V I0

AD-A234 741

ULTRAFAST TUNABLE MICROWAVE FILTERPHASE II, TASK 2

Adolph PresserRCA LaboratoriesDavid Sarnoff Research CenterPrinceton, New Jersey 08540

APRIL 1982

Final Report for Period 15 May 1980 to 30 April 1982

Prepared forDepartment of the NavyNaval Research LaboratoryWashington, D.C. 20375

DTIC FILE COPY 91 4 09 uII ~ ~ ~ ~ ~ 9 4ii~liHi | iliIH

SECURITY CLASSIFICATION OF THIS PAGE (When Data Entered)

READ INSTRUCTIONSREPORT DOCUMENTATION PAGE BEFORE COMPLETING FORM

1. REPORT NUMBER 2. GOVTACCESSIONNO. 3. RECIPIENT'S CATALOG NUMBER

4. TITLE (and Subtitle) 5. TYPE OF REPORT & PERIOD COVERED

Final Report

ULTRAFAST TUNABLE MICROWAVE FILTER (5-15-80 to 4-30-82)6. PERFORMING ORG. REPORT NUMBER

RCA-PRRL-81-CR-33

7. AUTHOR(s) B. CONTRACT OR GRANT NUMBER(s)

A. Presser N00173-79-C-0186

.. PERFORMING ORGANIZATION N&ME AND ADDRESS 10, PROGRAM ELP.AENT, PROJFCT, TASKAREA & WORK UNIT NUMBERS

RCA Laboratories

Princeton, New Jersey 08540 Phase II, Task 2

11. CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE

Department of the Navy November 1981

Naval Research Laboratory 13. NUMBER OF PAGES73______Washington, DC 20375 7

15. SECURITY CLASS. (of this report)

14. MONITORING AGENCY NAME &ADDRESS Unclassified(if different from Controlling Office)

15a. DECLASSIFICATION/DOWNGRADINGSCHEDULE

N/A

16. DISTRIBUTION STATEMENT (of this Report)

A~t'POV1 'FOR PUrBLIC RELEa-S

DISTRIBUTION U1NLIMITW

17. DISTRIBUTION STATEMENT (of the abstract entered in Block 20. if different from Report)

18. SUPPLEMENTARY NOTES

19. KEY WORDS (Continue on reverse side if necessary and identify by block number)

FilterVoltage tunableVaractor

FET two-section filter20.ABSTRACT (Continue on reverse side if necessary and identify by block number)

r' Varactor-tunable resonant filter elements that use the negatve

resistance of an FET to overcome the losses of the varactor and the

circuit were used to construct two two-section filters.

The two filters operate in the 8.2- to 10.3-CHz frequency range.

The tuning bandwidth of both filters is 2.0 GHz. Typically, the

midband insertion loss over the tuning band is 0 4:2 dB: the 3-dB

bandwith is 29 i 1 MHz.DD FORM 1473

1 JAN 73 "

SECURITY CLASSIFICATION OF THIS PAGE (When Data Entered)

PREFACE

This final report describes the work done at the Microwave Technology

Center of RCA Laboratories during the period from 15 May 1980 to 30 April

1982, in performance of a Phase II, Task 2 program, sponsored by the Naval

Electronics System Command, directed by the Naval Research Laboratory under

Contract No. N00173-79-C-0186. The work of Phase II, Task 1, covering

the period from 1 June 1979 to 15 May 1980, was described earlier in a

separate interim report. F. Sterzer is the Center's Director, E. F. Belo-

houbek was the Project Supervisor, and A. Presser was the Project Scientist.

I..-

iii/iv

TABLE OF CONTENTS

Section Page

I. INTRODUCTION......................................................... 1

II. FILTER DESIGN........................................................ 3

A. Design Approach.................................................. 3

B. Filter Components................................................ 5

1. Resonant Element............................................ 5

a. Varactor................................................ 5

b. Negative-Resistance FET Circuit......................... 5

C. Construction............................................ 8

2. Coupling Networks........................................... 9

a. Unsymmetric Divider..................................... 9

b. Symmetric Impedance Inverter............................ 10

C. History of Resonant-Element Development........................ 11

ID. Experimental Element Evaluation................................. 15

1. Selectivity................................................. 19

2. Tuning Range................................................ 21

3. Signal Input Effects........................................ 23

4. Noise Figure................................................ 25

5. Temperature Effects......................................... 27

6. Tuning Speed................................................ 27

E. Prototype Filters............................................... 29

1. Single-Section Filter....................................... 30

2. Two-Section Filter.......................................... 30

111. FINAL FILTER PERFORMANCE............................................ 45

A. Single Element.................................................. 45

B. Two-Section Filter.............................................. 47

IV. CONCLUSIONS AND RECOMMENDATIONS..................................... 63

REFERENCES................................................................ 64

V

LIST OF ILLUSTRATIONS

Figure Page

1. Block diagram of multisection bandpass filter with impedance

inverters .............................................................. 4

2. Idealized equivalent circuits of active resonant element ......... 4

3. Measured junction capacitance, resistance, and Q of typical

varactor ............................................................... 6

4. Equivalent of negative resistance producing FET circuit .......... 6

5. FET equivalent circuit for resonant element simulations .......... 7

6. Calculated impedance of FET circuit as shown in Fig. 4 ........... 7

7. Resonant element construction: (a) realizable configuration;

(b) equivalent circuit of configuration .............................. 8

8. Unsymmetric impedance divider ...... .............................. ... 10

9. Symmetric impedance inverter ...... ............................... ... 11

10. Earlier symmetric resonant element realization with two FETs

and two varactors ....... .............................................. 12

11. Equivalent circuit of symmetric element .............................. 13

12. Symmetric element diagram of experimental single-section filter

configuration (drain back to back): (a) tuning in drain circuit;

(b) tuning in gate circuit ...... .................................... 14

13. Symmetric element diagram of experimental single-section filter

configuration (gate back to back): (a) tuning in gate circuit;

(b) tuning in drain circuit ...... ................................... 14

14. Photographic close-up of final element circuit with one FET

and one varactor ....... ............................................... 16

15. Resonant element module ....... ........................................ 17

16. Test fixture for resonant element evaluation ..................... 18

17. Manual network analyzer test setup for transmission and

reflection evaluation ...... ......................................... 20

18. Measured 3-dB bandwidth and VSWR vs midband frequency of

single-section filter with high selectivity and small bandwidth

deviation over tuning band ...... .................................... 20

19. Varactor and gate voltages vs midband frequency in broadband

tuning experiment with two varactors in series ................... 22

vi

LIST OF ILLUSTRATIONS (Continued)

Figure Page

20. Measured 3-dB bandwidth and transmission gain vs midband

frequency (broadband experiment) ...... .............................. 22

21. Measured VSWR and transmission gain (broadband experiment) ....... 23

22. Measured transmission passband response at a given filter

setting for three different input power levels ................... 24

23. Test setup to evaluate IMD and other two-signal effects .......... 25

24. Measured two-signal intermodulation dist rtion (absoluLe

and relative) as a function of input power level ................. 25

25. Medsured noise figure at different frequency settings ............ 26

26. Noise figure test setup ...... ....................................... 27

27. Measured effects ot tF.nperature upon 3-dB bandwidth, midband

frequency, and transmission gain ......... ....................... 28

28. Test setup to evaluate filter tuning speed ........................... 29

29. Single-section filter prototype ...................................... 31

30. Varactor and gate voltage settings vs midband frequency ............. 32

31. Measured 3-dB bandwidth vs midband frequency ......................... 32

32. Measured input and notput VSWR vs midband frequency .................. 33

33. Measured transmission passband response of single-section

filter (Vv = 0 V) ....... ............................................. 33

34. Measured transmission passband response of single-section

filter (Vv = 5 V) ....... ............................................. 34

35. Measured transmission passband response of single-section

filter (Vv = 10 V) ........... .......................................... 34

36. Measured transmission passband response of single-section

filter (Vv = 20 V) ....... ............................................ 35

37. Measured transmission passband response of single-section

filter (Vv = 30 V) ....... ............................................ 35

38. Measured transmission passband response of single-section

filter (forward and reverse power flow) ...... ....................... 36

39. Measured input and output reflection coefficient in passband

of single-section filter ...... ...................................... 36

vii

LIST OF ILLUSTRATIONS (Continued)

Figure Page

40. Measured transmission response of a single-section filter

adjusted for high selectivity ........ ........................... 37

41. Measured phase response in passband of high-selectivity

filter ............................................................... 37

42. Measured group-delay passband response of high-selectivity

filter ............................................................... 38

43. Measured VSWR in passband of high-selectivity filter ............. 38

44. Two-section filter prototype ...... .................................. 39

45. Varactor and gate voltage vs midband frequency of each

element in two-section prototype filter ............................. 40

46. Measured 3-dB bandwidth and VSWR vs midband frequency of

two-section prototype filter ......................................... 41

47. Measured input and output reflection coefficient passband

response of two-section prototype filter ............................ 41

48. Measured transmission passband response for forward and

reverse power flow of two-section prototype filter ................... 42

A9. Measured passband response of experimental filter with

Chebyshev response .................................................... 42

50. Measured selectivity of experimental two-section filter

with Chebyshev response ...... ....................................... 43

51. Measured phase passband response of Chebyshev filter ............. 43

52. Measured group-delay passband response of Chebyshev filter ....... 44

53. Measured passband return loss of Chebyshev filter ................ 44

54. Measured midband frequency vs varactor voltage for four

different resonant elements ...... ................................... 45

55. Measured gate voltage for 0-dB transmission loss of four

different resonant elements at given varactor voltages ........... 46

56. Measured 3-dB bandwidth of four different elements for given

varactor voltage settings ...... ..................................... 46

57. Photographic close-up of final two-section filter circuit ........ 48

58. Photographic view of final two-section filter .................... 49

viii

LIST OF ILLUSTRATIONS (Continued)

