ParC Home

Modeling Capabilities

RF/Wireless

A lot of RF circuits use encoding schemes that are essentially a sequence of individual tones. From a simulation perspective this is usually difficult to handle as an analog simulation because it takes a really small time-step to simulate accurately, but because you are dealing with tones of multi-cycle duration, a lot of the simulation is redundant (repeated work). Event driven behavioral models can built in ParC that use spectral definitions (a complex real array) for signals on wires, with superposition (via resolution) and array functions to calculate transmitted spectra. This approach makes RF modeling considerably less compute intensive than doing it in (say) Verilog-AMS, and allows whole system verification in reasonable timescales.

Wireless communication can can also be modeled with spectra and use ParC's signal resolution scheme to model free-space effects by considering transmitting antennas as drivers into a signal, and receiving antennas as receivers. Antenna characteristics (e.g. position, orientation, polarization etc.) are stored in the driver/receiver classes.

Neural Networks

Unlike HDLs, modules in ParC can be created as desired and wired up dynamically, writing a model for a neuron is fairly straightforward, and the threading model supports having billions (see http://en.wikipedia.org/wiki/Neuron). That flexibilty allows ParC to be used as a base for Artificial Intelligence (of the kind found in mammals), as well as being capable of simulating the environment. How one programs that machine/model remains to be seen, but ParC can be reflective, so maybe with some genetic algorithms we'll get lucky.


ParC Home