ECE 304: Open- and Short-circuit Time Constants

See S&S, pp. 575-578, 497-503

Example Circuit
OUT
+ + + +
{C_1} {C_2} {R_3} {R_4}
VDB
+ V1
AC
Sweep 1V
{R_1} {R_2}

+
- {C_3} {C_4}

+
0
PARAMETERS:

R_1 = {R_L} C_1 = {C_LF} C_LF = 1uF pi = 3.1415926

R_2 = {R_H} C_2 = {C_LF} C_HF = 1pF
R_3 = {R_L} C_3 = {C_HF} R_H = {1E3/(2*pi)}
R_4 = {R_H} C_4 = {C_HF} R_L = {10/(2*pi)}
FIGURE 1
An RC-circuit that is a midband filter.
The first two sections in Figure 1 are high-pass filters and the last two are low-pass filters. The
factor 2π is introduced to make the frequencies easier to estimate.

(10.000M,-921.072u)

-50
(102.000K,-2.9688) (980.400M,-2.9686)

(980.390,-43.313) (102.000G,-43.312)
-100
100Hz 10KHz 1.0MHz 100MHz 10GHz 1.0THz
VDB(OUT) dB( 9.804E+08/Frequency) dB(9.804E+08/1.02E+11)+2*dB( 1.02E+11/Frequency)
dB( Frequency/1.0200E+05) 2*dB( Frequency/980.3921569)+dB(980.3921569/1.0200E+05)
Frequency
FIGURE 2
Bode magnitude plot for circuit of Figure 1. Midband region lies between the lower 3dB-
corner frequency of 102 kHz and the upper 3dB-corner frequency of 980 MHz; midband
gain is approximately zero dB

180d

(Midband,10.000M,0.000)
(980.392,134.724)
0d
(102.000K,45.272)
(980.400M,-45.270)
(102.000G,-134.722)
-190d
100Hz 10KHz 1.0MHz 100MHz 10GHz 1.0THz
P(V(OUT))
Frequency
FIGURE 3
Bode phase plot with phases marked at the pole frequencies for each section. The phases
are slightly affected by the nearness (proximity) of other poles. For isolated poles, the
phases at the pole frequencies taking midband as 0° are 135°, 45°, −45°, and −135°

Unpublished work ©3/7/2005 J R Brews Page 1 3/7/2005

Midband Analysis
At midband, all low-frequency capacitors are short-circuits and all high-frequency capacitors are
open circuits. That is, C1 and C2 are shorts, and C3, C4 are opens, making the gain VOUT/VIN = 1
V/V or 0 dB.
High-frequency analysis
At high frequencies, near or above the upper corner frequency of 980MHz, we can approximate
the two low-frequency capacitors C1 and C2 by short circuits. The result is shown in Figure 4: the
high-frequency gain of the two circuits is the same, but the low-frequency behavior is not,
because C1 and C2 are not present in the approximate circuit.

LF caps short-circuited
-50

Full circuit

-100
100Hz 10KHz 1.0MHz 100MHz 10GHz 1.0THz
DB(V(OUT)) DB(V(HF))
Frequency
FIGURE 4
Comparison of high-frequency approximation to circuit (capacitors C1 and C2 replaced by
short circuits) and the original circuit of Figure 1
+ + HF
{R_3} {R_4} VDB
+ V3
AC
Sweep 1V
+

-
{C_3} {C_4}
0
FIGURE 5
The approximate circuit valid at high-frequency
Circuit analysis of Figure 5 shows the gain to be
EQ. 1
VOUT 1
= .
VS 1 + jω[C 4 (R 3 + R 4 ) + C 3 R 3 ] + ( jω) 2 C 3 C 4 R 3 R 4