Figure Page

59. Measured varactor voltages vs midband frequency (final

filter No. 1) ....... ................................................. 51

60. Measured gate voltages at midband frequency settings (final

filter No. 1) ........................................................ 52

61. Measured 3-dB bandwidth and noise figure at midband

frequency settings (final filter No. 1) ...... ....................... 52

62. Measured transmission gain and VSWR at midband frequency

settings (final fiiter No. 1) ...... ................................. 53

63. Measured transmission passband response for forward and

reverse power flow (final filter No. 1) ...... ....................... 53

64. Measured phase passband response for forward and reverse

power flow (final filter No. 1) ...... ............................... 54

65. Measured input and output reflection coefficient in passband

(final filter No. 1) ....... .......................................... 54

66. Measured forward and reverse group delay in passband (final

filter No. 1) ....... ................................................. 55

67. Measured midband response for three input power levels (final

filter No. 1) ........................................................ 55

68. Measured group-delay midband response for three input power

levels (final filter No. 1) ...... ................................... 56

69. Measured varactor voltage for each element at given midband

frequencies (final filter No. 2) ...... .............................. 57

70. Measured gate voltages for each element at given midband

frequencies (final filter No. 2) ...... .............................. 57

71. Measured transmission gain and VSWR vs midband frequency

(final filter No. 2) ....... .......................................... 58

72. Measured 3-dB bandwidth and noise figure vs midband

frequency (final filter No. 2) ...... ................................ 58

73. Measured transmission passband response for forward and

reverse power flow (final filter No. 2). ............................ 59

74. Measured phase passband response, forward and reverse

(final filter No. 2) ....... .......................................... 59

ix

LIST OF ILLUSTRATIONS (Continued)

Figure Page

75. Measured input and output reflection coefficient frequency

response (final filter No. 2) ...... ................................. 60

76. Measured forward and reverse group-delay passband response

(final filter No. 2) ....... ........................................... 60

77. Measured midband response for three input power levels

(final filter No. 2) ....... ........................................... 61

78. Measured group-delay midband response for three input

power levels (final filter No. 2) ...... ............................. 61

LIST OF TABLES

Table Page

1. Two-Section Filter No. I ...... ..................................... 51

2. Two-Section Filter No. 2 ...... ..................................... 56

x

SECTION I

I NTRODUCT ION

Fast, electronically tunable high-Q filters are desirable components for

many EW applications. Tuning speed can be gained in these applications with

varactor tunable filters in place of the commonly used YIG filters, which are

inherently slow. At increasing frequencies above 1 GHz, however, varactor

losses in filters limit selectivity severely. The proposed approach to over-

come varactor losses with the negative resistance of GaAs FETs was shown to be

feasible during the Phase I effort under Contract No. N00173-77-C-02Ol. These

earlier efforts resulted in a single-section filter performance with a 3-dB

selectivity of 20 Mttz at 9 GHz over a 500-Nfiz tuning band; two-section opera-

tion with this approach was also demonstrated. The subsequent Phase II, Task I

effort of this program (Contract No. N00173-79-C-0186) used the approaches of

Phase I in the su(cessfiil design and construction of three-section filters

tunable over the 9.0- to 9 .6-G-lz frequency range. The results of this work

were reported i:i a separate interim report covering the period from 1 June 1979

to 15 May 1980.

The goals of the Phase II, Task 2 effort, reported here, were to extend

tuning bandwidth capabilities over a multigigahertz frequency band and to

demorlstate Ligh-selectivity performance in single-section filters and to

design, construct, and deliver two two-section prototype filters with the

following performance objectives:

Frequency 7-11 GHz

Filter Pass Bandwidth 30 MHz at 30-dB point

Passband Insertion Loss <2 dB

Tuning Voltage 0-30 V

Tuning Speed <30 ns

The major contributions of this program were:

* The redesign of the active resonant element to eliminate an inherent

coupling mechanism that is detrimental to multisection filter perfor-

mance.

* The demonstration of a basic 3.6-GHz tuning capability for a single

resonant element.

I

* The successful design and construction of two two-section filters

tunable over a 2.0-GHz ±20-MHz frequency band.

* The attainment of 29 ±11-MHz selectivity at the 3-dB points.

* The identification of problem areas, like noise figure and saturation

effects, to be solved (luring subsequent efforts.

Two prototype two-section filters were delivered at the end of this pro-

gram which had the following performance:

Unit 1 Unit 2

Midband Frequency, f 8.26-10.25 GHz 8.27-10.30 GHzo

3-dB Bandwidth, BW 27 ±9 MHz 36 ±4 MHz

Midband Insertion Loss, L 0 ±1 dB 0 ±2 dB

Rejection, f +-2 BW ............... 24 dB ................0

Noise Figure, F 21 ±2 dB 24 ±1 dB

Tuning Voltage ................ 0-30 V ..............

Size ........ 38 mm x 35 mm x 20 mm .......

2

SECTION II

FILTER DESIGN

A. DESIGN APPROACH

Electronically tunable filters require high-Q elements with tunable

reactances. Presently, the magnetically tuned YIG element is most commonly

,,sed in thc applications at microwave frequencies. The high-Q of YIG resona-

tors assures high selectivity, but the magnetic tuning circuitry limits their

tuning speed to the millisecond range. Varactors, which are frequently used at

UHF frequencies, can he tuned at rates that are about three orders of magnitude

higher than those of the Y[G, but their relative losses severely limit selectivity

in passive circuits at microwave frequencies.

The approach taken in this program was to overcome the inherent varactor

losses with the negative resistance of an active element. The negative re-

sistance of a GaAs VET was originally chosen because of broadband capabilities.

The design and results obtained during the Phase I effort [11 substantiated

feasibility of this approach generally; the Phase II, Task I efforts [21

demonstrated three-section-filter capabilities.

The negative resistance of the FET, the varactor, the necessary tuning

inductance, and dc connections are combined on a suitable carrier into a

resonant element. These elements are then employed in single- or multisection

handpass filters designed according to standard passive coupled-resonator

filter theory 13-51. The diagram in Fig. 1 depicts a multisection bandpass

filter in its basic form. The N resonant elements E are connected to the

terminating impedances R as shown, via (N + 1) coupling networks K. The0

coupling networks are designed to have impedance transforming properties to

1. A. Presser, "Ultra-Fast Tunable Microwave Filter," Final Report, ContractNo. N00173-77-C-0201, Aug. 1978.

2. A. Presser, "Ultra-Fast Tunable Microwave Filter," Final Report, Phase II,Task 1, Contract No. N00173-79-C-0186, Aug. 1980.

3. G. L. Matthaei et al., Microwave Filters, Impedance-Matching Networksand Coupling Structures (McGraw-Hill Book Co., 1964), p. 421.

4. J. F. Cline et al., "Tunable Passive Multi-Couplers Employing MinimumLoss Filters," IRE Trans. PGMTT-7, 121 (Jan. 1959).

5. S. B. Cohn, "Dissipation Loss in Multiple-Coupled Resonator Filter," Proc.IRE 47, 1342 (Aug. 1959).

give the required N-section filter response. The resonant element E in its ideal

form can be represented by the circuit diagram shown in Fig. 2. The varactor

and its losses are represented by the capacitor C and the resistor R, respec-

tively. The FET with its feedback configuration is represented by the impedance

Z, which has a negative real part. The inductor LT finally brings the combination

of all parameters into resonance. The following paragraphs of this section

describe the design, construction, and performance of the major filter com-

ponents, as well as element construction and performance.

INVERTER RESONATOR TERMINATION

Ro K N1 Ro

Figure 1. Block diagram of multisection bandpass filterwith impedance inverters.

R v v LT ZL _A CI

Figure 2. Idealized equivalent circuits of activeresonant element.

4

B. FILTER COMPONENTS

A filter as depicted in Fig. I consists of two major components, the

resonant elements E and the coupling networks K. These major components and

their subcomponents are described in this paragraph.

1. Resonant Element

a. Varactor

Generally, the most important varactor parameters are junction capacitance

(C.) variation with bias voltage, breakdown voltage, and series resistance.

These parameters are rather predictable and depend upon varactor construction

[6]. The particular varactor chosen for this program was an inexpensive

abrupt-junction Si-varactor in chip form (Part #GC-1500A-O0)* with a capaci-

tance ratio that is larger than 6:1 between 0 and 30 V (Cj(o) = 1.42 pF,

Cj(30 ) = 0.22 pF). The plot in Fig. 3 shows the measured junction capacitance

and series resistance of this varactor. The series resistance was derived from

Q measurements at 1 GHz. Since diode Q varies inversely with frequency, the

nee& to overcome all or most of the varactor losses (1.5 to 2.5 Q) at fre-

quencies near 10 GHz for high-selectivity applications is evident.

b. Negative-Resistance FET Circuit

The objective to attain an impedance element with a negative real part of

1.5 to 2.5 Q was met with an FET in an arrangement shown in Fig. 4. The

circuit element values were chosen to operate the FET in a common drain con-

figuration that provides negative resistance over the largest range of fre-

quencies.

