MICROWAVE MICROWAVE FILTERSFILTERS

A microwave filter is a two-port network used to control the frequency response at a certain point in a microwave system by providing transmission at frequencies within the passband of the filter and attenuation in the stopband of the filter. Typical frequency responses include low-pass, high-pass, bandpass, and band-reject characteristics. Applications can be found in virtually any type of microwave communication, radar, or test and measurement system. Microwave filter theory and practice began in the years preceding World War II, by pioneers such as Mason, Sykes, Darlington, Fano, Lawson, and Richards. The image parameter method of filter design was developed in the late 1930s and was useful for low-frequency filters in radio and telephony. In the early 1950s a group at Stanford Research Institute, consisting of G. Matthaei, L. Young, E. Jones, S. Cohn, and others, became very active in filter and coupler development. Today, most microwave filter design is done with sophisticated computer-aided design (CAD) packages based on the insertion loss method. Because of continuing advancements in network synthesis with distributed elements, the use of low-temperature superconductors, and the incorporation of active devices in filter circuits, microwave filter design remains an active research area.We begin our discussion of filter theory and design with the frequency characteristics of periodic structures, which consist of a transmission line or waveguide periodically loaded with reactive elements. These structures are of interest in themselves, because of the application to slow-wave components and traveling-wave amplifier design, and also because they exhibit basic passband-stopband responses that lead to the image parameter method of filter design.Filters designed using the image parameter method consist of a cascade of simpler two-port filter sections to provide the desired cutoff frequencies and attenuation characteristics, but do not allow the specification of a frequency response over the complete operating range. Thus, although the procedure is relatively simple, the design of filters by the image parameter method often must be iterated many times to achieve the desired results.A more modem procedure, called the insertion loss method, uses network synthesis techniques to design filters with a completely specified frequency response. The design is simplified by beginning with low-pass filter prototypes that are normalized in terms of impedance and frequency. Transformations are then applied to convert the prototype designs to the desired frequency range and impedance level. Both the image parameter and insertion loss method of filter design provide lumped element circuits. For microwave applications such designs usually must be modified to use distributed elements consisting of transmission line sections. The Richard's transformation and the Kuroda identities provide this step. We will also discuss

transmission line filters using stepped impedances and coupled lines; filters using coupled resonators will also be briefly described.

PERIODIC STRUCTUREAn infinite transmission line or waveguide periodically loaded with reactive elements is referred to as a periodic structure. As shown in Figure 1, periodic

structures can take various forms, depending on the transmission line media being used. Often the loading elements are formed as discontinuities in the line, but in any case they can be modeled as lumped reactance across a transmission line as shown in Figure 2. Periodic structures support slow-wave propagation (slower than the phase velocity of the unloaded line), and have passband and stopband characteristics similar to those of filters; they find application in traveling-wave tubes, masers, phase shifters, and antennas.

FIGURE 1 Examples of periodic structures. (a) Periodic stubs on a microstrip line. (b) Periodic diaphragms in a waveguide.

Analysis of Infinite Periodic StructureWe begin by studying the propagation characteristics of the infinite loaded line shown in Figure 2. Each unit cell of this line consists of a length d of transmission line with a shunt susceptance across the midpoint of the line; the susceptance b is normalized to the characteristic impedance, Zo. If we consider the infinite line as being composed of a cascade of identical two-port networks, we can relate the voltages and currents on either side of the nth unit cell using the· ABCD matrix:

VnIn=ABCDVn+1In+1where A, B, C, and D are the matrix parameters for a cascade of a transmission line section of length d/2, a shunt susceptance b, and another transmission line section of length d/2. we then have, in normalized form,ABCD=cosθ2jsinθ2jsinθ2cosθ210jb1cosθ2jsinθ2jsinθ2cosθ2

=cosθ-b2sinθjsinθ+b2cosθ-b2jsinθ+b2cosθ+b2cosθ-b2sinθ

where θ = kd, and 'k is the propagation constant of the unloaded line. The reader can verify that AD - BC = 1, as required for reciprocal networks.

1 1

2

Now for any wave propagating in the +z direction, we must haveV(z) = V(0)e-γz

I(z) = I(0)e-γz

For a phase reference at z = 0. Since the structure is infinitely long, the voltage and current at the nth terminals can differ from the voltage and current at the n+1 terminals.

only by the propagation factor, e-γd. Thus,Vn+1 = Vne-γd

In+1 = Ine-γd

Using this result in (1) gives the following:VnIn= ABCDVn+1In+1= Vn+1eγdIn+1eγd

,

A-eγdBCD-eγdVn+1In+1= 0

For a nontrivial solution, the determinant of above matrix must vanish:

AD+e2γd-(A+D)eγd-BC=0,

or, since AD - BC = 1,

1+e2γd-(A+D)eγd=0,

e-γd+eγd=A+D,

coshγd=A+D2=cosθ-b2sinθ

Where (2) was used for the values of A and D. Now if γ = α+jβ, we have that:

coshγd=coshαdcosβd=cosθ-b2sinθ.Case 1: α=0, β≠0. This case corresponds to a n6nattenuating, propagating wave on the periodic structure, and defines the passband of the structure. Then (8) reduces to

23b3a

4a4b

5

6

7

8

cosβd=cosθ-b2 sinθ,

Which can be solved for β if the magnitude of the right-hand side is less than or equal to unity. Note that there are an infinite number of values of β that can satisfy (9).

Case 2: α≠0, β=0, π In this case the wave does not propagate, but is attenuated along the line; this defines the stopband of the structure. Because the line is lossless, power is not dissipated, but is reflected back to the input of the line. The magnitude of (8) reduces to

coshαd=cosθ-b2sinθ≥1,Which has only one solution (α > 0) for positively traveling waves; α< 0 applies for negatively traveling waves. If cosθ - (bI2) sinθ ≤-1, (9b) is obtained from (8) by letting β = π; then all the lumped loads on the line are λ/2 apart, yielding an input impedance the same as if β=0.Thus, depending on the frequency and normalized susceptance values, the periodically loaded line will exhibit either passbands or stopbands, and so can be considered as a type of filter. It is important to note that the voltage and current waves defined in (3) and (4) are meaningful only when measured at the terminals of the unit cells, and do not apply to voltages and currents that may exist at points within a unit cell. These waves are sometimes referred to as Bloch waves because of their similarity to the elastic waves that propagate through periodic crystal lattices. Besides the propagation constant of the waves on the periodically loaded line, we will also be interested in the characteristic impedance for these waves. We can define characteristic impedance at the unit cell terminals as

ZB=Z0Vn+1In+1 ,Since Vn+1 and In+1 in the above derivation were normalized quantities. This impedance is also referred to as the Bloch impedance. From (5) we have that

(A-eγd)Vn+1+BIn+1 + 0,So (10) yields,ZB= -BZ0A-eγdFrom (6) we can solve for eγd in terms of A and D as follows:

eγd=(A+D)±(A+D)2-42

Then the bloch impedance has two solutions given by

9a

9b

10

ZB±=-2BZ02A-A-D (A+D)2-4∓

For symmetrical unit cells (as assumed in Figure (2)) we will always have A = D. In this case (11) reduces to

ZB±=±BZ0A2-1

The ± solutions correspond to the characteristic impedance for positively and negatively traveling waves, respectively. For symmetrical networks these impedances are the same except for the sign; the characteristic impedance for a negatively traveling wave turns out to be negative because we have defined In in Figure 2 as always being in the positive direction.From (2) we see that B is always purely imaginary. If α = 0, β≠0 (passband), then (7) shows that cosh γd = A ≤1 (for symmetrical networks) and (12) shows that ZB will be real. If α≠0, β = 0 (stopband), then (8.7) shows that cosh γd = A ≥1, and (12) shows that ZB is imaginary. This situation is similar to that for the wave impedance of a waveguide, which is real for propagating modes and imaginary for cutoff, or evanescent, modes.

Terminated Periodic StructuresNext consider a truncated periodic structure, terminated in a load impedance Z L, as shown in Figure 3. At the terminals of an arbitrary unit cell, the incident and reflected voltages and currents can be written as (assuming operation in the passband)Vn=V0+e-jβnd+V0-ejβnd,

In=I0+e-jβnd+I0-ejβnd=Vn=V0+ZB+e-jβnd+V0-ZB-ejβnd,

where we have replaced γz in (3) with jβnd, since we are interested only in terminal quantities.Now define the following incident and reflected voltages at the nth unit cell:

Vn+=V0+e-jβnd,

Vn+=V0-ejβnd

Then (13) can be written as:

11

12

13a

13b

14b

14a

Vn= Vn++Vn-,

In= Vn+ZB++Vn-ZB-

At the load, where n=N, we have

VN= VN++VN-=ZLIN=ZLVN+ZB++VN-ZB-

So the reflection coefficient at the load can be found as

Γ=VN-VN+=-ZLZB+-1ZLZB--1

If the unit cell network is symmetrical (A = D), then Z+B = -Z-

B = ZB, which reduces (17) to the familiar result that

Γ=ZL-ZBZL+ZB

FIGURE 3: A Periodic Structure in a Normalized Load Impedance ZL

So to avoid reflections on the terminated periodic structure, we must have ZL = ZB, which is real for a loss less structure operating in a passband. If necessary, a quarter-wave transformer can be used between the periodically loaded line and the load.