The dominant pole approximation

The denominator of EQ. 1 is quadratic in ω, which means there are two poles in the gain function
and two high-frequency break points. In Figure 3, these occur at 980 MHs and 102GHz, for this
example. In the case that these two poles are well separated (the dominant pole case) the linear
term in ω determines the corner frequency that marks the end of the midband region.
If the term linear in ω is examined more carefully, it is composed of two RC time
constants, one related to each of the two high-frequency capacitors. For example, the term
C4(R3+R4) uses the resistance seen between the nodes where C4 is attached if the applied signal
source is shorted out. This resistance can be found by putting a test current or voltage in place of
C4, shorting Vsh and open-circuiting C3 (see Figure 6)

Unpublished work ©3/7/2005 J R Brews Page 2 3/7/2005

 160.746V
+ +
{R_3} {R_4}

1A

+
{R_1} {R_2}
1.00000A
0
FIGURE 6
Circuit for finding the Thevenin resistance (R3+R4) = 160.7 Ω seen by C4
In Figure 6 a test current was used so we can read the resistance seen by C4 simply as the
voltage at the top of the test source. Likewise, the term C3R3 is found using Figure 7.
1.59155V

+ +
{R_3} {R_4}

1A 0A
+

{R_1} {R_2}
1.00000A 0A
0
FIGURE 7
Circuit for finding the Thevenin resistance R3 = 1.59 Ω seen by C3
In Figure 7 a zero-current source is added at the end of R4 because PSPICE does not allow
dangling resistors. The zero-amp current source tells PSPICE that there is no current in R4, as is
appropriate when the circuit ends with an open circuit because C4 has been replaced by an open
circuit.
The general approach to finding the high-frequency corner is then as follows:
1. Draw the small-signal circuit
2. Short all the low-frequency capacitors
3. Short all independent AC voltage sources and open all independent AC current sources
4. Select a particular high -frequency capacitor. Replace all the others by open circuits.
5. Put a test voltage at the site of the selected capacitor and find the resistance seen at this
site.
6. Multiply this resistance by the value of the selected capacitance → τ1, say
7. Do the same for all the other high-frequency capacitances.
8. The high-frequency corner frequency is fC = 1/(2π) × 1/(τ1 + τ2 + τ3 + …)

Low-frequency analysis

LF
+ +
{C_1} {C_2}
+
AC
V2
Sweep

-
1V {R_1} {R_2}
+

0
FIGURE 8
The approximate low-frequency circuit

Unpublished work ©3/7/2005 J R Brews Page 3 3/7/2005

 0

HF caps open-circuited
-50
Full circuit

-100
100Hz 10KHz 1.0MHz 100MHz 10GHz 1.0THz
DB(V(OUT)) DB(V(LF))
Frequency
FIGURE 9
Bode gain plots comparing the low-frequency circuit of Figure 8 with the complete circuit
of Figure 1
Figure 9 shows the low-frequency circuit and the original circuit agree at low frequencies, but not
at high frequencies, which is the reverse to Figure 4. Circuit analysis of Figure 8 shows the
voltage gain is given by
EQ. 2
VOUT 1
=
VS 1  1 1  1  1 
1+  +  +  
jω  C1 (R 1 // R 2 ) C 2 R 2  ( jω) 2 C C R R
 1 2 1 2

The dominant pole approximation

In complete analogy with the high-frequency case, the denominator of EQ. 1 is quadratic in ω,
which means there are two poles in the gain function and two low-frequency break points. In
Figure 3, these occur at 980 Hz and 102 kHz, for this example. In the case that these two poles
are well separated (the dominant pole case) the linear term in 1/jω determines the corner
frequency that marks the start of the midband region.

Looking at the linear term, it is composed of two RC time constants, one related to each of the
two low-frequency capacitors. For example, the term C1(R1//R2) uses the resistance seen
between the nodes where C1 is attached if the applied signal source is shorted out. This
resistance can be found by putting a test voltage or current in place of C1, shorting Vsh and short-
circuiting C2 (see Figure 10)
1.00000A 1.57579V R15 R16
+ +

I_X 1A
{R_3} {R_4}
R13 R14
0A
+

{R_1} {R_2}
0A
0
FIGURE 10
Circuit for finding the resistance R1//R2 = 1.58 Ω seen from the site of C1
Likewise, the resistance seen by C2 is found using Figure 11.