When an FET model [71 was used (see Fig. 5), calculations for the impedance

between terminals B and C of Fig. 4 without the capacitor CF show that this FET

circuit can be represented by a parallel circuit of a relatively large negative

resistance (50 to 200 Q) and a small capacitance (%0.07 pF).

6. P. Penfield, Jr., and R. P. Rafuse, Varactor Applications (MIT Press,Cambridge, MA, 1962).

7. P. Wolf, "Microwave Properties of Schottky-Barrier Field-Effect Transistors,"IBM J. Res. Dev. 14, 125 (March 1970).

*GHZ Devices, Inc., S. Chelmsford, MA.

5

LL .2- 600-

1.0 500 2 - ..a.2

I-l

z c;0.8- 400

0,6- 300-.

o 0.4- 200 - Cj ,,W

I.-

a-:

zS0.2- 00

0 0 1 1 0

0 10 20 30

VOLTAGE (V)

Figure 3. Measured junction capacitance, resistance,and Q of typical varactor.

C C C-CB

Cs L

RI

B - B BI B

Figure 4. Equivalent of negative resistance producing FET circuit.

The capacitor C therefore, serves two purposes: it Q-transforms the

negative resistance to a series value in the desired range, and also prevents

effective capacitance ratio reduction of the varactor, which eventually has to

be placed in series with the FET circuitry as shown in the element diagram of

6

Rg ,gd RD

cgs VI Rjj-_ ) cV

I Rs I

S

Figure 5. FET equivalent circuit for resonant element simulations.

Fig. 2. The results of a sample calculation for the complete circuit of Fig. 4

in the 7- to 12-GHz frequency range is shown in Fig. 6 for two values of C F

(0.4 and 0.6 pF).

0._ VDS=6.1 V 0.L 'VSG=2.0 V C -

REACTANCE .6PF

tOZ T " ' . . RESISTANCE...

I- - - -.. ) /

2.--20.

"-. .4PFL -40-

N --.0,,. N

FREQUENCY (MHZ)

Figure 6. Calculated impedance of FET circuit as shown in Fig. 4.

7P

c. Construction

The varactor and the negative-resistance circuitry are combined on a

common carrier to obtain a series-resonant high-Q element that serves as a

building block for multisection bandpass filter designs. The carrier must have

small parasitic effects upon resonance and permit easy access to the required

external bias and tuning voltages. Figure 7(a) shows a realizable assembly of

the resonant element that fulfills the essential carrier requirements. The

varactor, the negative-resistance circuitry, and the tuning inductance are

fabricated on an alumina wafer. Metallic thin-film circuits support the

varactor, the FET, and all associated circuit components as shown on the

sketch (Fig. 7). Wire bonds representing lumped-element inductors are used to

interconnect the individual components. Also shown on the sketch [Fig. 7(b)]

is an electrically equivalent circuit of this element.

C CS

\p

D

D LT

CV L T RV C LG G DLI

2bC

Figure 7. Resonant element construction: (a) realizable configuration;(b) equivalent circuit of configuration.

This realization differs electrically from the ideal-element representa-

tion in Fig. 2 by the carrier pad capacitances C2 to ground of the metallized

alumina wafer; this produces a four-terminal structure. The capacitances C

are, as will be shown later, absorbed into the coupling networks of the filter.

In the final realization of the element, thin-film technology is used to print

the necessary dc connections for the FET and varactor onto the carrier.

2. Coupling Networks

The coupled-resonator design approach requires the use of impedance

inverters. These inverters transform the terminating loads to the resonator,

and/or couple individual resonators as needed to attain the desired response

shape for multisection bandpass filters. In the following paragraphs, some

realizable impedance inverter circuits, found useful in element testing and

construction of the final two-section filters, are described.

The basic requirements for an ideal impedance inverter are a real image

impedance in the frequency range of operation and an image phase that is an odd

multiple of ±90'. One of the simplest inverter circuits, for instance, is a

quarter-wavelength transmission line. It has an inverter parameter K of Z 00

(the characteristic impedance of the line) and an overall phase shift of 90 at

some frequency f and its odd multiples. The use of such lines was precluded0

in the final design because of size and bandwidth limitations. Two other types

of inverters, an unsymmetric divider type and a symmetric type, were used in

this program.

a. Unsymmetric Divider

The general form of this divider is shown in Fig. 8. Very large down

transformation of the terminating resistor R can be obtained over large frequency

ranges if the reactance sum of X and X2 is large when compared to R. The

relationships between the given and equivalent parameters are also shown in the

figure. The specific unsymmetric type that was used is of the capacitive

variety in which X is the reactance of a small coupling capacitor C, and X2)

the reactance of the element carrier-pad capacitance C The transformed

series impedance R' loading the resonator is

RC) 2

9

X oA

R~ x2

_T G

A 1 R, << X1 + X2

XI 2

RI XI = X1 X2

X1+X 2

G

Figure 8. Unsymmetric impedance divider.

and the excess series capacitance to be absorbed by the resonator is

C" C I + C2. For specific values of R = 50 Q, C1 = 0.02 pF and C2 0.4 pF;

the transformed parameters are R' = 0.125 Q and C' = 0.42 pF.

b. Symmetric Impedance Inverter

The general form of a symmetric impedance inverter is shown in Fig. 9(a).

The shunt element Z is practically realizable in inductive or capacitive form.

The parameter values Z' are computable from a bisection of the network by use

of the basic inverter requirements with open- and shorted-circuit conditions

These parameters Z" must be absorbable by the resonators and/or the loads. In

the lossless case, in which the inverter element Z is purely reactive, the

elements Z' are also reactive and the inverter parameter K is real. If the

element Z is lossy, the Z" elements are also lossy and the inverter parameter K

is complex. The general values of Z' and K are also given in Fig. 9. The

specific realizable inverters in capacitive and inductive form are shown in

Figs. 9(b) and 9(c), respectively.

Experimentation with a variety of inverters of this type showed that at

X-band frequencies, 100:1 transformations are practical with capacitive elements.

10

ziI Z =-ZK

2=-Z

2

Z 2Z

(a)

-R -R

-C-C

R, C R R2

-R -L -L -R

R R2

K = wL-iR /RR

(C)

Figure 9. Symmetric impedance inverter.

Similar transformation ratios are feasible with inductive elements, even though

the experiments were conducted at a 10:1 ratio only. The importance of the

complex K-inverter was discovered in the analysis of an electrically symmetric,

negative-resistance resonant element in which excessive detuning and loading

effectb of the inverter caused unsymmetric frequency responses and limited

selectivity.

C. HISTORY OF RESONANT-ELEMENT DEVELOPMENT

During the earlier part of this program, we used a planar resonant element

in an electrically symmetric configuration with two varactors and two FETs.

Such an element is shown in Fig. 10, and its equivalent circuit in Fig. 11.

Printed rectangular rf-chokes in the dc-connecting circuits with cu1 s-Lerable

parasitic reactances to ground were responsible for instabilities occasionally

observed at frequencies in the 2- to 4-GHz frequency range. To increase tuning

11

Figuire 10. Earliter symmnetric resonant elemetnt r al izadonAS th wo FETs and two va ra ( tot .

12

VO

Se CV LT LT CV L8

CPT cj Cj CP~j CP cP

I;- VV

Figure 11. Equivalent circuit of symmetric element.

range, stability, and selectivity, variations in element configuration were

tried. While we were able to increase stability and tuning range merely by

increasing the alumina carrier height from 0.635 to 1.27 .nm, we could not

improve the selectivity. Thus, selectivity appears to be a function of the

basic circuit configuration. Figure 12(a) depicts the original symmetric

element (drain of FETs back-to-back) with the two varactors in the drain

circuits of the two FETs. An element on a 1.27-mm-high carrier embedded

between two impedance inverters K gave at best a tuning range of 3 GHz, and

that with a relatively poor selectivity of 200 MHz at 9 ±1.5 GHz. A variation

of the element with the varactors moved to the respective gate sides is shown

in Fig. 12(b). This type of element caused the tuning range to be reduced to

1.8 GHz and to become a pronounced function of external loading. Also, contrary

to expectations derived from calculations, selectivity did not improve.

Ii, a,,othei element variation (see Fig. 13), the FETs are connected with

gates back-to-back. Figure 13(a) shows the element with the varactors in the

gate circuitry, and Fig. (13b) shows it with the varactors in the drain cir-

cuitry. The aim with this type of element was to attain greater electrical

stability and increased freedom from spurious oscillations. The experimentally

13

C, C,

(a) TUNING IN DRAIN CIRCUIT

(b) TUNING IN GATE CIRCUIT

Figure 12. Symmetric element diagram of experimental single-section

filter configuration (drain back to back): (a) tuning

in drain circuit; (b) tuning in gate circuit.

(a) TUNING IN GATE CIRCUIT

C v

C v

(b) TUNING IN DRAIN CIRCUIT

Figure 13. Symmetric element diagram of experimental single-sectionfilter configuration (gate back to back): (a) tuning in

gate circuit; (b) tuning in drain circuit.

verified stability increase was offset by a poorer selectivity of 250 tl}z and a

smaller tuning range of 700 MHz.

