US 3634788 A
This invention relates to waveguide filters wherein waveguide sections functioning in the evanescent mode for frequencies in the desired passband are utilized to couple conventional cavity filter sections together. The evanescent sections operate in their normal modes at frequencies higher than the passband frequencies. Suppression devices are then coupled to one or more of the evanescent sections (instead of to the cavity filter sections) to suppress the parasitic harmonic waves while having negligible effect on the evanescent mode passband frequencies.
Claims available in
Description (OCR text may contain errors)
United States Patent  Inventor George Frederick Craven Sawbridgeworth, England  Appl. No. 671,046
 Filed  Patented  Assignee Sept. 27, 1967 Jan. 11, 1972 International Standard Electric Corporation New York, N.Y.
 WAVEGUIDE FILTER 4 Claims, 44 Drawing Figs.
 U.S. Cl 333/73, 333/83, 333/81  Int. Cl ..I-l03h 13/00, H03h 7/10  Field of Search 333/73, 73 C, 73 W, 10, 7, 9
 References Cited UNITED STATES PATENTS 3,215,955 11/1965 Thomas 333/7 2,866,595 12/1958 Marie... 333/10 2,849,683 8/1958 Miller 333/10 3,058,072 10/1962 Rizzi 333/73 W 2,626,990 l/1953 Pierce 333/9 2,623,120 12/1952 Zobel .1 3113/73 W 2,816,270 12/1957 Lewis 1133/) 2,106,768 2/1938 Southworth. 333/73 W 2,866,949 12/1958 Tillotson 333/11 3,451,014 6/1969 Brosnahan... 333/73 3,237,134 2/1966 Price 333/73 W Primary Examiner-Herman Karl Saalbach Assistant Examiner-C. Baraff Attorneys-C. Cornell Remsen, .lr., Rayson P. Morris, Percy P. Lantzy, Philip M. Bolton and Isidore Togut ABSTRACT: This invention relates to waveguide filters wherein waveguide sections functioning in the evanescent mode for frequencies in the desired passband are utilized to couple conventional cavity filter sections together. The evanescent sections operate in their normal modes at frequencies higher than the passband frequencies. Suppression devices are then coupled to one or more of the evanescent sections (instead of to the cavity filter sections) to suppress the parasitic harmonic waves while having negligible effect on the evanescent mode passband frequencies.
PATENTEU JAN] 1 I972 SHEET OlUF 15 1- JZO/ 1 591M Inventor GEORGE F. CRAVEN Agent PATENTED JAN! 1 1972 3534788 SHEET OBUF 15 iii F //M Inventor GEORGE E CRAVEN Byww Agent PATENTEI] JMI I I H12 SHEET USUF 15 Inventor GORG F. CRAVEN vi 74? Agent PATENTEUJmmsm 3634-788 SHEET 080F15 Fmguencyfl/cp In ventor GEORGE E CHM/51v Agent PATENTEI] mu 1 i972 SHEET OSUF 15 I nvenlor Geeks: F. CRAVEN By Agent PATENTED JAN] 1 i972 31634788 SHEET IUUF 15 v L l I l lnvenlor GEORGE F. CRAVEN y Agent amma PATENTEI] JAN! 1 i972 sum 11 0F 15 gam 0000 Fnequenqy Mc/S Inventor GEORGE F. CRAVEN Agent PATENTED mu 1 1912 315347 saw 12 0F 15 Frequenzy Frequency QM- lnvcnlor GEORGE F. CRAVN Agent PATENTED mu 1 m2 3.634.788
SHEET 1 3 OF 15 F/"equeng 54m fiqueng 92.
. lnuen or GEORGE F. CIRAVEN BYW% Agent PATENTEIJJANHIHYZ 3.634188 SHEET 1H0? '15 GEORGE F! CRAVEN MMM;
Agent 7 PATENTED m1 1 1972 SHEET lSUF 15 Q9. J/(ag.
Inventor GEORGE F. (RAVEN By Agent WAVEGUIDE FILTER BACKGROUND OF THE INVENTION This invention relates to waveguide band-pass filters.