k-β Diagrams And Wave VelocitiesWhen studying the passband and stopband characteristics of a periodic structure, it is useful to plot the propagation constant, β, versus the propagation constant of the unloaded line, k (or w). Such a graph is called a k-β diagram or Brillouin diagram (after L. Brillouin, a physicist who studied wave propagation in periodic crystal structures). The k-β diagram can be plotted from (8.9a), which is the dispersion relation for a general periodic structure. In fact, a k-β diagram can be used to study the dispersion characteristics of many types of microwave components and transmission lines. For instance, consider the dispersion relation for a waveguide mode:β=k2-kc2

15b

15a

16

17

18

k=β2-kc2

where kc is the cutoff wave number of the mode, k is the free-space wave number, and β is the propagation constant of the mode. Relation (19) is plotted in the k-β diagram of Figure 8.4. For values of k < kc, there is no real solution for β, so the mode is nonpropagating. For k > kc, the mode propagates, and k approaches β for large values of β (TEM propagation). The k-β diagram is also useful for interpreting the various wave velocities associated with a dispersive structure. The phase velocity is

vp=wβ=ckβ ,

FIGURE 4: k-β Diagram for a wave guide mode Which is seen to be equal to c (speed of light) times the slope of the line from the origin to the operating point on the k-β diagram. The group velocity is:

vg=dwdβ=cdkdβ,

Which is the slope of the k-β curves at the operating point. Thus, referring to Figure 4, we see that the phase velocity for a propagating waveguide mode is infinite at cutoff and approaches c (from above) as k increases. The group velocity, however, is zero at cutoff and approaches c (from below) as k increases. We finish our discussion of periodic structures with a practical example of a capacitive loaded line.

19

20

21

FILTER DESIGN BY THE IMAGE PARAMETER METHODThe image parameter method of filter design involves the specification of passband and stopband characteristics for a cascade of two-port networks, and so is similar in concept to' the periodic structures that were studied in previous section. The method is relatively simple but has the disadvantage that an arbitrary frequency response cannot be incorporated into the design. This is in contrast to the insertion loss method, which is the subject of the following section. Nevertheless, the image parameter method is useful for simple filters and provides a link between infinite periodic structures and practical filter design. The image parameter method also finds application in solid-state traveling-wave amplifier design.

Image Impedances and Transfer Functions for Two-Port Networks

We begin with definitions of the image impedances and voltage transfer function for an arbitrary reciprocal two-port network; these results are required for the analysis and design of filters by the image parameter method. Consider the arbitrary two-port network shown in Figure 5, where the network is specified by its ABC D parameters. Note that the reference direction for the current at port 2 has been chosen according to the convention for ABC D parameters. The image impedances, Zil and Zi2, are defined for this network as follows:

Zil = input impedance at port I when port 2 is terminated with Zi2,Zi2 = input impedance at port 2 when port I is terminated with Zil.

Thus both ports are matched when terminated in their image impedances. We will now derive expressions for the image impedances in terms of the ABCD parameters of a network.The port voltages and currents are related as

V1= AV2+BI2,I2 = CV2 +DI2

FIGURE5 : A two Port network terminated in its image impedanceThe input impedance at port 1, with port 2 terminated in Zi2, is

Zin1=V1I1=AV2+BI2CV2+DI2=AZi2+BCZi2+D ,

22a

22b

since V2 = Zi2I2

Now solve (22) for V2, I2 by inverting the ABCD matrix. Since AD – BC = 1 for a reciprocal network, we obtain

V2= DV1 – BI1,I2 = –CV1 + AI1.

Then the input impedance at port 2, with port 1 terminated in Zi1 can be found as

Zin2=-V2I2=-DV1-BI1-CV1+AI1=DZi1+BCZi1+A ,since V1= –Zi1I1

We desire that Zin1 = Zi1 and Zin2 = Zi2, so (23) and (25) give two equations for the image impedances:

Zi1 (CZi2+D) = AZi2+B,Zi1D-B = Zi2 (A-CZi1)

Solving for Zi1 and Zi2 gives

Zi1=ABCD

Zi2=BDACwith Zi2 = DZi1/ A. If the network is symmetric, then A = D and Zi1 = Zi2 as expected.Now consider the voltage transfer function for a two-port network terminated in its image impedances. With reference to Figure 6 and (24a), the output voltage at port 2 can be expressed as:V2=DV1-BI1=D-BZi1V1(since now we have V1 = I1 Zi1) so the voltage ratio is V2V1=D-BZi1=D-BCDAB=DAAD-BCSimilarly, the current ratio isI2I1=-CV1I1+A=-CZi1+A=ADAD-BC

FIGURE 6 A two-port network terminated in its image impedances and driven with a voltage generator.The factor DA occurs in reciprocal positions in (8.29a) and (8.29b), and so can be interpreted as a transformer turns ratio. Apart from this factor, we can define a propagation factor for the network as

23

24a24b

25

26a26b

27a27b

28

29a

29b

e-γ=AD-BCWith γ=α+jβ as usual. Since eγ=1AD-BC=AD-BCAD-BC=AD+BCandcoshγ=eγ-e-γ2, we also have thatcoshγ=ADTwo important types of two-port networks are the T and π circuits, which can be made in symmetric form. Table 1 lists the image impedances and propagation factors, along with other useful parameters, for these two networks.

Constant-k Filter SectionsNow we are ready to develop low-pass and high-pass filter sections. First

consider the T network shown in Figure 7; intuitively, we can see that this is a low-pass filter network because the series inductors and shunt capacitor tend to block high-frequency signals while passing low-frequency signals. Comparing with the results given in Table 1, we have Z1 = jωL and Z2 = 1/jωC, so the image impedance isZiT=LC1-ω2LC4If we define a cutoff frequency, ωc, as

ωc=2LC

And a nominal characteristic impedance, R0, as

R0=LC=kTable 1 Image Parameter for T and π Networks

30

31

32

33

34

Where k is constant, then (32) can be written asZiT=R01-ω2ωc2Then ZiT = R0 for ω = 0.The propagation factor, also from Table 1, iseγ=1-2ω2ωc2+2ωωcω2ωc2-1

FIGURE7: Low Pass constant k filter section in T and π form

Now consider two frequency regions:1. For ω < ωc: This is the passband of the filter section. Equation (35) shows that ZiT

is real, and (36) shows that γ is imaginary, since ω2ωc2-1 is negative and |eγ| = 1eγ2=1-2ω2ωc22+4ω2ωc21-ω2ωc2=1

35

35

2. For ω > ωc: This is the stopband of the filter section. Equation (35) shows that ZiT

is imaginary, and (36) shows that γ is real, since eγ is real and -1 < eγ < 0 (as seen from the limits as ω → ωc and ω → ∞. The attenuation rate for ω > > ωc is 40 dB/decade.Typical phase and attenuation constants are sketched in Figure 10. Observe that

the attenuation, α, is zero or relatively small near the cutoff frequency, although α → ∞ as ω → ∞. This type of filter is known as a constant-k low-pass prototype. There are only two parameters to choose (L and C), which are determined by ωc, the cutoff frequency, and R0, the image impedance at zero frequency.

The above results are valid only when the filter section is terminated in its image impedance at both ports. This is a major weakness of the design, because the image impedance is a function of frequency which is not likely to match a given source or load impedance. This disadvantage, as well as the fact that the attenuation is rather small near cutoff, can be remedied with the modified m-derived sections to be discussed shortly.

For the low-pass π network of Figure 7, we have that Zl = jωL and Z2 = 1/jωC, so the propagation factor is the same as that for the low-pass T network. The cutoff frequency, ωc, and nominal characteristic impedance, R0, are the same as the corresponding quantities for the T network as given in (33) and (34). At ω = 0 we have that ZiT = Ziπ = R0, where Ziπ is the image impedance of the low-pass π-network, but ZiT and Ziπ are generally not equal at other frequencies.

FIGURE8: Typical passband and stopband characteristics of the low-pass constant-k sections of Figure 7.

36

FIGURE9: High pass constant k filter sections in T and π form

High-pass constant-k sections are shown in Figure 9; we see that the positions of the inductors and capacitors are reversed from those in the low-pass prototype. The design equations are easily shown to be:R0=LC

ωc=12LC

m-Derived Filter SectionWe have seen that the constant-k filter section suffers from the disadvantages of a

relatively slow attenuation rate past cutoff, and non constant image impedance. The m-derived filter section is a modification of the constant-k section designed to overcome these problems. As shown in Figure 8.12a, b the impedances Z1 and Z2 in a constant-k T-section are replaced with Z1' andZ2', and we letZ1'=mZ1Then we choose Z2'to obtain the same value of ZiT as for the constant-k section. Thus, from Table 1,

ZiT=Z1Z2+Z124=Z1'Z2'+Z1'24=mZ1Z2'+m2Z124Solving for Z2'gives

Z2'=Z2m+Z14m-mZ14=Z2m+1-m24mZ1

Because the impedances Zl and Z2 represent reactive elements, Z2'represents two elements in series, as indicated in Figure 10. Note that m = 1 reduces to the original constant-k section.