14

The lack of selectivity of our symmetric active resonant element in actual

filter configurations with inverter circuitry prompted a detailed study of the

problem. The primary function of the inverter circuitry is to control the

external loading of the resonator caused by terminations and/or adjacent

resonators. We found that, as configured, the resonator alone is a detriment

to the transformation. The FET circuitry acts as a complex impedance inverter,

as shown in Fig. 9(b), in which R is a negative resistance. The magnitude of

the resistance is not negligible with respect to the FET reactance, and the

transformed reactive series component cannot be neglected either when compared

to the reactance of the varactor. The selectivity response under these condi-

tions is highly frequency-dependent and asymmetric.

Prevention of these undesirable effects relies upon low carrier parasitics

to decouple the negative-resistance element from the common ground of the

inverter circuitry. A different approach, as discussed previously in Section

II.B.l.c, was used to separate the grounds of the inverters and the negative-

resistance circuitry.

The element configuration finally used is shown in Fig. 7. This reali-

zation provides good electrical symmetry and requires only one FET and one

varactor per element. A true series resonance is established when one con-

siders that the parasitic pad capacitances C2 of the carrier can be absorbed by

necessary inverters. The negative-resistance element between the terminals B

and C in the diagram was realized electrirallv, as shown in Fig. 4 and described

in Section II.B.l.b. A 0.38-mm-thick, 5-mm x 5-mm alumina substrate was chosen

as the carrier material. Metallic thin-film technology was used to print onto

the carrier the supporting pads for the varactor, the FET circuitry, high-

impedance choke lines, and rf decoupting resistors. The photograph in Fig. 14

shows a close-up of the circuitcy with components mounted. The FET and varactor

circuitry is shielded from ground by the mounting pads. With the carrier and

rf bypass capacitors attached to a 10-mm x 5-mm x 1.5-mm flange, as shown in

Fig. 15, the assembly represents the basic filter building block in modular

form.

D. EXPERIMENTAL ELEMENT EVALUATION

During the program, various aspects of element and/or filter performance

were evaluated. The photograph in Fig. 16 shows a fixture in which individual

15

4kW4

16

17l

Figure 16. Test fixture for resonant element evaluation.

elements were tested. This fixture contains two 50-0 line sections terminated

by SMA-type connectors. The element is capacitively coupled to these lines via

adjustable tabs. Required dc voltages for varactor and FET of the element are

accessible over standoffs. Transmission and reflection responses of the

element are evaluated in a manual test setup as shown in Fig. 17. This setup

permits single and swept frequency filter evaluation at fixed and leveled input

powers. Single-frequency measurements with counter accuracies (less than 30

Hz) are made possible by both the "lockbox" feature of the counter used (EIP

371) and the FM capability of the signal generator (HP 8620C). In this para-

graph the results of individual evaluation aspects are highlighted.

1. Selectivity

The selectivity of a bandpass filter is defined by the 3-dB bandwidth and

the transmission rejection as a function of frequency. The rejection is mainly

determined by the basic filter design (number of elements, response shaping,

etc.) which, in this case, is that of a single-section, maximally flat filter.

The 3-dB bandwidth, on the other hand, is a function of the resonator para-

meters (slope parameter and Q) and the external loading (determined by the

coupling). The unloaded Q of the resonator can be adjusted over a wide range

with the gate voltage of the FET up to a point at which instabilities will take

over. The coupling at input and output is adjustable via the proximity capaci-

tance between the gap of a tab, which extends from each end of the resonator,

and the printed conductor of the 50-Q line sections. This range of adjustment

is also limited by stability considerations. Figure 18 shows the measured re-

sults for an element with coupling adjusted to assure stability while maintaining

a O-dB transmission loss over the tuning range. The 3-dB bandwidth, voltage

standing-wave ratio (VSWR), and gate and varactor voltages are shown as a func-

tion of midband frequency. The frequency is adjusted by varying the varactor

voltage while at each setting the gate voltage is adjusted to result in a O-dB

transmission loss. The high degree of uniformity of the 3-dB bandwidth (21 ±1

MHz) over a 2300-MHz tuning range is not typical, but demonstrates ultimate

feasibility. Typical reproducible minimum values of bandwidth at present lie

between 20 and 40 MHz, attainable over the entire 0- to 30-V tuning range.

19

FREQUENCYCOUNTER

NETWO~RK ANLZECEIP 371)(HP 8410)

REFLCTIO 5 1 dB Y LSWEEPTRANSMISION UNI OSCILLATOR

(HP 843A)(HP 8620C)

ATTENUATOR (AIL 220A)

Fiue 7 Mnalntwr nayertststuTo

tr n mi s o and re l cidev l a i n

P OWR- ~ -

VD = -6> 1 V

Ir~y U] -

BW ~ (

T 02

It ; _Vg

m7\/W>-

0.0- - - - - -- - - - - - J.

FRE0UENC, (MH-z) -

Figure 18. Measured 3-dB bandwidth and VSWR vs midband frequency ofsingle-section filter with high selectivity and smallbandwidth deviation over tuning band.

20

2. Tuning Range

The tuning range is a function of the capacitance ratio of the varactor

between useful minimum and maximum varactor voltage limits and of other capaci-

tive reactances in the resonator proper that affect this ratio. These unwanted

but unavoidable reactances are introduced by the carrier, the interconnections

between components, and the FET circuitry; they effectively reduce the capaci-

tance ratio and, with it, the tuning range. A varactor with high cutoff

frequency (high-Q) and large capacitance ratio is best suited for broadband

tuning. The reactance of such varactors is high, and the series resistance is

low.

Although the varactor and the FET chosen during this program were not fully

optimized, the maximum frequency tuning range for the particular combination

was found to extend over 2.5 ±0.1 GHz, while assuring stability and adequate

selectivity (30-60 MHz). In order to demonstrate wider-band tuning, the

varactor reactance was doubled by connecting two varactors in series. With the

FET circuit unchanged, the tuning range of such a combination increases because

of the smaller reactive loading effect of the FET on the varactor reactance.

The observed tuning range increased from 2.4 to 3.5 GHz. However, since the

series resistance of the varactor combination also doubled, the negative

resistance of the FET was not sufficient to maintain good selectivity over the

full varactor tuning range. Attempts to increase the negative resistance via

gate voltage adjustments were fruitless since instabilities near the upper

frequency limit occurred.

The plot in Fig. 19 shows the varactor and gate voltages at the cor-

responding frequencies (note the 0- to 60-V range necessary to drive the series

combination of two 30-V varactors). Figure 20 shows gain and 3-dB bandwidth vs

frequency. Selectivity decreases rapidly as transmission loss increases

because resonator Q decreases as a result of insufficient negative resistance.

Figure 21 shows the VSWR; as the gain deteriorates, so does the VSWR because of

mismatches caused by the unchanged coupling structure. It is recommended that,

in future phases of the program, the varactor-FET combination be optimized so

that simultaneous attainment of large tuning bandwidth and high selectivity is

possible.

21

0.F 220

50. 7 .

SVD-6. 1. V -1.5

- 40.-/-

LV /

M 30.-1.0

10.)IrI

0. AL.I.J. _I I.-i.. -L-A.i .I.L . L. .LJ., 0.

0 0 2 t 9 G C 0a 0) 0 0) 0 0 M M Is

W M M 0 ea

FREQUENCY CMHZ)

Figure 19. Varactor and gate voltages vs midband frequency in broadbandtuning experiment with two varactors in series.

300.. .... - 0./ GAIN

7Y"/ , I

/ 0V

-5.

/

I-r\ / _-,

M 100. /

/L

0. L. ..- .A. --.- I . _ _.L*. j. . .L . . .Li ._ .L...i_. L .L L..J..L--L_ J -10.

(0 0 9 0 0 9 0 0 0 0 009 0n G V) 0a ' 0 M I 0

FREQUENCY CMHZ)

Figure 20. Measured 3-dB bandwidth and transmission gain vsmidband frequency (broadband experiment).

22

0. 4.0

-5.5. 1\F30

0 .

I. -2.0

X.VSWR .f"

10. L 1-L-L-L I /J1

t9 UG UU U )

r- CD C 0) 0) (9 N

FREQUENCY (MHZ)

Figure 21. Measured VSWR and transmission gain (broadband experiment).

3. Signal Input Effects

During earlier efforts (Phase I and Phase II, Task 1), it was already

observed that power input affects the filter response. Generally the trans-

mitted power is reduced, and midband frequency shifts as input power is in-

creased. Whereas gain reductions of a fraction of a decibel can be observed

with power inputs as low as -15 dBm, noticeable frequency shifts of a few mega-

hertz occur at the -10-dBm level. As it was also observed, the effect is a

function of element loading; we therefore suspect that the rectified rf voltage

at the gate junction of the FET is producing power-dependent variations of

negative resistance as well as a reactance sufficient to cause the changes ob-

served in selectivity and transmission. The plot in Fig. 22 shows the response,

evaluated on an automatic network analyzer, of a tightly coupled element (with

95-MIz bandwidth) for three different power input levels. For a 20-dB input

level change from -30 to -10 dBm, the center frequency shifts by less than

10 MHz, the midband loss increases by 1.2 dB, and the 3-dB bandwidth increases

by 5 MHz (from 95 to 100 MHz). At higher selectivity adjustments (\20 MHz

bandwidth), frequency shifts can also be as large as 10 MHz, while loss can

increase by 4 dB, and bandwidth by 20 MHz.

