Computing and Enhancing Power Handling in Bandstop Filters

Transcription

1 Computing and Enhancing Power Handling in Bandstop Filters By R. V. SNYDER RS MICROWAVE BUTLER, NJ IMS2007 Workshop June 4, 2007

2 CONTENTS OF THIS TALK -Power handling in bandpass vs. bandstop similarities and differences -Passband vs. stopband -Peak power and average power -Some examples (quasi and full-elliptic bandstop with wide pass and stopbands) expensive failures noted, analyzed and cured -Distributed vs lumped -Mixed circuits (lumped and distributed) -Limitations of each -How to improve power handling -Performance (and other) costs associated with improvement -Generally warranted conclusions

3 -Similar problems: Bandpass and Bandstop For voltage breakdown (peak power)... air is the enemy avoid gaps For thermal breakdown (average power) dissipation is the enemy

4 Bandpass (peak power issues) -Peak power problems for ladder-type bandpass are near network center/high-power input, if power is applied in the passband/stopband region -Peak power problems for bandpass with finite frequency TZ s are near the resonators providing the TZ s. - Center and input refer to physical location in network

5 Bandpass (average power issues) -Passband area for bandpass is troublesome due to resonator impedance (reflections) and the effects of bandwidth (lower percentage BW means lower average power handling capability due to dissipation) -Stopband area for bandpass is usually sensitive to peak power (reflection issues) but not average power, because dissipation is low in the stopband area

6 Bandstop (peak power issues, traditional thinking)[1] Peak power problems for bandstop (Chebychev, narrow stopband) are near network high-power input/center, if power is applied in the stopband/passband region -[1]: Torgow and Collins, Bandstop Filters for High-Power Applications, Trans MTT, Sept, equations 12, 16, 17, 19 and 20 -Peak power problems for bandstop (quasi or full-elliptic, wide stopband) are distributed throughout the filter, with the worst problems in resonators effectively providing the frequency of a particular transmission zero (TZ) - This wasn t known when [1] was written because bandstop filters with wide stopbands were not possible - Center and input refer to physical location in network

7 Bandstop (average power issues) Passband area for the bandstop is usually the not a significant area of difficulty because dissipation and VSWR are minimized here -Stopband area for bandstop is difficult because of resonator impedance (reflections) and couplings to main through path

8 Bandpass and Bandstop (continued) -Stopband reflections can cause problems for both bandpass and bandstop -Passband edges for bandpass correspond to notch edges for bandstop These are trouble regions because dissipation losses are significant in these areas and thus average power capacity is most affected here Sounds the same, but passband for the bandstop is analogous to stopband for the bandpass

9 Bandstop example of interest Wide stopband, wide passband, quasi-elliptic [2]: R. Levy, R. Snyder and S. Shin, Bandstop Filters with Extended Upper Passbands, T-MTT-S, June 2006 Capacitively shortened coupled resonators [2] Short TL sections between resonators contribute to rejection and make response quasi-elliptic

10 Bandstop example of interest Wide stopband, wide passband, quasi elliptic

11 Bandstop example of interest Wide stopband, wide passband, quasi-elliptic This filter safely handles 1000 W in passband region BUT It was observed that power levels of 100 W in stopband area selectively destroyed loading capacitors. Depending on stopband frequency, different capacitors were damaged. This was an expensive failure (a failure basically means throwing away a very costly filter) and it was decided to investigate via simulation.

12 Electric Field Plots and Power capacity E-field plots are compared with 50 watts average power applied at each rejection pole frequency. Hot spots can be located from the E-field plots, showing high peak power at resonant frequency. Loading capacitors at high peak power area must be designed with dimensions below breakdown voltage limits. (Peak E-field intensity should be below the breakdown voltage of the loading capacitors) At passband frequencies nearest the rejection band, loading capacitors face higher peak power. Highest peak power occurs at passband edge transition frequency. Not at 1 st resonator must be analyzed using field solver. The resonator with the greatest sensitivity to the applied peak power determines the power capacity of the filter.