For a low-pass filter, we have Zl = jωL and Z2 = I/jωC. Then (39) and (41) give the m-derived components as

Z1'=jωLm

3738

39

40

41

42a

FIGURE 10: Development of an m-derived filter section from a constant-k section. (a) Constant k section. (b) General m-derived section. (c) Final m-derived section.Z2'=1jωCm+1-m24mjωLwhich results in the circuit of Figure 11. Now consider the propagation factor for the m-derived section. From Table 1,eγ=1+Z1'2Z2'+Z1'2Z2'1+Z1'4Z2'For the low pass m-derived filter,

Z1'Z2'=jωLm1jωCm+jωL1-m24m=-2ωmωc21-1-m2ωωc2

FIGURE11 : m-derived filter sectionwhere ωc=2LC as before. Then,1+Z1'4Z2'=1-ωωc21-1-m2ωωc2

If we restrict 0 < m < 1, then these results show that eγ is real and eγ > 1 for ω > ωc. Thus the stopband begins at ω· = ωc, as for the constant-k section. However, when ω= ω∞, whereω∞=ωc1-m2the denominators vanish and eγ becomes infinite, implying infinite attenuation. Physically, this pole in the attenuation characteristic is caused by the resonance of the series LC resonator in the shunt arm of the T; this is easily verified by showing

42b

43

44

that the resonant frequency of this LC resonator is ω∞. Note that (44) indicates that ω∞ > ωc, so infinite attenuation occurs after the cutoff frequency, ωc, as illustrated in Figure 8.14. The position of the pole at ω∞ can be controlled with the value of m.We now have a very sharp cutoff response, but one problem with the m-derived section is that its attenuation decreases for ω > ω∞. Since it is often desirable to have infinite attenuation as ω→∞, the m-derived section can be cascaded with a constant-k section to give the composite attenuation response shown in Figure 11. The m-derived T-section was designed so that its image impedance was identical to that of the constant-k section (independent of m), so we still have the problem of non constant image impedance. But the image impedance of the π-equivalent will depend on m, and this extra degree of freedom can be used to design an optimum matching section.

FIGURE12: Typical Attenuation response for constant-k, m-derived and composite filters.

FIGURE13: Development of an m-derived π-section. (a) Infinite cascade of m derived T-sections. (b) A de-embedded π-equivalent

The easiest way to obtain the corresponding π-section is to consider it as a piece of an infinite cascade of m-derived T-sections, as shown in Figure l5a, b. Then the image impedance of this network is, using the results of Table 1 and (35),

Ziπ=Z1'Z2'ZiT=Z1Z2+Z121-m24R01-ωωc2Now Z1Z2=L/C= R02 and Z12=-ω2L2=-4R02ωωc2 so (45) reduces to

Ziπ=1-1-m2ωωc21-ωωc2R0Since this impedance is a function of m, we can choose m to minimize the variation of Ziπ over the passband of the filter. Figure 14 shows this variation with frequency for several values of m; a value of m = 0.6 generally gives the best results.This type of m-derived section can then be used at the input and output of the filter to provide a nearly constant impedance match to and from Ro. But the image impedance of the constant-k and m-derived T-sections, ZiT, does not match Ziπ; this problem can be surmounted by bisecting the n-sections, as shown in Figure 15. The image impedances of this circuit are Zi1 = ZiT and Zi2 = Ziπ, which can be shown by finding its ABCD parameters:

A=1+Z1'4Z2'

B=Z1'2

45

46

47b

47a

C=1Z2'

D=1and then using (27) for Zi1 and Zi2:

Zi1=Z1'Z2'+Z1'24=ZiT

Zi2=Z1'Z2'1+Z1'4Z2'=Z1'Z2'ZiT=Ziπ,

Where (40) has been used for ZiT.

FIGURE14: Variation of Ziπ in the passband of a low-pass m-derived section for various values of m.

FIGURE15: A bisected π-section used to match Ziπ to ZiT

Composite Filters

By combining in cascade the constant-k, m-derived sharp cutoff, and the m-derived matching sections we can realize a filter with the desired attenuation and matching properties. This type of design is called a composite filter, and is shown in Figure 16.

47d

47c

48a48b

The sharp-cutoff section, with

FIGURE16: The final four stage composite filter

m < 0.6, places an attenuation pole near the cutoff frequency to provide a sharp attenuation response; the constant-k section provides high attenuation further into the stopband. The bisected-π sections at the ends of the filter match the nominal source and load impedance, Ro, to the internal image impedances, ZiT, of the constant-k and m-derived sections.

FILTER DESIGN BY INSERTION LOSS METHOD:

The perfect filter would have zero insertion loss in the passband, infinite attenuation in the stopband, and a linear phase response (to avoid signal distortion) in the passband. Of course, such filters do not exist in practice, so compromises must be made; herein lies the art of filter design.The image parameter method of the previous section may yield a usable filter response, but if not there is no clear-cut way to improve the design. The insertion loss method, however, allows a high degree of control over the passband and stopband amplitude and phase characteristics, with a systematic way to synthesize a desired response. The necessary design trade-offs can be evaluated to best meet the application requirements. If, for example, a minimum insertion loss is most important, a binomial response could be used; a Chebyshev response would satisfy a requirement for the sharpest cutoff. If it is possible to sacrifice the attenuation rate, a better phase response can be obtained by using a linear phase filter design. And in all cases, the insertion loss method allows filter performance to be improved in a straightforward manner, at the expense of a higher order filter. For the filter prototypes to be discussed below, the order of the filter is equal to the number of Reactive elements.

Characterization by Power Loss RatioIn the insertion loss method a filter response is defined by its insertion loss, or

power loss ratio, PLR:

PLR=Power available from sourcePower delivered to Load=PincPload=11-Γω2

Observe that this quantity is the reciprocal of S122if both load and source are matched. The insertion loss (IL) in dB is

IL = 10 logPLR.

49

50

We know that Γω2 is an even function of ω; therefore it can be expressed as a polynomial in ω2. Thus we can writeΓω2=Mω2Mω2+Nω2

where M and N are real polynomials in ω2. Substituting this form in (49) gives the following:

PLR=Mω2Nω2

Thus, for a filter to be physically realizable its power loss ratio must be of the form in (52). Notice that specifying the power loss ratio simultaneously constrains the reflection coefficient, Γ(ω). We now discuss some practical filter responses.Maximally flat. This characteristic is also called the binomial or Butterworth response, and is optimum in the sense that it provides the flattest possible passband response for a given filter complexity, or order. For a low-pass filter, it is specified byPLR=1+k2ωωc2N

where N is the order of the filter, and ωc is the cutoff frequency. The passband extends from ω = 0 to ω = ωc; at the band edge the power loss ratio is 1 + k2. If we choose this as the -3 dB point, as is common, we have k = 1, which we will assume from now on. For ω > ωc, the attenuation increases monotonically with frequency, as shown in Figure 21. For ω >> ωc, PLR ≈ k2(ω/ωc)2N, which shows that the insertion loss increases at the rate of 20N dB/decade. Like the binomial response for multi section quarter-wave matching transformers, the first (2N - 1) derivatives of (53) are zero at ω =0.

Equal ripple. If a Chebyshev polynomial is used to specify the insertion loss of an N-order low-pass filter as

PLR=1+k2TN2ωωcthen a sharper cutoff will result, although the passband response will have ripples of amplitude 1+k2, as shown in Figure 17, since TN(x) oscillates between ±l for |x|≤1. Thus, k2 determines the passband ripple level. For large x, TN(X) ≈ 1/2(2x)N, so for ω > > ωc the insertion loss becomes

PLR≈k242ωωc2Nwhich also increases at the rate of 20N dB/decade. But the insertion loss for the Chebyshev case is (22N)/4 greater than the binomial response, at any given frequency where ω >> ωc.

51

52

53

54

FIGURE17: Maximally flat and equal-ripple low pass filter response (N=3).

Elliptic function. The maximally flat and equal-ripple responses both have monotonically increasing attenuation in the stopband. In many applications it is adequate to specify minimum stopband attenuation, in which case a better· cutoff rate can be obtained. Such filters are called elliptic function filters, and have equal-ripple responses in the passband as well as the stopband, as shown in Figure 18. The maximum attenuation in the passband, Amax, can be specified, as well as the minimum attenuation in the stopband, Amin. Elliptic function filters are difficult to synthesize.

FIGURE18: Elliptic function low pass filter

Linear phase. The above filters specify the amplitude response, but in some applications (such as multiplexing filters for communication systems) it is important to have a linear phase response in the passband to avoid signal distortion. It turns out that a sharp-cutoff response is generally incompatible with a good phase response, so the phase response of the filter must be deliberately synthesized, usually resulting in an inferior amplitude cutoff characteristic. A linear phase characteristic can be achieved with the following phase response:

ω=Aω1+pωωc2Nϕwhere ω ϕ is the phase of the voltage transfer function of the filter, and p is a constant. A related quantity is the group delay, defined as

τd=d dω=A1+p2N+1ωωc2Nϕwhich shows that the group delay for a linear phase filter is a maximally flat function. More general filter specifications can be obtained, but the above cases are the most common. We will next discuss the design of low-pass filter prototypes which are normalized in terms of impedance and frequency; this normalization simplifies the design of filters for arbitrary frequency, impedance, and type (low-pass, high-pass, bandpass, or bandstop). The low-pass prototypes are then scaled to the desired frequency and impedance; and the lumped-element components replaced with distributed circuit elements for implementation at microwave frequencies. This design process is illustrated in Figure 19.