23

-30 dBm" L r" - 10 dam0, -20 dBm

.0o - /"

- 5 0/7 Vo- 6. 10V- , .// " 0

- VG- 2. 94V

i/

In U.,MIa

FREQUENCY (MHZ)

Figure 22. Measured transmission passband response at a given filter

setting for three different input power levels.

The effects on filter performance of power input from two signals were

briefly observed. In the evaluation, the test setup shown in Fig. 23 was used.

Two signals, adjustable in frequency and amplitude, were applied over a 3-dB

hybrid coupler to the input of the filter. The rf output of the filter was

observed on a spectrum analyzer. Amplitude responses of each signal within and

outside of the passband of the filter, amplitude and frequency of signals

generated by FET nonlinearities, could be determined. The filter was adjusted

with voltages as shown in Fig. 22. The observed amplitude and frequency shifts

were about the same for a combined power input of the two signals within the

passband as were those observed above for a single signal. The third-order

intermodulation (IMD) for two equal-amplitude signals in the passband (Af = 15

MHz) as a function of the power input of each signal is shown in Fig. 24. The

IMD product decreased while one signal remained in the passband and the differ-

ence frequency was increased. The power input of the second signal at the

response skirts was also increased to give identical power outputs of the two

signals; with a signal separation of 200 MHz, no third-order distortions were

observed when the inband signal was -27 dBm and the out-of-band signal +6 dBm.

However, as total power output cf the two signals was increased, response

shifts were about the same as for a single signal at a comparable power output.

24

U P OWE R

dcSUPPLY

(HP 8620C) R SPECTRUM@ -'ftOU T'l ANALYZER

rf GENERATOR

f2L3-dB HYBRID

(HP 8620B) (ANAREN 1NA0568-3)

Figure 23. Test setup to evaluate IMD and other two-signal effects.

-20.0 'o- 6. 1 OV 90

IM VG -- 2 . 4V

-30. 0 1M.

40

m 40.Oro 0

0

H tiF-15MHZ' 0

7 -0. 0.P FO-94OMHZ

0 10

-go. eF L 0

t9 t9 0 0

- de d d

INPUT POWER (DBM)

Figure 24. Measured two-signal intermodulation distortion (absoluteand relative) as a function of input power level.

4. Nois -Figure

Transmission loss is not the limiting parameter of noise figure in active-

filter applications; the filter loss is generally negligible or, with some opera-

ting voltage ajustments, can turn into gain. As with other active components,

25

output noise is a disturbing factor and noise figure is an appropriate measure

of performance. The noise figure is increasingly important in high-sensitivity

applications and those requiring large dynamic ranges. The plot in Fig. 25

shows the measured noise figure of a single-section filter over the 8.9- to

10.4-GHz frequency range. This noise figure was evaluated at discrete fre-

quencies in a test set shown in Fig. 26. The isolator between mixer and filter

to be tested is essential since it prevents local oscillator leakage from reach-

ing the filter; erroneous noise figure readings were observed without this

isolator. At each measured spot frequency, the filter was adjusted for a O-dB

transmission loss. The measured 16- to 20-dB noise figure is typical and con-

forms to values obtained at earlier phases of this program. The relatively

large values of noise figure were neither expected nor fully understood. We

expected the noise figure to be somewhat larger than that for an amplifier

with the same FET, which is rated 4 dB at 10 GHz, since the FET loading is far

from the optimum noise impedance values. The feedback configuration must be

responsible for a large portion of the noise figure. Future studies on noise

limitations and noise figure reductions are recommended.

20.

19.

G=ODB

0W

0 17.D

L 17. L

6 6N 9 ( D OD G

CD a) ) 9) 01 9 9 t9

FREQUENCY (MHZ)

Figure 25. Measured noise figure at different frequency settings.

26

AMPLIFIERTY I PE 1363)DR

V

LLJ (AIL TYPE 7 5)

MIXER NOISE SOUREE_(ANARENI15,050-30) (MSC 5112P)

(HP 8620C)

ISOLATORISOLATORdc (PBH LABS CI-XI6317)

Figure 26. Noise figure test setup.

5. Temperature Effects

The major effects of temperature on filter performance are caused by the

temperature sensitivity of the varactor and the FET. This sensitivity changes

the reactances and the negative resistance within the resonators. These

effecLs were already observed and reported during the Task 1 efforts of this

program [2]. Similar results were observed on an element evaluated during this

task and are shown in Fig. 27. With the operating voltages and the rf drive

level kept constant, and the temperature varied between 10 and 50'C, the center

frequency decreased by about 930 klz per degree Celsius; the 3-dB bandwidth

increased by 1.1 MHz per degree Celsius. The midband transmission decreased

by nearly 0.2 dB per degree Celsius at the higher test temperatures, and about

0.75 dB per degree Celsius at the lower temperatures. These relatively large

changes with temperature suggest the use of either oven-controlled operation or

voltage control to compensate partially for these changes.

6. Tuning Speed

The tuning operation of the filter type developed during this program

depends on accurate tuning voltage settings for attaining a predetermined

response at a predetermined frequency. The number of vai..ble voltages in the

27

890- 0 120 PIN- -20 dSm0 G VO-6. Iy 2VV-10.0V' , VG-2. 94V

9800 110 F

N -2

H NZ

2 9870. 0 10

o .W I

9850.0 8 0 M- 4

980. z -

9 0.0 10______0_ 200 3. 005.

TEMPERATURE (DEG C)

Figure 27. Measured effects of temperature upon 3-dB bandwidth,midband frequency, and transmission gain.

worst case are 2 N, where N is the number of elements in the filter. It is

envisioned that some of the voltages can be derived from a common variable, one

that is attainable under microprocessor control and uses preprogrammed voltage

look-up tables. This microprocessed addressing system activates the appropriate

driving circuitry for the filter. The tuning sneed of the filter is a combin-

ation of addressing system access time, slewing and settling times of the

driving system, and the time constants of the varactor and gate circuitry.

Since the time constants in the experimental element are only about 100 ns, the

ultimate speed of the operational filter will be determined mainly by the

addressing and driving circuitry. If need be, the 100-ns time constants could

be further reduced by altering the decoupling resistors in the dc lines and by

reducing the values of the feedthrough capacitors.

The tuning speed can be evaluated in a test set shown in Fig. 28. Two

different cw signals within the tuning bandwidth of the filter are injected

into the filter input over a 3-dB hybrid. The filter operating voltages, which

correspond to passband responses at the two injected frequencies, are provided

by the two levels of a zero-offset pulse. The detected output of the filter

together with the driving pulse are observed on a dual-trace oscilloscope. The

filter response time is estimated by the difference between the drive rise time

28

and the time the detected output has settled. A single-section test filter was

evaluated in such a test set. The two frequencies were separated by 180 MHz

(f, = 9.87 GHz, f2 = 10.05 GHz). The drive pulse in our experiment exhibited a

leading- and trailing-edge ringing of about 200 rs, making speed evluations

inaccurate. Excluding this ringing we estimated the filter to respond in less

than 200 ns.

IdcPOWERSUPPLY

(TEKTRONIX [TRIGGERTYP 11) PLSERI

(HP f DUAL-TRACE

~OSCILLOSCOPEDETECTOR

(HP 8472B) (TEKTRONIX

rf GENERATOR TYPE 585A)

f2

(HP 8620B) (ANAREN 1NA0568-3)

Figure 28. Test setup to evaluate filter tuning speed.

E. PROTOTYPE FILTERS

The step-by-step development of the filter element described in the above

sections led to an element design that promises a high degree of success to

meet program objectives of extended filter-tuning bandwidth. The particular

design was frozen at some time into the program to allow ample time for multi-

section design, experimentation, and evaluation. Some of the experimental

data that demonstrate electrical element symmetry with high selectivity (Section

Il.E.1) and freedom of element detuning effects in a two-section-filter design

(Section LI.E.2) are shown below. These two characteristics were lacking in

previous element designs in both Task I and earlier Task 2 efforts.

29

1. Single-Section Filter

The test fixture used to evaluate single-section-filter performance is

shown in Fig. 29. The sequence of plots in Figs. 30 through 37 sumarizes the

performance of an element in such a single-section-filter configuration. The

input and output coupling was adjusted to give, at each varactor voltage setting

a response with about 60-MHz bandwidth, while the gate voltage was adjusted for

0-dB transmission loss. Shown in these plots for each frequency setting are

the required voltages (Fig. 30), the 3-dB bandwidth (Fig. 31), and the input

and output VSWR (Fig. 32). The responses for five particular varactor voltage

settings (Figs. 33-37) weie obtained by use of an automatic nctwork analyzer.

The results demonstratc nearly ideal one-section-filter responses with loaded Q-

values of about 130 over a 1.4-GHz tuning range, while input and output matches

stay below a VSWR of 1.7:1. The plots in Figs. 38 ana 39 demonstrate symmetry

with respect to power flow and show the forward and reverse transmission re-

sponses and reflection coefficients for a particular filter adjustment.

Wb4en a different element was adjusted for higher selectivity (3-dB band-

width of 23 MlIz), the results with respect to symmetry were similar. The plot

in Fig. 40 shows the amplitude response, Fig. 41 the phase response, Fig. 42

the group delay, and Fig. 43 the VSWR of this filter. The phase and delay

responses are also as expected for a maximally flat single-section filter.

2. Two-Section Filter

Two experimental elements were arranged within a metallic enclosure for

two-section-filter operation. The photograph of Fig. 44 shows the two elements,

capacitively coupled to two 50-Q microstrip line sections that lead to the

input and output terminations, and the capacitive coupling between elements.

