/14/$ IEEE 939

Transcription

1 Electro-Mechanical Structures for Channel Emulation Satyajeet Shinde #1, Sen Yang #2, Nicholas Erickson #3, David Pommerenke #4, Chong Ding *1, Douglas White *1, Stephen Scearce *1, Yaochao Yang *2 # Missouri S&T EMC Laboratory, Missouri University of Science and Technology, Rolla, MO 6549, USA *Cisco Systems, Inc. 1 Research Triangle Park, NC, 2 San Jose, CA, USA Abstract Channel emulators are used to evaluate communication system performance either in absence of the real channel or to test the system s response for varying channel characteristics. For high speed differential digital channels bandwidths in excess of 2 GHz are common making it difficult to recreate the channel performance by electronic means such as FIR filters. An alternative solution is using a low loss short transmission line and having its properties modified by mechanical means. Passive structures are robust, have a frequency range only limited by the low loss trace, do not add noise, cannot be damaged by ESD and are very economical. This paper describes two electro-mechanical structures for introducing loss and nulling into the frequency response of a channel. The first part describes the design of a mechanically tuned quarter-wavelength stub filter that can be used to emulate the resonances of a channel. In the second part, an electromechanical structure, consisting of Bragg grating and lossy materials, is constructed to emulate the loss behaviour and the resonances of a channel. I. INTRODUCTION The performance of high speed digital communication systems can be measured using the real channel, or by emulating a channel. Emulation allows testing a broad range of channel characteristics. The response of an electrical channel is typically characterized by a smooth roll off with rising frequency caused by copper and dielectric losses and by nulling caused by reflections, or possibly resonant radiation losses. The easiest way to emulate the channel is to have a fixed channel, such as a long cable, or a fixed filter structure. However, this is inflexible if it comes to observing the system s performance for varying channel parameter. Wideband channel emulators and driver emphasis devices such as the Tektronix LE32, are implemented using finite impulse response (FIR) filters such as the Hittite HMC6545LP5E [1], [2]. The CLE1 uses lossy materials in close proximity to a trace to adjust the loss [3]. The MP1825B allows adjusting the general loss, however it cannot be used to create nulls [4]. However, all active emulation is limited by its added noise, the IC s frequency response, none linear distortion and the range of adjustability often given by the tap delay and the number of taps. Cost and ESD sensitivity maybe further considerations. An alternative approach is to use a low loss trace and to vary its frequency response by mechanical means. The bandwidth is then only determined by the low loss trace and its connectors and the structure is robust against overload and ESD. We have designed two systems to emulate the actual channel s response. These structures are constructed using low loss Megtron 6 PCB material. These structures form individual blocks which can be cascaded to construct a more complex channel. The first part describes the design of a mechanically tuned quarter-wavelength stub filter that can be used to emulate the resonances of a channel [5]. In the second part, an electro-mechanical structure, consisting of Bragg grating (periodic disturbances of a trace) and lossy materials, is constructed to emulate the loss behaviour and the resonances of a channel. II. MECHANICALLY TUNED BAND-STOP FILTER A. Concept A typical channel response may consist of one or multiple band stops or nulls. To emulate/introduce these band stops in the channel emulator, we design a mechanically tuneable quarter-wavelength open ended stub transformer. The tuning of the resonance frequency, in the range 1GHz to 2 GHz, can be achieved by changing the length of the stub by mechanical means. Nulls at odd multiples of the fundamental frequency are also present, however they are usually not of great concern. Nulls in the frequency range below one-half of the data rate strongly influence the eye-diagram, nulls between one-half and the data rate have a moderate influence, and nulls at frequencies greater than the data rate show little influence if the width of the null is not too large. Thus, the first harmonic dominates the effect of the null on the eye, while the third and subsequent odd harmonics, which are unavoidable in this concept, may already fall into a frequency, at which a null in the channel s response has little influence on the eye parameter. The quarter-wave transformer concept is shown in Fig. 1. 5ohm Trace Impedance = 5ohm Z in Trace Impedance = 5ohm Transmission Line Stub impedance = Z L R L (Open) Fig. 1: The quarter-wave open-ended stub transformer. 5ohm /14/$ IEEE 939