FIGURE19: The Process of filter design by the insertion loss method

Maximally Flat Low-Pass Filter PrototypeConsider the two-element low-pass filter prototype shown in Figure 20; we will derive the normalized element values, L and C, for a maximally flat response. We assume a source impedance of 1 n, and a cutoff frequency ωc = 1. From (53), the desired power loss ratio will be, for N = 2,

55

56

FIGURE20: Low Pass Filter prototype, N = 2PLR=1+ω4

The input impedance of this filter is

Zin=jωL=R1-jωRC1+ω2R2C2

SinceΓ=Zin-1Zin+1

The power loss ratio can be written as

PLR=11-Γ2=11-Zin-1Zin+1Zin*-1Zin*+1=Zin+122Zin+Zin*NowZin+Zin*=2R1+ω2R2C2 AndZin+12=R1+ω2R2C2+1 2+ωL-ωCR21+ω2R2C2 2So (59) becomesPLR=1+ω2R2C24R1+ω2R2C2+1 2+ωL-ωCR21+ω2R2C2 2

=14RR2+2R+1+ω2R2C2+ω2L2+ω4R2C2L2-2ω2LCR2

=1+14R1-R2+R2C2+L2-2LCR2ω2+ω4R2C2L2

Notice that this expression is a polynomial in ω2 Comparing to the desired response of (57) shows that R = 1, since PLR = 1 for ω= 0. In addition, the coefficient of ω2

must vanish, soC2+L2-2LC=C-L2=0or L = C. Then for the coefficient of ω4 to be unity must have

57

58

59

60

14L2C2=14L4=1orL=C=2or

In principle, this procedure can be extended to find the element values for filters with an arbitrary number of elements, N, but clearly this are not practical for large N. For a normalized low-pass design where the source impedance is 1 S1 and the cutoff frequency is ωc = 1, however, the element values for the ladder-type circuits of Figure 25 can be tabulated.

FIGURE21: Ladder circuit for low pass filter prototype and their element definitions. (a) Prototype beginning with a shunt element. (b) Prototype beginning with a series element.

TABLE 2 Element values for maximally flat low pass filter prototype (g0 = 1, ωc = 1, N =1 to 10)

FIGURE 22: Attenuation versus normalized frequency for maximally flat filter prototype

Table 2 gives such element values for maximally flat low-pass filter prototypes for N = 1 to 10. (Notice that the values for N = 2 agree with the above analytical solution.) This data is used with either of the ladder circuits of Figure 21 in the following way. The element values are numbered from g0 at the generator impedance to gN+1 at the

load impedance, for a filter having N reactive elements. The elements alternate between series and shunt connections, and gk has the following definition:

g0=generator resistance (Network of Figure) generator conductance (Network of Figure)

gk(k=1 to N)=inductance for series inductors capacitance fro shunt capacitors

gN+1=load resistance if gN is a shunt capacitor load conductance if gN is a series inductor

Then the circuits of Figure 21 can be considered as the dual of each other, and both will give the same response.Finally, as a matter of practical design procedure, it will be necessary to determine the size, or order, of the filter. This is usually dictated by a specification on the insertion loss at some frequency in the stopband of the filter. Figure 22 shows the attenuation characteristics for various N, versus normalized frequency. If a filter with N > 10 is required, a good result can usually be obtained by cascading two designs of lower order.

Equal Ripple Low Pass Filter Prototype:

For an equal ripple low pass filter with a cutoff frequency ωc = 1, the power loss ratio from (54) is

PLR=1+k2TN2ω,

where 1+k2is the ripple level in passband. Since the Chebyshev polynomials have the property that

TN0=0 for N odd1 for N even

equation(61) shows that the filter will have a unity power loss ratio at ω = 0 for N odd, but a power loss ratio of 1 + k2 at ω = 0 for N even. Thus, there are two cases to consider, depending on N.For the two-element filter .of Figure 20, the power loss ratio is given in terms of the component values in (60). We see that T2(x) = 2x2 - 1, so equating (61) to (60) gives:

61

1+k24ω4-4ω2+1=1+14R1-R2+R2C2+L2-2LCR2ω2+L2C2R2ω2

which can be solved for R, L, and C if the ripple level (as determined by k2) is known. Thus, at ω = 0 we have that

k2=1-R24Ror

R=1+2k2±2k1+k2

Equating coefficient of ω2 and ω4 yields the additional relations,4k2=14RL2C2R2-4k2=14RR2C2+L2-2LCR2

which can be used to find Land C. Note that (63) gives a value for R that is not unity, so there will be an impedance mismatch if the load actually has a unity (normalized) impedance; this can be corrected with a quarter-wave transformer, or by using an additional filter element to make N odd. For odd N, it can be shown that R = 1. (This is because there is a unity power loss ratio at ω = 0 for N odd.)Tables exist for designing equal-ripple low-pass filters with a normalized source impedance and cutoff frequency (ω’c = 1) , and can be applied to either of the ladder circuits of Figure 21. This design data depends on the specified passband ripple level; Table 3 lists element values for normalized low-pass filter prototypes having 0.5 dB or 3.0 dB ripple, for N = 1 to 10. Notice that the load impedance gN+1 ≠ 1 for even N. If the stopband attenuation is specified, the curves in Figures 23a, b can be used to determine the necessary value of N for these ripple values

Linear Phase Low Pass Filter Prototype

Filters having a maximally fiat time delay, or a linear phase response, can be designed in the same way, but things are somewhat more complicated

TABLE 3: Element values for equal Ripple low pass filter prototype (g0 = 1, ωc = 1, N =1 to 10)

62

63

because the phase of the voltage transfer function is not as simply expressed as is its amplitude. Design values have been derived for such filters, however, again for the ladder circuits of Figure 25, and are given in Table 5 for a normalized source impedance and cutoff frequency (ωc’ = 1). The resulting group delay in the passband will be Td = l/ωc’ = 1.

TABLE 4 Element values for maximally flat time delay low pass filter prototype (g0

= 1, ωc = 1, N =1 to 10)

FIGURE 23: Attenuation versus normalized frequency for equal-ripple filter prototypes. (a) 0.5 dB ripple level (b) 3.0 dB ripple level.

FILTER TRANSFORMATIONS

The low-pass filter prototypes of the previ.ous section were normalized designs having a source impedance of RS = 1Ωand a cutoff frequency of ωc = 1. Here we show how these designs can be scaled in terms of impedance and frequency, and converted to give high-pass, bandpass, or bandstop characteristics. Several examples will be presented to illustrate the design procedure.

Impedance and Frequency ScalingImpedance scaling. In the prototype design, the source and 'load resistances are unity (except for equal-ripple filters with even N, which have non unity load resistance). A source resistance of Ro can be obtained by multiplying the impedances of the prototype design by Ro. Then, if we let primes denote impedance scaled quantities, we have the new filter component values given by

L'=R0L

C'=CR0

RS'=R0,

RL'=R0RL

where L, C, and RL are the component values for the original prototype. Frequency scaling for low-pass filters. To change the cutoff frequency of a lowpass prototype from unity to ωc requires that we scale the frequency dependence of the filter by the factor l/ωc, which is accomplished by replacing ω by ω/ωc:

ω←ωωc

Then the new power loss ratio will be

PLR'=PLRωωcwhere ωc is the new cutoff frequency; cutoff occurs when ω/ωc = 1, or ω = ωc This transformation can be viewed as a stretching, or expansion, of the original passband, as illustrated in Figure 24.The new element values are determined by applying the substitution of (65) to the series reactance, jωLk, and shunt susceptances, jωCk, of the prototype filter. Thus,jXk=jωωcLk=jωLk'jBk=jωωcCk=jωCk'which shows the new element values are given by

64a

64b

64c

64d

65

66

Lk'=Lkωc

Ck'=Ckωc

When both impedance and frequency scaling are required, the results of (64) can be combined with (66) to give

Lk'=R0Lkωc

Ck'=CkR0ωc

Low Pass to High Pass Transformation: The frequency substitution where,

ω←-ωcω

can be used to convert a low-pass response to a high-pass response, as shown in Figure 28. This substitution maps ω = 0 to ω = ±∞, and vice versa; cutoff occurs when ω = ±ωc.

FIGURE24: Frequency scalling for low pass filters and transformation to a high pass response

The negative sign is needed to convert inductors (and capacitors) to realizable capacitors (and inductors). Applying (68) to the series reactances, jωLk, and the shunt susceptances, jωCk, of the prototype filter givesjXk=-jωcωLk=1jωCk'

jBk=-jωcωCk=1jωLk'

66a

66b

67a

67b

68

which shows that series inductors Lk must be replaced with capacitors C’k, and shunt

capacitors Ck must be replaced with inductors L’k. The new component values are

given by

Ck'=1ωcLk

Lk'=1ωcCk

Impedance scaling can be included by using 64 to give

Ck'=1R0ωcLk

Lk'=R0ωcCkBandpass and Bandstop transformationLow-pass prototype filter designs can also be transformed to have the bandpass or bandstop responses. If ω1 and ω2 denote the edges of the passband, then a bandpass response can be obtained using the following frequency substitution:

ω←ω0ω2-ω1ωω0-ω0ω=1∆ωω0-ω0ω

Where,

∆=ω2-ω1ω0

is the fractional bandwidth of the passband. The center frequency, ω0, could be chosen as the arithmetic mean of ω1 and ω2, but the equations are simpler if it is chosen as the geometric mean:ω0=ω1ω2Then the transformation of (71) maps the bandpass characteristics of Figure 31b to the low-pass response of Figure 31 a as fo1iows:

When ω = ω0

1∆ωω0-ω0ω=0

When ω = ω1

1∆ωω0-ω0ω=1∆ω12-ω02ω0ω1=-1

When ω = ω2

1∆ωω0-ω0ω=1∆ω22-ω02ω0ω2=1

69a

69b

70a

70b

71

72

73

The new filter elements are determined by using (71) in the expressions for the series reactance and shunt susceptances. Thus,

jXk=j∆ωω0-ω0ωLk=jωLk'-j1ωCk'

which shows that a series inductor, Lk, is transformed to a series LC circuit with element values,

Lk'=Lk∆ω0

Ck'=∆ω0LkSimilarly,

jBk=j∆ωω0-ω0ωCk=jωCk'-j1ωLk'

which shows that a shunt capacitor, Ck , is transformed to a shunt LC circuit with element values,Lk'=∆ω0Ck

Ck'=Ck∆ω0The low-pass filter elements are thus converted to series resonant circuits (low impedance at resonance) in the series arms, and to parallel resonant circuits (high impedance at resonance) in the shunt arms. Notice that both series and parallel resonator elementshave a resonant frequency of ω0.

