Design and Characterization of CPW Feedthroughs in Multilayer Thin Film MCM-D

Transcription

1 Design and Characterization of CPW Feedthroughs in Multilayer Thin Film MCM-D Design and Characterization of CPW Feedthroughs in Multilayer Thin Film MCM-D G. Carchon*, W. De Raedt +, B. Nauwelaers*, and E. Beyne + *K.U.Leuven, dep. ESAT-TELEMIC Kard. Mercierlaan 94, B-3001 Heverlee, Belgium Phone: Fax: carchon@imec.be + IMEC, div. MCP/HDIP Kapeldreef 75, B-3001 Heverlee, Belgium Abstract In this publication, the researchers report on the design and characterization of CPW feedthroughs for RF and microwave applications in multilayer thin film MCM-D. The feedthrough is based on an inverted multilayer microstrip line and two CPW-to-microstrip transitions. The bottom of the vertical metal wall is used as the ground-plane of the intrinsic feedthrough. Using 3-D simulations, it is shown that, for design and analysis purposes, the vertical metal wall can be replaced by a thin metal layer with only a very small impact on the performance. This equivalent structure can be more easily fabricated and measured. This allows for a faster design and characterization of the feedthrough. The transmission line properties (characteristic impedance and propagation constant) of the intrinsic feedthrough are extracted based on the measurement of two equivalent structures with different length. Two types of feedthroughs have been designed and realized. One design uses all-pass 50! lines and can be used up to at least 50 GHz. The other design is realized on a different metal-layer and is based on a low-pass structure. It has a superior performance (insertion loss) up to 25 GHz. Measurements indicate that a low-loss (<0.5 db), well matched feedthrough (return loss below -25 db), can be realized up to at least 50 GHz (intrinsic feedthrough length of 500 µm). Key words: MCM-D, Integrated Passives, Feedthrough, Shielding, and Coplanar Waveguide. 1. Introduction The multilayer thin film multichip module technology (MCM- D) offers a very high reproducibility of very small dimensions and is therefore promising for the low-cost integration of RF and microwave circuits. IMECs MCM-D technology (Figure 1) consists of alternating thin layers of photosensitive benzo-cyclobutene (BCB) dielectric (Cyclotene TM from Dow) and low loss copper metallizations deposited on a borosilicate glass carrier substrate (" r = 6.2, tan #! ). BCB has very low dielectric losses (tan #! ), a low dielectric constant (" r = 2.65), and a low moisture absorption. Various types of high frequency integrated passives (spiral inductors, TaN-resistors, Ta 2 O 5 -capacitors,... ) can be integrated 1,2. Integrating these passives directly on the low cost MCM-substrate gives a size and cost reduction and increases the packaging density. A coplanar waveguide topology is generally used in this technology to realize the transmission lines and integrated passives. Compared to a microstrip-based approach, this has the advantage that the ground plane is directly accessible at the front side of the wafer which makes it highly compatible with Flip Chip International Microelectronics And Packaging Society 353

2 Intl. Journal of Microcircuits and Electronic Packaging mounting. Coplanar waveguides also have a very low dispersion and the presence of the groundplane results in a decreased coupling between neighboring lines. Thin dielectric layers can be used which reduces the size and the series inductance of the vias. The main CPW-lines are located on the 3 µm Cu-layer as this offers the lowest loss. The standard 50!-line has the dimensions: width=77 µm, and slot=20 µm. Such an approach requires that a complete passive designlibrary is available to allow for an easy first-time right co-design between the active and the passive elements. This has recently been presented in References 2,3. Additional passive building blocks, such as quadrature couplers, have been presented in Reference 4. Some modules have been demonstrated in References 5,6. When different circuits are integrated in close proximity on the MCM substrate, it may be necessary to provide some shielding, for example to prevent radiative coupling between the receiver and the transmitter. It may also be required to protect the circuit from the external environment. This is accomplished by placing a metal box on top of the circuit. To connect the inside to the outside world, a specially designed, impedance controlled, low-loss feedthrough is required. Results on ceramic feedthroughs have already been presented in References 7-9. This paper describes how this component can be accurately designed and modeled in multilayer thin film MCM-D. 2. Equivalent Structure Ni/Au component layer 2 µ m Ti/Cu top metal 3 µ m Ti/Cu/Ti metal 1 µ m top Al contact metal 1 µ m bottom Al contact metal TaN resistor 5 µ m BCB dielectric 5 µ m BCB dielectric Ta O capacitor layer 2 5 Ta capacitor material 700 µ m Glass substrate Figure 1. Layer build-up of IMECs MCM-D technology. The technology is highly suited for the Sytem-on-a-Package (SoP) approach, schematically represented in Figure 2. The SoPapproach assumes that it is not possible to integrate complete front-ends on a single chip as there will always be some subblocks that are difficult to realize (such as filters, high-q spirals, among other components). This approach has the advantage that the most optimal technology (performance/cost) can be used to create each subblock. The passives are as much as possible integrated in the low-cost MCM-substrate, hereby, reducing the size and cost of the active chips. A detailed view of the feedthrough structure is given in Figure 3, it consists of a CPW input and output line that passes underneath a grounded vertical metal wall. The length of the feedthrough varies according to the thickness of the walls of the metal box (typical value 500 µm). For design and characterization purposes, it is rather cumbersome to work with a full metal wall. Thus, an equivalent structure (Figure 4) will be used, where the grounded metal wall is replaced by a thin (about 5 µm) grounded metal layer. Metal Hood Flipped CHIP Bonded CHIP Integrated passive e.g. a spiral inductor BCB dielectric Substrate CPW-line Feedthrough CPW-line Feedthrough Figure 2. System on a package (SoP) concept. Figure 3. Layout of the feedthrough with a vertical metal wall. Figure 4. Layout of the equivalent feedthrough structure: the metal wall is replaced by a thin metal strip. 354 International Microelectronics And Packaging Society