Conventional direct-coupled and quarter-wave-coupled filters each have advantages and disadvantages. The advantages of the quarter-wave-coupled filter lies in its small coupling susceptances and consequently, the ease with which a paper design can be realized with practical mechanical tolerances. Initially, this type of filter was preferred because the cavities could be made separately and tuned before assembly. However, now that filters are always made and tuned as a complete unit this advantage is of little value and general, if not complete, preference has passed tothe direct-coupled filter. The direct-coupled filter eliminates the frequency-sensitive quarter-wave couplings and, therefore, is applicable to wider bandwidth designs; it is also shorter (about 75 percent of the length of the quarter-wave filter). This is achieved by substituting one large susceptance for the two smaller susceptances and the quarter-wave coupling between cavities. This makes the direct-coupled filter potentially cheaper although this is offset by the much stricter tolerances, or additional adjustments, that are necessary. The difficulties arise from the highly critical nature of large susceptances which, in the example of a symmetrical diaphragm, for instance, varies as cot (1rd,/2a)(d,1 for large susceptances). Use has also been made of multipost arrangements but this alternative leads to manufacturing difficulties and the thin posts also increase loss.
SUMMARY OF THE INVENTION Therefore, the main object of this invention is to provide an improved waveguide filter wherein the higher order parasitic frequencies are suppressed while having negligible effects on the desired passband signals.
According to the invention there is provided a waveguide band-pass filter having main resonator cavities coupled together by waveguide sections which are evanescent at the passband frequency of the filter.
BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1(a) is the equivalent circuit of two cavities with a conventional quarter-wave coupling,
FIG. 1(b) is the equivalent circuit of two cavities with a conventional direct coupling,
FIG. 1(c) is the equivalent circuit of two cavities with an evanescent mode coupling,
FIG. 1(d) is the voltage transfer equivalent of an evanescent section,
FIG. 2(a) is the bisected circuit of FIG. 1(a).
FIG. 2(b) is the bisected circuit of FIG. 1(c).
FIG. 3 shows a typical attenuation/length characteristic for an evanescent mode coupling measured at 4,000 mc./sec.,
FIG. 4 shows the evanescent mode coupling giving the characteristic of FIG. 3,
FIGS. 5(a) and 5(b) show a symmetrical evanescent mode coupling its equivalent circuit, respectively,
FIG. 6 shows the reactance at a junction of a waveguide terminated in the symmetrical evanescent section of FIG. 5(a).
FIG. 7(a) and 7(b) show an asymmetrical evanescent mode coupling and its equivalent circuit, respectively,
FIG. 8 shows the reactance at a junction of a waveguide terminated in the symmetrical evanescent section of FIG. 7(a),
FIGS. 9(a), 9(b) and 9(0) are plan, side and end views, respectively, of a six-section resonant cavity waveguide bandpass filter with evanescent mode coupling sections,
FIG. 10 is a perspective view, partially cut away, of a modified form of an evanescent mode coupling section,
FIG. 11 shows the curves insertion loss and V.S.W.R. vs. frequency for the filter of FIG. 9,
FIG. 12 shows the measured standing wave pattern in a three-section filter with evanescent mode coupling sections throughout,
FIG. 13 shows the bisected equivalent circuit of a symmetrical network of imaginary characteristic impedance terminated for full energy transfer,
FIG. 14 shows the calculated standing wave pattern for the symmetrical half-section evanescent coupling between cavities of FIG. 13,
FIGS. 15(a) and 15(b) are plan and sectioned side views, respectively, of an evanescent coupling section with a harmonic rejection filter,
FIGS. 16(a) and 16(b) are plan and sectioned side views respectively of an evanescent coupling section with a constant resistance rejection filter,
FIG. 17 shows the transmission response of the filter of FIG. 16,
FIG. 18 is a perspective partially cutaway view of an evanescent coupling section with a resonant slot hybrid junction rejection filter,