The inverse transformation can be used to obtain a bandstop response. Thus,ω←∆ωω0-ω0ω-1where ∆ and ω0 have the same definitions as in (72) and (73). Then series inductors of the low-pass prototype is converted to parallel LC circuits having element values given by

Lk'=1∆ω0Ck

Ck'=∆Ckω0The element transformations from a low-pass prototype to a highpass, bandpass, or bandstop filter are summarized in Table 5. These results do not include impedance scaling, which can be made using (64).

TABLE 5: Summary of Prototype filter transformation

74a

74b

74c

74d

75

76a

76b

FILTER IMPLEMENTATION

The lumped-element filter design discussed in the previous sections generally works well at low frequencies, but two problems arise at microwave frequencies. First, lumped elements such as inductors and capacitors are generally available only for a limited range of values and are difficult to implement at microwave frequencies, but

must be approximated with distributed components. In addition, at microwave frequencies the distances between filter components is not negligible. Richard's transformation is used to convert lumped elements to transmission line sections, while Kuroda's identities can be used to separate filter elements by using transmission line sections. Because such additional transmission line sections do not affect the filter response, this type of design is called redundant filter synthesis. It is possible to design microwave filters that take advantage of these sections to improve the filter response; such non redundant synthesis does not have a lumped-element counterpart.

Richard’s TransformationThe transformation,

Ω=tanβl=tanωlvpmaps the w plane to the Ω plane, which repeats with a period of ωl/vp = 2π. This transformation was introduced by P. Richard to synthesize an LC network using open- and short-circuited transmission lines. Thus, if we replace the frequency variable ω with Ω, the reactance of an inductor can be written as

jXL=jΩL=jLtanβl

and susceptance of a capacitor can be written as

jBC=jΩC=jCtanβl

These results indicate that an inductor can be replaced with a short-circuited stub of length βl and characteristic impedance L, while a capacitor can be replaced with an open circuited stub of length βl and characteristic impedance 1/ C. A unity filter impedance is assumed. Cutoff occurs at unity frequency for a low-pass filter prototype; to obtain the same cutoff frequency for the Richard's-transformed filter, (8.77) shows that

Ω=1=tanβl

which gives a stub length of l = λ/8, where A is the wavelength of the line at the cutoff frequency, ωc At the frequency ω0 = 2 ωc, the lines will be λ/4 long, and an attenuation pole will occur. At frequencies away from We, the impedances of the stubs will no longer match the original lumped-element impedances, and the filter response will differ from the desired prototype response. Also, the response will be periodic in frequency, repeating every 4 ωc.

77

78a

78b

In principle, then, the inductors and capacitors of a lumped-element filter design can be replaced with short-circuited and open-circuited stubs, as illustrated in Figure 25. Since the lengths of all the stubs are the same (λ/8 at ωc), these lines are called commensurate lines.

FIGURE 25: Richard’s Transformation

Kuroda’s IdentitiesThe four Kuroda identities use redundant transmission line sections to achieve a more practical microwave filter implementation by performing any of the following operations:

• Physically separate transmission line stubs• Transform series stubs into shunt stubs, or vice versa• Change impractical characteristic impedances into more realizable ones

The additional transmission line sections are called unit elements and are λ/8 long at ωc; the unit elements are thus commensurate with the stubs used to implement the inductors and capacitors of the prototype design.The four identities are illustrated in Table 8.7, where each box represents a unit element, or transmission line, of the indicated characteristic impedance and length (λ/8 at ωc). The inductors and capacitors represent short-circuit and open-circuit stubs, respectively.

TABLE 6: The four Kuroda Identities

The two circuits of identity (a) in Table 6 can be redrawn as shown in Figure 26; we will show that these two networks are equivalent by showing that their ABCD matrices are identical. From Table 1, the ABCD matrix of a length e of transmission line with characteristic impedance Z1 is

ABCD=cosβljZ1sinβljZ1sinβlcosβl=11+Ω21jΩZ1jΩZ11

where 0 = tan βl. Now the open-circuited shunt stub in the first circuit in Figure 26 has an impedance of -jZ2cotβl = -jZ2/Ω, so the ABCD matrix of the entire circuit isABCD=10jΩZ211jΩZ1jΩZ1111+Ω2

=11+Ω21jΩZ1jΩ1Z1+1Z21-Ω2Z1Z2

The short-circuited series stub in the second circuit in Figure 35 has an impedance of j(Z1/ n2) tan βl = j(OZ1/n2), so the ABCD matrix of the entire circuit is

ABCD=1jΩZ2n2jΩn2Z211jΩZ1n20111+Ω2

79

80a

80b

The results in (80a) and (80b) are identical if we choose n2 = 1+Z2/ Z1. The other identities in Table 7 can be proved in the same way.

FIGURE 26: Equivalent circuits illustrating Kuroda identity (a) in Table 6

Impedance and Admittance Inverters

As we have seen, it is often desirable to use only series, or only shunt, elements when implementing a filter with a particular type of transmission line. The Kuroda identities can be used for conversions of this form, but another possibility is to use impedance (K) or admittance (J) inverters. Such inverters are especially useful for bandpass or bandstop filters with narrow (< 10%) bandwidths.

The conceptual operation of impedance and admittance is illustrated in Figure 28; since these inverters essentially form the inverse of the load impedance or admittance, they can be used to transform series-connected elements to shunt-connected elements, or vice versa. This procedure will be illustrated in later sections for bandpass and bandstop filters.In its simplest form, a J or K inverter can be constructed using a quarter-wave transformer of the appropriate characteristic impedance, as shown in Figure 8.38b. This implementation also allows the ABCD matrix of the inverter to be easily found from the ABCD parameters for a length of transmission line given in Table 4.1. Many other types of circuits can also be used as J or K inverters, with one such alternative being shown in Figure 8.38c. Inverters of this form turn out to be useful for modeling the coupled resonator filters of Section 8.8. The lengths, e/2, of the transmission line sections are generally required to be negative for this type of inverter, but this poses no problem if these lines can be absorbed into connecting transmission lines on either side.

FIGURE 28: Impedance and Admittance Inverters

STEPPED-IMPEDANCE LOW-PASS FILTERS

A relatively easy way to implement low-pass filters in microstrip or stripline is to use alternating sections of very high and very low characteristic impedance lines. Such filters are usually referred to as stepped-impedance, or hi-Z, low-Z filters, and are popular because they are easier to design and take up less space than a similar low-pass filter using stubs. Because of the approximations involved, however, their electrical performance is not as good, so the use of such filters is usually limited to applications where a sharp cutoff is not required (for instance, in rejecting out-of-band mixer products).

Approximate Equivalent Circuits for Short Transmission LineSectionsWe begin by finding the approximate equivalent circuits for a short length of transmission line having either a very large or very small characteristic impedance.

Z11=Z22=AC=-jZ0cotβl

Z12=Z21=1C=-jZ0cscβl

The series elements of the T-equivalent circuit are

Z11-Z12=-jZ0tanβl2

while the shunt element of the T -equivalent is Z12. So if βl< π/2, the series elements have a positive reactance (inductors), while the shunt element has a negative reactance (capacitor). We thus have the equivalent circuit shown in Figure 29a, whereX2=Z0tanβl2

B=1Z0sinβl

FIGURE 29 Approximate equivalent circuits for short sections of transmission lines. (a) T equivalent circuit for a transmission line section having βl< < π/2. (b) Equivalent circuit for small βl and large Z0 (c) Equivalent circuit for small βl and small Z0

Now assume a short length of line (say βl < π/4) and a large characteristic impedance. Then (83) approximately reduces to

X≈Z0βl

81a

81b

82

83a

83b

84a

B≈0

which implies the equivalent circuit of Figure 39b (a series inductor). For a short length of line and a small characteristic impedance, (83) approximately reduces to

X≈0

B≈Y0βl

which implies the equivalent circuit of Figure 29c (a shunt capacitor). So the series inductors of a low-pass prototype can be replaced with high-impedance line sections (Z0 = Zh), and the shunt capacitors can be replaced with low-impedance line sections (Z0 = Zl). The ratio Zh/Zl should be as high as possible, so the actual values of Zh and Zl are usually set to the highest and lowest characteristic impedance that can be practically fabricated. The lengths of the lines can then be determined from (84) and (85); to get the best response near cutoff, these lengths should be evaluated at ω = ωc

Combining the results of (84) and (85) with the scaling equations of (67) allows the electrical lengths of the inductor sections to be calculated as

βl=LR0Zh (inductor)

and the electrical length of the capacitor section as

βl=CZhR0 (capacitor)

where Ro is the filter impedance and L and C are the normalized element values (the gkS) of the low-pass prototype.

