ECE 209: Circuits and Electronics Laboratory

Review of Circuits as LTI Systems

Math Background: ODEs, LTI Systems, and Laplace Transforms

Engineers must have analytical machinery to understand how systems change over time. For example, springs and dampers in car suspension systems absorb kinetic energy after road disturbances and dissipate the energy gently over time. Electronic lters protect circuits from large input transients in a similar way. Hence, engineers need to know how to choose components in dynamical systems. LTI Systems and ODEs: The motion of many physical systems can be modeled knowing only a few of its derivatives. For example, a cars position over time can be described well knowing its velocity or acceleration. So we focus on the linear time-invariant (LTI) system x(t) LTI System y (t)

which has input x(t) (e.g., height of terrain under the car) and output y (t) (e.g., height of passengers seat). By our assumption that all we need are a few derivatives of the input and the output, then a typical ordinary dierential equation (ODE) that might model such a system is y + 3y + 2y = x x (1)

where y and y are the rst and second derivatives of signal y (t). Given a known input x(t), we would like to integrate this dierential equation to nd an expression of y (t) that is only in terms of t. Then we can see how to adjust our design parameters so that y (t) has a desirable shape (e.g., even with a bumpy x(t) road surface, the car seat y (t) stays in one place). So we need a convenient way to solve this dierential equation. Linear Decomposition: Assume that x can be broken into two parts so that x(t) = x1 (t) + x2 (t). By inspection, if I can solve Equation (1) for the response y1 to x1 alone and the response y2 to x2 alone, then the response y to x would be y1 + y2 . Further, imagine that I could nd a set of functions such that: When a function in the set is an input to Equation (1), the output is simple to nd. Every useful input x(t) can be expressed as a sum (or integral) of functions from this set. In this case, solving ODEs like Equation (1) would be trivial. Before even knowing the input x, I could nd the solution to the ODE for every function in this special set. Then when I nally do know my input, I just sum up the relevant prototypical solutions. Consider functions of the form t est where s = + j is a complex number with real part and imaginary part . Notice that for a function f (t) = Aest (where A is any constant complex number), f (t) = sAest = sf (t) and f (t) dt = 1 st 1 Ae = f (t). s s (2)

That is, for a complex exponential, dierentiation is identical to multiplication by s and integration is identical to division by s. So complex exponentials turn ODE calculus into simple algebra. Stability and the Characteristic Response: Because every x(t) = x(t) + 0, then the zero-input case is important to us, and we should handle it rst. These zero-input eects will be a part of every solution regardless of the input, and so the output trajectory for zero input is called the characteristic or intrinsic or transient response of the system. We will assume that this output is made up of complex exponentials, and so we let y (t) Aest and x(t) 0, which turns Equation (1) into Aest s2 + 3s + 2 = 0, (3)

and so we must solve for values of s that make this equation true. Fortunately, |Aest | = 0 for all s, and so the roots of the characteristic polynomial s2 + 3s + 2, s = 1
Document

and

s = 2,

(4)

from http://www.tedpavlic.com/teaching/osu/ece209/. Source code at http://hg.tedpavlic.com/ece209/.

Copyright 20072009 by Theodore P. Pavlic Creative Commons Attribution-Noncommercial 3.0 License