Most of the area within the enclosure is taken up by the two substrates for the

connecting lines. The coupling was adjusted by observing the skirt selecti-

vity, bandwidth, and input and output match while maintaining a true two-sec-

tion response.

The attainable tuning bandwidth and absolute frequencies were entirely

determined by the performance of the single elements. The tuning bandwidth of

thc filter is equal to the largest common frequency range of the two elements.

This range for the particular elements is indicated by dashed vertical lines in

Fig. 45, which shows at each frequency the necessary four voltages to obtain a

30

414

'VO1

30.0- ' 2.e

-0

-8 Ili 1 -. 7 0

1.0- V

2U-- 1.7

0B. Ol-.. .

-

-I t-Qf

FREQaUENCY (N14Z)

Figure 3 Meaorasurd gate volagewidtt vs midband frequency.

32 IV

In r) mn9N0 CD 0 ON a

FREQUENCYW(HZ)

Figure 3 1..Varctsrrad gate volagewstti vs midband frequency.

so. in32

GA IN-0 DBVa.". IV

S"OUT

".

i.0L- -J . I I.AL. f 8 . 1 t _.L L- I 8I t I I t

in In a 11SN Wt) ,QN* 0 0 6 (P a3 8

FREQUENCYMHZ]

Figure 32. Measured input and output VSWR vs midband frequency.

-5. 0.-

-i5. L '- 8 F I 8 8 , I I , , , I I , , ,. , I..i , ,I t L-,..L-_

C'~1 C'00FREGUENCYcMHZ)

Figure 33. Measured transmission passband response ofsingle-section filter (Vv = 0 V).

33

VO-1. 1 V

VG v

I n In I in

FREQaUENCYCMHZI

Figure 34. Measured transmission passband response ofsingle-section filter (Vv 5 V).

0.0 F

L

VG-1. 76 V

11 V) in 11

FREGUENCY(MHZJ

Figure 35. Measured transmission passband response ofsingle-section filter (Vv = 10 V).

34

-5.

VV-20. 0VVG.1. as V

-18. 0LLL-J-J-- L _L..L .L -L-.-A .L.- . .4- ... -~--- , J

m ae a ea a a a

U) 9 in a9 E-N III tI

FREGUENCY(MHZ)

Figure 36. Measured transmission passband response ofsingle-section filter (Vv 20 V).

0.0

-5. 0".

VO-8. 10 V

S-10. 0:_ ~ VG 1.B 6:V

-1 .OL IL _L _ .1L L LI .T- L- I JL. I. I- I

d d d fl, d daa i a a

FREGUENCY(MHZ]

Figure 37. Measured transmission passband response ofsingle-section filter (Vv = 30 V).

35

-0

VC-1. 78 V

FREQUENCY(MHZ)

Figure 38. Measured transmission passband response of single-sectionfilter (forward and reverse power flow).

1.0 REVERSE

U:Ug..VO 15 10 cic c

FREGUENCYDAHZ)

Figure 39. Measured input and output reflection coefficient inpassband of single-section filter.

36

V0 =6.1 V

V= 11 .3 VV3 .6 ..... .. .......... ..... ... ..6.. .... .... ..6. ..... .... ....... ... ..............

V G =3.07 V

zZ0 . 9 ............ .. .............. ....... ......................... .............

x-968.

9366.661 9318.666 9326.809 9329.999 9346.061 9366.89:

FREQ-MHZ

Figure 40. Measured transmission response of a single-sectionfilter adjusted for high selectivity.

VD1=6.616

Vv=113VG=3.7

W 0 0 . 9. ...... ..............- ... .....................................

9300.681 9310.080 9326.668 9329.999 9340.801 9366.091

FREQ-MHZ

Figure 41. Measured phase response in passband ofhigh-selectivity filter.

37

IA

.. .. ... .. ... .. ... .. ... .. ... .. ... . .. ... .. .... ....... ........ ........ ......aP-908 P .6

VC)8e = 61

VV=11.3V

VG =3.07VT6.00.00 - 9.0

1.000

9388.001 ~ ~ ~ ~ V 931.10 V3000 92.9 3691 96.9

8 . 00. .. .. ... .. ... .. ... .. ...F. ... .. ... .. ... .. .

Figre43 MasredVSR n asban o hghseectviy iler

7 .0 .. .. .... .... .... .... ... .... .... .... .... .... ..... .... .... .... ..3 8. ... ..

-J0

0

0

C.)

0

39

GAIN-0 09VO-o I V

30.&. rd7 .. b

VG2

I -

20.0'~ !1-

150rVl

I.- -I

100 - I

5. 0"/V

0. L_ L-L 11. L. A -. I I 1. 5

ID 11 11 0 CD I0D M 0 C 0 a a 0

FREGUENCY(MHZ

Figure 45. Varactor and gate voltage vs midband frequency ofeach element in two-section prototype filter.

0-dB transmission loss. Figure 46 shows the corresponding 3-dB bandwidth and

the input VSWR. The symmetric nature of this filter is demonstrated via Figs.

47 and 48, which show forward and reverse characteristics of reflection coef-

ficient and amplitude response, respectively, for a particular frequency

setting.

During filter evaluation a strong resonance of the enclosure was detected

at 11.8 GHz. This spurious resonance was eliminated with a layer of absorbing

material in the cover. In the final filter a metallic partition was used to

suppress enclosure resonances.

Control over coupling was demonstrated on a different two-section filter

in which coupling was adjusted to attain a Chebyshev response. Figure 49 shows

the amplitude response obtained near midband. The 3-dB bandwidth is 40 MHz,

and the midband ripple is about 2 dB. In Fig. 50 the response is plotted over

a larger frequency range; the increased selectivity of this type of filter over

that of a maximally flat response is noticeable. The nonlinear phase associated

with Chebyshev responses is evident in Fig. 51. The measured group delay and

return loss are shown in Figs. 52 and 53, respectively.

40

GAIN-0 0BvD-8. IV35.0 - 20

VSWR B A.j_ BANOW.

- 30.0" !S •

0

z<

25.01 .. 1(D t9 N 9 19

FREGUENCY(MHZ)

Figure 46. Measured 3-dB bandwidth and VSWR vs midbandfrequency of two-section prototype filter.

1.0

'- IN• 78 our

- :0vo-6. 10 Vu\\ VV1-10.4 Vw VV2-10. 0 V

3 " r VG -1.g9 Vz VG2-2. 37 V0 -

L3IL

0. - . . j j. J--Jj.1 U11 J 1.1J1JA I I LAJ

N1 f ) , n ' 9

0 C) 0 9 a) ) o)FREQUENCY(MHZ)

Figure 47. Measured input and output reflection coefficient passbandresponse of two-section prototype filter.

41

5. i

0

1 -0. or-

0c3 -1 z7 FWROV-1.

-30.10 0 V

-25. 0" LiFWRLO L. .LJL.L ~~ L-L J

tQ G9 M9 Gf Is t9ts 11) G V

F RE QUENCY (MHZ)

Figure 48. Measuicd transmissi ,n passband response for forward andre\verse power fluw of two-section prototype filter.

IOWVI-.V217669I 4,Q832. -. 1,-.

RP- 8.988 :RP2- S.0S6

9299.81 9290960 9300.661 9310.90 9320.800 9329.999 9340 .8ft1

FREQ-MHZ

Figure 49. Measured passband response of experimental filterwith Chebyshev response.

42

~0 . / ~FLAT

.. . ... .. .. .. .. .. .. .

9108.08 9288.88 9300.00 9400.00 9600.00

FREQ-MHZ

Figure 50. Measured selectivity of experimental two-section

filter with Chebyshev response.

I8UW-VVI.28 iVV2-17 .6 ,VGI-3.47 ,VG2-3 .28 .VD-6 .11,1-8.'

LL;

Lii 8.09

9280.801 9298.808 9388.881 9318.888 9320.808 9329.999 934S.881

FREQ-MHZ

Figure 51. Measured phase passband response of Chebyshev filter.

43

iRPI-, 8.889 RP2- 0.080

9288.081 9298.888 93s8..-31 9310.800 9328.808 9329.999 914e.891

FREQ-MHzFigure 52. Measured group-delay passband response of Chebyshev filter.

18UW-VV1-20,VV2-17.5,VG1*3.47VG23.28,.O6.1X.8.

(1) -2.8880

-j-3.088

zX -4.888

LU

9288.881 9298.88 9380.801 9318.88 9320.888 9329.999 9348.881

FREQ-MHZ

Figure 53. Measured passband return loss of Chebyshev filter.

44

SECTION III

FINAL FILTER PERFORMANCE

A. SINGLE ELEMENT

The elements to be used in the construction of the two final two-section

filters were constructed in modular form (see Fig. 15) as outlined in the

previous sections. Each element was tested and adjusted individually in a test

fixture, as shown in Fig. 16. The aim of the adjustments was to bring frequency,

selectivity, and timing voltages into a common range so that the largest range

of operation can be attained when combining two elements into a two-section-

filter arrangement. As indicated in the previous section, their range is

determined b the frequency overlap of the two elements.

The plot of Fig. 54 shows the frequency vs varactor voltage character-

istics of four elements that were prepared for the final filters. The fre-

quencies correspond to gate-voltage adjustments for O-dB transmission loss.

