GB1593486A - Microwave devices - Google Patents

Description

(54) IMPROVEMENTS RELATING TO MICROWAVE DEVICES (71) We, FERRANTI LIMITED a Company registered under the Laws of Great Britain of Hollinwood in the County of Lancaster, do hereby declare the invention for which we pray that a patent may be granted to us, and the method by which it is to be performed, to be particularly described in and by the following statement: This invention relates to microwave devices, and in particular to broadband microwave filtering devices, such as low-pass filters, high-pass filters, stopband filters, passband filters, diplexers and multiplexers.

A diplexer is obtained when a high-pass microwave filter is combined with a low-pass microwave filter, and there is normally a common cut-off frequency, or cross-over frequency.

A multiplexer comprises a combination of microwave passband or stopband filters, with different filters having different passbands or stopbands, respectively.

Broadband microwave filtering devices are defined as having some band which is in excess of an octave, and is possibly upto a decade. In terms of distributed elements, where the centre of this band is at an electrical length 0=900, the band edge will range between 9=j70 to 600 To have a reasonable selectivity without using an excessive number of constituent elements n will automatically rule out several types of microwave filter.

The interdigital microwave filter is difficult to realise for bandwidths in excess of 3:1, and has a significant loss in relative selectivity as this bandwidth is increased.

For example, for the Chebyshev response:

L= 1+62cosh2[(n-l)cosh~l + cosh cote ] (t) where E iS a constant, and OO is a constant dependent on the required bandwidth of the filter.

Since the variable cos0/cosO0 does not increase much above unity for 0O small, n has to be large if the loss L is to increase significantly for 0 just less than Oc.

If the interdigital filter is used in the low-channel of a diplexer another disadvantage is that the periodic passband may cause problems. If the interdigital filter is used in the high frequency channel of a diplexer the impedance values have to be very accurate to maintain the bandedge frequency to within 1% (i.e., approximately 0.0005 inch at 12 GHz), and any discrepancy is difficult to tune.

For most requirements the comb-line filter with distributed shunt, opencircuited stubs must be disregarded, because for bandwidths in excess of 1.5-tl they become almost impossible to realise in practice. This is due to the high impedance of the shunt inductive stubs.

A possible method of meeting the requirement of a broadband microwave filtering device with a reasonable selectivity is to employ a conventional ladder network using shunt stubs. The insertion loss L for the low-pass channel of a diplexer using such a filtering device is:

L = 1 + 62cosh 2 (n cosh~l ttdannS ) (2) Such a low-pass filtering device has a good selectivity if n is sufficiently large.

However, in such a filter the constituent elements are required to be located in close proximity, and adjacent elements will couple if adequate screening is not provided. Usually this isolation requires a relatively complicated mechanical arrangement, which is another disadvantage of such a filter. However, the most important point with respect to the ladder realisation is the method of maintaining the bandedge frequency to within 1%. As with the interdigital filter, the bandedge is controlled almost exclusively by the relative impedances of the elements. Hence, these elements must be realised very accurately, either directly, or by providing adequate tuning. For networks of greater than 20 elements, as are commonly used, the tuning can be difficult if the passband response is to be maintained whilst accurately providing the correct bandedge frequency. Thus, for a multiplexer the tuning process could represent a significant proportion of the production costs.

Since coupling naturally exists between adjacent elements of a filter network it is possible to provide a distributed network if the element values are reasonable, and as a realisation of a low-pass prototype with the coupling as an extra parameter. Hence, the optimum amplitude, or elliptic function, filter can be realised in this form, providing an optimum equi-ripple passband, and a stopband response with a significant reduction in degree for a very selective response. Digital elliptic filters in the bandwidth range 2:1 to 4:1 have been built based upon this lowpass prototype. However, even within this limited range of bandwidths the required impedance variations within the filter are difficult to obtain in a mechanical sense in the case of an air-line realisation. For a strip-line realisation of such a microwave filter device, and in the case of a diplexer in particular, such a response characteristic could not be obtained because of this impedance change.