Page 1 of 4

ECE 209

Review of Circuits as LTI Systems

both solve Equation (3). Hence, the zero-input transient response of Equation (1) is y0 (t) = Ae1t + Be2t where A and B are complex numbers that correspond to dierent initial conditions. Because et 0 as t for all < 0, then y0 (t) 0 as t . So as long as the roots in Equation (4) are very negative, then y0 (t) will decay quickly, and we can usually ignore its eect. In general, the roots of the characteristic polynomial are called eigenvalues (i.e., intrinsic values or characteristic values). When the real part of the eigenvalues are negative, then the characteristic response will continuously decay. Such LTI systems are called STABLE because aspects of the output that are not due to the input will always decay to zero. Unstable systems have components that grow and eventually dominate the output. Stable Response to an Input: Now we assume that the LTI system is stable (i.e., its eigenvalues have negative real parts) so that we can focus on the system after its transient response has decayed out. So we consider the input x(t) Xest where X is some complex number. It seems reasonable to guess that the st output y (t) = Y e where Y is some complex number. If our guess is wrong, there will be no solution, and we will try something else. So under these assumptions, Equation (1) becomes Y est (s2 + 3s + 2) = Xest (s2 1). Again because |est | = 0 for all s, s2 1 Y = 2 . (5) X s + 3s + 2 That is, for a given input Xest , we can nd the output Y est by scaling X by the ratio in Equation (5). This transfer function represents the impact of the system on any individual complex exponential. The denominator of the transfer function is the characteristic polynomial of the system. The systems eigenvalues are called poles because the transfer functions denominator is zero there, and so the transfer functions magnitude is like a tent being pulled toward the sky by each of its poles. So given a transfer function, we must make sure its denominator roots (i.e., poles) have negative real parts. For example, if x(t) e2t cos(3t), then the input can easily be expressed as a sum of complex exponentials using Eulers formula. That is, x(t) = e2t cos(3t) = e2t 1 1 ej 3t + ej 3t = e(2+j 3)t + e(2j 3)t , 2 2 2

which is a kind of generalized Fourier series of the signal. So the response to x(t) is the sum of (1/2)e(2+j 3)t and (1/2)e(2j 3)t scaled by the transfer function evaluated at s = (2 + j 3) and s = (2 j 3). Laplace Transforms and the Frequency Domain: All that is left is representing any input x as a sum or integral of complex exponentials. If x can be expressed this way, then mathematical theory states x(t) = 1 2j
+j

X (s)est ds
j

where

X (s) =
0

x(t)est dt.

(6)

That is, the Laplace transform X (s) of signal x provides the coecient for the est part of the signal, and we can scale X (s) by the transfer function to nd how much of est contributes to our output. Luckily, (i) Most inputs interesting to engineers do have Laplace transforms. (ii) Many common signals have well-documented Laplace transforms. (iii) Operations on Laplace transforms form other Laplace transforms. For example, the signal g (t) + h(t) has the Laplace transform G(s) + H (s). More importantly, the signal dg (t)/dt transforms to sG(s). (iv) As long as each of the poles of a real systems transfer function has a negative real part, the input X (s) and transfer function only need to be evaluated at s = j for all > 0. Because of item (iv), design work on lters focuses on the frequency (or Fourier) spectrum. That is, we think of the frequency content (e.g., fast changing or slow changing) of inputs and the ability of systems to shape that content (e.g., slows the signal or removes slow changes). Because of Eulers formula, we can think of sinusoids with angular frequency rather than ejt complex exponentials. Transfer functions then have a gain magnitude (e.g., from dissipative attenuation) and phase shift (e.g., from delay) for each .
Copyright 20072009 by Theodore P. Pavlic Creative Commons Attribution-Noncommercial 3.0 License

Page 2 of 4

ECE 209

Review of Circuits as LTI Systems

Laplace Representations of Circuit Elements

Resistors, capacitors, and inductors are important tools in electronic design. The following summary assumes zero initial conditions. + vC + vL + vR iC iL iR L C iC (t) = CvC (t) IC (s) = sCVC (s) ZC (s) vL (t) = Li L (t) R vR (t) = RiR (t) VR (s) = RIR (s)

VL (s) = sLIL (s)

1 VR (s) VC (s) VL (s) = = sL =R ZR (s) ZL (s) IC (s) sC IL (s) IR (s) The Laplace transformations of the voltage-to-current equations use the fact that derivatives in the time domain correspond to multiplication by s in the Laplace domain. The impedance of a circuit element is its voltage-to-current ratio at a given frequency. For example, assume that the voltage driving a capacitor is vC sin(2t). Then the current iC = Cv = 2C cos(2t). Hence, max{vC }/ max{iC } = 1/(2C ) which matches |ZC (j 2)| = |1/(j 2C )| = 1/(2C ). The larger the capacitance, the larger the current, and so bigger capacitors provide less impedance to high frequency current.