FIG. 19 shows the attenuation characteristic of the filter of FIG. 18,
FIG. 20 is a perspective partially cutaway view of an evanescent coupling section with a form of constant resistance rejection filter,
FIG. 21 shows the attenuation characteristic of the filter of FIG. 20,
FIG. 22 shows the equivalent circuit of the filter of FIG. 20,
FIGS. 23(a) and 23(1)) are side and end views, respectively, of an evanescent coupling section with a series resonant rejection filter,
FIG. 24 shows the attenuation characteristic of the filter of FIG. 23,
FIG. 25 is a perspective partially cutaway view of an evanescent coupling section with a form absorption filter,
FIG. 26 shows the standing wave pattern in the coupling section of FIG. 25,
FIG. 27(a) is an end view, and FIG. 27(b) is a half-sectioned side view along the line A-A of FIG. 27(a), of an evanescent coupling section with a low-pass rejection filter,
FIG. 28 shows the attenuation characteristic of the filter of FIG. 27,
FIG. 29 shows a typical frequency characteristic of a waveguide low-pass filter,
FIG. 30 shows the wide-band rejection characteristics of a waveguide low-pass filter, and
FIG. 31(a) is a plan view, and FIG. 31(b) a section on the line 8-8 of FIG. 31(b), of cascaded low-pass filter evanescent coupling sections with a dissipative section in between.
DESCRIPTION OF THE PREFERRED EMBODIMENTS Waveguide at frequencies below cutoff exhibits characteristics common to all nondissipative filter networks in their stop-band region. The characteristic impedance, which is real in the passband becomes imaginary in the stop-band. The propagation constant, which is imaginary in the passband becomes real in the stop-band. The transmission line analogue then has the properties shown in FIG. I(c), FIGS. 1(a) and I(b) represent the corresponding circuits for quarter-wave and direct-coupling, respectively. The conditions for virtual identity between the two circuits of FIGS. 1(a) and 1(b) have been established for example, in An Improved Design Procedure of the Multisection Generalized Microwave Filter," Levy R., I.E.E. Monograph Vol. 232R, Apr. 1957, in which it is shown that if the insertion loss and phase shift of the two networks are the same, the circuit of FIG. 1(a) may be replaced by that in FIG. 1(b). In a similar way the network of FIG. 1(c) may be replaced by that in FIG. 1(b). The input reactance of the network of FIG. 1(b) is equivalent to a certain line length in the network of FIG. 1(c) and thus, the cavity length will be generally less when evanescent couplings are used. The insertion loss of the two networks may then be made identical by appropriate choice of the parameters in the evanescent section.
The conditions for equivalence between the two networks of FIGS. l(b) and 1(c) can be readily derived from the standard transmission line equations. As the networks are symmetrical they may be bisected (FIGS. 2(a) and 2(b) nd the openand short-circuit admittances derived. If the respective open circuit and short-circuit admittances can be equated and solved for physically realizable values then the two networks may be made identical at a given frequency. The short-circuit admittances for the networks of FIGS. 2(a) and 2(b) are, respectively.
Y,,,.,=j Yo cot I) Y,,,. =jB j m" coth (pl/2 The open-circuit admittances for the same networks are:
'M/=1 (v+ (3) where d) cot (B ,2Yv)
Y =jB,jY0i tanh (Ill/2) Equating the short-circuit admittances.
Y0 cot 0=B,+Y0i coth (111/2) (5) Equating the open-circuit admittances.
Y0 cot (0+)=B,+Yoi tanh (Ill/2) Expanding the left hand side (6) and substituting for cot 6 (from (5)) we have, solving for cot 0,
B Yoi B Yet 71 B Yoi 71 Got Yet h 1 h w To (cot tan The extreme values that cot can have are when 71- wand l 0. In the first case both coth and tanh g tend to unity. Thus, cot w as the line length tends to infinity which is the expected result. In the second case as 'yl 0 the expression reduces to:
which is obviously true. Equation (5) states simply the second requirement for equivalence i.e., the relationship between the line length in FIG. 2(a) and the combined lumped susceptance including the junction susceptance and the (inductive) characteristic admittance ofthe line.
