Technology

Cybereda delivers on the technology challenge of having dramatically faster SPICE simulation without compromising accuracy. They are various ways to improve the speed of circuit simulation but can involve compromises that engineers are not aware of.

Dangers of Fast-SPICE
Cybereda technologists have extensive experience in analog circuit simulations. They understand the compromises that Fast-SPICE simulators use to cheat on accuracy and provide faster answers. Unfortunately, designers have to struggle with whether they are getting “more wrong answers per second.”

Fast-SPICE compromises:

  • Subcircuit and Model evaluation with look-up tables can deliver unrealistic results
  • Circuit partitioning tricks such as netlist cutting, latency, and multi-rate ignore analog behavior
  • Current accuracy is questionable because of partitioning tricks especially gate leakage calculation
  • Too many specialized simulator commands to make specific circuits run faster
  •  
    Accuracy Problems of Commercial SPICE
    Many designers are surprised that commercial SPICE simulators that have been in use since the 1980s still have accuracy issues. Because commercial SPICE is now under required to simulate faster because of circuit complexity, short-cuts are taken to speed analysis at the expense of accuracy. Cybereda technologists see analysis problems for even simple circuits.

    The following netlist produces incorrect simulation results in a popular commercial SPICE.

    m1 2 1 1 1 nr l=0.4u w=2.5u
    c1 2 0 1.e-15
    v1 1 0 pwl(0 0, 0.1n 2.97, 1.1n 2.97)
    .model nr nmos level=54
    .tran 1p 5n
    .option post delmax=1p
    .end

    This circuit was simulated with PCSIM (green trace) and commercial SPICE (red trace). At 3nsec into the simulation, the incorrect commercial SPICE result diverges by 50mv from the correct charging curve for the load capacitor.
     
    case1

     

    PCSIM CyberParallel™ Technology
    The PCSIM engine takes advantage of CyberParallel technology to accelerate circuit matrix calculations across multiple CPUs.

    Features of CyberParallel technology

  • Next-generation matrix solver delivers 3X speed-up per CPU
  • New algorithm with linear speed improvement with number of CPUs used
  • Unique solver decomposition to divide calculation across multiple CPUs
  • Guarantees exact equivalence with full-matrix solution
  • Minimizes CPU to CPU communication to enhance speed of processing
  • Minimizes memory usage to provide the greatest capacity for million device circuits
  • Balances computing task across all CPUs

  • Get in touch! get in touch

    Cybereda Corp.
    Suite 103
    4040 Moorpark Ave.
    San Jose, CA 95117

    Main +1 408 615 0586
    Fax +1 408 615 0596
    Click here to email us