These gate voltages for the four elements are plotted in Fig. 55. The par-

ticular 3-dB bandwidth that depends upon the input and output coupling is shown

in Fig. 56. In the final use of these elements, no additional adjustments, ether

than voltage adjustments, were made on the element circuits.

11000. 0

10500. 1 3

4 2>9500.0

zL r

a700. GAIN=OdB/, Vd=6.1 V

9500. 0

9000. 0 . . L L I _ J I -A 1 1 i _ I -I L i L I _ J0.0 5.0 10.0 15.0 20.0 25.0 30.0

VARACTOR (VOLT)

Figure 54. Measured midband frequency vs varactor voltage forfour different resonant elements.

45

2. 0,

1. - - GAIN=OdB 4Vd=6.1 V

1..

t0

1.7L - _1.0

0. 0 5. 0 10. 0 15.0o 20.0o 25.0o 30.0o

VARACTOR (VOLT)

Figure 55. Measured gate voltage for 0-dB transmission loss of fourdifferent resonant elements at given varactor voltages.

40.0 4

N -

T

I 30.01-

*GA IN-Od8

Z Vd-5. I V

~/- N , * / ..,-3

202. 0 .... I.. ...... I JT_ tT hiL.LL +W J--K[L ,lL

00

a)3 0) 0) 0 &

FREQUENCY (MHZ)

Figure 56. Measured 3-dB bandwidth of four different elementsfor given varactor voltage settings.

During the evaluation of the final filter elements it was noted that the

tuning bandwidth ranged between 1.9 and 2.0 GHz, far short of the expected

range of 2.4 to 2.6 GHz that was established in the prototype elements. The

46

reduced tuning range was traced to varactors received just before fabrication

of the final elements. A recheck was made by exchanging the varactor of a

prototype element with one of the newer ones. The 2.4-GHz tuning bandwidth

decreased to 1.9 GHz, and selectivity for field coupling decreased as well.

These measurements confirmed that the new varactors had higher resistive losses

and smaller capacitance ratios when biased between the rated 0- to 30-V range.

Since there was not sufficient time to order new varactors, the filter elements

were completed with the varactors at hand.

B. TWO-SECTION FILTER

The two-section filter consists of two modular elements that are capaci-

tively coupled at input and output to short 50-Q microstrip transmission lines

that terminate in SMA connectors. Coupling between elements is also via

capacitive proximity. Figure 57 is a photographic close-up of the assembled

two-section-filter arrangement within a 38-mm x 35-mm x 20-mm brass enclosure.

Capacitive feedthroughs provide access to the dc connections necessary for

filter operation. A metallic partition between elements with an iris around

the coupling tabs prevents both uncontrolled coupling between elements, and

waveguide resonances of the enclosure that cause spurious rf-feedthrough. The

photograph in Fig. 58 shows an overall view of a completely assembled filter

with partitions in place but with cover removed.

Adjustments of the filter are first made with the cover and the partition

removed. These adjustments entail capacitance changes obtainable from the

coupling tabs that extend from each end of each resonant element, and appro-

priate voltage changes. The operating voltages of the two elements are

adjusted to correspond approximately to those observed previously for each

element at the same frequency within the tuning band. Keeping the two voltages

of one element fixed, we made slight adjustments of the other voltage pair

while observing the swept response of the filter on a network analyzer.

Voltages are adjusted for best transmission response. The coupling to the

transmission lines and the tabs between elements are adjusted next, until a

true, well-matched, maximally flat response with desired 3-dB selectivity is

obtained. Coupling adjustments are limited by stability considerations, and

also have to be touched up subsequently to correct for effects caused by the

4j

C

C

48s

4 -)

0

49-.

partitions and the cover. It was found that coupling adjustments first made at

the upper-frequency end of the filter tuning band resulted in good filter

operation over the entire band. This also avoids instabilities that tend to

occur at the high-frequency end.

Two filters were fabricated, adjusted, and evaluated on both a test bench

and an automatic network analyzer. In the following paragraphs the measured

performance is described and some detailed results are shown. In the final

evaluation, the case temperature of the filter was maintained at 24.2 ±0.10C

by means of a circulating-bath heatsink. The most important characteristics

and effects of the filters were described previously in Sections II.D and II.E

of this report.

Table 1 shows the -ltages neces5ary for a give. frequency when filter

No. 1 was adjusted for a 0 ±1-dB transmission loss. The same data in graphic

form are shown in Fig. 59 for the varactor voltages v frequency and in Fig. 60

for the gate voltages. The maximum tuning range of 2 GHz is determined by the

mutually inclusive frequency range of the two elements in the 0- to 30-V tuning

range of the varactors. As shown in Fig. 61, the corresponding 3-dB bandwidth

over the tuning band varied between a low value of 18 MHz and a high value of

36 MHz. The bandwidth variations are mainly a function of varactor Q, a value

that is about four times larger at the high frequency than at the low frequency.

The coupling also varies as a function of frequency, but its effect upon band-

width in this case is a compensating one. Also shown in Fig. 61 is the noise

figure, which varies between 19.3 and 22.6 dB over the tuning frequency range.

Figure 62 shows the measured gain and the VSWR as a function of frequency for

the voltage setting corresponding to Figs. 59 and 60.

More detailed response results were obtained on the automatic network

analyzer. The plot in Fig. 63 shows the transmission response at a particular

set of voltages for both the forward and reverse directions of power flow. The

plot in Fig. 64 shows the corresponding phase responses in the passband; Figs.

65 and 66 illustrate the reflection coefficients and group delays, respectively.

The response symmetry is reasonably good. The effects of power input variations

become apparent in the plot of Figs. 67 and 68, in which amplitude and group

delays are shown for three different power input levels. A slight shift in

midband frequency, a widening of the response, and a fall-off in relative

transmitted power are noticeable.

50

TABLE 1. TWO-SECTION FILTER NO.1*

Frequency V VV VG V2 BW Gain

(M1Hz) (V MV (V MV (MHz) .(dB)

8256 0.00 0.8 1.63 1.40 20 0.0

8580 1.10 2.5 1.62 1.34 18 1.0

8922 2.90 5.0 1.62 1.34 18 0.0

9412 6.67 10.0 1.62 1.34 20 0.0

9767 10.34 15.0 1.70 1.34 31 -1.0

9987 13.23 20.0 1.70 1.34 21 1.0

10147 15.72 25.0 1.70 1.34 30 1.0

10246 17.52 30.0 1.74 1.34 36 0.0

*Pi -27 dBm; V D= 6.11 V; T =24.2 ±0.11C.

30.0-

25.0 Pi-,-- 27 d~m

T-24. 2Ct. 1

_J 20.0-0

15.0 I0

0.

11J111 L1 L t, A .t t9L L~1. to1 U2t

FREQUENCY (MHZ)

Figure 59. Measured varactor voltages vs midbandfrequency (final filter No. 1).

51

1. BF

F V911. 7r-

I- 1.5~> ,Pi,--27 dBmhLi 1.5 T-24. 2"C. 1

I- 1.5fr

G

0 0 0 0) t9 (FREQUENCY (MHZ)

Figure 60. Measured gate voltages at midband frequencysettings (final filter No. 1)

40. 0 440.0 40

L BW. i_P17--27 dBm

30.024.2*C±. 1 -S I n

0IX

2 . o L 2"0' W. .

z z NF .. ' -'2a 'm 0

0in in) 09 InO0 a1 0 0 GFREQUENCY (MHZ) -

Figure 61. Measured 3-dB bandwidth and noise figure at midbandfrequency settings (final filter No. 1).

52

2.0-N Pi,--27dBm 11.0

/ 7T-24. 2* C!.

I \ I ]N

1. \ 1.Nj0

1SW 191 1 ot

1 In IIi 1I

FREQUENCY (MHZ) I

Figure 62. Measured transmission gain and VSWR at midbandfrequency settings (final filter No. 1).

InnS na

I :FORWARD : .... :"= :..- -..-.. E VE R .E ......... . . .. ......... ,.. . ... ... . . .... ..'. . . ...

cc VV =1 15. Vfrequncy"stting (fal filtr. 1).-

VVG: = 1Od m'-FO WAR V1034.V

. RPI- 6.666? RP2- 6.9669700.006 9718.066 9728.666 9730.001 9746.606 97 0.eee 9760.001 9770.000 9786.660 9789.999 9866.661

FREQ-MHZ

Figure 63. Measured transmission passband response for forward

and reverse power flow (final filter No. 1).

53

0 8.86dB

FORWARD...... ............ REVERSE.......

C-

OIPI. 8.0 P2 0.80-200.09 .. . . . . . . . .. . . . . . -

9700.8900 9718.888 9728.880 9738.881 9748.888 9758.88 9760881 977800 9780.808 9789.999 9800.001

FREQ-MHZFigure 64. Measured phase passband response for forward and

reverse power flow (final filter No. 1).

z 1.88

I-z 1.8 8 ....... ....... .... ... .

.8.......................

0i. .....

Li .738....... ...... ........... ....-

C-

OPI. 8.886 PPZ- 8.8888.888

9798.088 9710.808 9728.880 9730.001 9748.888 9790.000 9760,801 9778.e88 97880 9709.999 9888.081

FREQ-MHZ

Figure 65. Measured input and output reflection coefficientin passband (final filter No. 1).

