respect the phase shifters based on LTCC technology [3] have advantages over thin film ferroelectric phase shifters, although the performance of the latter may ...
LTCC Compatible Ferroelectric Phase Shifters A.Deleniv1, S.Gevorgian1,2, H. Jantunnen3, T.Hu3 1 Chalmers University of Technology, SE 41296 Göteborg, Sweden 2 Microwave and High Speed Electronics Research Centre, Ericsson AB, Mölndal, Sweden 3 Microelectronics and Materials Physics Laboratory and EMPART Research Group of InfoTech, Oulu, Finland voltages lower than coplanar-plate designs, where the gap between the electrodes is typically more than 100 µm. The second device is a microstrip line periodically loaded by ferroelectric varactors. The latter’s are realized in the ferroelectric film deposited on the ground plane. Due to proximity of the ferroelectric film to the ground plane the former does not contribute to the losses of the main microstrip line and, hence, allows design of components with improved figure of merit.
Abstract — Two phase shifters with novel topologies are presented. The phase shifters are based on LTCC compatible ferroelectric films with εr=200, and tanδ=0.04 at 10 GHz. A Kuband phase shifter is designed to provide DC bias independent matching in the wide frequency range. It has microstrip design where ferroelectric film is used as a substrate and requires relatively low DC bias voltages. The phase shifter showed -15dB matching in ∼50% bandwidth and 15o/dB figure of merit. In Xband phase shifter the ferroelectric film is deposited on the low permittivity substrate. Using such two layered substrate a loaded line phase shifter based on coplanar-plate varactors is realized. The varactors are included in the circuit ground and are connected with the microstrip using vias. This designed allows reduction of overall loss due to more efficient use of ferroelectric. The figure of merit of this phase shifter is 20o/dB at 10 GHz. Index Terms — thick film, ferroelectric, phase shifter.
II. DESIGN AND EXPERIMENTAL PERFORMANCE OF THE PHASE SHIFTERS A. A phase Shifter with bias independent matching A common drawback of most tunable components is a degradation of input/output matching under applied voltage. The proposed design of the traveling-wave phase shifter, Fig.1, preserves a good matching under wide range of DC bias. The main part of the device is realized as a distributed
I. INTRODUCTION To a large extend the performance of phased arrays depend on the parameters of phase shifters [1]. Among different technologies the ferroelectric based phase shifters offer significant reduction of the driving (control) power consumption and improvement in the tuning speed [2]. The cost is also a critical issue especially where a phased array with a large number of components is concerned. In this respect the phase shifters based on LTCC technology [3] have advantages over thin film ferroelectric phase shifters, although the performance of the latter may be better. The main problem in utilizing thick ferroelectric LTCC films is associated with relatively high losses, as compared with the thin films. Additionally, the LTCC devices are fabricated by screen printing technique, where the size features are typically larger than 100 µm. With this size limitation the devices with coplanar electrodes require high voltages (up to 1.0 kV and more) to achieve reasonable tunabilities [4]. Two novel designs of the phase shifters based on LTCC ferroelectrics technology are presented in this work addressing some of the discussed above problems. The same ferroelectric material is used in both devices. The first device is designed to ensure a good matching in the entire range of the applied DC voltages. Due to the parallel plate electrodes this design allows utilization of the full tunability of the permittivity provided by the ferroelectric film, since the DC field is essentially homogeneous and coincides with the direction with the microwave field. Since the thickness of the LTCC film is typically 50 µm or less it requires tuning
0-7803-8846-1/05/$20.00 (C) 2005 IEEE
M2
LTCC
M1
a) Biased ground plane λ/4
Periodic structure with 25Ohm image impedance b)
Z odd ,θ odd Z even ,θ even Z 1odd ,θ 1odd
Z1odd,θ1odd
c) Fig.1 Cross-section (a), layout (b), and equivalent circuit of the unit cell (c) of the device.
