Semiconductor Spotlight

Design exploration in days, not months
Mach42’s platform sits at the intersection of analog verification and AI, delivering unmatched speed and accuracy in design verification by exploiting neural network models.

At its core, Mach42’s neural network architecture is dynamically reconfigurable, adapting to each simulation dataset to generate the most accurate prediction results. This flexibility ensures the platform delivers reliable, customized models tailored to the unique characteristics of every design scenario.

Unmatched confidence in your design’s performance—powered by AI

The Discovery Platform addresses challenges across the semiconductor design cycle to maximize design confidence and minimize time-to-market.
  • Accelerated Verification

    The Discovery Platform integrates with best-in-class EDA tools to accelerate the most resource-intensive phase of design—verification. In conjunction with industry-standard simulation software, it delivers faster, more accurate results and significantly reduces time-to-silicon.
  • Out-of-Spec Violation Analysis

    Before tape-out, the Discovery Platform scans the design space to catch out-of-spec issues, including package and PCB RLC effects. This ensures your design meets specs with high confidence and supports robust IP re-use across varied conditions.
  • AI-Powered Model Generation

    The Discovery Platform can auto-generate Verilog-A models based on its neural network surrogates. Fully SPICE-compatible, these auto-generated models eliminate the need for manual Verilog-A programming, boost quality, and save significant engineering time.
Out-of-Spec Analysis: In Action

Full Demo

Out-of-Spec Analysis: The Flow

Upload

The Discovery Platform communicates directly with your simulator of choice to build up a picture of your design. Getting started is simple: just upload your standard design files using the UI and define how you’d like to run your simulations.

Search

The platform intelligently samples simulation data to uncover regions of parameter space where your design may fall out-of-spec. By combining sampled simulation data with neural network predictions, it identifies areas that require more attention quickly and effectively.

Analyze

Following a final validation using your simulator, the platform creates a report on your device’s compliance. A visualization of the parameter space provides an overview of your device’s behavior, alongside case-by-case predictions for further analysis and design iteration.


Real designs. Real fast.

The Discovery Platform has been used to produce ultra-fast and accurate neural network models of real production designs including:
  • Power management devices
  • SerDes
  • Oscillators
  • ADCs