A Multek White Paper March 2017

Transcription

1 A Multek White Paper March 2017 Two Domas. Four Kds of Ratios. An Introduction to Characterizg the Electrical Properties of High Performance PCB Interconnects Usg Time and Frequency Doma Based Ratios. 1

2 Executive Summary Black Box Representation of a Simple Interconnect The Importance of Sufficient Characterization Bandwidth Output/Input Ratios for a Sgle-Ended Interconnect S and T-Parameter Ratios for Two Electrically Coupled Interconnects The Importance of Complementary S and T-Parameter Ratios Conclusions References

3 EXECUTIVE SUMMARY High performance prted circuit board (PCB) terconnects are rely used to re high speed digital, microwave, and RF/wireless signals from one location to another side a variety of products rangg from small wearable products to large rack-mounted storage devices cloud server farms. Unless they are designed properly, these terconnects modify the digital, microwave, and/or RF/wireless signals that propagate through them. The modifications are usually a combation of attenuation (the amplitude of the signal exitg the terconnect is lower than the amplitude of the signal enterg the terconnect) and distortion (the shape of the signal exitg the terconnect is different than the shape of the signal enterg the terconnect). Both of which are undesired and can, a worst case situation, render the terconnect non-functional. Quantifyg the attenuation and distortion properties of an terconnect can be accomplished by four kds of ratios the time and frequency domas. These four ratios are THRU (through), REFL (reflection), NEXT (near end crosstalk) and FEXT (far end crosstalk). This white paper describes the basis for why these two domas and four ratios are all that are needed to completely quantify the attenuation and distortion properties of a PCB terconnect. 3

4 Black Box Representation of a Simple Interconnect Referrg to Figure 1, we can conceptually envelop a simple PCB-based terconnect by a black box so that the only accessable electrical connections for measurg and characterizg the terconnect occur at each end. This is not as unrealistic as it sounds sce most PCB-based terconnects are accessible from any direct probg anyway because the conductive traces used to direct where the terconnect is red the PCB are covered by solder mask and/or red on ner layers. Usg this black box approach, we can completely encompass the entire physical construction of the terconnect cludg any connector and component mountg pads, via through/stubs and microstrip/striple transmission les structures. And by dog so we will also automatically capture all materials whose electrical properties impact how the signal propagates through the terconnect cludg anisotropic dielectric constant variations, conductor surface roughness variations, as well as frequency dependent variations dielectric and conductor conductivities. By their very nature, PCB terconnects are passive structures. They do not broadcast any formation ab their attenuation and distortion producg properties that we can sense, for example with some kd of antenna or probe. In order to quantify what is gog on side the terconnect, we need to ject some kd of signal to the terconnect and record the response of the terconnect to that jected signal. For the example shown Figure 1, that means sendg a predefed signal shape, V, to the terconnect at Port 1, and then measurg what comes at the other end, V, at Port 2. This poses a bit of a problem sce we may not always know what the actual signals are. For example, a signal passg through a midplane or backplane can vary dependg on what le cards or blades are plugged to it. A more universal approach that works around this problem is to characterize the terconnect by the ratio of the put signal to the put signal, V / V. Usg this approach, we can then multiply any shaped put signal - whether it s high speed digital, microwave or RF/wireless - by this ratio to get the shape of the put signal. It is worth notg that this ratio defition, everythg that goes on side the PCB terconnect that causes attenuation and distortion - dielectric losses, sk depth losses, surface roughness losses, oxide coatg losses and impedance mismatch losses - is cluded the ratio. Defg the ratio this black box manner results a comprehensive characterization that does not accidentally leave any detrimental phenomena of the measurement. Nothg falls through the cracks, so to speak. 4