3 Design and Characterization of CPW Feedthroughs in Multilayer Thin Film MCM-D This thin metal layer is directly available in the multilayer MCM-D process (top metal layer in Figure 1). This structure can therefore be more easily fabricated, measured, and analyzed. This allows for a faster design and characterization of the feedthrough structure. The equivalent structure can only be used as long as its behavior corresponds to that of the actual feedthrough (metal wall case). This assumption has been verified using a commercial 3- D simulator (HP-HFSS) which accurately accounts for all discontinuities present in the feedthrough and is capable of simulating the effect of a metal wall. The simulated S-parameters for the metal-wall and the metal-strip case are depicted in Figure 5. In these simulations, 250 µm long 50! feeding lines on the middle metal layer have been assumed (Figure 1). The intrinsic 500 µm long feedthrough is realized on the bottom metal layer using a 15 µm wide strip. Replacing the metal-wall by a thin metal strip only slightly influences the simulated S-parameters. Therefore, one can conclude that the extra capacitance-to-ground, induced by the presence of the metal wall, is sufficiently small. The equivalent structure can therefore be used to characterize the feedthrough structure. 3. Feedthrough Design For design and modeling purposes, the feedthrough can be divided into the following subsections (Figure 6): a CPW input and output line (subsection 1) on the middle (Cu) metal-layer, a via transition (if required) from the input/output line to the feedthrough-layer (subsection 2), a step-in-width or a taper section beween the CPW input/ output line and the feedthrough (subsection 3), the discontinuity at the edge of the metal wall (between subsection 3 and subsection 4), the intrinsic feedthrough (subsection 4) where the signal passes underneath the grounded metal-wall. Metal 1 Metal Metal 3 Return Loss (db) Figure 6. Layout of the feedthrough (only one half is drawn) subdivided into different sections. Models for the CPW input/output line, the via transition and the taper are available in the design library 2. The discontinuity at the edge of the metal wall can be assumed to be very small (this discontinuity also occurs at every bridge in a CPW based technology and has a negligible effect on their performance). One can therefore further focus on the design and modeling of the intrinsic feedthrough. Insertion Loss (db) Design of the Intrinsic Feedthrough At first sight, there are several parameters that can be separately optimized for the intrinsic feedthrough: the feedthrough can be realized on the bottom or the middle metal layer. This determines the distance between the feedthroughstrip and the grounded metal-strip (respectively 10 µm and 5 µm), the width of the feedthrough-strip, and the slot of the CPW feedthrough: distance between the strip and the CPW ground plane. However, for a high yield design, the mimimum slot-dimension Figure 5. Simulated (using HP HFSS) return loss and is taken to be at least 20 µm. This implies that the CPW insertion loss for a feedthrough on the bottom metal layer ground plane is located much farther away from the feedthroughstrip (width 15 µm, length 500 µm), together with 250 µm 50! than the overlaying grounded metal-strip (only 5 to 10 µm). CPW feeding lines on the middle metal layer: (-o-) met$l The influence of the CPW ground-plane can therefore be neglected strip case, (-V-) metal wall case. and the intrinsic feedthrough will behave as a multilayer microstrip line. The performance of the feedthrough can be very well predicted by only considering the transmission line sections (feeding line, via section, and intrinsic microstrip feedthrough) while International Microelectronics And Packaging Society 355

