Bill Kaune, W7IEQ

160 Cedarview Dr, Port Townsend, WA 98363: [email protected]

Quality Factor, Bandwidth,

and Harmonic Attenuation
of Pi Networks
The author looks at several definitions of Q, and makes some
interesting discoveries about what these calculations can
tell us about the design of pi networks..

I recently became interested in pi networks, a type of impedance- where f0 is the design frequency (in which the output and input
matching resonant network that provides harmonic attenuation impedances are matched), and dissipated power is the electrical
and is often used in tube-type Amateur Radio transmitters. I was energy dissipated per second, that is, converted into some other form
particularly interested in the relationship between the quality factor, of energy, such as thermal or radiation. (Dissipated power divided by
Q, and the bandwidth of these networks. In various editions of frequency is the energy dissipated per cycle.)
The ARRL Handbook for Radio Communications, I found three Equation 2 gives a second definition of Q.
definitions of Q, all different. None of the three were very good
predictors of the bandwidths of pi networks for most values of the f
load and source impedances. During these explorations, however, I Q= 0
discovered that a modification of one of the three provided a much BW [Eq 2]
better predictor of bandwidth. Subsequently, I was able to derive this
where BW is the 3 dB bandwidth, which is the width between the
modified form of Q theoretically.
upper and lower frequencies at which the response of the circuit is
While the bandwidth of pi networks is interesting, a perhaps more down 3 dB from its response at f0.
important characteristic of these networks is their ability to attenuate
Circuit-analysis textbooks show that these two definitions are
harmonics present in the signals passing through them. This paper
exactly equivalent for simple series and parallel RLC circuits. For
presents data on the harmonic attenuation of these networks as a circuits using more than two reactive components, such as the
function of their quality factors.
pi network, it is unclear (to me at least) that the definitions given
Finally, I investigated whether the methods used in this paper by Equations 1 and 2 are even approximately equivalent. This is
could be extended to more complex networks. For pi-L networks,
important because the Q of a more-complicated matching network is
they yield a value for Q that fairly accurately predicts bandwidth but most easily calculated using Equation 1, because this calculation can
not harmonic attenuation. They also yield quality factors for more only be performed at one frequency, whereas the bandwidth and the
complicated networks, but these values do not seem to have much
network’s ability to attenuate harmonics of the design frequency are
relationship with bandwidth. usually the more important parameters in the design process. That is,
it seems to me that Q is useful for practical design work in so much
Quality Factor of Impedance-Matching Networks as it provides information about the bandwidth and/or harmonic
Impedance-matching networks are characterized by, among other attenuation of a circuit.
things, their design frequency (the frequency at which the input and Wes Hayward, W7ZOI, in his ARRL book Introduction to Radio
output impedances are matched) and by the quality factor, Q. Quality Frequency Design has the following discussion of Q.1
factor is defined in two different ways. The first, and perhaps most “Q is not well defined for networks with three or more
common way, to define Q is given by Equation 1. reactive components. Still, Q is a frequently used parameter in
the design equations for more complex networks. The meaning
Average energy stored in reactive elements of Q is different when applied to such networks. It is the ratio of a
Q = 2π × f0 ×
Power dissapted by lossy elements
[Eq 1] 1
Notes appear on page 35