The main problem comes from the part of the filter which realises the transmission zero remote from the bandage. In addition, in the diplexer case, where normally even degree networks with parallel connections are used, the first series inductor in the low-pass channel can cause problems, even in the air-line realisation.

It. is an object of the present invention to provide a broadband microwave filtering device with good selectivity and comprising a distributed ladder network with shunt stubs, and based upon the distributed, low-pass prototype with the coupling between adjacent constituent elements as an extra parameter, as referred to above, in such a device the impedances being realisable in air-line form, and possibly can be realised in strip-line form.

According to the present invention a broadband microwave filtering device includes a distributed ladder network as a realisation of a prototype with coupling between adjacent constituent elements, the distributed ladder network having shunt stubs presenting element values in accordance with formulae representing approximations to the Generalised Chebyshev function of equation (3) herein, and with all the transmission zeros at, at least approximately, the same frequency close to the bandedge.

Since the bandedge frequencies are very sensitive to the resonant frequencies of the series arms of the network, such a filter network desirably has series arms which can be adjusted, very accurately, by tuning.

It is required to provide a good broadband match between the input and output of the filtering device over the major part of the passband of the filter. This latter requirement is because any realisation based upon the distributed low-pass prototype with coupling between adjacent constituent elements as an extra parameter, theoretically, should have the impossible feature of zero physical distance between the input and output. Thus, the path along which the major energy content of signals with frequencies in the passband must be physically constructed to provide the minimum amount of reflection at any junction within the device.

According to another aspect the present invention comprises a diplexer in which the high-pass microwave filtering component, and the low-pass microwave filtering component, each comprises a broadband microwave filtering device of one of the forms referred to above, and at least the equivalent of a series feed is provided for the diplexer.

According to a further aspect of the present invention comprises a multiplexer having a combination of microwave passband and stopband filters, at least one constituent filter comprising a broadband microwave filtering device of one of the forms referred to above.

The present invention will now be described by way of example with reference to the accompanying drawings, in which: Figure 1 is of a conventional ladder network, Figure 2 is of a low-pass prototype with coupling between adjacent constituent elements as an extra parameter, Figure 3 is of a diplexer based on the prototype of Figure 2.

Figure 4 is a doubly terminated prototype corresponding to the prototype of Figure 2, but with the Generalised Chebyshev response, in accordance with the present invention, Figure 5 shows the overall response of the prototype of Figure 4, Figure 6 shows impedance inverters in the prototype of Figure 4, Figure 7 is of an interdigital filter with a Chebyshev response, Figure 8 shows the filter of Figure 7 with shunt short-circuited stubs extracted from the unit elements, Figure 9 is of pairs of the unit elements of the filter of Figure 8, and transforming a transformer of frequency dependence

where t is j tan w, through the common shunt element, Figure 10 corresponds to Figure 9 using a particular frequency transformation, Figure 11 is obtained from the network of Figure 10 by transforming the inverters out of the network, Figure 12 is a simple ladder low-pass-high-pass diplexer, Figure 13 is a modification of the first two elements in each channel of the diplexer of Figure 12, Figure 14 is a unity impedance network including an impedance of the diplexer of Figure 13, Figure 15 is a more complicated diplexer arrangement than that shown in Figure 13, Figure 16 shows the network termination part of the arrangement of Figure 15, Figure 17 is a modification of the network termination part of Figure 16, Figure 18 is of a diplexer according to one embodiment of the present invention, and shows normalised stub admittance values, Figure 19 is of a C-section, such C-sections being used in parallel, together with capacitive blocks, to provide the low-pass side of the diplexer of Figure 18, Figure 20 is a section of an air-line physical realisation of the diplexer of Figure 18, Figure 21 indicates the manner of construction of a triple layer strip-line basic section of the low-pass side of the diplexer, Figure 22 shows the terminal connections for the basic section of Figure 21, Figure 23 shows the equivalent circuit for the basic section of Figure 21, Figures 24, 25 and 26 correspond, respectively, to Figures 21, 22, and 23, but instead relate to the basic section of the bandpass filter of the diplexer, and Figure 27 indicates the coupling and even mode fringing capacitances of the basic section of Figures 24 to 26.