4 Intl. Journal of Microcircuits and Electronic Packaging the discontinuities (step-in-width and the discontinuity at the CPW-to-microstrip transition) are neglected. This is depicted in Figure 7 where full 3-D simulations (metal wall case) are compared with the simulation where only the transmission line sections are taken into account. A good correspondence between the simulated return losses and the insertion losses are obtained. gives another reason to take a sufficiently large separation between the feedthrough-strip and the CPW ground plane: a small slot would result in an even lower Z c of the intrinsic feedthrough. Table 1. Characteristic impedance of the intrinsic feedthrough as a function of the selected metal-layer and the stripwidth. 0 STRIPWID STRIPWIDTH TH BOTTOM METAL LAYER MIDDLE METAL LAYER Return Loss (db) Insertion Loss (db) Figure 7. 3-D metal-wall simulation (-V-) versus transmission line model (-o-) for the feedthrough outlined in section 2. An excellent correspondence can be observed. The parameters of the multilayer microstrip line can be determined using an in-house developped quasi-tem program able to calculate the transmission line parameters of a multi-conductor transmission line in a multilayer substrate. The thickness of the metal layers is also accurately taken into account 10. The charactersitic impedances, of the intrinsic feedthrough as a function of the selected metal-layer and the stripwidth, are summarized in Table 1. From this data, one can see that a 50! line can only be realized on the bottom metal layer. On the middle metal layer, it would require a very small stripwidth. This would become difficult to process with sufficient accuracy and the associated conductor losses would also become excessively high. This 5 µm µm Z C =74 Ω Z C =60 Ω 10 µm µm Z C =60 Ω Z C =45 Ω 15 µm µm Z C =51 Ω Z C =36 Ω 20 µm µm Z C =44 Ω Z C =30 Ω Based on the above parameters, it was decided to design a feedthrough on the bottom and the middle metal layer. This also illustrates the two design-approaches that can be followed. The feedthrough on the bottom metal layer is based on an all-pass structure as it uses 50! lines for the feeding lines and the intrinsic feedthrough section. It should therefore have a good return loss up to very high frequencies. This design has, however, two major drawbacks: a narrow strip (15 µm) is used and the feedthrough is realized on the bottom, 2 µm thick Alumina layer. This will lead to higher losses for the low-frequency applications. The second feedthrough is realized on the middle metal layer. It uses a 20 µm wide line. To increase the operating frequency, two 290 µm long standard high-impedance CPW lines (width: 16 µm; slot: 50.5 µm; Z c =106!) are added before and after the intrinsic feedthrough section. These high impedance lines compensate for the low impedance of the intrinsic feedthrough. The total structure behaves as a low-pass filter. This design has an improved insertion loss due to the wider strip, the 3 µm Cu-metal and the absence of via sections. The measured results for both designs will be provided in section Non-Idealities Predicting the performance of an actual feedthrough is complicated by BCB-planarization effects: the surface profile of the BCB is not perfecly flat as assumed by electromagnetic simulators. This can be seen in Figure 8, where the surface profile of a standard 16 µm wide CPW line on the middle metal level is depicted (the ground-to-ground spacing is 117 µm). This complicates the accurate prediction of the exact distance between the intrinsic feedthrough (middle or bottom metal layer) and the groundplane (located on the top metal-layer). This distance, however, has a direct impact on the characteristic impedance of the intrinsic feedthrough, and hence, the obtained return loss. 356 International Microelectronics And Packaging Society

5 Design and Characterization of CPW Feedthroughs in Multilayer Thin Film MCM-D Figure 8. Non-ideal planarization of the BCB: measured surface profile of a 16 µm wide transmission line on the middle metal level. the ground-to-ground distance is 117 µm. Units are in micron. The effect of the BCB-planarization will be less pronounced when the feedthrough is realized on the bottom metal-layer as a thicker BCB-layer is being used. To evaluate the performance of the quasi-tem program, used to predict the transmission line parameters of the intrincic feedthrough, the researchers have fabricated and measured several feedthrough-pairs realized on the bottom and middle metal layer. One feedthrough has an intrinsic length of 240 µm, the other has an intrinsic length of 2070 µm. Based on these measurements, it is possible to extract the transmission line properties of the intrinsic feedthrough using Reference 11. This method directly extracts the characteristic impedance and propagation constant while the influence of the discontinuities at the edges of the feedthrough (e.g. an extra capacitance-to-ground where the line goes underneath the overlaying ground-plane) are removed. The results for a feedthrough realized on the middle metal layer (width of 20 µm) are given in Figure 10. The extracted characteristic impedance is about 25!, the effective dielectric constant is about 3 and the intrinsic feedthrough shows an intrinsic loss of about -0.6 db/ 50 GHz. The difference between the measured (Z c =25!)and the simulated results (Z c =29!) can be attributed to BCBplanarization effects Measurement and Characterization Z c (Ω) Characterization of the Intrinsic Feedthrough An illustration of a feedthrough realized on the bottom metal layer (stripwidth: 15 µm, length: 240 µm), is given in Figure 9. One clearly observes the transition from the main (middle) CPW layer to the bottom metal layer and the tapered line on the bottom layer. ε eff α (db/mm) Figure 9. Picture of a feedthrough (intrinsic length of 230 µm) realized on the bottom metal layer. One clearly observes the transition from the middle metal layer to the botton metal layer and a taper on the bottom layer Figure 10. Extracted characteristic impedance, " eff and losses (db/mm) for a feedthrough on the middle metal layer. w=20 µm, ground-to-ground spacing=117 µm. International Microelectronics And Packaging Society 357

