J-fraction Route to the Cauer Ladder
Polynomial long-division and Lanczos both produce the Cauer (Stieltjes) continued fraction directly. A third family of algorithms—due to Khatwani and Sobhy—first casts the impedance into a J-fraction and converts that to the desired S-fraction. This detour avoids explicit high-order divisions and relies only on scalar recurrences.
Starting point: rational impedance
After deconvolution the driving-point impedance is known in rational form
The goal is to rewrite (63) as the Cauer (S-) fraction of length 2 N whose coefficients are the thermal resistances and thermal capacitances.
J-fraction representation
Both Khatwani’s and Sobhy’s schemes build the continued fraction
known as a J-fraction. The auxiliary parameters \(a_i^2\) and \(b_i\) are obtained by recursively expanding the simpler form
with \(H_i , h_i\) computed either from the Markov parameters (Khatwani) or by Sobhy’s direct iteration. This is referred to as the \(H\)–\(h\) continued fraction.
For \(i > 1\) the conversion is
From J– to S-fraction
The J-fraction (64) must be rearranged into the Stieltjes or S-fraction
whose \(c_k\) are exactly the Cauer \(R'_k,\,C'_k\) in alternating order (\(c_{2k-1}=C'_k\), \(c_{2k}=R'_k\)).
Initial step
Recursive update for \(i>0\)
Iterate until \(i=N-1\) to obtain all \(2N\) coefficients.
Algorithm outline
J-fraction Algorithm for Foster-to-Cauer Conversion
Input: \(\{\alpha_k\}\), \(\{\beta_k\}\) (numerator / denominator of \(Z(s)\))
Output: \(\{R'_k, C'_k\}\) for \(k = 1 \ldots N\) (Cauer ladder)