QEX September/October 2015 29

 resistance to a reactance when looking into one end of the network I first looked into the Amateur Radio literature to see what others
at one frequency. Network reduction methods are always used. had learned about Q and bandwidth for pi networks. I discovered a
The user should not deduce the bandwidth of the network by the 1983 paper by Elmer Wingfield, W5FD, titled “New and Improved
Q used for design.” Formulas for the Design of Pi and Pi-L Networks,” published in
After puzzling about this for some time, I decided to look into the QST.2 In this paper, the author notes that earlier ARRL publications
relationship between Q and bandwidth for a specific matching circuit, had defined the pi network quality factor as the expression given in
the pi network (which includes three reactive components). Equation 3.3
RS
Quality Factor and Bandwidth of Pi Networks
QI =
X C1 [Eq 3]
Figure 1 shows a pi-network circuit configuration often used to
match a load impedance, RL, to a source impedance, RS. A common where RS and |XC1| are the source resistance and the magnitude of the
application of pi networks is in the output circuits of tube-type power reactance of the input capacitance. (Note that capacitive reactances are
amplifiers, where relatively large source impedances have to be negative so, in this case, XC1 = – |XC1|.) Since this definition of Q refers
matched to relatively low antenna impedances. The configuration only to the input section of the pi network, I will refer to it as QI in
shown in Figure 1 forms a low-pass filter and provides substantial the remainder of this paper. Wingfield argued in his paper that a better
attenuation of harmonics of the design frequency. There is also a value for Q would be one that accounts for both the source and load
high-pass form of the pi network, but this configuration is not often resistances, and proposed that Q be defined by Equation 4.
used in Amateur Radio equipment, so I will concentrate solely on the RS R
low-pass form. QW = + L [Eq 4]
The transfer function of a pi network is defined as VL / VS, where X C1 X C2
VL is the voltage across the load resistor and VS is the source voltage.
Figure 2 is a typical transfer function for a pi network, as a function This expression, which I will refer to as QW, includes both input
of frequency. In this example, the design frequency was 1 MHz, the and output resistances and reactances.
source and load resistances were 500 and 50 W, respectively, the 3 dB The next documents I looked at were the 2003 and 2007 editions
bandwidth was 100 kHz, and the transfer function was normalized of The ARRL Handbook for Radio Communications, where the results
to 0 dB at the design frequency. Notice the peak that occurs in the reported by Wingfield were essentially reproduced.4, 5 I then looked at
transfer function at the design frequency: It is the 3 dB width of this the 2013 and 2015 editions of the Handbook, where I found a new,
peak that defines bandwidth. and slightly more complicated, definition for Q, given in Equation
5.6, 7
 RS
 if RS > RL
 XC1
QM = 
[Eq 5]
Rs
L
 RL
if R S ≤ RL
 XC 2
Vs C1 C2 RL I will call this definition QM because it refers to the latest, most
modern definition of Q that appears in The ARRL Handbook.
Finally, I am going to define a fourth expression, QBW, based on
QX1509-Kaune01
the actual bandwidth of the circuit, as calculated using Equation 2.
As discussed earlier, the Amateur Radio literature uses at least three
different definitions for the quality factor of a pi network. Are these
Figure 1 — This schematic diagram shows a low-pass pi network.
definitions equivalent, and if not, which is the best predictor of the
actual bandwidth of the circuit, that is, QBW?
To answer this question, I wrote a computer program (using Visual
Basic 6.0) that, for a given pi-network design, calculates QI, QW, and
0
QM, and also the transfer function from which the bandwidth and QBW
–10 can be determined. I then used this program to consider four cases, in
which RS = 5, 50, 500, and 5000 W, while RL = 50 W. In each case,
Transfer Function (dB)

–20
I selected values for XC1, XL, and XC2 so that the bandwidth of the
–30
resulting circuit was exactly 1/10 of the design frequency, that is, with
–40 QBW = 10. Table 1 shows my results. This table shows very clearly that
–50 the four definitions of Q are in substantial disagreement. For example,
–60 Q1 varies between 4.8 and 19.0 as the source resistance is varied from
5 to 5000 W, even though the bandwidth is held constant. QM, on
–70
the other hand, consistently indicates a smaller bandwidth than the
–80 actual value; when RL = 5000 W, the indicated bandwidth is 53% of
0.1 1 10 the actual value. Obviously, these two definitions of Q would be poor
Frequency (MHz) predictors of bandwidth (except when RS ≈ RL).
QX1509-Kaune02
The Wingfield value for Q remains consistently in the range 20.4
to 20.6 as the source resistance is varied. If we were to divide this
Figure 2 — Here is a plot of the transfer function (T) of a pi network. value by 2, we would obtain values nearly equal to Q defined by
The design frequency for this example was 1 MHz. T is normalized to bandwidth. This suggests that the definition for Q given by Equation
0 dB at the design frequency. 6 might be more useful for predicting the bandwidth of a pi network.