13 Bandstop example of interest Wide stopband, wide passband, quasi-elliptic (Movie clip follows)

14 JTIDS/MIDS Elliptic Band Rejection Filter

15 JTIDS/MIDS Elliptic Band Rejection Filter (movie clip follows)

16 JTIDS/MIDS Elliptic Band Rejection Filter (some slides from the movie clip now follow for purposes of visualization in the handout)

17 E-field at 1 st Rejection Pole 50 W average power applied at 964 MHz, 1 st rejection pole Hot area: Resonators close to applied frequency and close to power imposed port, face high peak power. 50 W 50 W

18 E-field at 2 nd Rejection Pole 50 W average power applied at 973 MHz, 2 nd rejection pole Hot area: Resonators close to applied frequency and close to power imposed port, face high peak power. 50 W 50 W

19 50 W average power applied at 999 MHz, 3 rd rejection pole Hot area: Resonators close to applied frequency and close to power imposed port, face high peak power. E-field at 3 rd Rejection Pole 50 W 50 W

20 50 W average power applied at 1081 MHz, 4 th rejection pole E-field at 4 th Rejection Pole Hot area: Resonators close to applied frequency and close to power imposed port, face high peak power. 50 W 50 W

21 E-field at 5 th Rejection Pole 50 W average power applied at 1182 MHz, 5 th rejection pole Hot area: Resonators close to applied frequency and close to power imposed port, face high peak power. 50 W 50 W

22 E-field at 6 th Rejection Pole 50 W average power applied at 1203 MHz, 6 th rejection pole Hot area: Resonators close to applied frequency and close to power imposed port, face high peak power. 50 W 50 W

23 E-field at 7 th Rejection Pole 50 W average power applied at 1212 MHz, 7 th rejection pole Hot area: Resonators close to applied frequency and close to power imposed port, face high peak power. 50 W 50 W

24 E-field at lower Passband edge 50 W average power applied at 959 MHz, lower passband edge Hot area: Resonators (6 th ) close to applied frequency face the highest peak power 50 W 50 W

25 E-field at upper Passband edge 50 W average power applied at 1226 MHz, upper passband edge Hot area: Resonators close to applied frequency face high peak power 50 W 50 W

26 E-field at Passband (1-1) 1000w average power applied at 750 MHz, passband area. Hot area: At 750 MHz, 1 Kw, tight coupled line spacing area faces high peak power 500 W 1000 W

27 E-field at Passband (2-1) 1000w average power applied at 1400 MHz, passband area. Hot area: At 1400 MHz, 1 Kw, tight coupled line spacing area faces high peak power 500 W 1000 W

28 E-field at Passband (3-1) 1000 W average power applied at 1700 MHz, passband area. Hot area: At 1700 MHz, 1 Kw, tight coupled line spacing area faces high peak power 50 W 1000 W

29 Distributed vs. Lumped (and mixed together) -Similar problems: For voltage breakdown (peak power)... air is the enemy avoid gaps For thermal breakdown (average power) dissipation is the enemy (The problems listed are common to all filter types) It is easier to avoid air gaps with most distributed circuits than with lumped -Coiled inductors are difficult to fully insulate -Component connections are weak points -Distributed element Q tends to be greater than lumped element Q -Bandwidths can be more narrow and thus breakdown is a bigger problem -Because bandwidths are typically more narrow, dissipation remains as an issue even with higher Q -Very difficult to model in E-M domain; experimental results match expectation from [1], because rejection bands are typically narrow -Easy to model in circuit domain but breakdown doesn t show up without making assumptions as per [1]

30 Distributed vs. Lumped (and mixed together) Lumped-Distributed mixture lumped series lowpass mainline capacitivelycoupled to short circuited coax resonators-updated version of [3] [3] R. V. Snyder, Quasi-elliptic Compact High-Power Notch Filters Using a Mixed Lumped and Distributed Circuit, TMTT-S, April 1999

31 Distributed-Lumped-1030/1090 MHz dual notch filter Dual 1030/1090 MHz notch filter; 1000 W peak, 5% duty in stopband, 300 W peak, 40% duty in passband. We have not had a power failure in over 1000 units tested, and so no damage photos are available. These are simply too expensive to force into failure!

