SPICE runs per minute
to evaluate yield and performance throughout both the
product design space and the expected process space.

Stochastic Integration

Stochastic integration can be used in combination with
the surrogate models to evaluate yield of selected designs
across the expected process window

The methods we use allow yield evaluation of circuits
beyond 3 sigma -- a necessity where yield is required and
there is little or no repair available in the design.
Once we have surrogate models of a circuit the following
is enabled:

Answers to key questions:
  • Is it possible to meet the design goals in this
    technology?  If so,  what is the design?
  • Is the design sensitive to the process?
  • Is a spec failure possible in the process window ?
  • What are the real corners for the circuit?
  • What is the expected yield across the process?
  • How can the process be shifted for optimal bin 1
    yield?

Understand the performance impact of:
  • Local temperature variation
  • Local voltage variation
  • Local current variation
Visualize the
complexity
See the
Sensitivities
Find the
optimum design
Counter
We recognize that time to working samples and time to
volume are the goals.  

Using our methodology it is possible to significantly
shorten the product design cycle, developing products
which meet or exceed multiple design goals,  and which
are simultaneously optimally robust to process variation.