6 Intl. Journal of Microcircuits and Electronic Packaging 4.2. Measurements of the Feedthroughs The measured results for the two 500 µm long feedthroughs are given in Figure 11. The feedthrough realized on the bottom metal layer can be used up to at least 45 GHz, having a return loss superior to -30 db and an associated insertion loss lower than 0.5 db over the entire frequency band. Return Loss (db) Insertion Loss (db) Conclusion and Future Directions In this paper, the authors have reported on the design, characterization and modeling of CPW feedthroughs implemented in a multilayer thin film MCM-D technology. Using 3-D simulations, the researchers have shown that the vertical wall may be replaced by a thin metal layer with a very small impact on the performance of the feedthrough. This equivalent structure can be more easily realized and therefore allows for a faster design and characterization of the feedthrough. It is shown that the discontinuities and the CPW-to-microstrip transition only have a small impact on the S-parameters such that a good approximation of the feedthrough s behavior can be obtained by a cascade of three transmission line sections corresponding to the feeding lines, and the intrinsic microstrip-based feedthrough. The transmission line parameters of the intrinsic feedthrough can be accurately extracted based on the measurement of two equivalent structures with different length. Two feedthroughs have been designed using different metal layers and different design-approaches. The all-pass structure can be used up to very high frequencies (at least 50 GHz), whereas the low-pass design is limited up to 25 GHz. It has, however, the advantage of a superior insertion loss for the lower frequencies. Measured results have been presented for both feedthroughs. They demonstrate that a low loss, well matched feedthrough, can be realized up to at least 50 GHz (insertion loss below -0.5 db and return loss below -25 db for a typical length of 500 µm). The feedthroughs will now be used to build complete microwave circuit-modules using bare-die active components and integrated passives on the MCM-substrate. This will be achieved using the custom design-library. Future work will also be directed towards the development of a chip-to-next level interconnection. Acknowledgments Figure 11. Measured return loss and insertion loss for the two 500 µm long feedthroughs: (-%-) feedthrough reallized on the middle metal layer, (-o-) bottom metal layer. The feedthrough realized on the middle metal layer has a superior insertion loss up to 25 GHz and is therefore more suited for the low-frequency applications. Naturally, the performance of the feedthrough can be further improved using a thinner metal wall, which results in a shorter intrinsic feedthrough length. Geert Carchon was supported by a scholarship granted by the Flemish Institute for the Advancement of Scientific-Technological Research in Industry (IWT). The author also acknowledges the support of the European Space Agency under contract number 13627/99/NL/FM(SC). References 1. G. Carchon, S. Brebels, W. De Raedt, and B. Nauwelaers, Accurate Measurement and Characterization up to 50 GHz of CPW-based Integrated Passives in Microwave MCM-D, 358 International Microelectronics And Packaging Society