5 The Importance of Sufficient Characterization Bandwidth This ratio concept works quite well as long as the put signal used to obta the ratio has a bandwidth equal to or greater than the bandwidth of any signal that is propagated through the terconnect durg actual or proposed use. This is relatively easy to accomplish an electrical characterization laboratory equipped with modern terconnect characterization struments such as the Vector Network Analyzer (VNA) and Time Doma Reflect/Time Doma Through (TDR/TDT). A VNA characterizes terconnects by jectg a variable frequency se wave to the terconnect and observg how the se waves of varyg frequency are attenuated and phase shifted by the terconnect. As long as the start and stop frequency of the variable frequency source the VNA are lower and higher than the necessary bandwidth of the terconnect, accurate characterization is achieved. A measurement made usg se waves is classified as a frequency doma measurement. A frequency doma ratio is also called an S-parameter. A TDR/TDT accomplishes the same goal by jectg a leadg edge of a pulse to the terconnect. As long as the rise time of the leadg edge of the TDR/TDT pulse is faster (shorter time) than the fastest rise time of any signal propagated through the terconnect durg actual use, accurate characterization is achieved. A measurement made usg the leadg edge of a pulse is classified as a time doma measurement. A time doma ratio is also called a T-parameter. Note that both VNA and TDR/TDT cases, two measurements need to be made before a ratio can be computed. One measurement that captures the amplitude and shape of the put signal, V, and one measurement that captures the amplitude and shape of the put signal, V. Output/Input Ratios for a Sgle-Ended Interconnect In Figure 1, we jected the signal to the left side Port 1, while we looked at what came at the right side Port 2. We may then ask: Does the terconnect behave differently if we reverse the ports, and send the put signal, V, to Port 2 and monitor what comes of Port 1? It is a valid question to ask sce many PCB-based terconnects are not symmetrical ab the center of the terconnect. Bends, twists and via locations on the put side of the terconnect s center are not mirror images of the bends, twists and via locations on the put side of the terconnect s center. Thus if we want to properly characterize the terconnect, we need to also characterize the terconnect with the signal propagatg the opposite direction. To keep track of which direction the signal is propagatg when we measure/calculate one of the S-parameter or T-parameter ratios, we use a subscriptg notation where the first subscript is the port associated with the put port, while the second subscript is associated with the put port. For example, the S and T-parameters for the terconnect Figure 1, where the put signal, V, is sent to Port 1 and the put signal is measured at Port 2 would be defed as S and T respectively. However, if the put signal is sent to Port 2, then the subscripts would be reversed to S 12 and T 12 5

6 respectively. S and T-parameters associated with an put port that is at the other end of the terconnect from the put port are also called THRU (through) parameters. One of the nice properties of these two THRU parameters is that for passive lear PCB terconnects, they are equal, meang S S12 and T T12 If they are not, then it dicates a re-measurement is required. If there are impedance mismatches side the terconnect, then a portion of the signal that is sent to Port 1 comes back of Port 1. Similarly, if we send a signal to Port 2, a portion of the signal comes back of Port 2. This is obviously not a desirable situation, for if part of the put signal comes back of the put port, it cannot also be part of the signal present at the put port. In many cases, impedance mismatches create structural resonances that troduce a significant amount of distortion. Usg the same subscript notation scheme defed above, we can then defe four new ratios: S 11 and T 11 for the case where the signal is put to Port 1; and S 22 and T 22 for the case where the signal is put to Port 2. Ideally, we would like the terconnect to be direction-agnostic, meang S11 S22 and T 11 T 22. But unlike the THRU parameters, this is rarely the case. Ratios associated with an put port that is the same as the put port are also called REFL (reflection) parameters. Note that an ideal terconnect, the magnitude of the THRU parameters must equal 1 (the entire put signal appears at the put), while the magnitude of the REFL parameters must be equal to 0 (nothg from the put signal is reflected back of the put port). From these two properties of an ideal terconnect one can easily gauge the overall quality of the terconnect by simply observg how far the magnitudes of the THRU parameters are from 1 and how far the magnitudes of the REFL parameters are from 0. Referrg to Figure 2, we can thus defe the four VNA and four TDR/TDT measurement-based ratios tabulated Table 1. Figure 2: VNA and TDR/TDT Parameters for a Sgle-Ended (One Trace) Interconnect 6

7 Table 1: VNA and TDR/TDT Parameters for a Sgle-Ended (One Trace) Interconnect Freq Doma S-Parameter Time Doma T-Parameter S S S S Type Parameter Ratio Defition T THRU V (Port 2) / V (Port 1) T THRU V (Port 1) / V (Port 2) T REFL V (Port 1) / V (Port 1) T REFL V (Port 2) / V (Port 2) S-Parameter and T-Parameter Ratios for Two Electrically Coupled Interconnects If we have two sgle-ended terconnects that are red close to each other, energy can flow between the two terconnects. One example of such an terconnect is shown Figure 3, where two coupled traces are designed to carry a differential pair set of signals. In this case, we have two additional kds of ratios called NEXT (near end crosstalk) and FEXT (far end crosstalk) that get added to the THRU and REFL parameters defed above. Referrg to Figure 3, we can thus defe sixteen VNA and sixteen TDR/TDT measurement-based ratios for two electrically coupled terconnects. Figure 3: S-Parameter and T-Parameter Ratios for Two Coupled Interconnects 7