32 Dual notch filter example W peak, 5% duty (stopband), 300 W peak, 40% duty (passband)

33 1030 MHz notch example (accidental failure) Similar single notch, F0=1030 MHz, 800 W peak (25% duty)

34 1030 MHz notch example (before failure) Similar single notch, F0=1030 MHz, 800 W peak (25% duty) Simulated and measured data, before applying too much power in reject band

35 1030 MHz notch example, high power in reject band 1200 W applied, 1030 MHz, at 80,000 feet.first two resonators burned air gap between resonator end and insulator caused arc, which then propagates

36 1030 MHz notch example, high power in reject band 1200 W applied, 1030 MHz at 80,000 feet.first two resonators burned, air gaps visible

37 1030 MHz notch example, high power in passband Intentionally forced to failure 1030 MHz single notch filter, 1 Kw peak (5% duty) in passband, 300 W (40% duty)

38 1030 MHz notch example, high power in passband Basic structure, shown prior to applying excessive high power note it is another mixed lumped-distributed network

39 1030 MHz notch example, high power in passband 1030 MHz single notch filter, 1 Kw peak (5% duty) in passband, 300 W (40% duty) measured response prior to high power

40 1030 MHz notch example, high power in passband 1250 W peak, 100 W average applied at 1000 MHz, 80,000 feet (specification is 1000 W peak, 50 W average)

41 1030 MHz notch example, high power in passband 1250 W applied, 1000 MHz at 80,000 feet.center resonators burned, air gaps visible

42 How to improve power handling for bandstop (or any other) filters 1. Analyze problem using field solver to identify hot spots in structure. These will depend on the frequency of the high power energy. 2. Breakdown voltage for elements in hot spots must be greater than applied RF voltage found in step (1). 3. Eliminate air gaps much easier said than done, because insulators expand and contract as temperature changes, causing gaps to form. However, shrink fitting insulators, taking advantage of temperature rather than being subject to thermal changes, and preferentially insulating those areas most subject to breakdown.all helpful -It is important to note that little air gaps are worse than big air gaps-electrons, like fleas, jump more readily over the small gap 4. Be sure to heat sink to remove thermal flux resulting from dissipation of high average power 5. All the above are obvious, but the failures are so expensive and embarrassing that it pays to pay attention!

43 Performance (and other) Costs for power handling enhancement 1. Increased insertion loss is a typical penalty. Elimination of air gaps sometimes requires inclusion of compliant fill material, with loss tangent higher than that of the insulators. 2. Volume.it usually takes increased dmensions because achieving loading capacitance or required resonator diameter or space from housing or other, simply takes volume. 3. Design time field solvers take time and money, both to validate appropriate models and simply to translate results into practical dimensions. 4. Manufacturing cost: This is an always penalty, because eliminating gaps, ensuring flat surfaces for heat sinking these are labor-intensive processes and thus are costly.

44 Generally warranted conclusions 1. Failures based on application of high power are usually terminal for the device under test. The DUT becomes a throw-away. 2. Design for high power requires understanding of the network on a detailed, internal examination basis. Where are the regions of sensitivity to high voltage, high current, thermal expansions, etc? 3. Given validated models, field solvers find the problems but overcoming the problems requires careful design and manufacture. 4. Obtaining validated models requires some controlled failures to ensure that predicted failure modes are real. Prepare to spend money! 5. Finally, bandstop and bandpass filters have very similar failure modes, but some critical differences. These have to be appreciated in order for successful design and test phases to be completed.

45 Acknowledgement Thanks to my colleague, Sanghoon Shin, for his excellent work in model testing and model validation. Thanks to the technical staff at RS Microwave for the design and fabrication of excellent high power notch filters. And thanks to you, the audience for sitting here so patiently.