Lower-loss Filters on Langasite

The advantages of langasite as a substrate material for SAW filters are not realized entirely till now because of the natural unidirectivity of most cuts available for practical applications. This paper presents new results that allow the utilization of langasite for SAW filters.

The mathematic model, SPUDT cell structures and design method were developed for naturally unidirectional orientations. This approach allows designs of slanted and RSPUDT filters on yxlt/48.5o/26.6o langasite with 7dB – 10dB insertion losses and bandwidths 1% – 3%.

INTRODUCTION

Langasite (LGS) has great potential as substrate material for SAW filters [1-3]. One of the more promising material orientations and propagation directions for SAW devices are defined near the Euler angles [0°, 140°, 22o-25°]. This double rotated Y cut is characterized by the following advantages.

  • Higher coupling coefficient comparing with quartz;
  • Low diffraction;
  • Low sensitivity to fabrication process parameters;
  • Temperature stability, i.e. TCF ≅ 0;
  • Low level of excitation of bulk acoustic waves;
  • Zero power flow angle.

Furthermore, these LGS cuts show a natural directivity enabling the design of natural single phase unidirectional transducers (NSPUDTs) [4,5]. An NSPUDT is a normal finger transducer exhibiting a natural unidirectivity owing to anisotropic properties of the substrate crystal. It seems that the advantages of such an effect would have been taken for the development of low loss SAW filters. However, for a long time, the applications of langasite have been limited to medium-loss filters, resonators and impedance element structures.

Development of state-of-art SPUDT or RSPUDT based filters on LGS is hindered because they need different types of cells owning to its direct and reverse unidirectivity. In addition, for RSPUDT structure, the phase difference between transduction and reflection centers in the cells has to be close to 45 degree in order to provide resonant cavities inside the transducer. Because of the effect of natural unidirectivity, a general solution to meet these requirements for LGS has not been found so far.

In this paper we describe a new approach to the synthesis of low-loss filters similar to RSPUDT based filters. Design principles as well as theoretical and experimental results are presented.

MATHMATIC MODEL

Equivalent circuit model was modified and adapted for the simulation of different types of SAW filters on LGS. The basic section of a transducer is shown in Fig.1. It extends between centers of neighboring fingers and includes three basic cells connected with electrical circuit by transformers r.

Each cell is described by the following 3x3 matrix:


Pic56

where Z=Zo or Zm is acoustical resistors for  free or metalized surface, respectively; L v i θ =ω / is the angle shift under wave propagation in the cell.;  r is transformer ratio; and v is SAW velocity. The circuit element B superposed with electrode edges is a complex number. It describes the second order effects such as propagation loss Re(B) and energy storage effect Im(B).

Fig.1. Equivalent circuit model representation.

Fig.1. Equivalent circuit model representation.

In order to adapt the model for simulations on LGS and describe the effect of natural unidirectivity, an additional angle φ was included into the phase shift for cells under fingers, i.e. θ =θ +ϕ m . The sign of the angle depends on SAW propagation direction at finger edges. It should be mentioned that the angle is equivalent to a shift between reflection and transduction centers in a transducer.

The angle φ and substrate parameters depending on relative film thickness were determined by experiments. The test structure is shown in Fig. 2- a. Typical responses for different film thickness are shown in Fig.2-b, when transducers 1 and 3 are used as in put and output, respectively. The responses for direct and reverse directions on LGS, when the  transducer 2 is input and transducers 1 or 3 are used as output, are shown in Fig.2-c.

The analysis of responses for different film thicknesses allows the determination of the angle φ and the dependence of reflection and energy storage effects as the first approximation. The average φ value turns out to be around 60o. It should be noted that the angle depends on film thickness. The angle change is around 10% with thickness variation from 0.4% to 3%. More precise tuning of these parameters has been carried out with the development of different types of filters on LGS substrates.

Fig.2. Test structure (a) and typical responses for 1-3 transducers (b) and 1-2 and 2-3 transducers (c).

Fig.2. Test structure (a) and typical responses for 1-3 transducers (b) and 1-2 and 2-3 transducers (c).

