Eisenhart and Khan [1] waveguide probe configuration. Borrowing from Withington's anal- yses [5], the real part of the probe's input impedance is influenced in a ...
A Full-Height Waveguide to Thinfilm Microstrip Transition J. W. Kooi, G. Chattopadhyay, F. Rice, J. Zmuidzinas California Institute of Technology, 320-47, Pasadena, CA 91125, USA.
S. Withington, G. Yassin University of Cambridge, Dept. of Physics, UK. Abstract Excellent progress in the development of Submillimeter-wave SIS and HEB mixers has been demonstrated in recent years. At frequencies below 800 GHz these mixers are typically implemented using waveguide techniques, while above 800 GHz quasi-optical (open structure) methods are often used. The choice of waveguide offers several advantages over quasi-optical techniques, such as the use of broadband corrugated feed-horns with well defined on axis (Gaussian) beam patterns. However, it has in practice been a problem to efficiently couple energy to the mixer (DUT) over a large fractional RF bandwidth. Many waveguide probe transitions have been proposed over the years, most of which with RF bandwidth less than 35%. Some of these probe designs require significant reductions in waveguide height (lower impedance). Unfortunately, reducing the height makes machining of mixer components at terahertz frequencies rather difficult. It also increases RF loss as the current density in the waveguide goes up, and surface quality is degraded. We propose a variation on the “single sided” rectangular microstrip probe that is found in common use at the lower microwave and millimeter wave frequency bands. The proposed constant radius thinfilm microstrip transition couples with better than 99% efficiency (VSWR ≤ 1.20) to an entire waveguide band. Extensive HFSS simulations, backed by scale model measurements are presented in this paper. The radial probe is seen to give vastly improved performance over existing designs discussed in literature. It is found that by selecting the appropriate substrate and radius any real impedance between 10-60 Ω can be achieved. Keywords Radial probe, full-height waveguide to thinfilm microstrip transition, capacitive waveguide tuning step, split-block, hot electron bolometer (HEB), superconducting-insulating-superconducting (SIS) tunnel junction.
I. Introduction The majority of SIS mixers to date employ planar circuit probes that extend all the way across the waveguide [1]-[4]. An important reason for the popularity of this kind of probe is the convenience at which the active device can be biased and the IF signal extracted. Unfortunately this kind of “double-sided” probe exhibits a rather poor RF bandwidth (≤ 15%) when constructed in full-height waveguide. When the waveguide height is reduced by 50%, the probe’s fractional bandwidth improves dramatically to a maximum of about 33%. However, reducing the height can result in significant fabrication problems (cost) and increased RF loss, especially at frequencies near or above a teraherz. These results can be understood in that the popular “across the waveguide” probe is essentially a variation on the well known Eisenhart and Khan [1] waveguide probe configuration. Borrowing from Withington’s analyses [5], the real part of the probe’s input impedance is influenced in a complex way by
the parallel sum of individual non-propagating modal impedances, and as such is frequency dependent. The way to minimize this problem is to reduce the waveguide height, which has been done very successfully by for example Tong and Blundell et al. [2]. An alternative approach would be to use a probe that does not extend all the way across the waveguide. For this kind of probe, referred to from now on as a “one-sided” probe, the modal impedances add in series. The real part of the input impedance only comes from the single propagating mode, and is relatively frequency independent. These probes are typically implemented in full-height waveguide which minimizes conduction loss and eases fabrication complexity. Though a rectangular version of the “one-sided” probe is used quite extensively by microwave engineers [8][9] and was introduced to the submillimeter community by Kerr et al. [10] in 1990, it is seen to be fundamentally differerent from the proposed radial shaped probe. The radial probe described here represents an attempt to extend the use of thin-film radial modes, which are known to give superior broad band performance in microstrip radial tuning stubs as compared to retangular stubs, to the microstrip to waveguide coupling problem. Indeed analytical and numerical work has shown that whereas the transverse component of surface current does not have much influence on the behaviour of rectangular probes, it is central to the operation of radial probes. In addition, the electric field lines of the commonly used rectangular probe extend all the way thru the substrate (thick-film probe). Whereas in the case of the proposed radial probe design, the fields lines are eventually confined to a very thin dielectric layer sandwiched between the RF ground and a metal wire layer on top, hence the nomenclature: “thin film microstrip probe”. Experimentally we have found a constant radius probe implemented in the described thin-film configuration to give vastly superior performance over the more tradional approach. In this paper an effort has been made to understand all aspects of the “one-sided” thinfilm radial probe waveguide transition. It is extended from a novel concept inside the waveguide
Fig. 1. Withington’s radial probe setup. The orientation is parallel to the E-field of the dominant TE10 waveguide mode. Though convenient for split block design, it is not critical to the probe’s performance.