7 Design and Characterization of CPW Feedthroughs in Multilayer Thin Film MCM-D Presented and Published at the Electronic Components and Technology Conference, ECTC 2000, Las Vegas, Nevada, pp , G. Carchon, P. Pieters, K. Vaesen, S. Brebels, D. Schreurs, S. Vandenberghe, W. De Raedt, B. Nauwelaers, and E. Beyne, Design-oriented Measurement-based Scaleable Models for Multilayer MCM-D Integrated Passives. Implementation in a Design Library offering Automated Layout, Presented and Published at International Conference and Exhibition on High Density Interconnect and Systems Packaging, Denver, Colorado, pp , G. Carchon, S. Brebels, K. Vaesen, P. Pieters, D. Schreurs, S. Vandenberghe, W. De Raedt, B. Nauwelaers, and E. Beyne, Accurate Measurement and Characterization of MCM-D Integrated Passives up to 50 GHz, Presented and Published at International Conference and Exhibition on High Density Interconnect and Systems Packaging, Denver, Colorado, pp , G. Carchon, S. Brebels, P. Pieters, K. Vaesen, D. Schreurs, S. Vandenberghe, W. De Raedt, B. Nauwelaers, and E. Beyne, Design of Microwave MCM-D CPW Quadrature Couplers and Power Dividers in X-, Ku- and Ka-band, Presented and Published at International Conference and Exhibition on High Density Interconnect and Systems Packaging, Denver, Colorado, pp , K. Vaesen, P. Pieters, G. Carchon, W. De Raedt, E. Beyne, A. Naem, and R. Kohlmann, Integrated Passives for a DECT VCO, Presented and Published at International Conference and Exhibition on High Density Interconnect and Systems Packaging, Denver, pp , K. Vaesen, S. Donnay, P. Pieters, G. Carchon, W. Diels, P. Wambacq, W. De Raedt, E. Beyne, M. Engels, and I. Bolsens, Chip-Package Co-Design of a 4.7 GHz VCO, Presented and Published at International Conference and Exhibition on High Density Interconnect and Systems Packaging, Denver, pp , M. P. Goetz and G. Mancias, Measurement of a 24-GHz Broadband Multilayer Ceramic Feedthrough for Microwave Packaging, IEEE Microwave and Guided Wave Letters, Vol. 2, pp , E. Holzman, R. Teti, B. Dufour, and S. Miller, An Hermetic Coplanar Waveguide-to-HDI Microstrip Microwave Feedthrough, Presented at IEEE MTT-S Digest, pp , P. Monfraix, T. Adam, C. Devron, G. Naudy, B. Cogo, and J. L. Roux, A New Concept of RF Feedthrough Applied to Multichip Modules for Space Equipment, Presented at IEEE MTT-S Digest, Boston, Massachusetts, P. Pieters, S. Brebels, E. Beyne, and R. P. Mertens, Generalized Analysis of Coupled Lines in Multilayer Microwave MCM-D Technology - Application: Integrated Lange Couplers, IEEE Transactions on Microwave Theory and Techniques, Vol. 47, pp , G. Carchon, B. Nauwelaers, W. De Raedt, D. Schreurs, and S. Vandenberghe, Characterizing Differences Between Measurement and Calibration Wafer in Probe-Tip Calibrations, Electronics Letters, Vol. 35, pp , About the authors Geert Carchon received the M.Sc. Degree in Electrical Engineering from the Katholieke Universiteit Leuven, Belgium in As a Research Assistant of the IWT, he is currently working towards a Ph.D. Degree at the K.U.Leuven in close cooperation with the High-Density Interconnect and Packaging group of IMEC, Leuven, Belgium. His main interests include the measurement, characterization and modeling of passive devices and the design of RF and microwave circuits (LNAs, modulators, ) in MMIC and multilayer MCM-D. Walter De Raedt received the M.S. Degree in Electrical Engineering at the Katholieke Universiteit Leuven, Leuven, Belgium, in Subsequently, he joined the ESAT laboratory as a Research Assistant and worked on Direct Write Electron Beam Technology. From 1984, he is with IMEC where he started research on MMICs and submicron technologies for advanced HEMT devices. Since 1997, he joined the High Density Interconnect and Packaging group where he is working on integrated passives and interconnections for RF front-end systems. Bart Nauwelaers received the M.S. and Ph. D. Degrees in Electrical Engineering from the Katholieke Universiteit Leuven, Leuven, Belgium, in 1981, and 1988, respectively. He also holds a Master Degree from ENST, Paris, France. Since 1981, he has been with the Department of Electrical Engineering (ESAT) of the K.U.Leuven, where he has been involved in research on microwave antennas, microwave integrated circuits and MMICs, and wireless communications. He teaches courses on Microwave Engineering, Analog and Digital Communications, Wireless Communications, and Design in Electronics and Telecommunications. International Microelectronics And Packaging Society 359

8 Intl. Journal of Microcircuits and Electronic Packaging Eric Beyne received the M.S. and Ph.D. Degrees in Electrical Engineering from the Katholieke Universiteit Leuven, Leuven, Belgium, in 1983, and 1990, respectively. From 1983 to 1985, he was a Research Assistant at the K. U. Leuven. In 1986, he joined IMEC, where he worked towards his Ph.D. Degree on the interconnection of highfrequency digital circuits. He is presently responsible for projects on multichip modules and advanced packaging at IMEC. Dr. Beyne is a member of the IMAPS-Benelux Committee. 360 International Microelectronics And Packaging Society