A conventional ladder network with shunt tabs of a microwave filtering device has the equivalent circuit shown in Figure 1. The insertion loss L for the low-pass channel of such a filter is:

L = 1 + 62cosh 2 (n cosh~l ttadnn 0 ) (2) where E iS a constant, n is the number of constituent elements of the filter, o is the electrical length at the centre of the required bandwidth of the filter, and UO is a constant dependent on the required bandwidth of the filter. Such a filter has good selectivity if n is sufficiently large. The dual higfp-pass version may readily be formed.

Since the elements are required to be located in close proximity coupling naturally exists between adjacent constituent elements of the network of Figure 1, e.g. C1 to C3, and this coupling can be introduced as an extra parameter to give the low-pass prototype shown in Figure 2. In principle this prototype is easier to be realised as a distributed network if the element values are reasonable. For a diplexer it is normal to employ an even degree network with parallel connections, as shown in Figure 3, and comprising a modification of the prototype of Figure 2.

There are, however, problems in obtaining such a realisation of the prototype, and especially in obtaining a diplexer. In particular, there is difficulty in providing impedances in strip-line form, and in airline form.

In accordance with the present invention there is provided a microwave broadband filtering device as a realisation of the prototype of Figure 2, and with the "Generalised Chebyshev" response; having all transmission zeros at the same frequency close to the bandedge; and, for a diplexer, a series feed or the equivalent. Thus, the insertion loss L is given by:

L = 1 + 62cosh2 [(n-l)cosh~/( cosh''w (3) /7@ ) + cosh~ i] (3) where a is a constant, and w is the normalised frequency value based on the frequency at the bandedge being taken as unit i.e., 1 radian per second.

The transmission zeros are of order n-1αt#=#1/#a, with a single one at #=α, and n is odd. Such a prototype, which is doubly terminated, is shown in Figure 4, For the direct evaluation of equation (3), for different ranges of w, there are the following equivalent forms:

~ jF-a L = 1 + 2cos2 [(n-t) cosh fi + cos-tt"] (4) OwJ L =1+ 2cosk2kn-f)coSh-I g)jw-a f cosh - (5) m7i-aw and L = 1 + 62 cosh2 [(n-t) sinh~l + cosh it (6) 6 0c with the overall response shown in Figure 5.

The first point of the important points of the characteristic is the minimum loss in the stopband Lmin indicated in Figure 5 at #=#m; In this region, and as obtained from equation (6), the insertion loss L is: L = I + # Fn (#) (7)

with Fn (t) = (n-l) sinh 8 cosh~/ -I + cosh a) (8) Hence JFn -dFn / I t ~ 1ffMn4)Pa at Fn(a)) AT r (9) and tom is deemed as

yielding

and at this point Lmin = 1+ Fn ( m) (12) As shown in FIgure 5, at a frequency #a, below the frequency 1/#a, the insertion loss is again equal to Lmin. Hence, L#Lmin for all ###a. Thus, as also is shown in Figure 5, for a stopband filter in which L#LA for ###c, it is required taht #a##c. An optimum solution to a given stopband filter specification will occur if 0) ca, resulting in a particular, calculable, value for n and a. However, for practical realisation considerations it is more desirable to retain control of a in order to obtain reasonable impedance values, and for tuning purposes, even at the possible expense of increasing n. Thus, it is desirable to assume that, in addition to the passband specification of the filter defined by E, and the stopband specification give by LA and uc, a is prescribed, leaving the specification solely to be met by increasing n. From practical considerations it is desirable that a should be between 0.4 and 0.6.