to a much more complex geometry in a final application. In the simulations we have assumed metals to have perfect conductivity, i.e. no loss. An analytical expression at this point is not possible. Toward the end of the paper we present scalemodel measurements that confirm the results obtained by HFSS [11]. II. Radial Probe Basic Setup In 1999 Withington et al. [7] presented extensive theoretical analyses of a “one-sided” rectangular probe in full-height waveguide [9][10]. In his paper he looked into the use of alternative shapes of metalization and presented very promising scale-model measurements on a probe with a constant radius. In these measurements a 90◦ radial fan was connected to a SMA connector, mounted to the side of the waveguide as shown in Fig. 1. The radial probe orientation was parallel to the electric field of the waveguide. Since then we have used extensive HFSS analyses to extend the idea to include a broadband waveguide to thinfilm microstrip transition. In order to take advantage of splitblock fabrication techniques, we too have oriented the probe parallel to the electric field in the guide. The performance of the probe appears not senstive to it’s orientation [5][6]. If a SMA connector were to be connected to a radial probe at the waveguide wall, we would measured a probe impedance as depicted in Fig. 2. Several things are observed: First, the radius of the probe determines both the RF bandwidth and real part of the probe
Fig. 2. Input impedance of the probe as simulated inside the waveguide for silicon (25 µm) and quartz (50 µm) substrates. Respective radii for the quartz and silicon probe are 112 µm and 85 µm. The quartz and silicon backshort distances are 0.089 λg and 0.053 λg . The quartz/silicon probe impedance ratio is ≈ 1.75 given that r quartz = 3.80 and r silicon = 11.7 (see text).
Fig. 3. Simulated input return loss for the quartz and silicon radial probe as depicted in Fig. 1. The respective normalizing impedances for quartz and silicon are 31+j0 Ω and 17-j5 Ω.
impedance. Optimal bandwidth is achieved with a “tear drop” shaped frequency dependent input impedance on the Smith chart. Second, the real part of the probe’s input impedance is related to the phase velocity of the launched substrate wave. For this particular probe design it is of crucial importance that the substrate be extended all the way thru (or beyond) the guide. Third, the probe reactance is infuenced by the energy stored in the substrate. Changing the height, width or dielectric constant of the substrate provides a convenient and predictable way to “null” the reactive part of the probe impedance. This is demonstrated with the choice of a 50 µm thick quartz substrate in Fig. 2. Finally, the lower dielectric substrate (quartz) material provides an inherent larger RF bandwidth as less reactance needs to be tuned out. Note that improved performance (S11 ≤ 25 dB) is easily obtained by a slight adjustment of the probe radius if less than optimal bandwidth is required (Fig. 6). If both bandwidth and performance are to be optimized, then a small capacitive tuning step can be used in the waveguide, which we will discuss in Section VIII. III. Probe Impedance as a function of Angle and Waveguide Height The effect of the probe opening angle and waveguide height were investigated. Keeping the overall RF bandwidth fixed (by making small adjustments to the probe radius) we found that the real part of the probe impedance is linearly proportional to the change in opening angle from the 90◦ reference. For example, if the probe’s opening angle were to be changed 20% from 90 degrees, then the real part of the probe impedance is seen to change by a similar 20%. In addition, it has been noticed that the imaginary part of the radial probe impedance is proportional to the change in waveguide height to the third power. In other words, when the waveguide height is reduced by 13% the imaginary part of the probe impedance is seen to decreases by approximately 45%. IV. Introduction of a channel into the Waveguide Wall In Fig. 4 we show a parallel to the E-plane oriented, radial probe with it’s substrate extending out of the waveguide. In this particular configuration there is a perfect conducting
ground plane that extends all the way up to the waveguide wall, commonly referred to as a beam-lead contact. Not surprisingly, creating such a large opening in the waveguide wall has a significant effect on the electric field distribution in the vicinity of the probe. This affects the probe impedance in a significant way, as shown in Fig. 5. To minimize the effect we found it useful to add an airgap directly underneath the substrate. Not only does it help diminish the change in input impedance, it also increases the cutoff frequency of the dielectric loaded channel. To maximize the cutoff frequency of the TE10 and TE01 modes in the channel, we find it important to keep the substrate thickness to channel height ratio ≤ 0.5, and the substrate thickness to channel width ratio ≤ 0.4 . As far as the TE10 and TE01 mode cutoff frequencies of the channel are concerned, the optimal size appears to be about square. Referring to Fig. 5, adding a channel to the waveguide wall increases the silicon the probe impedance from 18-j6 Ω → 26-j1 Ω. For quartz the impedance is seen to change from 31+j0 Ω → 41+j8 Ω. To maximize the RF bandwidth, we found it useful to optimize the impedance shape to that of a “tear drop”. This usually requires a slight increase in probe radius, and quite possibly an increase in the substrate width. As we have seen, enlarging the radius of the probe increases the real part of the probe impedance while enlarging the substrate width increases the reactive component of the probe. In this particular situation we have increased the probe radius by 2.5% and the substrate width by 12%. The resulting change in impedance is 41+j8 Ω → 43+j12 Ω, quite small unless optimal RF bandwidth and/or performance is required. V. Probe Impedance as a Function of Radius and Throat Dimension Apart from knowing the exact probe impedance is it constructive to understand the effect of probe radius and throat dimension on the probe’s input impedance and RF bandwidth.