The notion of energy propagating freely through a nondissipative line (or guide) of imaginary characteristic impedance, with the propagation constant real and assuming large values, may conflict with the intuitive notions held by many. This con cept amounts to the belief that the ratio of input voltage, E to output voltage, E is given to good accuracy by the equation:
(E,/E )=ev where y propagation constant I =length However, this is only true if the network is terminated in its characteristic impedance (imaginary). With other terminations the results can be very different. For instance, if the termination is capacitive (FIG. 1d) the ratio E /E is given by For large values of y], cosh 71 and sinh y! are nearly equal. Thus, for example, if Z0 -x E /E is quite small and may be zero. The circuit then behaves as a series resonant circuit with a large resonant rise in voltage across the inductance and capacitance. Clearly, a wide range of values including E,/E =l is obtainable. A general engineering conception of the behavior is simply to say that the reactive insertion loss of evanescent guide may be tuned out" by a suitable termination. Such evanescent waveguide sections are generally described on my copending application, Ser. No. 643,279. It roughly corresponds to neutralizing the effect of a shunt susceptance by adding one of equal value (but opposite sign) in shunt with it.
The design of direct-coupled filters is described in Direct Coupled Filters, Cohn S. B., Proc. I.R.E., Feb. 1957. The
which for large values of b is, to good accuracy,
10 Iog The insertion loss of a section of evanescent guide interposed between matched guides will be a function of the guide length and junction susceptance. The guide length will normally be the principal factor, the insertion loss being:
and M cutotf wavelength A0 free space wavelength.
The junction susceptance is generally quite large but the insertion loss from this effect is less than might be expected. The reason for this is that the field from the junction spreads into the evanescent section causing the initial rate of decay to be less than e1 FIG. 3 shows the insertion loss, as a function of length, of a typical section shown in FIG. 4 of l.00OXO.667- inch waveguide 41 coupling 2.000XO.667-inch waveguide 42. Except when the evanescent section is very short (and ultimately degenerates into a thin iris) the curve is a straight line. Thus, only two properly chosen measurements are essential in order to establish a design curve. Comparing the calculated (eq. (10)) and measured curves it will be seen that they are approximately 2 db. apart and parallel within about I db. over the range. Therefore, eq. (10) will give a very accurate guide to the effect that minor variations in A, will have on the measured attenuation curve. It is of interest that the difference between the two curves is roughly one-half the insertion loss of an iris of the same dimensions as the cutoff guide. This is a useful rule of thumb correction, if it is desired to use the theoretical curve for ratios a/a (a'la is defined in FIGS. 5 to 8).
The practical design of a filter is carried out in the following way. The values of the prototype sections are calculated using the equations in FIG. 2 of Cohn's paper. The susceptance values are then determined from FIG. 5 of the same paper. The insertion loss is then calculated from eq. (9) of this specification. A width dimension, which is well beyond cutoff at the passband frequency is chosen, and the input reactance of the section is then determined (FIG. 6 or 8). The halflength of the cavity is then found from:
Where tan (l,/Z0) (see FIG. 6 or8 for value of(x/Zo). If the cavity is physically symmetrical about the electrical center i.e., the input reactance at each end is identical) the total length will be twice the value found in eq. (ll); otherwise the second electrical length is calculated. If the total cavity length is unreasonably short as a consequence of the end loading, a second trial using a smaller guide width for the evanescent section will be necessary. The length of the section which will yield the desired insertion loss is then found from a curve such as that shown in FIG. 3.
Where the precise value of the filter bandwidth is not very important an approximate design technique using eq. 10) with an approximate correction for the discontinuity, can be employed. The desired characteristic maximally flat or Tchebycheffwill be obtained even if the theoretical attenuawhere tion curve is in error. This follows because the correct ratio between the insertion lossvalues will be realized. The overall I error in bandwidth can be gauged roughly by noting that an error of 0.5 db. in all of the intercavity coupling sections will produce an error of about 5 percent in the bandwidth of a typical filter.