599
periodic microstrip structure on top of the ferroelectric film (metal layer M2, Fig.1a). The microwave ground plane, realized in the bottom metal layer M1, Fig.1b, is used for DC biasing, which is supplied through the contact pad and a via. The impedance of the main part of the device is rather low, since the dielectric permittivity of the 50µm thick LTCC ceramic layer is high, ε F ≈ 200 . To increase the impedance the microstrip lines are meander shaped providing an increased the effective per unit length inductance. The input and output of the device are connected to differential twin strip lines. Considering this periodical structure as the synthetic line, its effective parameters may be defined as: Leq C eq
(1)
Z eq = Leq C eq
(2)
Insertion Losses S21, dB
v f =1
0
S21(dB)0V S21(dB)80V S21(dB)150V S21(dB)220V
-5
-10 -15 -20 -25 -30
The equivalent per unit length parameters Leq and Ceq are expressed in terms of the voltage dependent impedances and electrical lengths shown in Fig.1c. The equivalent inductance Leq is designed in the way that under applied DC field it
a) 5
10 15 20 Frequency, GHz
0
Matching S11, dB
changes result in larger tunability in phase velocity under DC field, as it follows from (1). Additionally, if the rate of tuning of the Leq and Ceq under the same DC field is similar, the
30
S11(dB)0V S11(dB)80V S11(dB)150V S11(dB)220V
-5
decreases similarly to the equivalent capacitance Ceq . These
25
-10
impedance of the component is constant, as it follows from (2). The device is designed to operate at central frequency 25GHz. The thickness of LTCC ferroelectric layer is ∼55µm. The sizes are about ∼2x0.5mm which, to the best of authors knowledge this is one of the most compact LTCC phase shifter demonstrated so far. The circuit is made using gravure printing technology providing close to 20µm feature resolution. The dimensions of the unit cell are chosen so to provide week sensitivity of matching to DC biasing. The equivalent impedance (“odd” mode) of the periodic structure is 25Ohm. Two quarter-wavelength transformers are used for matching with 50 Ohm feed lines. The experimental characterization of the device is made using a probe station and a pair of SG and GS probes (“Picoprobes”). The DC bias is supplied to the ground electrode using the pad and is changed in 0-220V range what result in ∼4V/µm field (this field is close to the breaking one for the utilized ceramic). The measured tunability of LTCC at this field is nearly ∼25%. The measured transmission losses, matching and differential phase shift are shown in Fig.2 a-c. In comparison with expected from simulation performance the measured data are slightly shifted down on frequency. This is explained by higher, in comparison with the simulations, dielectric constant of the dielectric layer. Fig.2b indicates better than -15dB matching in the range ∼15-25GHz (in about 50% bandwidth). One may note that the matching range does not shrink while increasing the DC bias, which was one of the main objectives. The insertion losses are rather high. It was experimentally verified that this is caused by the losses of the
-15 -20 -25
Differential phase shift φ, deg
-30
b) 5
10 15 20 Frequency, GHz
250
25
30
25
30
ph_shift_80V ph_shift_150V ph_shift_220V
200 150 100 50 0
c) 5
10
15
20
Frequency, GHz
Fig.2 Measured performance of the phase shifter in Fig.1.
600
interfacial layer located between microwave ground and LTCC layer [6]. The loss contribution of the latter was measured indicating effective loss tangent tan δ eff ≈ 0.2 of the
Normalized line Q-factor
1
ferroelectric layer at 25GHz, which is nearly twice of the intrinsic ferroelectric layer losses. Although this value seems far from the practically required, the potential for the improvement is high provided that the interface problem is solved. The differential phase shift is obtained in the range 0240deg. The figure of merit, using the definition (phase shift)/(geometrical mean of the losses) [7], is ∼15deg/dB at 15GHz, which is better than the reported value for the similar devices [3], [5]. B. A phase shifter with reduced losses While the technology of LTCC ferroelectrics with the bottom electrode is now under optimization, a second design of the phase shifter is proposed using the same material. The loss tangent of the material measured at 10GHz is tan δ ≈ 0.04 . The new design of the phase shifter intended to reduce the overall losses by using the lumped ferroelectric varactors. Fig.3 explains how the reduction of the losses is achieved. The microstrip line in Fig.3 is comprised of two layers – 0.5mm thick alumina ( ε A = 9.2 ) and 20µm thick LTCC film with ε F = 220 . The strip is 0.5mm wide. The losses of such the structure are analyzed using spectral domain technique for different loss factor of LTCC layer and are given in the normalized form in Fig.4. It is assumed that the strip is made of an ideal conductor, while a conductivity of σ g = 3.5 × 10 7 (S/m) is assigned to the ground electrode. As it
0.95 0.9 w
0.85
M2
0.5mm thick Alumina
0.8 0.75 0
M1 20µm thick LTCC ferroelectric
0.02 0.04 0.06 0.08 Loss tangent, tand
0.1
Fig.3 Normalized quality factor of the line versus losses in LTCC layer.