Fig. 4. Rendering of the substrate extruding out of the waveguide and into the IF channel. In the simulations, the airgap to substrate height ratio is one. Note that the since the backshort position λg the backshort acts as an inductive (shunt) tuning element.
Fig. 5. Progression of impedance circles as a result of adding a channel to the waveguide wall. Smith chart is normalized to 31+j0 Ω. The respective silicon and quartz substrate heights are 25 µm and 50 µm.
We have defined the probe’s throat as being the width of the microstrip line that drives the radial probe. Simulations indicate that the input impedance and RF bandwidth of the probe is not a sensitive function of the width of the throat (Fig. 6), but is indeed quite sensitive to
Fig. 6. Probe impedance sensitivity to throat size (r=80 µm). In the simulations the throat size was changed from 1.3 µm to 4 µm without significant effect. On the other hand, changing the probe radius by 12% (r=80 µm → r=70 µm) has a major effect on both RF bandwidth (shape) and impedance (26+j1 Ω → 20-j1 Ω). The dielectric material is silicon (r =11.7), the backshort distance is 0.041 λg , and the normalizing impedance 26-j1 Ω (See also Fig. 4).
the actual radius. Fortunately, the exact probe radius can be precisely set by lithographic means. The throat can thus be sized to match into a microstrip or cpw transmission line. Unless optimal bandwidth is of no great concern, the probe radius should be held to a tolerance of a few percent. VI. Addition of a 4-Section RF choke Extending the RF ground all the way to the waveguide wall, as in Fig. 4, is in many instances problematic. This is especially the case when operating frequencies approach the terahertz regime. Beam-lead techniques, as are often used in GaAs Schottky diode multiplier processes, are a viable option [12]. However this scheme requires advanced, and therefore expensive, processing. Furthermore the beam-lead process is not readily extended to quartz substrates, which due to their low (r =3.8) dielectric constant and RF loss is often the substrate of choice in the submillimeterwave regime. It should be noted that membrane techniques[4][6] offer a good alternative to quartz substrates at frequencies near or above a teraherz. In Fig. 8 we show the progression of the radial probe input impedance as a 4-section RF choke, and small airgap above the substrate are incorporated. We observe that: 1) Extending the substrate out of the waveguide, moves the probe impedance from 31+j0 Ω → 41+j8 Ω. In this case the substrate is covered with a conducting ground all the way to the waveguide wall. There is no airgap above the substrate. The height of the “air” gap underneath the substrate to the substrate height ratio is one. 2) Adding a 50 µm “air” channel above the substrate results in another impedance shift, 41+j8 Ω → 51+j16 Ω. This airgap should be minimized as it limits the high frequency response of the probe. 3)The introduction of a RF choke results in a significant, but always predictable trend toward lower input impedance. The high frequency tail is due to evanescent fields directly above the first section of the RF Choke. This effect can be eliminated by minimizing the airgap above the substrate. Some loss of bandwidth is evident from the Smith chart. If this is a concern, the RF
Fig. 7. RF Choke providing a ground potential at the waveguide wall. The IF and bias lines run on top of the choke’s metalization, which serves as a ground layer for thinfilm microstrip or CPW transmission lines as shown on the inset (not to scale). The ribbon(bond) wires provide the bias return.