30 QEX September/October 2015

Table 1
Calculated Q Values and Harmonic Attenuations for Selected Source and Load Resistances, With Constant Bandwidth.
Q1 = RS / |XC1|, QW = RS / |XC1| + RL / |XC2|, QM = RS / |XC1| if RS ≥ RL or RL / |XC2| if RS < RL, and QBW = f0 / BW, where f0 = Design
Frequency and BW = 3 dB Bandwidth.
Resistance (W) Reactance (W)

RS RL XC1 XL XC2 Q1 QW QM QBW

5 50 1.03 4.19 3.21 4.8 20.4 15.6 10.0
50 50 4.91 9.73 4.91 10.2 20.4 10.2 10.0
500 50 32.1 41.9 10.3 15.6 20.4 15.6 10.0
5000 50 264 285 31.0 19.0 20.6 19.0 10.0

QW 1  RS R 

rapidly, falling to –8.6% when Qnew = 5 and –2.2% when Qnew = 10.
Qnew = =  + L 
[Eq 6] These results show that Qnew is a good measure of the true bandwidth
2 2  X C1 X C 2  of a pi network, certainly for Qnew ≥ 5. Qnew is a better indicator of
bandwidth than any of the three Q equations that appear in The ARRL
To test this hypothesis, I used the computer program described Handbook.
earlier to determine the error estimating bandwidth using Qnew defined Of course, it would be desirable to also have a confirmation, based
by Equation 6. I did this for a fixed load resistance of 50 W and source in theory, of this definition of Q [Equation 6]. In the next section, I
resistances of 5, 50, 500, and 5000 W. I found that the errors using present such a derivation.
Qnew to estimate bandwidth were reasonably similar across all the
load resistances, so I selected the maximum error across all source Theoretical Derivation of Qnew
resistances. The result is shown in Figure 3. Qnew is plotted on the Various textbooks on circuit analysis show (Norton’s Theorem)
horizontal axis and the vertical axis is the error using Qnew to estimate that a voltage source, characterized by an open-circuit voltage, VS,
bandwidth. The error estimating bandwidth using Qnew is about –31% and a series impedance, RS, can be replaced by a current source, IS,
for Qnew = 3. As Qnew is increased, the maximum error decreases shunted by an impedance RS, where IS = VS / RS. Thus, the pi network
drawn in Figure 4 is equivalent to the pi network in Figure 1. Next,
convert the parallel combination RS and XC1 in Figure 4 into a series
0 combination RS´ and XC1´, and similarly convert XC2 and RL into XC2´
and RL´, resulting in the circuit configuration shown in Figure 5. The
–5 relation between the original and converted quantities is given in
Equations 7 and 8.
Bandwidth Error (%)

–10
XC2 1 RS2
RS′ = RS and XC′ 1 = XC1 [Eq 7]
–15 RS2 + XC2 1 RS2 + XC2 1

–20 XC2 2 RL2

RL′ = RL 2 and X ′
C2 = X C2
[Eq 8]
RL + XC2 2 RL2 + XC2 2
–25

At first sight, Figure 5 appears to be a simple series RLC circuit,

–30
and, this is true so long as only one frequency is considered. Things
0 5 10 15 20 25 30 become more complicated, however, as frequency is varied, because
Qnew the resistances, RS´ and RL´, and the capacitances, C1´ and C2´ vary
QX1509-Kaune03 with frequency according to Equations 7 and 8. We can use Figure 5
as long as we restrict ourselves to the design frequency.
Figure 3 — This graph estimates the error in the predicted bandwidth
using the quantity Qnew, as defined in the text.