So far only the intercavity couplings have been discussed as it is in these sections that evanescent couplings have their chief application. Susceptances at the end of the filter are usually of a comparatively low value and may, conveniently, be of the conventional type. However, where it is desirable to preserve the same type of construction throughout the filter it is possible to use evanescent sections at the end.
FIGS. 9(0), 9(1)) and 9(0) show a six-section maximally flat direct-coupled filter with a 3 db. bandwidth of 37 mc./sec. The 3 db. limit frequencies are 3,983 mc./sec. and 4,020 mc./sec. the construction being in 2=2/3 in guide. There are evanescent mode coupling sections 1 between the resonant cavities 2. Each cavity 2 has a 2-BA coarse-tuning screw 3 and an 8-BA fine-tuning screw 4, and each evanescent section 1 has A O-BA coupling adjustment screws 5. Inductive arises 6 at the end of the filter are conventional.
The physical dimensions of the filter are given below:
The first step in designing the filter was to calculate the values of the prototype sections using Cohns paper. For a maximally flat filter, using the same notation as Cohn, the values are:
The guide wavelength tenns A, and A are the values occurring at the 3 db. band limits.
Taking x,,,=11.207 cm. and A =l0.981 cm. F0.032l. Substituting for L and g in 12) the values of B,, H, are:
The value of the outer susceptances B and 13,, are obtained by substituting g,,=L in 12) which gives The insertion loss of 8, B and B are from (9):
L, =22.49 db=L L =28.23 db=L L ,=29.57 db.
The dimensions of the evanescent guide section may now be calculated. Choosing a symmetrical junction with guide widths of l in. (evanescent section) and 2 in. (propagating section) respectively, the value XZ,,/2a)t may be obtained from FIG. 6. For a'/a=0.5;)t,,=l1.l cm. X(Z,, normalized) is given by X=0.3l. From (II) the electrical half-length of the cavity or a total length, for identical junctions at each end of the cavity, of 145.6". This gives a value of l 145.6Xl I.l/360)=4.49 cm. (or 1.77 in.)
Reducing this value by percent to allow for tuning screws we have:
The details of the inductive irises at the end of the filter remain to be settled. These are' conventional symmetrical irises the dimensions of which may be obtained from any standard text.
The performance of the filter is shown in FIG. 11.
The field distribution in evanescent modes conveying full power through the section is of interest. FIG. 12 shows the measured field distribution in a typical three-section filter using evanescent couplings throughout. The filter was constructed with a slotted line section and measurements made with a probe (with very small insertion) in the usual way. When measuring the field in the various sections it was necessary to close up the slot in order to prevent radiation from the asymmetrical end sections. This was achieved by fitting a number of tongued sections (similar to the probe section) into the slot. These were linked to each other and attached to the probe so that they travelled along the guide with it.
FIG. 14 gives the calculated field along a transmission line of imaginary characteristic impedance which is terminated for full energy transfer (HG. 13). The network is symmetrical and it is, therefore, sufficient to bisect the network and calculate E lE as a function of 1 (equation (8)). The values of y and x, (x, is the equivalent reactance presented by the cavity at the line terminals) are chosen for a typical example. It will be seen that a standing wave exists on the line reaching zero in the dissipationless case at the center and comparatively small values if the loaded Q is high. The general character of results is the same at that shown in the experimental results of 12.
It is of interest that the magnitude of the electric field behaves in much the same way as if an iris of very high susceptance were located at the center of the intercavity coupling section. The end sections behave differently owing to their asymmetrical nature. The semiresonant rise in voltage predicted by eq. (8) can be clearly seen in FIG. 12.