The advantage of this model is its flexibility with varieties of finger structures because the parameters of cells are determined by edge positions of electrodes. The model has been applied to the analysis of single-phase unidirectional transducers (SPUDT) on LGS in order to estimate their efficiency for SPUDT filter designs. Analysis of different electrode structures allows the selection of basic SPUDT structures, that can be utilized for the development of lower-loss SAW filters on substrates with natural unidirectivity, such as LGS. These SPUDT sections are shown in Fig.3. It should be mentioned that the efficiencies of the section’s unidirectivity for the forward direction (coinciding with the natural unidirectivity) and the reverse direction are close. This feature simplifies the development and allows better results of SPUDT filters.

Fig.3. Sell structures for forward (a) and reverse (b) directions and with reflection removed (c).

Fig.3. Sell structures for forward (a) and reverse (b) directions and with reflection removed (c).

SYNTHESIS OF RSPUDT FILTERS

Because the phase between transduction and reflection centers for SPUDT sections shown in Fig.3 is not optimal (45o), it is a problem to use these sections for resonant SPUDT (RSPUDT) filters. The method based on variations of electrode structures inside a transducer region [6], has been used in order to reach a total phase of reflections required to realize resonant conditions in the transducer region.

The method consists the division of an electrode section by elements with equal widths. Usually a section’s width is equal to the wave length λ. The element width is λ/N, where N is the number of elements in the section. Every element can be represented as a strip connected with opposite transducer busbar or a free space element. So the maximal number of layout variants for the current section is 3N. The number of simulation variants can be reduced significantly if some restrictions are imposed in the synthesis algorithm. These restrictions are listed as follows:

  • Variants including neighboring strips connected to bars with opposite polarity are excluded because in this case the transducer is shorted;
  • An electrode is composed of a few elementary strips connected to a common bar. Maximal andminimal electrode widths are limited.
  • A gap is the sum of free space elements. Maximal and minimal gap widths are limited.

The developed design algorithm is based on an ordinary routine applied for synthesis of RSPUDT filters. A filter structure is represented as a combination of basic elements. Responses of the synthesized filter are simulated for every current structure and compare with requirements for the filter. The error Eg between the current response and requirements is estimated.

If simulated Eg is reduced for a current section structure the structure corresponding to a minimal error is fixed. Then a next section of the transducer is analyzed. The optimization algorithm allows minimizing the error after a few iterations. In this case the simulated response becomes close to the required.

Analysis of responses corresponding to current  filter structure is carried out for tuned filter. Values of elements of matching circuits are determined automatically for fixed circuit configuration and termination impedances. The reliability of synthesis results depends on both accuracy of model and charge distribution calculation. Charge distribution on electrodes is determined by solving the electrostatic task for each electrode configuration.

FILTERS DESIGN

Figs. 4 and 5 refer to slanted filters on LGS substrate with orientation yxlt/48.5o/26.6o. The filter shown in Fig.4 used NSPUDT transducer and reversed one with 1/6 λ fingers. The fractional bandwidth of the filter is 2.6%, and insertion loss is 12.4dB.

Fig.4 Responses of slanted filter based on NSPUDT

Fig.4 Responses of slanted filter based on NSPUDT

The second filter is based on SPUDT structures  shown in Fig.3. It consists of unweighted input transducer and withdrawal weighted output transducer. Weighing functions for reflectors were optimized in order to allow minimal passband ripples based on simulations. The lengths of transducers are 78 λ and 120 λ. respectively, with aperture of 44 λ. 3 dB bandwidth of the filter is 3.6 MHz (3.1%), insertion loss is less than 10dB (9.6dB) and out of band rejection is about 50dB.

Experimental results (Fig.5) agree with simulations (dotted line) very well. The comparison confirms that the model we developed is adequate for the analysis of SPUDT filter on LGS. It should be mentioned that the use of novel SPUDT structures developed in the work allows to improve the SAW filters parameters over NSPUDT. Insertion loss has been reduced by 2.5 to 3dB, even with wider bandwidth and lower passband ripple.

Fig.5. Responses of slanted filter based on SPUDT cells shown in Fig.3.

Fig.5. Responses of slanted filter based on SPUDT
cells shown in Fig.3.