XL

X'C1 X'C2

Is
Is Rs C1 C2 RL R's R'L

QX1509-Kaune04
QX1509-Kaune05

Figure 4 — This schematic diagram shows a pi network with a Figure 5 — Here is a redrawn version of the pi network shown in
Norton Equivalent current source replacing the voltage source Figure 4, obtained by converting parallel resistor and capacitor
shown in Figure 1. combinations into equivalent series combinations.

QEX September/October 2015 31

 In a matched system, the power transferred to the load resistance, This is just the expression for Q that we earlier identified as Qnew in
RL´ in Figure 5, is maximized. It is easy to show (and is well known) Equation 6. We previously found that Qnew is a better predictor of the
that Equation 9 expresses the conditions for maximum power transfer. bandwidth of a pi network than other definitions of Q that have been
given in the Amateur Radio literature, and we have now shown that
X L + XC′ 1 + XC′ 2 =
0 there is theoretical plausibility for this form of Q.
[Eq 9]
RS′ = RL′
These two equations determine the values of two of the three Calculation of Components of a Pi Network
adjustable components (C1, L, and C2) in a pi network. We are thus Now I will show how to calculate XC1 given a value of Q. Equation
free to choose the value of the remaining component. (Ultimately, 16 can be rewritten as Equation 17.
we will use a specification of Q to determine the value of this 2
  2
component.) Thus, for the time being, assume that a value for XC1  2Q − RS  = RL2 [Eq 17]
has been specified. We will determine XL and XC2 as functions of XC1.  X C1  X C2
 
Use the second equation in each set shown as Equations 7, 8 and 9.
Use Equation 11 to eliminate X2C2 in the denominator on the right-
X2 X2 hand side of Equation 17. This gives an equation that involves only RS,
RS′ = RL′ ⇒ RS 2 C 1 2 = R L 2 C 2 2
RS + X C 1 RL + X C 2 RL, XC1, and Q. Solving for XC1 gives the result shown in Equation 18.

[Eq 10]  2QR + 4Q 2 R R − R − R 2 

( S L)
XC1 = − RS   [Eq 18]
S S L

Solve this equation for XC2, yielding Equation 11. 


2
(
RS 4Q + 1 − RL ) 