Series Impedance
Reducing a complicated network of impedances to a few equivalent impedances is a common circuit analysis task. Consider the equivalent impedance of n impedances in series. Apply the (Laplace-domain) loop rule. I I I Vin Vout Z1 Z2 Zn Vin Zseries Vout

The equivalent impedance Zseries is the total voltage drop divided by the series current. That is, Zseries (s) = I (s)Z1 (s) + I (s)Z2 (s) + + I (s)Zn (s) Vin (s) Vout (s) = = Z1 (s) + Z2 (s) + + Zn (s), I (s) I (s)

and so the equivalent series impedance is the simple sum of the constituent impedances.

Parallel Impedance
Next, consider the equivalent impedance of n impedances in parallel, which is often denoted Z1 Z2 Zn . Z1 Vin Z2 . . . Zn The equivalent impedance Zparallel is the voltage drop divided by the total current. That is, Zpar (s) = Vin (s) Vout (s)
Vin (s)Vout (s) Z1 (s) Vin (s)Vout (s) Z2 (s) Vin (s)Vout (s) Zn (s)

Vout

=
+ + =

Vin

Zpar

Vout

1 1 1 + + + Z1 (s) Z2 (s) Zn (s)

In the special but very common case of only two branches (i.e., n = 2), Z1 2 (s) Z1 (s) Z2 (s) = 1 1 + Z1 (s) Z2 (s)
1

Z2 (s) + Z1 (s) Z1 (s) Z2 (s)

Z1 (s) Z2 (s) . Z1 (s) + Z2 (s)

Copyright 20072009 by Theodore P. Pavlic Creative Commons Attribution-Noncommercial 3.0 License

Page 3 of 4

ECE 209

Review of Circuits as LTI Systems

The Voltage Divider

Most passive electronic circuits can be reduced to a voltage divider which splits a driving signal between two impedances.
Vin I ZB ZA I Vout

0V

The (Laplace-domain) current through the entire circuit is I (s) = and the output Vout (s) = I (s) ZB (s), and Vout (s) ZB (s) = . Vin (s) ZA (s) + ZB (s) So at each frequency , the output is the same proportion of the input as ZB is of the total ZA + ZB . However, the value of impedances ZA and ZB vary with frequency. Vin (s) , ZA (s) + ZB (s)

Example: An RC Low-Pass Filter

Vin R C Vout
1 1 Vout (s) sC = 1 = sRC + 1 Vin (s) R + sC

0V

The DC (i.e., constant) parts of Vin are passed directly through the lter and show up at the output. Faster components of the input induce displacement current within the capacitor, and so they get attenuated through the resistor. So Vout is a slower version of Vin .

First-Order Filters: Shortcut via Th evenin Equivalent Source

When an RC circuit contains only one capacitor and an output across that capacitor, its helpful to convert the rest of the circuit into its Th evenin equivalent; the result is a simple voltage divider. That is, 1. Remove the capacitor C from the system (i.e., replace it with an open circuit ). 2. Find the constant gain from input to output. Call it K . 3. Place a short circuit on the input and nd the equivalent resistance across the output. Call it Req . Then the transfer function will then be K sReq C + 1 For example,
Vin R1 R2 R3 + C Vout
3 Vout (s) R1 +R2 +R3 +R4 = Vin (s) s (R3 (R1 + R2 + R4 )) C + 1

(compare to RC divider low-pass example above).

=
R4 0V

R3 sR3 (R1 + R2 + R4 ) C + R1 + R2 + R3 + R4

Simplifying circuits this way (i.e., by using Th evenin equivalents) is useful in many situations.
Copyright 20072009 by Theodore P. Pavlic Creative Commons Attribution-Noncommercial 3.0 License

Page 4 of 4