MicronPresentationppt

1、micron engineering clinicfall 08 spring 09 welcome to micron engineering clinic analysis and optimization of multi gb/s chip- to-chip communication micron engineering clinicfall 08 spring 09 team members documentation lead raheem alhamdani - ce technical leads bryson kent - ee jordan kemp - ee lucas

2、 loero - ee team lead ben meakin ce micron engineering clinicfall 08 spring 09 introduction and motivation for modeling and verification of interconnects raheem alhamdani documentation lead computer engineer a micron engineering clinicfall 08 spring 09 introduction the sponsor: ma

3、nufacturer of dram, flash memory, and image sensor integrated circuits what have they asked us to do? design a software application for modeling and verification of chip-to-chip interconnects micron engineering clinicfall 08 spring 09 introduction what is verification? proving through tests and form

4、al methods that a design does what it is intended to do what are chip-to-chip interconnects? electrical systems for communication between two integrated circuits micron engineering clinicfall 08 spring 09 motivation background memory and i/o devices operate much slower than cpu access to off-chip re

5、sources is expensive 300 cycles - usually cycles are wasted demand for low-power yet high-performance - cant have wasted cycles! goal: speed up devices and speed up interconnect micron engineering clinicfall 08 spring 09 as devices move towards being smaller, faster, lower power interconnects become

6、 slower, noisier, and unreliable issues: inter-symbol interference (isi) co-channel interference timing jitter voltage noise problem conventional testing methodologies are not feasible or sufficient micron engineering clinicfall 08 spring 09 eye diagram voltagvoltag e e timtim e e what is an eye dia

7、gram? a useful tool for the qualitative analysis of signal used in digital transmission. micron engineering clinicfall 08 spring 09 eye diagram timetime voltagevoltage voltagevoltage timetime bitsbits superimposedsuperimposed 1 unit interval (ui)1 unit interval (ui) how is it created? micron enginee

8、ring clinicfall 08 spring 09 eye diagram (noise) timetime voltagevoltage voltagevoltage timetime bitsbits superimposedsuperimposed 1 unit interval (ui)1 unit interval (ui) bit-streambit-stream voltage noisevoltage noise what causes noise? interference from neighboring wires (co- channel interference

9、) electromagnetic interference link resistance, capacitance, and inductance micron engineering clinicfall 08 spring 09 eye diagram (jitter) timetime voltagevoltage voltagevoltage timetime bitsbits superimposedsuperimposed 1 unit interval (ui)1 unit interval (ui) bit-streambit-stream jitterjitter wha

10、t causes jitter? clock variation (skew) reflection general timing uncertainty micron engineering clinicfall 08 spring 09 real eye diagram data jitterdata jitterclock clock jitterjitter signal noisesignal noise vrefvref data signaldata signal clock signalclock signal vref noisevref noise + + receiver

11、receiver sensitivitysensitivity how to interpret it? micron engineering clinicfall 08 spring 09 solution our objective is not to solve these problems through better design, but to provide designers with a tool that correctly models and verifies interconnects with these problems deliverables: cross p

12、latform app with graphical user interface provide worst-case and statistical based link analysis spice compatible correctly model co-channel interference and tx/rx jitter micron engineering clinicfall 08 spring 09 project documentation meeting minutes, time-line, progress report , presentations, pro

13、posal and links are all on the teams website: /alhamdan/micron/micron.html my roles graphical user interface and software development gui software skeleton plotting code code documentation micron engineering clinicfall 08 spring 09 teams website micron engineering clinicfall 08 sprin

14、g 09 meeting minutes micron engineering clinicfall 08 spring 09 bibliography b. k. casper, m. haycock, and r. mooney, “an accurate and efficient analysis method for multi-gb/s chip- to-chip signaling scheme”, in digest of technical papers from the ieee symposium on vlsi circuits, june 2002, pp. 5457

15、. b. k. casper , g. balamurugan, j. e. jaussi, j. kennedy, m. mansuri, “future microprocessor interfaces: analysis, design and optimization”, in proceedings of the ieee custom integrated circuits conference, sept. 2007, pp. 479-486. p. k. hanumolu, b. k. casper, r. mooney, g. y. wei, and u. k. moon,