Hence, if LA=20 log is the passband return loss, and LA is also in dB:-

LA + LR 4 20 log [cosh(n-l)cosh fie ,42 + COSk'a)cJ (13) N.B. #c

or LL cosh4 w( A2+0R) COSh4a)c (14) cosh-l EkJH v77-aa)c2 For example, if LA=60 dB, L,=20 dB, c=1.2 and a=0.6, n7.9, in order to satisfy equation. It is also necessary to check the Lmjn > LA i.e:-

20 log [coshjin-ri sinh-' t C s*~tBm)0 > LA f LR cosh (n-l) > /-msn"f(' LR)j - cosb 1rn sinh -I Wml1 with n exceeding the value in equation (14) and

For the example given above, for #m=3.18, and for n=9 (since n must be odd), equation (15) yields n#11. Hence an 11th degree network must be used.

If a is varied, which is an approximation to the essential feature of the tuning process, of particular interest is any rapid change in insertion loss L near the frequency #c, which essentially is varying the bandedge frequency without a significant change in Lmin.

If a contiguous diplexer is required the 3 dB frequency as to be computed:

E cosh [(nOcosh-{J;2l=1 ~~~~~ = x = cosh cosh or m7t-aw (ni) (Is) i.e. #3#B is approximately

Element values may be scaled to give the 3 dB point at n)--l. In the previous example this scaling is 1.5%.

Having determined the appropriate value for n synthesis of the network must be obtained. An exact synthesis requires the use of numerical factorisation. Also since high degree networks are envisaged difficulties might be encountered in obtaining accuracy. Thus, it is desirable to have explicit design formulae, particularly since the levels of the impedance values are important, and it is very important to know how the levels of the impedance values vary with a.

In order to obtain explicit formulae for element values in "Generalised Chebyshev" filters, for convenience, impedance inverters are used in the prototype of Figure 4, to form the modified prototype of Figure 6.

Kr, r+1 (r=1 to n-1) are the characteristic admittances of the inverters.

With the insertion loss L given by eqation (5), the element values are given by:

Cr,r = Cn+1-r,n+1-r =

r = 1 # (n+1)/2 a Lr,r = r even Cr,r (20)

t= sinh (ksinh'j) These element values are obtained by considering the frequency transformation: cot # # # (21) cor #0

Then L cosh L (n-I) cosh-1 cot8 + cosh1cotO (22) I-a cot 0 cot26" and if d is chosen such that -a = cot#0 (23) then

- sh-l COS t cosh L =1+ 2cosh2 L(n-l)cosh~l cos Icotol (24) where L is the insertion loss response for an interdigital filter with the Chebyshev response, and as shown in Figure 7, (neglecting the 1: -I transformers). In relation to Figure 7:

and Yr,r = Yn-r+1, n-r+1 =

r=1 # 2 and α = cos #0 Extracting shunt short-circuited stubs from the unit elements of the network of Figure 7 provides the network shown in Figure 8, where the coupling elements possess transfer matrices of the form:

Taking pairs of these coupling elements and transforming a transformer of frequency dependence

through the common shunt element, and assuming n to be odd, results in tne network shown in Figure 9. The coupling elements are now ideal inverters of characteristic admittance y, r+1 and the even shunt elements are given by:- t Yr = -Yr,r . (28) t-t Using the frequency transformation implied by equation (21):

where p is the lumped complex frequency variable, there is obtained the network shown in Figure 10. In relation to this network the element values are:

Cr Yr,r z r = I n (30) r even on2 Cr and Lr with Kr,r+1 = Yr, rtl From equation (23):- or

From equations (30), (25) and (26) are obtained the desired design formulae given in equations (18), (19) and (20).

In one particular embodiment n=l l, a=0.6 and E=0.I.

#=0.27595.

From these values the element values are:

K12 = @ 4290 K23 = 2. 1996 K34 = 2. 9155 K45 = 3 4446 K56 = 3. 7237, and # (32) C1 = 0 559 C2 = 1. 579 C3 = 3. 105 C4 = 4 079 C5 = 4 682 C6 = 4. 896, with physical symmetry about the central element.

The ladder form of network shown in Figure Il is obtained by transforming the inverters out of the network of Figure 10.