RS X C21 [Eq 11] There is actually a second solution, obtained by replacing the plus
X C 2 = − RL
RL RS2 + (RL − RS )X C21 sign in front of the square root symbol with a minus sign. I have been
able to show, however, that this solution always leads to values of XC2
Then, the first equation in each set shown as Equations 7, 8 and 9 can and/or XL that have the wrong sign.
be combined to yield Equation 12 for XL. Once XC1 is determined, the values of XC2 and XL can be calculated
using Equations 11 and 12. In order for XC1 to be real and negative,
RS2 RL2
X L = −( X C′ 1 + X C′ 2 ) = − X C1
the quantity in Equation 18 under the square root sign must not be
− X C2 negative and the denominator must be positive. In addition, for XC2,
RS2 + X C21 RL2 + X C2 2
defined by Equation 11, to be a real number, the denominator of
[Eq 12] the fraction under the square root sign must be a positive number.
Analysis shows that these three conditions are met if the following
What is the Q of the circuit in Figure 5? Treating it as a simple inequalities are satisfied
series RLC circuit with total resistance RS´+ RL´ and inductive 1 RS
reactance XL, we get Equation 13. Q> − 1 if RS > RL
2 RL
XL
Q= [Eq 13] [Eq 19]
RS′ + RL′
1 RL
Then, using Equations 7, 8, and 12, we get Equation 14. Q> − 1 if RS < RL
2 RS
RS2 RL2
− X C1 − X C2
RS2 + X C21 RL2 + X C2 2 [Eq 14]
Q= Harmonic Attenuation of Pi Networks
X2 X2 Figure 2 shows that pi networks provide substantial attenuation of
RS 2 C 1 2 + R L 2 C 2 2
RS + X C 1 RL + X C 2 the harmonics of the design frequency. Indeed, impedance matching
and harmonic attenuation are the usual reasons for employing a pi
Equation 10 can be rearranged to yield the identity shown as network. I used the computer program described earlier to calculate
Equation 15. the attenuation of the second through tenth harmonics of the design
[Eq 15] frequency of the network. The results for the second, third, and fourth
RL2 RL X C21 RS2 harmonics are shown graphically in Figure 6 for values of Q defined
= by Equation 6, ranging from 1 to 20, and for source impedances of
RL2 + X C2 2 X C2 2 RS RS2 + X C21 50, 500, and 5000 W. Because of the reciprocal nature of pi networks,
harmonic attenuation for a network with a given ratio of source to
load resistances will be the same as network whose ratio of source
Placing this expression into Equation 14, we obtain Equation 16. to load resistances is the inverse of the first. Thus, for example, the
RL X C21 harmonic attenuation of a network with source and load resistances
− X C1 − of, say, 500 and 50 W, respectively, will be the same as that of a
RS X C 2 1  RS R 
 [Eq 16]
Q= 2
=  + L 
network with source and load resistances of 5 and 50 W, respectively.
X 2  X C1 X C 2  The data in Figure 6 show that harmonic attenuation increases as
2 C1 Q increases, but also that the rate of this increase decreases for larger
RS values of Q. For a Q of 10, the second, third, and fourth harmonics

32 QEX September/October 2015

 Second Harmonic Third Harmonic Fourth Harmonic
0 0 0
RS = 50
–10 –10 –10
RS = 500 RS = 50
–20 –20 –20
RS = 500 RS = 50
Attenuation (dB)

Attenuation (dB)

Attenuation (dB)
RS = 5000
–30 –30 –30
RS = 500
RS = 5000
–40 –40 –40
RS = 5000

–50 –50 –50

–60 –60 –60

–70 –70 –70

0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Q Q Q
QX1509-Kaune06

Figure 6 — These graphs plot the attenuation of second, third and fourth harmonics by a pi network as a function of Q. Data are shown for
load resistance of 50 W and source resistances, RS, of 50, 500, and 5000 W.