Fig. 8. Progression of probe input impedance as the substrate is extended out of the waveguide and a 4-section RF choke with 25 dB of isolation is added. Normalizing impedance is 31+j0 Ω, and substrate height (quartz) is 50 µm. See text for details.
bandwidth can very easily be recovered with the addition of a small capacitive tuning element in the waveguide (section VIII). VII. Sensitivity to Misalignment in the Waveguide
Fig. 9. Progression of probe input impedance and bandwidth as the substrate (50 µm quartz) is misaligned in the waveguide. The normalizing impedance is 31+j0 Ω.
In a practical system alignment of the probe cannot be guaranteed. As such we ran a set of simulations to better understand the effect of misalignment. As can be observed from
Fig. 9, misalignment of the probe essentially increases or decreases the “effective” radius of the probe, altering the shape (bandwidth) of the probe. Misalignment errors should be kept less than 4% of the waveguide height. A straightforward way to minimize allignment error for thick substrates is to precisely cut the substrate to the correct dimension and have it butt up against the top of the waveguide. VIII. Addition of a Capacitive Tuning Element in the Waveguide With the addition of a capacitive step in the waveguide, the characteristic “tear drop” impedance of the probe collapses into a “star” as shown in Fig. 11. The improvement in both the probe’s input return loss and bandwidth is dramatic. Fig. 10 shows a rendering of the probe and waveguide tuning step. Typically it is noticed that a 15% reduction in the waveguide height is adequate to tune out most of the probe’s residual reactance. Though the exact dimensions of the step are probe and substrate specific, it appears that to a first order, the capacitive step and backshort are symmetrically centered around the center of the probe. A typical length of the step is on the order of 0.13-0.14 λg . Because some of the reactance in the probe is tuned out by the step in the waveguide, the distance between the substrate and backshort can actually be slightly increased (∆ = 0.03-0.04 λg ). This is an additional advantage of using a waveguide step in that it actually eases circuit and machining tolerances. Though not shown here, we have ran simulations with the waveguide step and backshort corners chamfered. The chamfered corners have no effect on the overall performance of the probe, though the actual backshort position will have to be compensated by a tiny amount.
Fig. 10. Rendering of a capacitive tuning step in front of the radial probe substrate. Though the physical size of the wg constriction is minimal (15%), the reduction of the probe’s input return loss and increase in bandwidth are dramatic.
Fig. 11. Input return loss of the radial probe with and without the capacitive waveguide step. The characteristic “tear drop” impedance of the probe collapses into a star. The simulation here used the perfectly conducting (beam-lead) ground plane. Normalizing impedance is 31+j0 Ω.
IX. Scalemodel Verification of the Presented Results To verify the simulations and obtained results, we decided to run a series of S-band scalemodel measurements. The setup is shown in Fig. 12. For calibration standards we used 1, 50, and 100 Ohm chip resistors. After making a small adjustment for the physical length of the chip resistors we established the phase reference at the edge of the waveguide. The substrate material was Stycast with a measured dielectric constant of r =4.05 and a loss tangent δ=0.020. In Fig. 13 we present the results which verify the accuracy of the HFSS [11] computer simulations. The reactive component of the probe impedance can be
Fig. 12. Scalemodel (3.4-5.4 GHz) of the radial Probe with a 2-section RF choke and capacitive tuning step (not shown). The reference plane for the measurements is at the waveguide wall. Waveguide dimensions are 47.6 x 22.2mm, the probe radius 9.5mm, and the substrate dimensions 13.8 x 4 mm. The substrate height to airgap (below substrate) ratio is a half.
Fig. 13. Measured and simulated returnloss and input impedance of the discussed radial probe. The results include a two section RF choke with a capacitive tuning step in the waveguide. Smith chart is normalized to 50 Ω.
reduced by thinning the substrate and/or increasing the airgap under the first section of the RF choke. Based on scale model data, below the cutoff frequency of the waveguide the radial probe is modeled very precisely as a capacitor. Normalized to 1 GHz, it has a capacitance of 7.0 pF. X. Example of a fixed tuned 270-430 GHz Probe Design In Fig. 14 we present results from the final 270-430 GHz probe design, for an SIS mixer application. The mixer block consists of a full-height fixed tuned waveguide which uses as active device a superconducting (SIS) tunnel junction. The mixer block employs a capacitive tuning step as discussed in Section VIII. The waveguide and quartz substrate dimensions are 600 x 280 µm and 200 x 50 µm respectively. Substrate height to airgap ratio is one, and the probe radius 118 µm.
