Quantcast
Channel: Active questions tagged transfer-function - Electrical Engineering Stack Exchange
Viewing all articles
Browse latest Browse all 234

LC filter synthesis, given transfer function (by expressing driving point impedance as continued fraction)

$
0
0

I want to know how to synthesise a passive LC circuit to implement an arbitrary known transfer function. (I understand that not all functions are suitable, so let's assume \$T(s)\$ is a positive real function; all poles and zeros are in the left half plane; in conjugate pairs where appropriate, etc. etc. If it simplifies things to assume \$T(s)={1\over{D(s)}}\$ for polynomial \$D(s)\$; \$R_S=R_L\$; or similar, that's fine. I assume Cauer ladder-style networks throughout for concreteness, but I'm not attached to that topology if an alternative is easier to answer.)

I am following Electronic Filter Design Handbook 4th Ed. (Williams and Taylor), and by performing the procedure described there for expanding Equation (1-4):$$Z_{11} = {{D(s)-s^n}\over{D(s)+s^n}}$$as a continued fraction, I can successfully replicate Table 11-2 (LC element values for Butterworth filters of arbitrary order). This is exactly what I want, but I am interested in other transfer functions in general and not only Butterworth.

An example of a transfer function I am unable to synthesise is the 5th order Chebyshev low-pass filter (0.5 dB ripple) in Example 2-20. I can reproduce the pole positions listed in the book (\$-0.105699\pm0.954967i, -0.276724\pm0.590202i, -0.34205\$), and if I apply Equation (1-4) then I obtain:$$Z_{11}={{1.1069s^4+1.72666s^3+1.10185s^2+0.597734s+0.13417}\over{2s^5+1.1069s^4+1.72666s^3+1.10185s^2+0.597734s+0.13417}}$$Following the same Example 1-1 method that seems to works perfectly for all Butterworth transfer functions, I invert the expression to find the continued fraction, and get as far as:$$Z_{11}={1\over{1.80685s-1.81853+{{2.87577s^3+2.02559s^2+1.44231s+0.378163}\over{1.1069s^4+1.72666s^3+1.10185s^2+0.597734s+0.13417}}}}$$At this point I'm stuck. I know the \$1.80685s\$ term is correct, since it agrees with the first element in the solution to Example 2-20, but I do not understand how to handle the \$-1.81853\$ constant term in the denominator. Such terms had never appeared when transforming Butterworth functions, where each intermediate level of the continued fraction involved only a multiple of \$s\$, indicating a pure reactance? I can absorb the constant into the continuation of the fraction, but that makes little difference, since then the subsequent quotient is not a multiple of \$s\$.

I would appreciate any advice on how to proceed.

(Please note that I am looking for more than just the values for the Chebyshev example; I realise it would be very easy to find those in any standard table, and also that for those particular functions, closed-form solutions are known. I would like to understand how to compute components for transfer functions that are neither Butterworth nor Chebyshev, but are too obscure to appear in tables or to have names.)


Viewing all articles
Browse latest Browse all 234

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>