The essential difference between this type of band-pass filter and conventional types lies in two features: the physical length of the coupling sections and the nature of the field existing within the sections. The insertion loss of an evanescent section depends both on its width and length and, therefore, the length may vary over a wide range of values. This means that the filter will be somewhat longer than the conventional direct-coupled type although, in general, it need not be longer than the quarter-wave coupled filter. This is a disadvantage where the shortest possible filter is desired. However, where a modest increase in length can be tolerated the filter has a number of advantages.
a. The freedom of choice in the length of the coupling sections means that the overall length of a given type of filter can be maintained irrespective of its center frequency. This follows, because of the longer wavelengths the greater cavity length is obtained by reducing the length (and if necessary, the width) of the evanescent section. Thus, only one fixed body size, with appropriate drillings, needs to be retained for a complete waveguide band. This simplified storage problems in production. The inserts, which form the evanescent sections, and are considerably more robust than conventional irises, are then the only variables.
b. The mechanical tolerances for large values of B, by whatever method that are obtained, are very severe. For example,
if the coupling secfions in the center of the filter of FIG. 9 are I replaced by symmetrical inductive irises of the appropriate value, the tolerance on the width dimension (i.e., the gap between the irises) is only one-third of that which may be permitted with the corresponding evanescent guide dimension. Other types of obstacles, such as holes and multipost arrangements require even stricter tolerances. Therefore, filters in which the bandwidth is significantly narrower than about I percent (the above example) will require considerably less severe tolerances if evanescent coupling sections are employed.
c. The style of construction considered so far can be produced by inserting milled blocks in waveguide in order to produce the evanescent sections. Another method is to spray metal on a suitable mandrel. The filter can be produced in two halves by milling from solid material. This method of construction can be expected to yield very high precision and has the advantage of eliminating soldered joints.
d. Variable bandwidth filters can be produced by using variable-cutoff evanescent coupling sections such as that shown in FIG. 10, which incorporates a movable block 7 locked in any desired position by a screw 8 and additional to fixed blocks 9 and 10.
e. The most important feature of this filter is its potential for eliminating the parasitic passbands that are characteristic of all band-pass filters at microwave frequencies. Normally, suppression of these passbands requires an additional filter which then affects the midband insertion loss and reflection. However, with evanescent coupling sections this disadvantage is avoidable. One or more of the coupling sections can be so designed that the frequency it is desired to suppress propagates in the coupling section instead of being an evanescent wave, like the desired passband. It is then possible to insert parasitic suppressors in the section concerned which will have a very large effect on a progressive wave, but none on an evanescent wave.
Conventional resonant-cavity waveguide band-pass filters suffer from unwanted transmission bands above the desired passband. These passbands occur for two reasons. Firstly, in direct-coupled resonant-cavity filters resonance occurs at frequencies for which the electrical length of the cavity is n. Ag/Z (where n is an integer and Ag is guide wavelength). In addition to these multiple resonances, which are approximately in harmonic relationship, further resonances can be expected as a result of higher order modes propagating. In general, individual cavities tend to resonate at slightly different frequencies and interaction between these resonances occurs. As a result, multiple narrow passbands occur at unpredictable frequencies and, therefore, there is no way of knowing whether a filter will provide satisfactory harmonic rejection, for example, before it is built. If, in addition, a specific filter design is to be tuned over a wide range of frequencies the danger of harmonics penetrating the filter becomes great. In these circumstances, where suppression of high-power harmonics is essential, it is necessary to cascade a second filter giving broad band suppression. This may take the form of a corrugated (or waffle-iron) low-pass filter or a leaky wall filter. The former is based on the classical lumped circuit analogue and is a reflection filter; the second permits the harmonics to leak through the wall of the filter to auxiliary guides where they are dissipated. The difficulty with both of these filters is that they affect the passband performance adversely. Both filters add unavoidable dissipation and reflection that cannot be matched out (because the phase of the reflection varies rapidly with frequency).
The behavior of filters using evanescent mode couplings is completely different from conventional iris-type filters at frequencies will above type filters at frequencies well above the main transmission band. Above a critical frequency the coupling section propagates and the resonant cavities formed by the evanescent coupling sections virtually disappear. Instead, the filter behaves, practically, as a straight through device of nearly zero attenuation and therefore, only has minor internal reflections. In this form it is worse than the conventional filter, but the elimination of large internal reflections in the filter is a prerequisite to providing predictable suppression of unwanted frequencies. The second necessary condition is that the coupling waveguide should behave differently for the passband and unwanted frequencies, respectively. In this way devices which affect the unwanted frequency will have little or no effect on the desired frequency. The behavior of evanescent couplings, as described above, suits them to this application.