Fig. 14. Simulated input impedance for the final fixed tuned full-height waveguide 270-430 GHz SIS mixer. Fractional RF bandwidth is ≈ 50%. The probe input impedance is 51+j7 Ω, normalizing impedance is 50 Ω.
XI. Conclusion We have presented detailed simulation results on an inherent low impedance “one-sided” full-height waveguide to thinfilm microstrip probe design. Extensive computer simulations suggest that the optimal geometry for this kind of transition is one with constant radius, i.e. a radial fan. It is seen that the proposed thinfilm transition couples with better than 99% efficiency to an entire TE10 mode dominated waveguide band. Though the orientation of the probe is not critical, the results presented are solely for a split block design with the substrate parallel to the electric field. Simulations and scale model measurements suggest that a small capacitive step in the waveguide is able to tune out most of the residual reactance the probe may have left. Insignificant as the waveguide step may seem, it affords a significant improvement in bandwidth, performance, while easing machining and circuit tolerances. Though the probe has been developed for use with broadband fixed tuned SIS and HEB mixers in the submillimeter and THz regime, applications at microwave and millimeter wavelength are numerous. Examples include low loss waveguide to coaxial or microstrip transitions and broad bandwidth waveguide terminations. By selecting appropriate substrate materials and probe radii, any real impedance between 10 and 60 Ω can be achieved. XII. Acknowledgments We wish to thank Tony Kerr of NRAO and Sander Weinreb of JPL for very helpful discussions on “one-sided” waveguide probes. This work was supported in part by NSF grant# AST-9980846. References [1] R. L. Eisenhart and P. J. Khan, “Theoretical and experimental analyses of a waveguide mounting structure”, IEEE, Microwave Theory and Techniques, Vol MTT-19, pp. 706-717 (1971) [2] Tong C-Y. E., Blundell R, Paine S, “Design and characterization of a 250-350-GHz fixed-tuned superconductorinsulator-superconductor receiver”, IEEE, Microwave Theory and Techniques, Vol MTT-44, pp. 1548-1556, Sept. 1996 [3] J. W. Kooi , M. Chan, B. Bumble, and T. G. Phillips, “A low noise 345 GHz waveguide receiver employing a tuned 0.50 µm2 Nb/AlOx /Nb tunnel junction,” Int. J. IR and MM Waves, vol. 15, No. 5, May 1994. [4] J.W. Kooi, J. Pety, B. Bumble, C.K. Walker, H.G. LeDuc P.L. Schaffer, and T.G. Phillips, “A 850 GHz Waveguide Receiver employing a Niobium SIS Junction Fabricated on a 1µm Si3 N4 Membrane,” IEEE Transactions on Microwave Theory and Techniques, Vol. 46, No. 2, pp151-161, February 1998. [5] S. Withington, and G. Yassin, “Analytical expression for the input impedance of a microstrip probe in waveguide, newblockInt. J. IR and MM Waves, Vol. 17, pp. 1685-1705, Nov. 1996. [6] J. W. Kooi, C.K. Walker, J. Hesler, “A broadband suspended membrane waveguide to microstrip transition for THz Applications,” 9th International Conference on Therahertz Electronics, University of Virginia, Oct. 15-16, 2001 [7] S. Withington, G. Yassin, J. Leech, and K. G. Isaak, “An accurate expression for the input impedance of one-sided microstrip probes in waveguide”, Tenth International Symposium on Space Terahertz Technology, Charlottesville, March 1999 [8] Y-C Leong, and S. Weinreb “Full-band Waveguide-to-microstrip probe transitions” IEEE, Microwave Theory and Techniques, Digest of Papers, Anaheim, CA, June 13-19, 1999. [9] J.H.C. van Heuven “A new integrated waveguide-microstrip transition”, IEEE, Microwave Theory and Techniques, Vol MTT-24, pp. 144-147, March 1976 [10] A. R. Kerr and S. K Pan, “Some recent developments in the design of SIS mixers,”, Int. J. IR and MM Waves, Vol. 11, No. 10, pp. 1169-1187, Nov. 1990. [11] Ansoft Corporation “ Four Station Square, Suite 200, Pittsburgh, PA 15219-1119, USA [12] E. Schelecht, G. Chattopadhyah, A. Maestrini, A. Fung, S. Martin, D. Pukala, J. Bruston, and I. Mehdi, “200, 400, and 800 GHz Schottky diode substrateless multipliers: Design and Results” 2001 IEEE, MTT-S International Microwave Symp. Digest, Phoenix, Az, pp1649-1652, May 2001.