Narrower filters have been designed by the optimization routine described above as well. The width of the basic synthesis elements is 0.125λ. Maximal gap and electrode widths are restricted as 0.25λ and 0.375λ, respectively. The structure of these filters is shown in Fig.6. The central grating in these filters is used both as feedthrough screen and additional structure element for the synthesis of the filter responses. A fragment of electrode structure is shown in Fig.6-b. As indicated in the picture, because of irregular electrode structure, the pitches between transduction centers are variable.

Fig.6. Schematic of in-line RSPUDT filters (a) and fragment of electrode structure (b).

Fig.6. Schematic of in-line RSPUDT filters (a) and fragment of electrode structure (b).

Performances of RSPUDT filters on the same langasite cut are shown below.

3dB bandwidth of the 129 MHz filter (Fig.7) is 3.1 MHz or 2.4%, with 8.7dB insertion loss. The  filter fits in 3.8mm square package.

Responses of 251 MHz filter with 3dB bandwidth of 3.26MHz or about 1.3% are shown inFig.8. The pass band ripple is only 0.1 dB, with insertion loss of 7.7 dB. The filter also fits in 3.8mm square package.

Fig.7. Experimental responses of 2.4% RSPUDT filter comparing with simulation (dotted line)

Fig.7. Experimental responses of 2.4% RSPUDT filter comparing with simulation (dotted line)

Impulse response of the 251MHz filter is shown in Fig.9, which is typical for RSPUDT filters. The impulse duration T1 is about 2.7 times of the one way propagation time of SAW T0.

Fig.9. Impulse response of 251MHz RSPUDT filter

Fig.9. Impulse response of 251MHz RSPUDT filter

It should be mentioned that comparing with experimental results for identical filters on quartz substrate, insertion losses of LGS filters are 5 to 7dB lower. Experimental results agree with simulations for different types of filter designs very well.

CONCLUSION

Modified equivalent circuit model has been applied for the analysis of SAW filters on materials with natural unidirectional effect such as langasite. Basic SPUDT cells for forward and reverse directions have been developed for LGS cuts with Euler angles [0°,140°,22o-25°]. Synthesis method based on the optimization of electrode structures by varying basic structure elements has been development. This method has shown effective for the designs of SAW filters on LGS substrate, including RSPUDT filters.

Experimental results agree with simulations very well. Slanted and RSPUDT filters on yxlt/48.5o/26.6o langasite have been designed and fabricated. Insertion losses of these filters are 7dB – 10dB with bandwidths 1% -3%.

REFERENCES

  1. P.G. Ivanov, V.M. Makarov, V.S. Orlov, V.B. Chvets, , “Design of SAW filters on langasite”, Proc.IEEE Ultrasonic Symp.., 1999, pp.51-54.
  2. D.C. Malocha, M.P.da Cunha, D. Puccio and K. Casey “Investigations of Langanite and Langatate materials for use in SAW device applications” Proc. IEEE Ultrasonic Symp. pp 231-234, 2001.
  3. S.N. Kondratiev, T. Thorvaldsson, S.A. Sakharov, O.A. Buzanov, A.V. Medvedev “Extraction of COM parameters on langasite substrates and the application to design of a SAW filter”, Proc. IEEE Ultrasonic Symp., pp 53-56, 2001.
  4. S. A. Sakharov , A. N. Zabelin, O. A. Buzanov, A. V. Medvedev, V. V. Alenkov, S. N. Kondratiev “Nondestructive investigation of 4-inch langasite wafer acoustic homogeneity” Proc. IEEE Ultrasonic Symp. 2002, pp 218-221.
  5. S. Zhgoon, A. Shvetsov, O. Shteynberg, D. Morgan, P. Ivanov “Single port SAW resonators design for arbitrary reflection phase”, Proc. IEEE Ultrasonic Symp. 2006, pp 1883-1886.
  6. P.G. Ivanov, V.M. Makarov, A.L. Danilov and J.D. Dai, “RSPUDT Filters Based on Different Width Split Fingers”, IEEE Ultrasonic Symp. Proc., pp.2081-2084, 2003.
ООО «БУТИС»
Предприятие «полного цикла» и на 100% локализовано в России в части разработки и производства фильтров на ПАВ.
Разработка и изготовление осуществляется на собственном производстве в Москве в кластере «Микроэлектроника» ОЭЗ «Технополис «Москва».
Информация
2024 © ООО «БУТИС»
crossmenu