84b

85b

85a

INTRODUCTION TO PRINTED CIRCUIT BOARD (P.C.B)

The Printed Circuit Board (P.C.B) is an instantly recognizable symbol of both the beauty of electronics design and its overwhelming sophistication. A fine P.C.B is a mixture of high art and solid engineering. It is a synthetic, laminated insulating material to which copper tracks have been added.A typical P.C.B consists of 1.5mm (1/16th inch) epoxy-glass substrate bonded on both sides of a 1.0 ounce copper foil (approx. 1.4mils or 0.03mm thickness). The signal, return and interconnect etched traces are about 1.0mm in width.A P.C.B is an integral part of electronic devices. Electronic components mounted on a P.C.B big or small. Besides keeping the components in place P.C.B provides electrical connection between the components mounted on it. Printed circuit board can be divided into four types:

1) Rigid board2) Flexible and Rigid-Flex Boards3) Metal Core Boards 4) Injection Molded Boards

The board that is most widely used is the rigid board. The boards can be further classified into single sided, double-sided or multi-layered boards. The ever-increasing packaging density and faster propagation speeds, which stem from the demand for high performance systems, have forced the evolution of the boards from single-sided, double-sided to multi-layered boards. On Single-Sided Boards, all the inter connections are on one side. Double Sided Boards have connections on both sides of the board and allow wires to cross over each other without the need of jumpers. This was accomplished at first by Z-wires, then by eyelets and at present by P.T.H (Plated Through Hole). The increased pin count of the ICs has increased the routing requirements, which led to multi-layer board.

The necessity of controlled impedance for the high-speed traces, the need for bypass capacitors, the need for low inductance values for the power and ground distribution networks have made the requirements of power and ground planes a must in high performance boards. These planes are possible only in Multi-Layer Boards. In multi-layer boards, the P.T.H can be buried, semi-buried, or through vias.

Material used in p.c.b:-Properties: - The material used in P.C.B manufacturing should be strong• Mechanically so as to bear mechanical shocks and vibrations.

• Chemically so as to bear environmental effects.

• Electrically so as to withstand high frequency signal.

• Thermally so as to withstand temperature changes due to heat loss by components

The properties that must be considered when choosing a P.C.B substrate material are their mechanical, electrical, chemical, and thermal properties. Earlier used was copper clad phenolic & paper laminate material.Some special properties of certain material are mentioned below:Polyamide has long-term thermal rest, low C.T.E (Coefficient of thermal expansion), high reliability. Cyanate Ester has low ε and is used in high-speed circuits.Teflon has lower €, lower dissipation factor, high temperature stability.FR-4 Epoxy resin impregnated for glass cloth. Rolls of glass cloth are coated with liquid resin (A-stage). Then the resin is partially cared to a semi stable state (B-stage). The rolls are cut into large sheets and several sheets are stacked to form the desired final thickness.For a board green is the standard color. While selecting a board material, C.T.E of material must be matched to that of component material or else the mechanical connections may break and cause malfunctioning.

Ground Plane

Need for ground plane:

○ Reduces interference due to high frequency nets

○ Reduces the capacitive effect existing between two electric layers

○ Provides return path

○ Provides shield to critical signals

○ Provides termination for traces

○ Reduces ground bounce

Grounds should be run separately for each section of the circuitry on a board and brought together with the power ground at only one point on the board. On multilayer boards, the ground is frequently run in its own plane. This plane should be interrupted no more than necessary, and with nothing larger than a via or a through hole. An intact ground plane will act as a shield to separate the circuitry on the opposite sides of the board, and it will also not act as an antenna-unlike a trace, which will act as an antenna. On single-sided analog boards, the ground plane may be run on the component side. It will act as a shield from the components to the trace/solder side of the board.

A brief illustration of the issue of inter-plane capacitance and crosstalk is presented in figure, which shows the two common copper layouts styles in circuit board design (cutaway views). To define a capacitor C where copper is the conductor, air is an insulator, and FR-4 is an insulator,

C=A εo /d

A = area of the "plates"

d = distance between the planes

For a circuit board, the area A of the plates can be approximated as the area of the conductor/trace and a corresponding area of the ground plane.

The distance d between the planes would then be a measurement normal to the plane of the trace.

The impedance of a capacitor to the flow of current is calculated as:

XC =1/2πfC

f= frequency of operation

C= value of the capacitor or the capacitance of the area of the circuit board trace.

Analog Design

The analog design world relies less on C.A.D systems auto routes than the digital design. Analog design also follows the stages of schematic capture, simulation, component layout, critical signal routing, non-critical signal routing, power routing and a "copper pour" as needed. The copper pour is a C.A.D technique to fulfill an area of the board with copper, rather than having that area etched to bare F.R-4. This is typically done to create a portion of a ground plane.

The first rule of design in the analog portion of the board is to have a ground that is separate from another ground such as digital, RF, or power. This is particularly important since analog amplifiers will amplify any signal presented to them, whether it is legitimate signal or noise, and whether it is present on a signal input or

d

on the power line. Any digital switching noise picked up by the analog circuitry through the ground will be amplified in this section.

Digital DesignThe first rule of design in the digital portion of the board is to have a ground that is separate from any other grounds such as analog RF or power. This is to primarily protect other circuit from the switching noise created each time a digital gate shift state. The faster the gate, and the lower its transition time, the higher the frequency components of its spectrum. Like analog circuits, digital circuits should either have a ground plane on the component side of the board or a separate "digital” ground that connects only to the other grounds at one point of the board. At some particular speed (>40MHz), the circuit board traces need to be considered as transmission lines. The normal rule of thumb is that this occurs when the two way delay of the line is more than the rise time of the pulse. The "delay" of the line is really the propagation time of the signal. Since we assume that electricity travels down a copper wire or trace at the speed of light, we have a beginning set of information to work with.

R.F DesignRF circuits cover a broad range of frequencies from 0.5 KHz to >2 GHz. Generally RF designs are in shielding enclosures. The nature of RF means that it can be transmitted and received even when the circuit is intended to perform all of its activities on the circuit board. As in the case of light the RF circuits lead to very short band widths, which means that the RF can leak through smaller openings .This is the reason for shielded enclosures and the reason that RF I/O uses trough pass through filters rather than direct connections.

What Is E.M.C?E.M.C or Electromagnetic Compatibility is the ability of the system, component, P.C.B, to function as designed or without malfunctioning, in real conditions.

E.M.I or Electromagnetic Interference is any signal, which contains unwanted conductor or radiator signal, which can degrade the performance of the system, sub-system, component or P.C.B.

Worldwide E.M.C regulations set strict limitations on electromagnetic emissions from electronic equipment, and the susceptibility of electronic equipment to external electromagnetic interference. Interference is either conducted or radiated:

Why Analyze For E.M.C?

The ever-increasing clock speeds, which improve the performance of electronic equipment, have the unwelcome side effect of producing increased levels of electromagnetic emissions. International guidelines have been established in order to keep electromagnetic emissions down to an acceptable level.

Rules for Analyzing E.M.I / E.M.C

Impedance Profile Rule○ Areas of high characteristic impedance give higher emissions.

In order to reduce E.M.I and Susceptibility for a design, it is good practice to reduce the inductance and increase the capacitance of the signal tracks.The characteristic impedance of a transmission line is approximately:

So, to reduce the impedance, one needs to increase the capacitance (C). One can increase the capacitance of a track by shielding it, or by altering the dielectric between the track and the Power plane

XY Tracking RuleRoute segments on the adjacent layers of multi-layer boards should run at right angles to each other. This helps to reduce electromagnetic interference by reducing capacitive coupling between routes on adjacent layers:The Layer Stack defined for the design affects the following rules:

• Closed Loops• Open Loops

Layer Stack

• Impedance Profile• XY Tracking• Track Stubs• Track Resonance• Return Loops• Layer Stack• Overlapping Planes• Termination• Track Mitering• Crosstalk• Impedance• Field Solver

The Layer Stack is set up by the Layers option on the Settings menu.The Layer Stack is the system of layers you have set up in your design technology. In particular it considers the physical construction of your board, considering the materials, permittivity, and thickness of the layers making up the board.For example:

Overlapping Power Planes RuleSome of the designs may make use of multiple Power planes at different voltages.

It is not good practice to overlap the planes of different voltages (this can lead to undesirable noise current distribution throughout the system):

It is also good practice to overlap corresponding Power planes (this gives additional capacitive decoupling)... For example, the 5 Volt plane and its corresponding Ground plane):

Overlapping Ground planes are capacitively coupled to each other - so noise can transfer from one to another.In addition, large copper areas are excellent radiators of E.M.I so it is good practice to prevent high frequency signals from appearing on Ground planes.