Unpublished work ©3/7/2005 J R Brews Page 4 3/7/2005

 1.00000A 159.155V
+ +

IX 1A {R_3} {R_4}

0A

+
{R_1} {R_2}
0A
0
FIGURE 11
Circuit for finding the resistance R2 =159.2 Ω seen from the site of C2
The general approach to finding the low-frequency corner is then as follows:
1. Draw the small-signal circuit
2. Replace all the high-frequency capacitors with open circuits
3. Short all independent AC voltage sources and open all independent AC current sources
4. Select a particular low -frequency capacitor. Replace all the others by short circuits.
5. Put a test voltage at the site of the selected capacitor and find the resistance seen at this
site.
6. Multiply this resistance by the value of the selected capacitance → τ1, say
7. Do the same for all the other low-frequency capacitances.
8. The low-frequency corner frequency is fC = 1/(2π) × (1/τ1 + 1/τ2 + 1/τ3 + …)

Finding the other pole

We've discussed finding the high- and low-frequency corner frequencies using the dominant pole
approximation. How can the higher pole at 102 GHz (or the lower pole at 980 Hz) be found? A
technique based on the dominant pole assumption is described next (see S&S, pp. 852-853).
The case of the high frequency pole will be discussed here. The low frequency case is
similar.
We suppose that only two poles matter, so the denominator of the high-frequency gain
expression is of the form
EQ. 3
D(jω) = (1 + jωτ1)(1+jωτ2) = 1 + jω (τ1+τ2) +(jω)2τ1τ2

In general, we won't have factored this expression so neatly; and we will have instead of EQ. 3
EQ. 4
D(jω) = 1 + jω a1 +(jω)2 a2

If one time constant, say τ1, is much bigger than the other, we can make the approximation

EQ. 5
τ1 ≈ a1,

which means we can simply read τ1 as the coefficient of the linear term in jω. Having made this
approximation, the second time constant is simply

EQ. 6
τ2 = τ1τ2/τ2 = a2/τ2 ≈ a2/a1,

that is, the ratio of the coefficients. We can check after doing this to see if the assumption that τ1
>> τ2 is accurate. How does this approach differ from a more exact approach? What is a more
exact approach? Suppose we set s = jω. Then, like EQ. 4,
EQ. 7
D(s) = 1 + s a1 +s2 a2,

which can be written like EQ. 3 in terms of its zeros (say, s= −1/τ1 or s = −1/τ2)

Unpublished work ©3/7/2005 J R Brews Page 5 3/7/2005

EQ. 8
D(s) = (s+1/τ1)(s+1/τ2)τ1τ2 = 1 + s(τ1+τ2) + s2(τ1τ2)

Using the quadratic formula, the zeros of D(s) (the poles of 1/D(s)) are given by

− a1 ± a12 − 4a 2 − ( τ1 + τ 2 ) ± ( τ1 + τ 2 ) 2 − 4τ1τ 2  
− ( τ1 + τ 2 )  ( τ1 − τ 2 ) 2 
s= = =  1± 
2a 2 2τ1τ 2 2τ1τ 2  ( τ1 + τ 2 ) 2 
 

If τ1>>τ2, the square root is nearly 1. For example, suppose it is 0.99. Then the bracketed term is
either 1 + 0.99 ≈ 2 or 1−0.99 = 0.01. In this last case, unless we have very good accuracy, the
round-off error in this estimate for the shortest time constant will be large. Under these
circumstances, EQ. 6 is more accurate than the quadratic formula. However, if τ1 and τ2 are
close in value, the quadratic formula is better.

Unpublished work ©3/7/2005 J R Brews Page 6 3/7/2005