Issue link: https://resources.pcb.cadence.com/i/1180526
PSpice A/D User Guide Analog behavioral modeling October 2019 378 Product Version 17.4-2019 © 2022 All Rights Reserved. responses. It also reduces the number of opportunities for numerical artifacts that might reduce the accuracy of your transient analyses. Laplace transforms can contain poles in the left half-plane. Such poles will cause an impulse response that increases with time instead of decaying. Since the transient analysis is always for a finite time span, PSpice A/D does not have a problem calculating the transient (or DC) response. However, such poles will make the actual device oscillate. Non-causality and Laplace transforms PSpice A/D applies an inverse FFT to the Laplace expression to obtain an impulse response, and then convolves the impulse response with the dependent source input to obtain the output. Some common impulse responses are inherently non-causal. This means that the convolution must be applied to both past and future samples of the input in order to properly represent the inverse of the Laplace expression. For example, the expression {S} corresponds to differentiation in the time domain. The impulse response for {S} is an impulse pair separated by an infinitesimal distance in time. The impulses have opposite signs, and are situated one in the infinitesimal past, the other in the infinitesimal future. In other words, convolution with this corresponds to applying a finite-divided difference in the time domain. The problem with this for PSpice A/D is that the simulator only has the present and past values of the simulated input, so it can only apply half of the impulse pair during convolution. This will obviously not result in time-domain differentiation. PSpice A/D can detect, but not fix this condition, and issues a non-causality warning message when it occurs. The message tells what percentage of the impulse response is non-causal, and how much delay would need to be added to slide the non-causal part into a causal region. {S} is theoretically 50% non-causal. Non-causality on the order of 1% or less is usually not critical to the simulation results. You can delay {S} to keep it causal, but the separation between the impulses is infinitesimal. This means that a very small time step is needed. For this reason, it is usually better to use a macromodel to implement differentiation.