8 It is also worth notg that this ratio scheme is scalable to multiple (more than two) coupled terconnects. One example where this is commonly done is the characterization of parallel data bus architectures where the dividual traces are red close to each other. The Importance of Complementary S and T-Parameter Ratios One of the major sources of a PCB terconnect s undesired attenuation and distortion properties is conversion of electromagnetic energy (associated with the signal that is propagatg through the terconnect) to thermal energy. This conversion is frequency dependent with more conversion occurrg at higher frequencies than at lower frequencies. Conversion mechanisms that fall to this category clude dielectric loss tangent and conductor sk depth, surface roughness and oxide coatg related losses. Structures such as via stubs are also prone to resonate, troducg additional signal distortions. These sources of attenuation and distortion are best characterized the frequency doma usg the S-parameter ratios. However, discrete changes the characteristic impedance of the terconnect, usually occurrg at via locations or along the terconnect at locations where the cross section of the conductors and/or the dielectric construction that comprise the terconnect are not physically or electrically uniform, are usually best charactersized the the time doma usg the T-parameters. Conclusion High performance prted circuit board (PCB) terconnects are rely used to re high speed digital, microwave, and RF/wireless signals from one location to another side a variety of products rangg from small wearable products to large rack-mounted storage devices cloud server farms. Unless they are designed properly, these terconnects can attenuate and distort the digital, microwave, and/or RF/wireless signals that propagate through them. Complete characterization of these PCB terconnects can be accomplished by four kds of ratios the time and frequency domas. These four ratios are THRU (through), REFL (reflection), NEXT (near end crosstalk) and FEXT (far end crosstalk). For a sgled-ended terconnect, a total of eight ratios are required: four the frequency doma and four the time doma. For a coupled pair terconnect, a total of thirty two ratios are required: sixteen the frequency doma and sixteen the time doma. 8

9 References [1] David M. Pozar, Microwave Engeerg, Second Edition, John Wiley and Sons, Inc., [2] Lawrence N. Dworsky, Modern Transmission Le Theory and Applications, John Wiley and Sons, Inc., 1979 [3] Matthew N. O. Sadiku, Elements of Electromagnetics, 3 rd Edition, Oxford University Press, [4] C. W. Davidson, Transmission Les for Communications, A Halsted Press Book, Ab the Author: Franz Gis is Director of Signal Integrity at Multek s Interconnect Technologhy Center Milpitas, California. Franz Gis s core focus is the electrical characterization of PCB-based high performance digital, RF and microwave terconnects. Prior to Multek, Franz Gis has worked for over 40 years electromagnetics cludg EMC (Electromagnetic Compatibility), signal tegrity and the characterization and modelg of high performance terconnects. [5] Fred E. Gardiol, Lossy Transmission Les, Artech House, [6] Richard E. Matick, Transmission Les for Digital and Communication Networks, McGraww-Hill, [7] Mike Resso, Eric Bogat, Executive Editors, Signal Integrity Characterization Techniques, IEC (International Engeerg Consortium). Franz Gis has a BS Electrical Engeerg and an MS Applied Mathematics. Ab Multek s Interconnect Technology Center (ITC): The Interconnect Technology Center (ITC) is Multek s advanced technology development organization. We engage with customers early the design process to create novative solutions to pressg technical challenges. Our technical core competencies are aligned to meet the challenges of trends around creasg data rates, creasg density of PCBs, and new shape requirements. 9

10 Two Domas. Four Kds of Ratios. March 2017 Author: Franz Gis Multek 17 th Floor, Na Tower (Tower II) 8 Yeung Uk Road, Tsuen Wan New Territories, Hong Kong Worldwide Inquiries: Phone: Fax: multek.com 2017, Multek and/or its affiliates. All rights reserved. This document is provided for formation purposes only and the contents hereof are subject to change with notice. This document is not warranted to be errorfree, nor subject to any other warranties or conditions, whether expressed orally or implied law, cludg implied warranties and conditions of merchantability or fitness for a particular purpose. We specifically disclaim any liability with respect to this document and no contractual obligations are formed either directly or directly by this document. This document may not be reproduced or transmitted any form or by any means, electronic or mechanical, for any purpose, with our prior written permission. Multek is a registered trademark of Multek and/or its affiliates. Other names may be trademarks of their respective owners. 10