54

29.090

~19.99 P~~-3OdBmLaJPi

FORWARD

OPI- 0- . ... RP21, 0.000

9796.989 9719.898 9729.000 9730.001 9740.SSe 975e.000 9760.861 9778.8e8 9788.080 9789.999 9899.091

FREQ-MHZ

Figure 66. Measured forward and reverse group delay inpassband (final filter No. 1).

0 .0

-2920999

-10 0.09 RZ- 9.9.....99 ..1.9 2..........9.9 79.9 76.9 77.9 78.0 98.9 98.8

z VVF=1-34HV

Figure ~ ~ ~ ~ ~ V241. 67Vesrdmdadrsos o he nu

power ~ ~ VG level (fnlfleV o )

- .............. . ~ j-553

vd6.1 V

Vv,=10.34 V

! / . -oda " -a eee VG2=I.3 4 -V : ........ ..... ........ 4LdJ

~--10 dBm ;i ' '

Rpi . 88 RP2- 8.0

9700.06 9716.68 9720.000 9730.801 9740.800 97S8.006 9760.e61 9770.000 9780.800 9789.999 900.001

FREQ-MHZ

Figure 68. Measured group-delay midband response for three

input power levels (final filter No. 1).

Table 2 shows the voltages necessary at any given frequency for filter

No. 2. The filter was adjusted for 0 ±2-dB transmission loss. Figures 69 and

70 show the required varactor and gate voltages; the corresponding gain and

VSWR are shown in Fig. 71; the 3-dB bandwidth and noise figure are shown in

Fig. 72. These measured characteristics were similar to the ones observed in

filter No. 1.

TABLE 2. TWO-SECTION FILTER NO. 2*

Frequency VVI VV2 VG1 VG2 BW Gain

(MHz) (V) (V) (V) (V) (MHz) (dB)

8274 0.26 0.0 1.71 2.04 36 -4.0

8842 2.95 2.5 1.69 1.96 35 -1.0

9185 5.48 5.0 1.73 1.94 32 0.0

9661 10.36 10.0 1.76 1.94 34 0.0

9941 14.33 15.0 1.82 1.94 36 -0.5

10118 17.70 20.0 1.82 1.94 36 -1.0

10224 20.60 25.0 1.82 1.94 36 -1.0

10297 23.20 30.0 1.81 1.94 39 -1.5

*Pin = -27 dBm; VD 6.11 V; T = 24.2 +O.10 C.in5

56

30.0 * V-2

25.0 /~Pi.,--27 d~m

I20.0- T-24.2'C:.1 I

0

W 15.00

5.0

0. 0 1. -1 I -- L -- I A- L - ~ L L -A -L--.

6i 6

w M 0 0)FREQUENCY (MHZ)

Figure 69. Measured varactor voltage for each element at givenmidband frequencies (final filter No. 2).

2. 1

2.0L ~

2V-

0 1.P±,--27d~m

w T-24.2C!.1

1., V!91

L L-.i L L A -I. I- A, -l 2- L..-2 --

in0 0 02 6 V

FREQUENCY (MHZ)

Figure 70. Measured gate voltages for each element at givenmidband frequenLies (final filter No. 2).

57

/ A A IN0.

GAIN

- \Pi,- -, 27:. "/r, jl -\"/ T-24. 2'C. I

2. / -2.!

> 1z

F erVSWR

.- 4 .

1 . 0-L- - .1 .L.I .i -..L.±. _L.-.4.. L. J .LA .LL AJJ_ i .

( Go 0 0FREQUENCY (MHZ)

Figure 71. Measured transmission gaii and VSWR vs midbandfrequency (final filter No. 2).

40.0 F30.e

L Pi,---27d~mBW ~T-24. 2*Ct. 11

_T 35.0_' 25. 1Z

NF_ t L

0 0

(D 0D m 0) fa0FREQUENCY (M:HZ,

Figure 72. Measured 3-dB bandwidth and noise figure vsmidband frequency (final filter No. 2).

The sequence of pints in Figs. 73 to 76 demonstrates symmetric performance

of the filter with respect to direction of power flow. Figures 77 and 78 show

the effect of power input on transmission and delay responses. Symmetry and

power input sensitivity were similar to the values observed in filter No. 1.

P-3OdBmjCD FORWARD

- -- *--.REVE RSE

Figure 73. Measured transmission passband response forfowr

and reverse power flow (final filter No. 2).

2o t -2 q I

W

W -10088.in. . .'1

I FORWARDREVERSE

0.

9E66.0e81 610.0e3 '46-6.00e 630.001 9640.86: 96TO.e9 9 .1K Q670.061 9680.001 969e801 9799.00@

FREQ-MHZ

Figure 74. Measured phase passhand response, forward

and reverie (final filter No. 2).

59

I. I a e

0) .. . ... .. .... .

.... P~n -3OdBm

. se- FORWARD /

UJ .480 ~~~ REVERSE ..... .....w

..6 ... .. ..

0.81 0.00 82 8.9.. .........

9600.081 9610.OeO 9620.080 9630.881 9640.801 9650.000 9659.999 96?8.001 9680.081 969e.081 9788.996

FREQ-MHZ

Figure 75. Measured input and output reflection coefficientfrequence response (final filter No. 2).

VVI19.25.V219.9,YI-1.74,2.n.96,I.28MA

/P.. -3OdBm

-FORWARD

REVERSE

0.0@0~8P ...... 6.... RP 2 .. . 9 8 e9666.991 9612.909 9620.00@ ?630.81 9640.001 36L;0.000 9659.19) 9678.091 9680.801 9698.981 3'99.896

FREQ-MHZFigure 76. Measured forward and reverse group-delay passband

response (final filter No. 2).

60

9.9 I1.[ I~h 5 . 4

* rV

1.8-OdBm

------- -- OdBm-2e.eee

0PP 00.f0 RP2= 0.000

9608.881 9610.08 9620.000 9630.001 9640.Oee 96SO.000 96S9.999 9,70.001 9680.901 9690.901 9768.000

FREQ-MHZ

Figure 77. Measured midband response for three input powerlevels (final filter No. 2).

26.869

III

20 0 .......-

/ - -30 dBr-20dBm-10 dBm -

PFI 0..000 PP2 8....96000.01 9610.000 9620.030 ?630.8C1 9640.001 96SO.000 %89g.?q9 36-0.001 9680.801 9698.991 9700.000

FREQ-MHZ

Figure 78. Measured group-delay midbarid response for threeinput power levels (final filter No. 2).

61

The relatively similar performance of the two filters was attained without

special prescreening of parts and without elaborate matching techniques. Thus,

the reproducibility is good, and better uniformity and better overall filter

performance can be expected with improved fabrication controls. Resettability

of obtained results is a function of case temperature, device dissipation, and

instrument accuracy.

62

SECTION IV

CONCLUSIONS AND RECOMMENDATIONS

The Phase II, Task 2 program achievements, findings, and recommendations

can be summarized as follows:

* The tunable resonant element developed during Task 1 was redesigned

to eliminate a variety of detrimental effects. This new element uses

only one FET and one varactor while maintaining good electrical

symmetry.

* Filter performance is essentially symmetric with respect to power

flow. Symmetry suffers with "near instability" adjustments.

* Element stability was improved by eliminating critical rf-choke

lines.

* Tuning bandwidth capabilities of 3.6 GHz were demonstrated on a

single resonant element.

* Two two-section filters that are tunable over a frequency band of

2.0 GHz ±20 MHz in the 8.2- to 1O.3-GHz frequency range were completed

and tested.

* A 29 ±11-MHz 3-dB bandwidth was demonstrated over the entire tuning

range.

* Response adjustments from maximally flat to Chebyshev with 2-dB

ripple could be achieved with proper coupling control.

* Stability considerations limit the response bandwidth of multisection

filters to about 20 MHz.

• Power input level sensitivity of responses increases with increasing

selectivity. For a 3-dB bandwidth of 30 MHz, a power input level of

-10 dBm causes only minot response degradations. We recommend study

of this problem aiming at a level increase of 20 dB.

* Noise figure is a function of FET feedback arrangement and is about

18 dB for a single element. There seems to be no straightforward

cascading relationship for multiple-element filters. We recommend

further work to improve the noise figure and to investigate multiple-

element noise relationships.

" Future tasks should concentrate on improvements with respect to

power input level, noise figure, and addressing and driving circuits

that determine frequency setting and ultimate tuning speed.

63

REFERENCES

1. A. Presser, "Ultra-Fast Tunable Microwave Filter," Final Report,

Contract No. N00173-77-C-0201, Aug. 1978.

2. A. Presser, "Ultra-Fast Tunable Microwave Filter," Final Report,

Phase II, Task 1, Contract No. N00173-79-C-0186, Aug. 1980.

3. G. L. Matthaei et al., Microwave Filters, Impedance-Matching Networks

and Coupling Structures (McGraw-Hill Book Co., New York 1964), p. 421.

4. J. F. Cline et al., "Tunable Passive Multi-Couplers Employing

Minimum Loss Filters," IRE Trans. PGMTT-7, 121 (Jan. 1959).

5. S. B. Cohn, "Dissipation Loss in Multiple-Coupled Resonator Filters,"

Proc. IRE 47, 1342 (Aug. 1959).

6. P. Penfield, Jr., and R. P. Rafuse, Varactor Applications (MIT Press,

Cambridge, MA, 1962).

7. P. Wolf, "Microwave Properties of Schottky-Barrier Field-Effect

Transistors," IBM J. Res. Dev. 14, 125 (March 1970).

64