Track Resonance RuleAs the time-length of a route (given by the delay along the track) approaches a multiple of a quarter of the signal wavelength (or a harmonic of that signal), its efficiency as a radiating antenna increases.This increases the electromagnetic radiation from the route, and also enhances the route's susceptibility to externally generated electromagnetic interference. Thus track

resonance should be avoided.

Component Decoupling RuleAppropriate decoupling of active components reduces component noise and power plane transients.Reducing the voltage transients on power supply connections reduces emissions from the signal loops formed by the power connections.

Although circuit conditions affect the choice of decoupling capacitor to a certain extent, the technology of the device being decoupled is the most important factor in the choice of capacitor and method of decoupling.Hence, we normally use the same decoupling capacitor each time you use a particular component in the design. Termination RuleHigh frequency nets with long routes should be terminated to avoid extra high frequency harmonics.Each track has a critical length above which reflections are a maximum, and below which they are reduced. This critical length occurs when the round-trip delay time along the track is equal to the rise-time of the signal. It is recommended practice to terminate lines whose length is close to or above this critical value.A typical termination is shown below:

Track Length RuleHigh frequency signal tracks should be as short as possible.Long routes, particularly those that operate at high frequencies and edge rates, generate E.M.I. Long routes are also more susceptible to E.M.I.

Track Mitering RuleElectromagnetic emissions are more concentrated on the right angled corners, especially on high speed nets. Mitering is replacing right angled corners with angled segments at 45 degrees. Mitering reduces trace lengths, reduces high frequency signal reflections and minimizes the effect of acid traps. Track mitering helps to reduce the number of sharp corners and therefore the electromagnetic emissions. Two examples of mitering are shown below:

Closed Loop Rule

A tracking loop constitutes a loop antenna for E.M.I. Radiation is generated in the plane of the loop. Closed loops always use more than one layer and the tracks cross

over another segment of the same net, as shown in the diagram below. The intervening plane acts as a local return path for the loop signals, greatly reducing radiation from the loop.

Open Loop Rule

A tracking loop constitutes a loop antenna for E.M.I. Radiation is generated in the plane of the loop. The area and frequency of operation, and signal amplitude on each loop, are determined.

Track Stub Rule

Stubs (branches from the main routing path) cause unwanted reflections and hence H.F harmonics if they exceed a critical, maximum stub length. This generates additional electromagnetic interference, causing signal distortion. Thus the stub length should be minimized.

Return Loop Rule

There should be an appropriate ground return within a specified distance of each signal (the default is 2 millimeters).

To reduce costs, a two-layer board is preferable to a four-layer board.

If transmission lines are not an issue, you can carefully arrange the ground return paths to allow two-layer boards to be used:

However, when E.M.C or high-speed signals need to be considered, it is better to introduce Ground planes for the purposes of screening and controlling the impedance. The use of planes inevitably means a minimum of four layers.

Current Rule

The amount of current that can be carried by the design depends on the width of the tracks used on that design. There must be sufficient track width on your board to support the current it is expected to carry.

Thus, optimum current should be passed through the conductor.

Power Plane Impedance Rule

A Ground or Power plane may become excessively perforated by through-hole pins, viasand thermal relief cut-outs in a particular area. This has the effect of increasing the resistance of the plane in that area, and the impedance of the total plane. This can result in transient conditions on the board, leading to increased E.M.I.

Component Placement Rule

High-speed components should be nearest to power sources to reduce transients.

For boards without Power-planes, the power connections form a signal loop in the design. If you keep high speed and high power components as close to the power source as possible, you will reduce these signals; loop areas and hence reduce the radiation from the power routes.

Components with a high-speed factor should be placed close to power sources and components with lower speed factors further away from power sources:

CROSSTALKCrosstalk is caused by parallel, or near parallel, tracks on the same board layer or on adjacent board layers. A change of state on one line causes interference on another due to mutual capacitance and inductance between the lines. The net causing crosstalk is known as the active net. The net receiving (picking up) the crosstalk is called the passive net. The active signal is the signal on the active net. The E.M field couples adjacent transmission lines.Coupling depends on:

○ Distance between the traces ○ Distance to the reference plane○ Dielectric material○ Thickness, width of the traces

○ Coupled length○ Rise/fall time of the signal

Crosstalk can be Capacitive, Inductive or Radiative.

➢ Capacitive (electrostatic) when transmission lines are parallel on adjacent planes

It can be minimized by:○ Restricting routing on adjacent planes to right angles➢ Inductive (magnetic) when transmission lines are parallel on same planeIt can be minimized by:○ Restrict parallelism to < the critical length○ Adjust board stack-up○ Adjust trace-trace clearance and trace width○ Shielding the critical net

There are two types of crosstalk:

1. Backward Crosstalk:

This is usually the dominant form. The backward crosstalk signal travel down the passive line in the opposite direction to the active signal. It is always of the same polarity as the active signal:

2. Forward Crosstalk:

This is a secondary effect, which occurs predominately on the outer board layers. The forward crosstalk signal travels in the same direction as the active signal, and is usually of opposite polarity.

Isolated Copper Area Rule

A common way to reduce emissions is to fill spare areas of each layer with copper, thus providing an effective screen to all signal tracks on the layer. However, these areas of copper can act secondary radiators of E.M.I and should always be connected to a power plane:

Secondary radiators can couple to a nearby net purely by virtue of its proximity, and resonate in sympathy with the signals on that net.

So, while the nets itself may not be of the correct dimensions to radiate, it may couple to a secondary radiator whose dimensions are a multiple of the wavelength of the signal on the net. In this situation, the secondary radiator acts as a antenna and transmits the signal as a electromagnetic wave.

Hence we should check every copper area in the design to see if it is connected to a signal.

Net Shield Rule

Net Shielding should be assigned to high-speed nets that need shielding from interference by other signals. The critical net can be shielded by ground as shown in the following fig.

NEED OF HIGH SPEED CIRCUIT

Not long ago, the major T.T.L logic families all had typical rise and fall times that, by modern standards were very slow on the order of 10ns. Products that use these old chips have few problems with ringing or crosstalk. Today’s digital logic operates much faster with rise and fall time of the order of 1ns or less. As chips continue to shrink, and as switching speed continue to improve by a factor of two every few years, we will soon see rise and fall times slink down into the deep sub-nanosecond realm, a territory previously reserved for U.H.F and micro-wave engineers. This relentless, shrinking trend towards sub-nanosecond rise times brings us to the crux of the problems: Faster switching exaggerates problems with ringing and crosstalk. Over the years, as logic outputs have continued to switch faster and faster, problem with ringing and crosstalk have gotten progressively worse.

TRACE BANDWITHEvery trace has a signal bandwidth, where signals are transferred with acceptable results, and a cutoff region, where undesired effects may occur; the curve here shows the frequency response for an ideal trace. As shown the response is flat in the operating region, then rolls off gradually in the cutoff region. In a typical P.C.B trace (bottom graph) resonance can occur at high frequencies. This uncontrolled resonance can create havoc in a wideband circuit and can be difficult to curve.

Possible solutions are:• Eliminate branches and nodes on P.C.B traces. This may be impossible on

clock nets, which have to be distributed to several devices. In this event, buffer may be needed, which add to circuit delay.

• Ensure all traces are terminated in there characteristic impedance and run over unbroken ground plane. This goes without saying in high-speed design, and it

requires careful attention to trace dimension and fabrication processes.• Keep traces as short as possible. This leads to the same considerations as in

trace delay discussed above.• When necessary, use coax for critical signals instead of P.C.B traces.

Parasitic OscillationsTransistors, Op-Amp, F.E.T and analog I.C including simple voltage regulators, can exhibit parasitic oscillations. This is usually a layout problem caused by the invisible inductance and stray capacity in each trace. The graph bellow shows the effect of a parasitic oscillation in an amplifier. The input signal is shown in first figure and the output is shown in second figure.

Note the oscillation can come and go with different operating conditions. Some symptoms of parasitic oscillations are:

• Erratic operation when a hand is waved nears the circuit, or cables are moved or touched.

• Excessive noise or distortion in the output.• Spurious responses in signal processing circuits such as spectrum analyzer,

network analyzer, and communication circuits.• Erratic circuit behavior when signal parameters are changed.

Trace Delay

Earlier we ignored time the time delay as signals propagated along PCB traces. Today, this delay can account for over half the total delay in a design. Trace delay may make it impossible to use a design, component placement may be limited by physical size, and if the components can't be mounted close together, excessive trace delay may cause timing failure in some circuits.

Comparing the two graphs given above for a simple "D-latch” where the input signal D, clock (clk) and Q the output, the circuit fails due to extra delaying in the second case.

When a design fails due to excessive trace delay some alternate solution may be -

• Simplify the design This is the preferred solution. If the design goals can be met by eliminating

critical timing specifications, it usually reduces costs and givers a more robust design.

• Add logic elements to compensate for the trace delay. If a separate gate is available, and if it is located in a suitable package, it

could be used to add delay to a critical signal to help meet timing requirements

Motorola E.C.L I.C. E.C.L logic is a temperature compensated for propagation delay, but other logic families may not have this advantage. Guaranteeing the design will meet specs over the temperature and supply voltage may be difficult.

• Choose smaller packages so they can be moved closer together. This may mean a redesign since smaller logic devices may not have the

same functions that are available in larger ones. It also may increase junction temperatures and reduce reliability, since

the heat is dissipated in a smaller region.