2 Since for the open ended stub, R L is infinity, Z in transforms to zero, or a short at the stub resonance frequency. B. Structure The movable rod is supported by placing it inside a metal (brass) tube that is soldered onto and along the length of the micro-strip trace. The metal rod and the brass tube are chosen such that the outer diameter of the metal rod matches the inner diameter of the brass tube. Conductive grease is applied between the movable rod and the brass tube. This ensures sufficient contact between the brass tube and the metal rod. Further, in order to allow for tuning two independent nulls at two different frequencies, two pairs of metal rods and brass tubes are used. Two tubes are connected on the two top and bottom side traces with a via-transition in between. metal tube above the trace, gives rise to additional fringing capacitance from the tube to the ground plane as described in Fig 4. The combined characteristic impedance the trace-tube combination becomes lower than that of a micro-strip having the same trace width. Thus, the trace width of the micro-strip under the tube is reduced to compensate for the additional capacitance due to the tube and match the characteristic impedance to 5-ohms. The S 21 magnitude is simulated for different lengths of the movable metal rod to obtain different resonance frequencies for the band-stop filter. Modified Microstrip Trace Width =.8mm Original Microstrip Trace Width = 1mm Metal Tube (Outer Diameter=.8mm) Fringing fields due to the metal tube PORT 1 Metal Tube (Dia=.8mm) Low Loss Substrate Metal Rod Stub (Dia=.5mm) Signal Via Airgap (.15mm) Substrate Thickness=2mils Fields due to the microstrip Plane Plane Via Microstrip Trace PORT 2 Fig. 2: Diagram describing the structure of the mechanically tuned bandstop filter. Its main structure comprises two microstrip lines on opposite sides of a PCB that are connected by a via. C. Simulation Model The structure is simulated in Ansoft HFSS 15. using the frequency domain solver as shown in Fig. 3. Fig. 4: Fringing fields due to the metal tube and the micro-strip trace. D. Design and Construction The optimized dimensions obtained from the full-wave simulations are used to construct a 2-layer, printed circuit board on the low loss substrate Megtron 6. The layout is shown in Fig. 5. Brass tubes are soldered onto the microstrip trace and SMA connectors are mounted. The assembled structure is shown in Fig. 6 below. Fig. 5: Printed circuit board layout. Signal-via Trace Fixed tube mounted on this transition Top ground for movable Width=.82mm portion of the micro-strip Movable Rod rod impedance transformer trace (Stub) SMA Connector Trace Width=1mm Via cage along the signal line SMA Connector Fig. 3: Full-wave simulation model of the mechanically tuned band-stop filter in Ansoft HFSS 15. Lumped ports terminate the microstrip traces. The structure requires a metal tube soldered onto the micro-strip trace. Therefore, characteristic impedance of the tube-over-trace combination must be matched to 5-ohms. The addition of the Fig. 6: Photo of the structure. The board is mounted on a metal sheet for mechanical support. E. Measurement setup and results The S 21 magnitude of the structure is measured using a Vector Network Analyzer up to 2 GHz, using two ports connected to the two SMA connectors. The measurement setup is shown in Fig. 7 below. Only one movable rod is used 94