Table 2 equivalent series pair that I was able to make progress. In this way, I
Fitted Values for Parameters in Equation AN = a0 + a1Q + a2Q 2 was able to deduce a value for the Q of a pi network that effectively
predicted bandwidth and harmonic attenuation, at least for larger
Harmonic a0 (dB) a1 (dB) a2 (dB) Accuracy (dB) values of Q. I began to wonder if the same approach would work for
2 –22.4 –1.58 0.0316 ±0.4 more complicated networks.
3 –34.0 –1.62 0.0328 ±0.5 My first effort was to add a second inductor between the load
4 –41.8 –1.64 0.0333 ±0.5 resistor and the top of capacitor C2 in Figure 1 to form a pi-L network.
By dividing the capacitance C2 into two parallel capacitors, C2A and
C2B, the pi-L network can be transformed into a cascaded pair of
networks, a pi network followed by a L network; this is illustrated in
Figure 7. Furthermore, by selecting C2B appropriately, the impedance
L1 L2
RS
RV looking into the input of the L network can be made a pure resistance,
RV; this “virtual” resistance is drawn in dotted lines in Figure 7 and is
the load resistance of the pi network and the source resistance for the
L network. Analysis of the L network is simple, and we can use the
Vs C1 C2A C2B RL
results earlier in this paper for the pi network.
Let Q2 be the quality factor for the L network. By transforming the
series pair L2 and RL into an equivalent parallel pair, the L network is
QX1509-Kaune07 transformed into a simple parallel RLC network, from which we can
show that the reactances of L2 and C2B, and the resistance of RV are
Figure 7 — Here, the pi-L network is divided into a pi network given by the expressions of Equation 20.
followed by an L network. The resistance looking into the input of the
L network is RV; this is also the load resistance for the pi network.  4Q 2 + 1 
X L 2 = 2Q2 RL , X C 2 B = − 2  RL , and
 2Q2 
of the design frequency are attenuated by about –35 dB, –47 dB, and RV = (4Q22 + 1)RL [Eq 20]
–54 dB, respectively, and are nearly the same for all three source
resistances. The data for Q ≥ 7.5 are described with an accuracy of
±0.5 dB or better by the form AN = a0 + a1Q + a2Q2, where AN is the
attenuation of the Nth harmonic, and values for the constants a0, a1, and Next, let Q1 be the quality factor for the pi network. Equations 11,
a2 are listed in Table 2. 12 and 18, with L, C2, and RL replaced by L1, C2A, and RV, respectively,
can then be used to calculate XC1, XC2A, and XL1. XC2 can then be
calculated by combining XC2A and XC2B in parallel.
More Complex Networks What, then, is the quality factor, Q, for the composite network?
When I started this investigation, I had little idea how to simply According to Equation 1, it is the sum of the energies stored in the
analyze a pi network, and my first efforts led to quite complicated pi and in the pi-L networks, divided by the power dissipated in the
equations that were not very illuminating. It was only when I happened source and load resistors (assuming losses in C1, C2, L1, and L2are
on the idea of transforming parallel resistances and reactances into an small enough to be neglected).

QEX September/October 2015 33

 is 10, so the predicted bandwidth would be 100 kHz, substantially
0
less than the actual value. Each pi network, taken alone, would have
–10 had a single peak at 1 MHz, but the two together yield two peaks.
Evidently, there are interactions between the two pi networks as the
–20 frequency is varied away from 1 MHz. From this and other examples
Transfer Function (dB)

–30
I have worked out, I conclude that, in general, Q is no longer related
in any simple way to the network bandwidth and level of harmonic
–40 attenuation. This is consistent with the comments of Wes Hayward
quoted earlier in this paper.
–50

–60
Discussion and Conclusions
–70 This article has found that three existing definitions for the quality
factor of pi networks found in the Amateur Radio literature are not
–80 good predictors of the bandwidths (and harmonic attenuations) of
0.1 0.1 10 these circuits. Both empirical and theoretical analysis suggest that a
Frequency (MHz) better definition of Q, at least for predicting bandwidth and harmonic
QX1509-Kaune08 attenuation, is given by Equation 24.

