Introduction
T
HE TIME-DOMAIN IMPEDANCE PROFILE of cables and connectors has remained for decades an
important indicator of their characteristics such as characteristic impedance and delay, or trans-
mission problems that include reflections caused by impedance discontinuities, attenuation, skew
and some others. Historically, time-domain reflectometry (TDR) impedance first was a measurement
technique. Later on, when operational frequencies advanced, the need for detailed electrical simulation
necessitated full S-parameter measurements, and the connector vendors supplied either TDR results and
S-parameters, or – more frequently - only S-parameters. Today, TDR impedance characteristics remain of
great importance for designers, but more and more often these should be found directly from S-parameters
by computations. To save processing time and disk space, the sampling of measured S-parameter data
is often made barely enough to correctly represent the connector’s delay, and the upper frequency comes
short to what is required by the Nyquist criteria for a given data rate. In TDR impedance computations,
we deal with two times the connector delay, which makes given S-parameters even more under-sampled
considering this particular purpose. The latter creates some computational challenges which we are going
to address in the paper. The second issue, perhaps even of greater importance, is the fact that what we
observe in the impedance plot may not be the actual characteristic impedance of the transmission line, if it
is masked by a segment with different characteristics.
A commonly used method to obtain the impedance profile from S-parameters is by using inverse dis-
crete Fourier transform (IDFT) of the return loss. This method does not take into account the multiple
reflections due to impedance discontinuities. These impedance discontinuities introduce multiple reflec-
tions, which act as secondary sources. If not accounted for, these multiple reflections can give incorrect
result for impedance profile. In this paper we propose an alternate way of computing the impedance profile
by modeling the line as that obtained from the cascade connection of small transmission line structures.
The method involves calculating the impedance for small section at a time and then undoing the effects of
this small section before proceeding to calculate the impedance for the following sections. As a by-product
of the algorithm, we describe the assumption involved in calculating the impedance profile by performing
the IDFT of the return loss. It turns out that for situations where the characteristic impedance of the line
is close to the reference impedance, the approximate method of using the IDFT does not introduce large
errors and can be used. Nevertheless appropriate care needs to be taken to compute the IDFT of return loss
to account for the discrete nature of the S-parameters, band-limited data and the frequency spacing of the
data.
The paper is organized as follows. First, we show some basic relations between characteristic impedance,
voltage transfer functions and S-parameters. Then, we consider two methods of converting frequency-
domain characterization into time domain: by inverse Fourier transformation of the sampled dependence
and then by rational fitting represented in a form of rational fraction expansion (RFE). The two approaches
are applied to the same connector model. We show that different termination conditions on the far end may
affect the initial portion of the impedance plot if the data is not perfect. In the next half we describe how
the impedance profile can be recovered by modeling the line as that obtained from a cascade connection
of small transmission line structures. An algorithm is provided which can be easily incorporated into a
computer program.
1