Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.
Although the S-parameter was used for several decades in RF applications, it was used mainly in the frequency domain. The recent application of the S-parameter to time-domain simulations has raised several numerical issues. For example, in a typical frequency-domain analysis, the knee frequency Fknee, (defined as 0.5/trise) covers sufficient energy for digital signals [6]. However, as demonstrated in a later example, this knee frequency could still lead to significant inaccuracy in a time-domain simulation. This section covers several issues associated with broadband modeling and presents several tips for building accurate broadband S-parameter models.
The signal-integrity community has had many debates about comparing the accuracy of the S-parameter model, based on time-domain and frequency-domain measurement methods [7] [8]. Many of these discussions focused on the dynamic range of the measurement instruments used to compare the accuracy of measured data. This type of accuracy comparison is useful in RF applications, but it is less useful for digital applications, because both TDR and VNA devices can provide sufficiently accurate measurements. With reasonable instrument bandwidths, and proper calibration procedures, any error introduced by either of these two devices is negligible, when compared to the process or manufacturing variations typically observed in digital systems. A more critical issue, and one often neglected, is the model’s numerical stability in transient time-domain simulations.