#1 = 0 . 559 #2 = 0 . 773P/(1+0.6p) #3 = 1 . 311p #4 = 1 . 137p/(1+0.6p) # (33) #5 = 1 . 416 #6 = 1 . 168p/(1+0.6p) If n=9, and retaining the same values for a and @ #,

#@ #1 = 0 . 562p #2 = 0 . 777p/(1+0.6p) #3 = 1 . 317p # (34) #4 = 1 . 140p/(1+0.6p) #5 = 1 . 415p The variation in corresponding element values is less than 0.5%. This compares with greater than 10% in similar degree Chebyshev filters. Hence, if higher degree networks are required, the central elements which form a basic section, and which vary very little, simply may be repeated.

If transformation is made to an appropriate bandpass or low-pass microwave network the limiting elements on realisation are the first two, as can be seen from the particular #@ values in the two examples given above.

Figure 12 shows the simple ladder network of a low-pass-high-pass diplexer.

Since the first elements in each filter are shunt elements, at the common port there is a series feed. In theory, a shunt feed can be used with the dual network. For continuous diplexing it is possible to produce a perfect match at the common port at all frequencies by using complementary filters designed exactly on a singly terminated prototype. For the simple ladder network this can only be done with a maximally flat response, as the equi-ripple response requires elliptic function filters. However, even in this case, the return loss in the passbands must equal the insertion loss in the stopband, which is not a normal requirement. Hence, normally pseudo-complementary filters are used, such as Chebyschev ladder networks, but these only approximate to a good match at the common port, and necessaril.y require a 3dB crossover point. A technique which is more general, which can be applied to channels crossing over at any level, and which attempts to provide the correct match at all ports of the appropriate passbands of the channels, is as follows.

The element values in the simple ladder network of Figure 12 are normalised such that the crossover frequency is to be at 0)=l. Further, a pair of isolated channels designed on a doubly terminated prototype, and frequency scaled to give the required insertion loss at the crossover frequency =0, or the appropriate bandedge frequency with respect to cut-off, is considered. If #=#i are the perfect match frequencies, then in the lowpass channel |#i| < 1, with the reciprocal being applicable in the high-pass channel. Further, the first two elements in each channel are considered to be modified as shown in Figure 13, in which Z1 is the impedance at any point #=#i. Then the impedance of the network shown in Figure 14 is unity at 0)=0)j Hence,

Hence, J()j l ) (35) + + JWj L2 = (35) or Z1 2 jwi j I-jtu C1 12 Hence, by modifying C1 and L2 to C,' and L2, at #=#i the input impedance of the low-pass channel is:

Since I el)j < 1 we may expand this as a power series as:

At #=# the input impedance of the high-pass channel is:

and expanding as a power series:

Thus, for a match at the common port at #=# we must have: ZL+ZH=1 (#=#i) (40) and from equations (37) and (39), for errors of the order of wi : L2' - L2 + C1 - C1' + L1' = 0 (41) 2C1'(L2'-L2)-(C1-C1') = 0 (42) Applying a similar argument to the highpass channel there is obtained the following equations: I ~ I 1 1 1 C2' C2 L1 L1' C1'

Solving the four sets of simultaneous equations in the four unknowns gives:

In the common case, where the highpass channel is obtained from the low-pass channel by replacing # by 1/#, then the design equations simplify to:

From these formulae it can be shown that the network response deteriorates very little even up to the contiguous case.

If the technique described above in relation to the simple ladder network of Figure 13 is applied to the more complicated prototype shown in Figure 15 there is obtained:

where |xi| I This may be expanded as a power series in x12. Since the modifications to C1 and L2 are contained in the real and imaginary parts of the leading coefficients it is possible to evaluate at x=0, at which ZH+ZL=l, to give:

from the real part:-

which gives:-

from the imaginary parts:

which gives:

The final design equations (52) and (54) are valid for any crossover level greater than 3dB.

