WMB-7: MTT-S Workshop on Filter II: Practical Aspects of Microwave Filter Design and Realization Passive Intermodulation in Microwave Filters: Experimental Investigation Giuseppe Macchiarella (+), Alessandro Sartorio (+), (+) Dipartimento di Elettronica e informazione del Politecnico di Milano (++) Forem (An Andrew Company) Outline Overview of physical mechanisms in PIM generation PIM in microwave cavities and filters Experimental investigation of PIM produced in coaxial cavities Evaluations of PIM in filters and duplexers through non-linear circuit simulations Some case histories Conclusions
Generation of PIM in microwave cavities 1) Due to the materials Ferromagnetism Contacts between metals (even identical!) Galvanic Silver plating (?) ) Due to the cavity structure Shape and dimensions Tuning structure Input/output coupling system Input/output connectors 3 Passive Intermodulation (PIM) and base station combiners Passive intermodulation in cellular base station duplexers arises from the very weak non-linearity produced in the filters cavities. Intermodulation generated by the TX filter may fall into the RX band; then, reaching the LNA input, it worsen the overall performances of the system. Duplexer TX Filter RX Filter PA LNA GSM Frequencies TX Band: 935 96 MHz RX Band : 89 915 MHz 4
Passive Intermodulation Typical Requirements GSM PCS 19 DCS 18 UMTS Input power x 43dBm x 47dBm x 46dBm PIM in RX band -1dBm (3 rd order) -118dBm (3 rd order) -1dBm (7 th order) 5 Experimental study on PIM generated by microwave coaxial cavity Purpose: To investigate the dependence of PIM on the cavity parameters Methodology: Fabrication and measurements on suitably designed test cavities Question: How to design the test cavities? 6
A simple circuit model for describing PIM generation in resonators Heuristic Assumption: non-linearity associated to the cavity dissipation Resonant R R Cavity R L eq C eq R p Equivalent Circuit R 1 X 1 L =, C =, R = R i eq eq p ω ω= ω ω L eq ( ) Q ω = r L eq ωleq ωl QL = R + r R eq R p is a non-linear resistor described by the following I-V characteristic: ( ) v r i r i r i x i = 3 + 3 +... 1+ x=r 3 /r 7 Two-tone characterization V(f 1 ) V(f ) jx(f) R P I k(i) R Generator f X f X f 1 X f Fn f f ( ) ( ) = ( ) ( ) R Load Input tones: Frequencies: f 1, f Amplitude (volt.): V Input available power (P av ): V 8R 8
Simplified circuit analysis Output power at f 1, f : Output power at f 1 + f, f + f 1 (PIM): P L k k 3 9 P Q av k (3) 6 4 Px Q L P = P + Q Q α av k (1) ( 1 ) L Q α A 1 + Q k k=1, 1 1 r α =, P =, A = 1+ 1+ α α 1 k x k x ( QLFn ( fk (1) )) ( QLFn ( fk (3) )) 9 Evaluation on PIM with the simplified model Model parameters: Resonant frequency f Loaded and unloaded Qs Intrinsic PIM (P x ) Input power 3 Q Q L L ( P ) = 3( P ) ( P ) + 4log 6 log 1+ + log ( α A ) IM dbm disp dbm x dbm k k Q Q For an ideal cavity (r 3 =): P x, P IM 1
Dependence of P IM on the model parameters Decreases of db/db with P x Increases with Q L /Q for Q L /Q << 1, then decreases Increases of 3 db/db with P in PIM (dbm) PIM (dbm) -1-7 -11-8 -1-9 -1-13 -11-14 -1-15 Slope: db/db Slope: 3dB/dB -13-16 -14-17 -15-18 -16 6 65 7 75 8 85 9 95 1-19 5 (dbm) 3 35 P x 4 45 P av (dbm) 11 Parameters to be varied in the test cavities Unloaded Q (cavity volume) Loded Q (Input/Output coupling level) Intrinsic non-linear parameter P x : - cavity shape - coupling structure - tuning element - Silver plating 1
Physical structure of the test cavities (GSM TX band) Cavity type: Coaxial with capacitive loading Outer cross section: square Tuning element: screw into the inner conductor 13 Dimensions of the built cavities and varied parameters 15 mm ~ 5-65 mm 45/.5 mm ~ 55-7 mm Width of the cavity:.5 or 45 mm (changes Q ) Section of inner conductor: circular or square Input/Output coupling: capacitive or inductive (tap). Loaded Q: 15, 5, 4 (by changing the in/out coupling) Tuning: with or without the screw Total built cavities: 3 Inductive Capacitive 14
Does filter plating affect PIM? Skin depth in silver around 1 GHz is about 1µ; so 5µ of silver tickness should be sufficient (no contact between different conductors). However: The realizable silver tickness is not uniform inside the cavity (less at the bottom) Irregularities and impurities on the surface are possibly generated by the plating process Contacts between silvered and not-silvered parts become unavoidable (tuning screws) Conclusion: PIM has been measured before and after silver plating the test cavities (5µ tick.) 15 PIM Measurement (two-tone test) Measurement test set: : Summitek mod. SI 9A (transmission set-up) Tone f 1 TX 5 Ohm Tone f SUM DUT DPX RX PIM power meter Input power meter (max: 43 dbm/tone) Instrument noise floor: -13 dbm Choice of the two-tone frequencies Tone f 1 must be near the lowest end of the TX band (935 MHz) Tone f is around 956 MHz, in order to obtain the lower 3 th order intermodulation product inside the RX band (< 915 MHz) Cavities are tuned at about (f 1 +f )/ 16
Measurement results (sample) Circular Circular Square cross cross Cross section, section, Section, capacitive capacitive Tap coupling, coupling, Silver-plated not silver silver plated plated -1-11 -6 dbm dbm dbm -11-15 -7 with -114 screw with screw -116-8 -11-118 -9-115 -1-1 -1-1 -14-11 -16-15 -1-18 with screw without without screw without screw screw -13-131 15 5 3 35 4 45 1 15 5 3 35 4 QL 45 QL QL Circular Circular Square cross cross section cross section section -- Capacitive -- Capacitive -- Tap coupling coupling -- -- Not Silver Silver silver plated plated 17 Comments on measurement results (1) PIM dependence on Q L estimated by the model seems similar to that The increase of the cavity size reduces PIM (as Q increases) The rod cross section (square or circular) does not seem to affect PIM Capacitive coupling has better performances than tap coupling 18
Comments on measurement results () Tuning screws strong increase PIM (5-15 db) Measured PIM after silver plating seems to increase (especially with tap coupling and tuning screws). 19 Dependence of PIM on input Power -114-116 -118 15 3 PIM (dbm) -1-1 -14-16 -18-13 -13 13 Dashed lines: Slope db/db -134 35 36 37 38 39 4 41 4 43 P (dbm) 3 th th order Model polynomial prevision: model slope not sufficiently = 3dB/dB accurate!