3 Magnitude S 21 Magnitude S 21 for the measurements. The measurement results are compared with the simulation results for two different lengths of the movable rod. Two band stops corresponding to the two different lengths can be observed in Fig. 8. introducing lossy material close to the stub is shown in the Fig. 9. It must be noted here that the lossy material also introduces a shift in the resonant frequency; however the desired resonance frequency can be obtained by tuning the stub length Lossy material Without Lossy Material With Lossy Material Fig. 7: Measurement setup to measure the S 21 of the band-stop filter Different lengths of the movable rod: mm, 2mm, 4mm, 7mm Meas-Rod-len-mm Sim-Rod-len-mm Meas-Rod-len-1mm Sim-Rod-len-1mm Meas-Rod-len-3.4mm Sim-Rod-len-3.4mm Frequency (MHz) Fig. 8: Measurement and simulation comparison for different lengths of the movable rod length. III. METHODS TO CHANGE THE Q-FACTOR AND DEPTH OF THE BAND STOPS The mechanical structure described above can be used to introduce band stops at different frequencies by changing the position of the movable rod. For a practical application of such a filter for channel emulation of a channel, there is a requirement to be able to change the Q-factor and depth of the band stops, such that the desired shape of the resonance can be emulated. We investigate some of the methods to change the Q-factor of the band stops. A. Effect of lossy materials in close proximity to the stub: The Q-factor of the band stop can be reduced by placing lossy materials in close proximity to the movable rod/stub. The measurement result comparison before and after Frequency (MHz) Fig. 9: Simulated S 21 magnitude for proximity of lossy material to the movable rod. B. Effect of increasing the height of the stub (movable rod) above the ground plane: On increasing the height of the stub above the ground plane, the characteristic impedance of the stub is increased. This results in the reduction of the null depth. The simulation result comparison for two different heights 1.5mm and.25mm is shown in Fig. 1. The rod length is kept at 7.3mm and 6.6mm for the rod height of.15mm and 1.5mm respectively. The length is changed to correct for the slight shift in resonance frequency caused due to the change in height. The simulation result shows a reduction in the null depth, caused due to an increase in the characteristic impedance of the stub Movable Rod: 1.5 mm above ground plane -3 Rod-height-.15mm Movable Rod:.25 mm Rod-height-1.5mm above ground plane Fig. 1: Simulated S 21 magnitude for different heights of movable rod above the ground plane 1.5mm and.25mm. 941

4 C. Effect of loss of the Rod Material: Brass and Graphitecomposite The material of the movable rod has an influence on the Q- factor of the resonance. The comparison between the results of the measurement carried out using a metal rod and a graphitecomposite rod, in Fig. 11, shows the difference in the null depths. The loss factors of the rods affect the null depth of the resonance. The null depth can be adjusted from -1dB to - 35dB using different lossy materials for the rod. Fig. 12: Simulation model with the modified ground plane and copper patch connected to the ground plane with a variable resistance boundary ohm 1-ohm 3-ohm 1-ohm 5-ohm Fig. 13: Simulated S 21 magnitude of the modified ground structure for different resistance values. Fig. 11: Measured S 21 magnitude for different movable rod materials Graphite and Brass. The null depth is reduced by about 2dB. D. Effect of ground plane impedance variation: The depth of the null can be tuned by changing the impedance of the current return path under the quarter-wave stub. In the simulation model, the ground plane structure under the stub was modified by designing a copper patch and connecting the patch with a resistive boundary to the surrounding ground plane as described in Fig. 12. The resistance of this boundary was parametrically varied to obtain the S 21. The simulated S 21 for different resistance values are shown in Fig. 13. The results show that the null depth reduces as the resistance value is increased. For practical implementation, PIN diodes, used as variable resistors, can be used to change the impedance of the current return path. E. Effect of Rod diameter: Changing the diameter of the movable rod influences the width of the resonance. The simulation result comparison for rod diameters.5mm and.1mm is shown in Fig. 14. The rod length in both cases is kept at 7.3mm. The result shows that a smaller rod diameter results in a reduction in the width of the null. This can be explained as result of a smaller rod diameter and an increase in the characteristic impedance of the stub. Copper Patch Variable resistance boundary Plane Plane Fig. 14: Simulated S 21 magnitude for different diameters of the movable rod.5mm and.1mm. IV. TARGET S-PARAMETERS VS EMULATED S-PARAMETERS The mechanical band-stop filter is used to emulate the s- parameters of a measured channel. The S 21 magnitude of the measured channel shows a notch at around 4.4 GHz and general loss behaviour. Fig. 15 shows the target and the emulated S 21 magnitude. The notch is tuned by tuning the movable rod of the mechanical band stop filter and the general loss behaviour is emulated by placing lossy materials on the trace. 942