From a detailed study of equation (50) it is of interest to note than in the passband of the low-pass filter the only element in the high-pass filter which affects the response in a substantial manner is the first element. Thus, as far as the passband performance is concerned, only the single inducator1/C,' needs to be present. Hence, at the non-common port of the low-pass filter it is possible to modify the filter from that shown in Figure 16 to that shown in Figure 17 without a significant change in performance, and design equations (52) and (54) are used.

A similar result applied to the high-pass channel. In the following example, for an ultra broadband microwave diplexer, such a result is obtained in order to create physically realisable dimensions, from an unrealisable situation if the network termination in Figure 16 were to be used:- n=, 9 a=0.439, t=O.l, and cl)c=1.28 #1=0.714p #2=1.004p/(1+0.439p) #3=1.525p #4=1.305p/(1+0.439p) =1.614p Applying diplexer design equations for common junction gives: #1'=1.475 and 2' 1 .200p/( I +0.439p2) For a diplexer from 0'0.5+5 GHz, 2.75 GHz is the quarter wavelength frequency, and: - for low-pass, and 0.285 0.302 p# for bandpass, t if the 3dB point is at #=16.36 .

Thus, it is necessary to use the modified input to each filter since 17, becomes unrealisable physically. Hence, the network is as shown in Figure 18, with the normalised stub admittance values a the conductor and the associated digit. After the appropriate adjustments the tubes are secured by adhesive within the digits, and any excess lengths of the tubes are cut off. The central part 30 of the central digit, between the two conductors, is of brass, and has a projection 31 extending between, but not connected to, the adjacent strips 26. The common web 25 is also connected to the R.F. shield 23. The digits 24 of the comb-line filter extend in a common plane coincident with the axis of symmetry of the common part of the space within the casing, the digits being substantially uniformly distributed along the symmetrical axis of the high pass side.

The low pass side comprises five brass blocks 32 insulated from each other, and from the ground planes of the casing 20, by a polytetrafluoroethylene body 33.

The body 33 also serves to locate the brass blocks 32 within the enlargement 21, so that surfaces 34 of the blocks serve partially to define the part of the space within the casing of the low pass side, but not of the enlargement 21. Thus, the blocks 32 have mutual capacitance, and also capacitance to ground. The blocks 32 and the body 33 cross the central part of the enlargement 21, parallel to the symmetrical axis of the remainder of the space within the casing. The blocks 32 are interconnected by looped conductors 35 forming the C-sections of Figure 19, and comprising transmission line sections. The conductors 35 extend into the part of the space within the casing coaxial with the high pass side. The common plane of the conductors 35 is co-extensive with the common plane of the digits of the high pass side, and are substantially uniformly distributed along the high pass side part of the symmetrical axis of the common part of the space within the casing. The conductors 35 initially are slidable in bores 36 in the blocks 32 for tuning purposes, and subsequently the conductors 35 are secured in the bores 36. The casing is divided into its two parts at the common plane.

The input to the low pass side is from the common input terminal 22 via the associated conductor 27 of the adjacent end digit 24, of the high pass side, which conductor 27 extends beyond the digit 24 to contact the adjacent end block 32 of the low pass side. The output of the low pass side is via an output terminal 37, also comprising a coaxial socket, and located in a part of the appropriate end wall of the casing 20 defining partially the part of the space within the casing coaxial with the high pass side. The output terminal 37 is connected to the adjacent end block 32 via a conductor 38.

Over the broadband of the bandpass filter it is particularly important to maintain a 50 ohm line along the connecting edge of the stubs if a reasonable return loss is to be maintained.

The equivalent shunt fed diplexer has physically unrealisable elements adjacent to the junction.

