Transcript ppt
ESE370: Circuit-Level Modeling, Design, and Optimization for Digital Systems Day 36: December 6, 2010 Transmission Lines 1 Penn ESE370 Fall2010 -- DeHon Last week + today • • • • • • • • General wire formulation Lossless Transmission Line Reflections at ends of line See in action in lab Where arise? Termination Implications Discuss Lossy Penn ESE370 Fall2010 -- DeHon 2 Transmission Lines • This is what wires/cables look like – Aren’t an ideal equipotential – Signals do take time to propagate – Shape and topology of wiring effects how signals propagate • …and the noise effects they see 3 Penn ESE370 Fall2010 -- DeHon Transmission Lines • Need to understand – How to model how to reason about – What can cause noise – How to engineer high performance communication 4 Penn ESE370 Fall2010 -- DeHon Transmission Lines • Cable: coaxial • PCB – Strip line – Microstrip line • Twisted Pair (Cat5) 5 Penn ESE370 Fall2010 -- DeHon Coaxial Cable • Inner core conductor: radius r • Insulator: out to radius R • Outer core shield (ground) R L ln 2 r 1 R Z 0 ln 2 r • RG-58 Z0= 50W – networking, lab • (RG-59 Z0= 75W – video) Penn ESE370 Fall2010 -- DeHon 6 Printed Circuit Board • Stripline – Trace between planes w t b r W 1 1 b Z 0 ln t W 4 b b 7 Penn ESE370 Fall2010 -- DeHon Printed Circuit Board • Microstrip line – Trace over single supply plane 0 r w t h 1 4h ln Z 0 2 0.475 r 0.67 0 0.536W 0.67t 8 Penn ESE370 Fall2010 -- DeHon Twisted Pair • Category 5 ethernet cable – 100W – V=0.64c0 9 Penn ESE370 Fall2010 -- DeHon Termination / Mismatch • Wires do look like these transmission lines • We are terminating them in some way when we connect to gate – Need to be deliberate about how terminate, if we care about high performance 10 Penn ESE370 Fall2010 -- DeHon Where Mismatch? • • • • • Vias Wire corners? Connectors Board-to-cable Cable-to-cable http://www.fpga4fun.com/Hands-on_Flashy.html http://wiki.altium.com/display/ADOH/An+Overview+of+Electronic+Product+Development+in+Altium+Designer 11 Penn ESE370 Fall2010 -- DeHon Visualization • http://www.williamson-labs.com/xmission.htm 12 Penn ESE370 Fall2010 -- DeHon End of Line Reflection R Z 0 Vr Vi R Z 0 2R Vt Vi R Z 0 13 Penn ESE370 Fall2010 -- DeHon Termination Cases • Parallel at Sink – Pix • Series at Source – pix 14 Penn ESE370 Fall2010 -- DeHon Pipeline Bits • For properly terminated transmission line – Do not need to wait for bits to arrive at sink – Can stick new bits onto wire 15 Penn ESE370 Fall2010 -- DeHon Limits to Bit Pipelining • What limits? – Risetime/distortion – Clocking • Skew • Jitter – For bus • Wire length differences between lines 16 Penn ESE370 Fall2010 -- DeHon Eye Diagrams • Watch bits over line on scope – Look at distortion – “open” eye clean place to sample • Consistent timing of transitions • Well defined high/low voltage levels http://en.wikipedia.org/wiki/File:On-off_keying_eye_diagram.svg http://focus.ti.com/analog/docs/gencontent.tsp?familyId=361&genContentId=41762&DCMP=ESD_Solutions&HQS=Other+OT+esd 17 Penn ESE370 Fall2010 -- DeHon Bad “eye” http://archive.chipcenter.com/knowledge_centers/asic/todays_feature/showArticle.jhtml?articleID=12800254 18 Penn ESE370 Fall2010 -- DeHon Impedance Change • What happens if there is an impedance change in the wire? 19 Penn ESE370 Fall2010 -- DeHon Z0=75, Z1=50 • At junction: – Reflects • Vr=(50-75)/(50+75)Vi – Transmits • Vt=(100/(50+75))Vi R Z 0 Vi Vr R Z0 2R Vi Vt R Z 0 20 Penn ESE370 Fall2010 -- DeHon Impedance Change Z0=75, Z1=50 21 Penn ESE370 Fall2010 -- DeHon What happens at branch? 22 Penn ESE370 Fall2010 -- DeHon Branch • Transmission line sees two Z0 in parallel – Looks like Z0/2 23 Penn ESE370 Fall2010 -- DeHon Z0=50, Z1=25 • At junction: – Reflects • Vr=(25-50)/(25+50)Vi – Transmits • Vt=(50/(25+50))Vi R Z 0 Vi Vr R Z0 2R Vi Vt R Z 0 24 Penn ESE370 Fall2010 -- DeHon End of Branch • What happens at end? • If ends in matched, parallel termination – No further reflections 25 Penn ESE370 Fall2010 -- DeHon Branch Simulation 26 Penn ESE370 Fall2010 -- DeHon Branch with Open Circuit? • What happens if branch open circuit? 27 Penn ESE370 Fall2010 -- DeHon Branch with Open Circuit • Reflects at end of open-circuit stub • Reflection returns to branch – …and encounters branch again – Send transmission pulse to both • Source and other branch • Sink sees original pulse as multiple smaller pulses spread out over time 28 Penn ESE370 Fall2010 -- DeHon Open Branch Simulation 29 Penn ESE370 Fall2010 -- DeHon Open Branch Simulation 30 Penn ESE370 Fall2010 -- DeHon Bus • Common to have many modules on a bus – E.g. PCI slots – DIM slots for memory • High speed bus lines are trans. lines 31 Penn ESE370 Fall2010 -- DeHon Multi-drop Bus • Ideal – Open circuit, no load 32 Penn ESE370 Fall2010 -- DeHon Multi-Drop Bus • Capacitive load (stub) at drop? – If tight/regular enough, change Z of line L Z0 C 33 Penn ESE370 Fall2010 -- DeHon Multi-Drop Bus • Long wire stub? – Looks like branch • may produce reflections 34 Penn ESE370 Fall2010 -- DeHon Lossy Transmission Line • How do addition of R’s change? 35 Penn ESE370 Fall2010 -- DeHon Lossy Transmission Line • R’s cause signal attenuation (loss) – Voltage, energy – Reduced signal swing at sink – Limits length of transmission line before need to restore signal 36 Penn ESE370 Fall2010 -- DeHon Transmission Line Noise • Frequency limits • Imperfect termination • Mismatched segments/junctions/vias/connectors • Loss due to resistance in line – Limits length 37 Penn ESE370 Fall2010 -- DeHon Admin • Proj3b due Friday 38 Penn ESE370 Fall2010 -- DeHon Idea • Transmission lines – high-speed – high throughput – long-distance signaling • Termination • Signal quality 1 c0 w LC r r L Z0 C R Z 0 Vr Vi R Z 0 Penn ESE370 Fall2010 -- DeHon 39