• Usehybrid. It will increase cost, as hybrids are more expensive and difficult to

manufacture. Temperature may become a serious problem. Yields may suffer, since limited facilities are available to test devices

before they are mounted in the hybrid.

• Design a custom integrated circuit. It is the most expensive option, but, depending on the circuit topology

and clock speed, there may be little choice. However, this raises a host of new problems, such as floor planning, electro migration, crosstalk and timing delay.

SIGNAL INTEGRITY

Introduction to Signal Integrity (S.I)

The term Signal Integrity (S.I) addresses two concerns in the electrical design aspects – the timing and the quality of the signal.

• Does the signal reach its destination when it is supposed to? • And also, when it gets there, is it in good condition?

The goal of signal integrity analysis is to ensure reliable high-speed data transmission. In a digital system, a signal is transmitted from one component to another in the form of Logic 1 or 0, which is actually at certain reference voltage levels. At the input gate of a receiver, voltage above some reference value is considered as logic high, while voltage below the reference value considered as logic low.

Where S.I Problems Happen?Since the signals travel through all kinds of interconnections inside a system,

any electrical impact happening at the source end, along the path, or at the receiving end, will have great effects on the signal timing and quality. The chip package could be single chip carrier or Multi-Chip Module (M.C.M). Through the solder bumps of the chip package, signals go to the Printed Circuit Board (P.C.B) level. At this level, typical packaging structures include daughter card, motherboard or back plane. Then signals continue to go to another system component, such as an A.S.I.C (Application Specific Integrated Circuit) chip, a memory module or a termination block. The chip packages, printed circuit boards, as well as the cables and connecters, form the so-called different levels of electronic packaging systems. In each level of the packaging structure, there are typical interconnects, such as metal traces, vias, and power/ground planes, which form electrical paths to conduct the signals. It is the packaging interconnection that ultimately influences the signal integrity of a system.

S.I Issues in Design“Timing” is every thing in a high-speed system. Signal timing depends on the delay caused by the physical length that the signal must propagate it also depends on the shape of the waveform when the threshold is reached. Signal waveform distortion can be caused by different mechanisms but the following are mostly considered.

Rise Time and S.ISince many SI problems are directly related to dV/dt or dI/dt, faster rise time significantly worsens some of the noise phenomena such as ringing, crosstalk, and power/ground switching noise. Systems with faster clock frequency usually have shorter rise time; therefore they will be facing more SI challenges.

Transmission Lines, Reflection, CrosstalkIn S.I analysis, since the electric models for many interconnects can be treated as transmission lines, it is important to understand the basics of transmission line theory and get familiar with common transmission line effects in high-speed design such as crosstalk, resonance, reflection, ground bounce, termination, impedance mismatch, etc.

Power/Ground NoiseTransient currents drawn by a large number of devices (core-logic, off-chip

drivers) switching simultaneously can cause voltage fluctuations between power and ground planes, namely the simultaneous switching noise (S.S.N), or Delta-I noise, or power/ground bounce. S.S.N will slow down the signals due to imperfect return path constituted by the power/ground distribution system. It will cause logic error when it couples to quiet signal nets or disturbs the data in the latch. It may introduce common mode noise in mixed analog and digital design.

S.I in Electronic PackagingTechnology trends toward higher speed and higher density devices have pushed the package performance to its limits. The clock rate of present personal computers is approaching gigahertz range. As signal rise time becomes less than 200ps, the significant frequency content of digital signals extends up to at least 10 GHz. This necessitates the fabrication of interconnects and packages to be capable of supporting very fast varying and broadband signals without degrading signal integrity to unacceptable levels. While the chip design and fabrication technology have undergone a tremendous evolution: gate lengths, having scaled from 50 µm in the 1960’s to 0.18 µm today, are projected to reach 0.1 µm in the next few years; on-chip clock frequency is doubling every 18 months; and the intrinsic delay of the gate is decreasing exponentially with time to a few tens of Pico-seconds. However, the package design has lagged considerably. With current technology, the package interconnection delay dominates the system-timing budget and becomes the bottleneck of the high-speed system design. It is generally accepted today that package performance is one of the major limiting factors of the overall system performance. Advances in high performance sub-micron microprocessors, the arrival of gigabit networks, and the need for broadband Internet access, necessitate the development of high performance packaging structures for reliable high-speed data transmission inside every electronics system. Signal integrity is one of the most important factors to be considered when designing these packages (chip carriers and P.C.B) and integrating these packages together.

Principles of S.I AnalysisA digital system can be examined at three levels of abstraction: logic, circuit theory, and electromagnetic (E.M) fields. The logic level, which is the highest level of those three, is where SI problems can be easily identified. E.M fields, located at the lowest level of abstraction, comprise the foundation that the other levels are built upon. Most of the S.I problems are E.M problems in nature, such as the cases of reflection, crosstalk and ground bounce. Therefore, understanding the physical behavior of S.I problems from E.M perspective will be very helpful. Even though E.M full wave analysis is much more accurate than the circuit analysis in the modeling of

packaging structures, currently, common approaches of interconnect modeling are based on circuit theory, and SI analysis is carried out with circuit simulators. This is because field analysis usually requires much more complicated algorithms and much larger computing resources than circuit analysis, and circuit analysis provides good S.I solutions at low frequency as an electrostatic approximation. Typical circuit simulators, such as different flavors of SPICE, employ nodal analysis and solve voltages and currents in lumped circuit elements like resistors, capacitors and inductors. In SI analysis, an interconnection sometimes will be modeled as a lumped circuit element. For instance, a piece of trace on the printed circuit board can be simply modeled as a resistor for its finite conductivity. With this lumped circuit model, the voltages along both ends of the trace are assumed to change instantaneously and the travel time for the signal to propagate between the two ends is neglected. However, if the signal propagation time along the trace has to be considered, a distributed circuit model, such as a cascaded R - L - C network, will be adopted to model the trace. To determine whether the distributed circuit model is necessary, the rule of thumb is – if the signal rise time is comparable to the round-trip propagation time, you need to consider using the distributed circuit model.In summary, as the current high-speed design trend continues, fast rise time reveals the distributed nature of package interconnects. Distributed circuit models need to be adopted to simulate the propagation delay in SI analysis. However, at higher frequencies, even the distributed circuit modeling techniques are not good enough, full wave electromagnetic field analysis based on solving Maxwell’s equations must come to play. As presented in later discussions, a trace will not be modeled as a lumped resistor, or an R-L-C ladder; it will be analyzed based upon transmission line theory; and a power/ground plane pair will be treated as a parallel-plate wave-guide using radial transmission line theory.

P.C.B SimulationA traditional board-level digital design starts with the creation of a schematic at the level of components and interconnects. The schematic is passed to the board designer in the form of a net-list. The design is laid out, prototyped, and then verified. Simulation is intended to precede the prototype stage and find problems before they are committed to hardware. In this scenario, the schematic design/net-list is passed to the simulator, along with definitions of power supplies and required stimulus. The simulator has libraries of parts that ideally will include all parts used in the schematic. If not, the design team must create the part specifications and provide criteria such as amplification, frequency response slew rate, logical function, device timing constraints such as set up and hold times and line and load regulations.

Simulation results are presented in both graphical and textual forms. Simulator helps to design correctly even in presence of low power supplies. One of the most

difficult things is timing delays and effects of parasitic inductance and capacitance. Digital circuits involve fault simulation, which aims to determine all possible faults such as solder short between traces and ground. So as to support the high-speed model circuitry the I.B.I.S model (I/O Buffer Information Specification) can now be purchased. The I.B.I.S standard also includes transmission line descriptions to perform signal integrity simulation. This part of I.B.I.S is termed as E.B.D (Electrical Board Description).

I.B.I.S

The Input/output Buffer Information Specification (I.B.I.S) is an emerging standard used to describe the analog behavior of the Input/output (I/O) of a digital Integrated Circuit (I.C). I.B.I.S specifies a consistent software-passable format for essential behavioral information. With I.B.I.S, simulation tool vendors can accurately model compatible buffers in SI simulations. Improvement of chip and package design

technology accompanying industrial competition has resulted in the need for new descriptive models of integrated circuit drivers and receivers. These models should be nonproprietary and capable of maintaining suitable accuracy and speed in the simulation of transmission lines and signal integrity related effects such as crosstalk and power/ground bounce (noise).

Simulation of digital I/O buffers, together with their chip packages and printed circuit boards, can mainly be done in two ways. The traditional approach is to use transistor level models, which is useful when small-scale simulations or analysis of some particular network is the objective of the simulation. This approach would be very time consuming for simulations of large number of buffers and their interconnections. Transistor level models may also reveal vendor's proprietary device information. As a solution to this problem, behavioral models of devices such as I/O Buffer Information Specification (I.B.I.S) are introduced. The behavioral I.B.I.S modeling data can be derived from measurements as well as circuit simulations. Simulations with behavioral models can be generally executed faster than the corresponding simulations with transistor level models. A behavioral device model does not reveal any detailed and sensitive information about the design technology and the underlying fabrication processes, so the vendor’s intellectual property would be protected.The behavioral I.B.I.S based models of a device provide the D.C current vs. voltage curves along with a set of rise and fall time of the driver output voltage and packaging parasitic information of the I/O buffer. It should be noted that the I.B.I.S modeling data itself does not provide explicit information on driver transient state transitions beyond the steady-state I-V curves. The extraction of the transient state transition of buffers is necessary for correct S.I simulations.