λ/4
Varactor
follows from Fig.4 the quality factor of the microstrip line shows week dependence on the losses in LTCC layer. For our case ( tan δ ≈ 0.04 ) only 5% reduction in quality factor is observed as compared to ideal loss free ferroelectric. This is explained by the proximity of the ferroelectric to the ground electrode, which allows only a small part of the microwave energy to flow in it. Hence, the ferroelectric has a week affect on the total losses of the microstrip line. This effect is utilized in the design of the phase shifter shown in Fig.4a-c. The top (microstrip, M2) and bottom (ground plane, M1) layouts of the device are given Fig.4. The close-up view of the varactor is shown in Fig.4c. The component is made using∼200µm resolution screen printing technology. The 200µm wide microstrip line (∼60Ohm) is loaded by a number of varactors. The coplanar-plate varactors, Fig.4c, are integrated in the ground plane. Vias connect the top microstrip with the square plate of the varactor formed in the opening in the ground plane. Photographs of the microstrip in M2 with the via, and the varactor in the ground plane M1 are shown in Fig.5c. For the propagating microstrip mode the varactor is seen as a lumped element. The image impedance of the loaded line section is ∼25Ohm. Two quarter-wavelength transformers are added at the input and output to provide matching with the 50Ohm line. The measurements are made using two external
c) Via hole
a)
b)
Fig.4 Top (a), bottom (b) views of the device and the coplanar-plate varactor in the bottom ground plane (c).
Contact pad
Via
a)
b)
Fig.5 Photo of the top microstrip with a via (a), and coplanarplate varactor embedded in the ground plane (b).
601
-5
bias tees. The DC voltage was changed in the range 0-750V. The obtained experimental data are presented in Fig.6. The matching is ∼10dB over entire frequency range and is worse than expected from simulations. The obtained figure of merit is about 20deg/dB at 10GHz. Expected from simulation figure of merit is 30deg/dB. Both above discrepancies are explained by the larger slot between the rectangular slot and ground plane, Fig.5b, which leads to the lower capacitance of the varactor. The latter results in increased impedance of the periodically loaded section, which is responsible for the mismatching. Additionally, the larger slot decreases the tunability and requires higher DC biasing voltages. As compared to the first component we obtained higher figure of merit under lower driving DC fields. In this case the opening is about 200µm which result in ∼3.75V/µm. III.
Matching S11, dB
-10 -15 -20 -25 S11(0V) S11(250V) S11(500V) S11(600V) S11(700V)
-30 -35 -40
CONCLUSION
a) 2
4
6 8 10 12 Frequency, GHz
0 Insertion Losses S21, dB
The design of two phase shifters is proposed that realizes periodically loaded structure. The experimentally obtained data demonstrate the potential of LTCC and screen printing techniques as applied to design of cost effective agile components. With available potential for improvement the components with more competitive parameters may be expected. ACKNOWLEDGEMENT
14
16
14
16
14
16
-2 -4 -6
The authors wish to acknowledge the European Commission for founding project “MELODY” under Framework 5 and support of Swedish Science Foundation (VR).
S21(0V) S21(250V) S21(500V) S21(600V) S21(700V)
-8
-10
REFERENCES
b) 2
140
Differential phase shift φ, deg
[1]. S. K. Koul, B. Bhat, Microwave and Millimeter Wave Phase Sifters, Artech house, 1991 [2]. http://www.agilematerials.com/ [3]. P. Scheele, S. Muller, C. Weil, and R. Jakoby, „Phase shifting coplanar stub line filter on ferroelectric thick films”, Proc. EuMC 2004, pp. 1501-1504. [4]. A. Deleniv, S.Gevorgian,”Tuneable power splitters and matching networks based on LTCC ferroelectrics”, Proc. of Workshop “Tuneable ferroelectric materials and devices for microwave applications, EuMC 2003, pp.45-52. [5]. K. S. K. Yeo, W-F. Hu, M. J. Lancaster, B. Su and T. W. Button, “Thick film ferroelectric phase shifters using screen printing technology”, Proc. EuMC 2004, pp. 1489-1492. [6]. A. Deleniv and S. Gevorgian, “Open resonator technique for measuring multilayered dielectric plates”, unpublished. [7]. V. Sherman, K. Astafev, N. Setter, A. Tagantsev, O.Vendik, I. Vendik, S. Hoffman-Eifert, U. Böttger, and R. Waser, ”Digital reflection type phase shifter based on a ferroelectric planar capacitor”, IEEE Microwave and Wireless Comp. Letters, vol.11, pp. 407-409, 2001.
4
6 8 10 12 Frequency, GHz Diff(250V) Diff(500V) Diff(600V) Diff(700V)
120 100 80 60 40 20 c) 0
2
4
6 8 10 12 Frequency, GHz
Fig.6 Measured performance of the phase shifter.
602