In general, the suppression of unwanted frequencies presents itself in two possible forms. One occurs when the filter functions in a closed circuit e.g., it provides filtering in the output of an oscillator. The unwanted frequencies, virtually entirely harmonics, are known. If the device is a highpower oscillator a very large amount of suppression may be necessary, but it will be required at specific frequencies i.e., narrow-band rejection circuits of very high rejection will be required. The second example occurs when the filter is located in an open-circuit system i.e., a system connected, for example, to a wide-band aerial. Signals over a wide band of frequencies will be likely to be received and, thus, wide-band suppression is necessary. In general, the degree of suppression required will be required to be less than the first example.
FIGS. 15(a) and 15(b) shows an evanescent coupling section 11 between resonant cavities 12 incorporating a narrowband harmonic rejection filter. At the desired transmission band the coupling section is designed to be evanescent but the dimensions are chosen so that propagation occurs at the frequency to be suppressed. A three-section rejection filter is shown, the stubs 13 containing semiconductor tuning pistons 14 being conveniently connected to the guide by the broadband resonant slots 15. Approximately 35 db. rejection per section can be obtained in a typical example over a relatively narrow band. Thus, a three-section filter built in the coupling sections as shown gives rejection, at a given frequency, in excess of I00 db. If it is desirable to absorb the undesired harmonic, rather than reflect it back to the source, the filter can be converted into a constant resistance filter (British Pat. No. 1,018,923 G. F. Craven 7) and the energy absorbed into the matched load 16, FIGS. 16(a) and 16(b) in which like references have been used as for FIG. 15. The rejection of one section is largely lost but the advantage of being able to absorb the energy is often significant. The performance of a filter using one of the sections of FIG. 16 is shown in FIG. 17. The midband insertion loss of this section (0.4 db.) compared with a section in which no rejection sections were included was identical within the limits of measurement. The reason for the absence of additional loss is that the guide section incorporating the load is beyond cutoff at the main transmission frequency and can easily be made sufficiently long to prevent measurable loss. In this respect, this form of the rejection filter has something in common with leaky wall filters which dissipate the energy by coupling to a large number of guides (which are beyond cutoff at the main transmission frequency). However, this filter avoids the main disadvantage of the leaky wall filter: the main transmission frequency propagates in the dominant mode and therefore, harmonics propagate in several modes. Multiple coupling to all possible modes is then necessary and this produces measurable insertion loss at the main frequency. In many applications the present filter is likely to be superior to the leaky wall filter, because the guide size of the coupling section may be chosen so that the harmonic to be suppressed propagates in the dominant mode only (the main frequency propagating in an evanescent mode). Thus, the present filter will be much simpler, have less insertion loss and combine the functions of band-pass filter and harmonics suppression filter.
Since the filter of FIG. 15 is essentially a reflection filter, only one suppression filter of this type is pennissable per complete band-pass filter. This is because, if additional evanescent coupling sections are converted into suppression filters, mutual cancellation can be expected with further parasitic passbands resulting. Thus, the amount of suppression that can be obtained is limited to the number of suppression sections that can be contained in one coupling section.
However, this limitation does not apply if the suppression filters are dissipative. Ordinarily, dissipation in the main coupling arm is not acceptable, but if the suppression filter is designed as a constant resistance type a considerable resistive attenuation is obtainable at the parasitic frequency. At the desired passband frequency the coupling guide is evanescent and therefore no dissipation can occur.
- Another form of narrow-band suppression filter is shown in FIG. 18. Two resonant cavities 17 are coupled by an evanescent coupling section I8. Resonant coupling slots 19 .couple the evanescent section 18 to a length of circular