Due to the limitations upon coupling admittance in a broadband strip-line realisation no planar geometry solution has been found. Two layer structures are releasable for a very restricted range of bandwidths. However, much more flexibility is obtained by using triple layers without having connections between the layers. The basic section used in such a low-pass strip-line filter is formed from the parallel connection of a C-section and a capacitive block. The c-sections are realised in planar geometry, whereas the capacitive blocks are realised in triple layers as shown in Figure 21. The terminal connections are shown in Figure 22, and the equivalent circuit is shown in Figure 23. For the dimensions shown in Figure 21, if the term Y2(C2) is large then:

y L i43x + 0.9 and (55) j = 4 65x + 0 2 t For the bandpass filter the basic section is shown in Figures 24, 25 and 26, corresponding to Figures 21, 22 and 23, where:

YC is approximately 14x + 0.5 Y = 4.65x + 2Cfe # (56) YL = CC Cc and Cfe are the coupling and even mode fringing capacitances of the section shown in Figure 27.

A diplexer with nIl, a=0.6 and E=0. I and with a bandwidth of 5:1 is obtainable. For bandwidths broader than this the bandpass filter becomes difficult to realise, and for narrower bandwidths the low-pass filter becomes difficult. This is because, in equations (55) and (56), for a given coupling. there must be at least a minimum susceptance to ground to separate physically adjacent sections.

Thus, according to the present invention a broadband microwave filtering device with good selectivity includes a distributed ladder network as a realisation of the low-pass prototype, with the coupling between adjacent constituent elements as an extra parameter, as shown in Figure 2, and comprises a distributed ladder network with shunt stubs representing approximations to the "Generalised Chebyschev" function. The transmission zeros are at at least approximately, the same frequency close to the bandedge.

A broadband microwave filtering device according to the present invention has impedances which readily can be realised in air-line form, and possibly can be realised in strip line form. Further, since the bandedge frequencies will be very sensitive to the resonant frequencies of the series arms of the network, such a filter network has series arms which can be adjusted very accurately by tuning. In addition, it is required to provide a good broadband match between the input and output of the filtering device over the major part of the passband of the filter. This latter requirement is because any realisation based upon the distributed low-pass prototype with coupling between adjacent constituent elements as an extra parameter, and as shown in Figure 2, theoretically, should have the impossible feature of zero physical distance between the input and output. Thus, the path along which the major energy content of signals with frequencies in the passband must be physically constructed to provide the minimum amount of reflection at any junction within the device.

A multiplexer having a combination of microwave passband and stopband filters, and in accordance with the present invention, has at least one constituent filter comprising a broadband microwave filtering device according to the present invention.

Claims (10)

WHAT WE CLAIM IS:

1. A broadband microwave filtering device including a distributed ladder network as a realisation of a prototype with coupling between adjacent constituent elements, the distributed ladder network having shunt stubs presenting element values in accordance with formulae representing approximations to the Generalised Chebyshev function of equation (3) herein, and with all the transmission zeros at, at least approximately, the same frequency close to the bandedge.

2. A filtering device as claimed in claim 1 having a network with series arms which can be adjusted by tuning.

3. A filtering device as claimed in claim 1 or claim 2 having the path along which the major energy content of signals with frequencies in the passband provides the minimum amount of reflection at any junction within the device.

4. A filtering device as claimed in claim I or claim 2 or claim 3 in which the impedances are of air-line form or of strip-line form.

5. A filtering device as claimed in any one of the preceding claims with having shunt stubs representing values in relation to equations 18, 19 and 20 herein.

6. A diplexer in which the high-pass microwave filtering component, and the loss-pass microwave filtering component, each comprises a broadband microwave filtering device as claimed in any one of claims I to 5, and at least the equivalent of a series feed is provided for the diplexer.

7. A diplexer as claimed in claim 6 with element values in relation to equations (52) and (54) herein.

8. A multiplexer comprising a combination of microwave passband and stopband filters, at least one constituent filter comprising a broadband microwave filtering device as claimed in any one of claims 1 to 5.

9. A broadband microwave filtering device substantially as described herein with reference to Figures 4 to 19 in combination with either Figure 20 or Figures 21 to 27 of the accompanying drawings.

10. A diplexer substantially as described herein with reference to Figures 4 to 19 in combination with either Figure 20 or Figures 21 to 27 of the accompanying drawings.