1 R R 
Figure 8 — This graph shows the transfer function for a cascade of Q =  S + L 

[Eq 24]
two pi networks. 2  X C1 X C 2 
where RS and RL are the source and load resistances, respectively, and
XC1 and XC2 are the input and output capacitive reactances of the pi
S1 + S 2 S1 S1 [Eq 20] network, respectively.
Q = 2π f 0 = 2π f 0 + 2π f 0 It appears to me that the methods employed in this paper can be
PS + PL PS + PL PS + PL extended to all networks containing three reactive components, with
good results.
where S1 and S2 are the average energies stored in the two networks, I was able to extend my analysis to pi-L networks with good
and PS and PL are the powers dissipated in the source and load results for bandwidth prediction but not harmonic attenuation.
resistances, respectively. Now look at each network separately. Analysis of networks more complex than pi-L networks yielded
Network 1, the pi network, has as its load resistance RV, which values for Q that were not predictive of bandwidth; I am not sure that
dissipates a power PV. Then we obtain Equation 21. Q has much significance for these more complex networks.
S1 There are, of course, limitations to what is presented in this paper.
Q1 = 2π f 0 [Eq 21] The two main limitations are:
PS + PV 1) My analysis assumes that the reactive elements are lossless, but
For network 2, the L network, we obtain Equation 22. all real inductors and, to a lesser extent, capacitors have loss.
2) The source and load resistances are assumed to be constant
S2 independent of frequency, but this is probably seldom the case. For
Q2 = 2 π f 0 [Eq 22] example, pi networks are often used in power amplifiers to match
PV + PL the output impedance of the amplifier to an antenna, and antenna
Since RV is matched to RS by network 1, and RL is matched to RV impedances vary with frequency.
by network 2, the powers dissipated in all three resistances are equal. It would be well in any actual design to check the performance
With this result, we have proven Equation 23. of circuits that include pi networks using one of the modern
sophisticated computer circuit modeling programs, such as SPICE.
Q = Q1 + Q2 [Eq 23] During the process of the work reported here, I could not perform
an exhaustive search of the technical literature on pi networks. The
material in this paper is new to me, but I would not be surprised to find
I modified the computer program described earlier to model pi-L a paper somewhere that made similar remarks to the ones here. I hope
networks and found that Q calculated in this way is a good predictor the results are of interest to hams and other electronics experimenters.
of bandwidth (at least for larger values of Q). Q was not a good Finally, I would like to express my appreciation to the hams that
measure of harmonic attenuation, however, which varied significantly reviewed an earlier version of this paper. They identified several
as the source resistance was varied. errors and made very helpful suggestions.
I have explored even more complicated networks, such as two
cascaded pi networks using the techniques described in this paper.
Bill Kaune, W7IEQ, is a retired physicist (BS, PhD). He is married
With two pi networks, there are three independent parameters that
and has two grown daughters and four grandchildren. Bill spent most of
have to be specified in order to uniquely define the network. For his career collaborating with biologists and epidemiologists researching
these three I chose Q1 and Q2, the quality factors of the individual pi the biological effects of power-frequency electric and magnetic fields.
networks, and RV, the “virtual” resistance looking into the input of the Along with Amateur Radio, Bill spends his time hiking, backpacking,
second network. Figure 8 shows the transfer function for one network and doing some volunteer work. Bill was first licensed in 1956 as a
where RL = 50 W, RS = 1800 W, RV = 300 W, Q1 = Q2 = 5, and with a novice and then a general, but became inactive while in college. He
design frequency of 1 MHz. was licensed again in 1998 and upgraded to the Amateur Extra class in
Note that there are now two peaks rather than one in the transfer 2000. Bill is a member of the Jefferson County Amateur Radio Club and
function, leading to a considerably broadened response. In fact, the the ARRL.
actual 3 dB bandwidth is 342 kHz. The overall Q of this network

34 QEX September/October 2015

Notes
1
Wes Hayward, W7ZOI, Rick Campbell,
KK7B, and Bob Larkin, W7PUA,
Introduction to Radio Frequency Design,
ARRL, Newington CT, 1996, p 139.
2
Elmer A Wingfield, “New and Improved
Formulas for the Design of Pi and Pi-L
Networks,” QST, August 1983, pp 23 – 29.
3
ARRL, The Radio Amateur’s Handbook, The
American Radio Relay League, Inc., The
Rumford Press, Concord, NH, p 51.
4
Dana Reed, W1LC, Ed, The 2003 ARRL
Handbook for Radio Communications,
ARRL, 2002, Newington, CT, pp 13.6 –
13.7.
5
Mark Wilson, K1RO, Ed, The 2007ARRL
Handbook for Radio Communications,
ARRL, 2006, Newington, CT, pp 18.6 –
18.7.
6
H. Ward Silver, NØAX, Ed, The 2013ARRL
Handbook for Radio Communications,
ARRL, 2012, Newington, CT, pp 5.25 –
5.26.
7
H. Ward Silver, NØAX, Ed, The 2015 ARRL
Handbook for Radio Communications,
ARRL, 2014, Newington, CT, pp 5.25 –
5.26.

QEX September/October 2015 35