5 -1-2 lossy material under the trace is controlled using the motor drivers. Trace PCB Bragg on top of the trace -3 Target S-parameter Emulated S-parameters GND plane with slot Lossy material Slot Fig. 15: Target and emulated S 21 magnitude by tuning the movable rod and lossy materials on the microstrip trace. Fig. 16: Diagram describing the structure with the Bragg sliders and the lossy material. V. ELECTROMECHANICAL STRUCTURE FOR CHANNEL EMULATION The mechanical structure for channel emulation utilizes a combination of two structures, for emulating a given channel response. The Bragg introduces band stops, based on the concept of periodic and non-periodic discontinuities on a transmission line. The lossy material lifter emulated the smooth loss function of the channel. Motor Potentiometer Slider A. Structure and Construction: The structure mainly consists of two parts, as mentioned above. The mechanical Bragg sliders are placed on the top of a differential microstrip pair and the lossy material lifter on the bottom. The schematic is shown in Fig. 16. The mechanical Bragg structure, shown in Fig. 17, drives five identical and equally spaced sliders on top of the trace. These sliders function as periodic discontinuities for the transmission line. The sliders over the trace are shown in Fig. 19. By changing the distance between each slider, the notch frequency on S 21 can be changed. By setting them in nonperiodic distances other perturbations of the channel can be achieved. The microstrip has a slotted ground. This way, perturbations of the field can be introduced from the top and from the bottom. The lossy material lifter, shown in Fig. 18, lifts the lossy material to the underside of the PCB and introduces losses by attenuating the fields that pass through the slotted ground plane. By varying the space between the lossy material and the slot, the general loss function of the channel response can be emulated. A prototype of the structure was built using copper-clad PCB structure. DC motors are used to control the position of individual sliders over the differential trace pair. One motor is used to control the height of the lossy material under the slot. The control circuit for the DC motors consists of motor drivers which are controlled by a microcontroller having a USB interface. The position of the sliders and the height of the Fig. 17: The mechanical structure showing the sliders of the Bragg structure. Lifter Lossy material Fig. 18: The lossy material lifter to lift the loss material towards the slot under the trace. B. Measurement Setup and Results: The S 21 magnitude of the structure is measured using a Vector Network Analyzer up to 2 GHz, using two ports connected to the two SMA connectors. This is a single ended measurement; however the structure can support differential microstrip lines. The measurement setup is shown in Fig. 2. Two measurements are recorded in the first case, the separation between the Bragg grating is changed while keeping the height of the lossy material at a fixed height. In the second measurement, the height of the lossy material under the slot is changed while keeping the Bragg sliders in a fixed position. The S 21 measurement result comparison for the first case is shown in Fig. 21 which shows the variation of the 943

6 Magnitude S 21 band stops due to the change in the separation between the Bragg sliders. Bragg sliders over trace the S 21 magnitude response with the change in the height of the lossy material under the slots in the ground plane of the board No Loss Loss-1 Loss USB Controller Fig 19. Bragg sliders over trace. DC Power Supply Control Circuit VNA Mechanical structure Lossy Material Lifter Mechanical Bragg Fig. 2: Measurement Setup showing the mechanical structure and the control circuitry No Bragg Bragg Spacing-1 Bragg Spacing Fig.21: Measured S 21 magnitude for different spacing between the Bragg grating showing different band stops. Fig. 22 shows the measurement comparison results for the second case which shows the change in the roll-off function of Fig. 22: Measured S 21 magnitude for different distances of proximity of the lossy material to the slot under the trace. VI. CONCLUSION This paper describes two electro-mechanical structures for emulating different characteristics of the frequency response of a channel. The structures when constructed using low loss substrate (eg. Megtron 6) for the printed circuit boards can be cascaded. The designs described here are relatively low cost and simple to implement as compared to other methods that use integrated circuit based channel emulators. The upper limit of usable frequency range for the mechanically tuned band-stop filter is determined by the via-transition, edgelaunch connectors, and the diameter of the fixed and movable tube. Band stops at the higher order odd harmonics of the stub are also present, which place a lower limit on the resonance frequency that can be emulated. For the motor controlled mechanical structure, the accuracy of emulating the frequency response of a given channel depends on the precision with which the Bragg grating and lossy material lifter can be positioned using the control circuit. VII. ACKNOWLEDGEMENT This material is based upon work supported by the National Science Foundation under Grant No We also thank Cisco Systems Inc., USA for their support towards this work. VIII. REFERENCES [1] [2] [3] [4] Solutions/Products/MP1825B.aspx. [5] D. M. Pozar, and D. H. Schaubert, Microstrip Antennas, IEEE Press