Matching PIM vs. P Non-linear function for the I/V characteristic: I g ( 1 k1 tanh ( kv )) = + g =B eq /Q 13 k = 6.36 1 1 4 k = 3.4 1 Simulator: ADS (Harmonic Balance) 1 Conclusions (first part) For reducing intrinsic PIM of a cavity: - no tuning screws - no soldering in the structure (avoid tap coupling) - thickness of silver plating at least 3-5 times the skin depth (for avoiding the influence of the adhesive layer, typically realized with nickel) The dependence of PIM on Q L and Q has been demonstrated The shape of the inner rod (square or circular) does not seem to affect PIM A 3 th order model is not sufficient to match PIM vs. P o
Overall PIM generated in a duplexer Reference Filters TX FILTERS Lower passband frequecy (MHz): 94. Upper passband frequecy (MHz): 96.5 Return Loss (db):. Number of resonators: 1 Unloaded Q of resonators:. External Q: 4.65 Number of Transmission Zeros: Frequencies (MHz): 917. 918.5 RX FILTERS Lower passband frequecy (MHz): 878. Upper passband frequecy (MHz): 918.5 Return Loss (db): 16. Number of resonators: 1 Unloaded Q of resonators:. External Q: 7.18 Number of Transmission Zeros: Frequencies (MHz): 94.5 96.3 3 Equivalent circuit for the resonators PLC ID= LC1 L= Leq nh C= Ceq pf + NLRES ID= R1 A1= g A3= g B 1 R = Q, C =, L = C g L eq eq B ω ω B g =, g = Q Px eq f =935 MHz f 1 =931.65 Q =7 f =938.35 B =1.55 S P x =1 dbm 4
Duplexer linear response TX 1 1 RX PORT P= 1 Z= 5 Ohm TLINP ID= TL Z= 54.83 Ohm L=.77 mm Eeff= 1 Loss= F= GHz TLINP ID= TL1 Z= 6.47 Ohm L= 4.9 mm Eeff= 1 Loss= F= GHz PORT P= Z= 5 Ohm PORT P= 3 Z= 5 Ohm Antenna -1 - -3-4 -5-6 -7-8 -9 Duplexer response -1 port.85.87.89.91.93.95.97.985 Frequency (GHz) N.B. Ideal inverter has been employed in the filters equivalent circuit 5 Non linear analysis (Harmonic Balance) TX input: two-tone (93 MHz, 95 MHz), 43 dbm/tone HB analysis: 5 harmonics/tono, max mix order 9 Non-linearity in all resonators (both TX and RX filters) 6 4 - -4-6 -8-1 -1-14 -16 Out spectra.88.89.9.91.9.93.94.95.96.97.98 Frequency (GHz) TX RX DPX ANT Blue curve, RX out Purple curve, ANT out 6
PIM at RX out vs. frequency -1 PIM Contribution from RX filter -15-11 -115-1 -15-13 -135-14 -145-15 9 9.5 95 97.5 91 91.5 915 917.5 9 9.5 Blue curve: non linearity in both filters Purple curve: non linearity in TX filter only PIM is due to the TX filter only 7 Contribution to PIM from TX resonators -1 PIM contribution from TX resonators -15-11 -115-1 -15-13 -135-14 -145 : Res. 1 X: Res. 1+ : Res. 1++3 : Res. 1++3+4-15 9 95 91 915 9 9.5 Only the first 4 resonators from ANT node give a contribute to PIM 8
PIM vs. Filters topology -9-95 -1-15 -11-115 -1-15 -13-135 -14-145 PIM vs Filters Topology : Triplets (input TX side) X: Box section : Quadruplet : Triplets (out TX side) -15 9 9.5 95 97.5 91 91.5 915 917.5 9 9.5 Triplets close to the ANT node produce lower PIM! 9 Some Case Histories 3
Problems with silver plating Poor silver plating: PIM 1dBm x45dbm Solution: lateral hole to improve silver plating PIM 115dBm x45dbm 31 Resonators design Resonator too loaded and with thin ring PIM 15dBm x43dbm Solution: increase depth of cavity and resonator ring PIM 1dBm x43dbm 3
Dissimilar metals Stainless steel screws in silver-plated cavity PIM 11dBm x45dbm Solution: shorter screws PIM 15dBm x45dbm 33 Conclusions: Good Rules for making low PIM filters - Wide resonator ring - Deep cavities and not too loaded resonators - No sharp corners - No contact between dissimilar metals - Attention to the position of cross couplings - Short tuning screws in the cavities - Cleaning - Attention to solder joints - Good silver plating - Ensure good contacts 34