16、 “jitter in high-speed serial and parallel links”, in proceedings of the ieee international symposium on circuits and systems, may 2004, pp. 425428. pavan kumar hanumolu, bryan casper, randy mooney, gu-yeon wei, and un-ku moon, “analysis of pll clock jitter in high-speed serial links”, in ieee trans

17、actions on circuits and systems, november 2003, pp.879-886 micron engineering clinicfall 08 spring 09 questions? micron engineering clinicfall 08 spring 09 worst case verification of high speed interconnects bryson kent technical lead electrical engineer b micron engineering clinicfall 08 spring 09

18、introduction what is worst case analysis why is the worst case important how to calculate the worst case what are the results conclusion and implementation micron engineering clinicfall 08 spring 09 worst case analysis summation of all negative effects good representation of what can happen if certa

19、in conditions arise verification of error free transmission classic analysis of 1 trillion bits (1*1012 bits) * (10-6sec) = over 10 days micron engineering clinicfall 08 spring 09 worst case eye diagram wc1 wc2 wc3 voltage vs one period of time distortion sources add to close the eye from the eye di

20、agram we can calculate a system pass fail tim hollis, micron senior project proposal pass/fail micron engineering clinicfall 08 spring 09 inter-symbol interference inter-symbol interference is the main source of interference data dependent jitter and co-channel interference add to signal degradation

21、 tim hollis, micron senior project proposal micron engineering clinicfall 08 spring 09 c(t) = transmitter symbol response p(t) = impulse response of the channel worst-case computation worst case eye diagram due to inter-symbol interference worst case eye diagram due to inter-symbol interference and

22、cochannel interference j. g. proakis, “digital communication”, mcgraw-hill, 3rd ed., 1995. b. k. casper, m. haycock, and r. mooney, “an accurate and efficient analysis method for multi-gb/s chip-to-chip signaling schemes”, in digest of technical papers from the ieee symposium on vlsi circuits, june

23、2002, pp. 5457. micron engineering clinicfall 08 spring 09 89110911291 14911691 18912091 22912491 26912891 3091 0 0.5 1 picoseconds voltage worst case eye due to inter-symbol interference data-rate = 10 gb/s -100-50050100 0 0.5 1 isipp = 815.9948 mv ddjpp = 34.3117 ps picoseconds voltage pulse respo

24、nse pre-cursor cursor post-cursor calculating the eye diagram 89110911291 14911691 18912091 22912491 26912891 3091 0 0.5 1 picoseconds voltage worst case eye due to inter-symbol interference data-rate = 10 gb/s -100-50050100 0 0.5 1 isipp = 815.9948 mv ddjpp = 34.3117 ps picoseconds voltage pulse re

25、sponse pre-cursor cursor post-cursor 89110911291 14911691 18912091 22912491 26912891 3091 0 0.5 1 picoseconds voltage worst case eye due to inter-symbol interference data-rate = 10 gb/s -100-50050100 0 0.5 1 isipp = 815.9948 mv ddjpp = 34.3117 ps picoseconds voltage pulse response pre-cursor cursor

26、post-cursor 89110911291 14911691 18912091 22912491 26912891 3091 0 0.5 1 picoseconds voltage worst case eye due to inter-symbol interference data-rate = 10 gb/s -100-50050100 0 0.5 1 isipp = 815.9948 mv ddjpp = 34.3117 ps picoseconds voltage pulse response pre-cursor cursor post-cursor 89110911291 1

27、4911691 18912091 22912491 26912891 3091 0 0.5 1 picoseconds voltage worst case eye due to inter-symbol interference data-rate = 10 gb/s -100-50050100 0 0.5 1 isipp = 815.9948 mv ddjpp = 34.3117 ps picoseconds voltage pulse response pre-cursor cursor post-cursor 89110911291 14911691 18912091 22912491

28、 26912891 3091 0 0.5 1 picoseconds voltage worst case eye due to inter-symbol interference data-rate = 10 gb/s -100-50050100 0 0.5 1 isipp = 815.9948 mv ddjpp = 34.3117 ps picoseconds voltage pulse response pre-cursor cursor post-cursor 89110911291 14911691 18912091 22912491 26912891 3091 0 0.5 1 pi

29、coseconds voltage worst case eye due to inter-symbol interference data-rate = 10 gb/s -100-50050100 0 0.5 1 isipp = 815.9948 mv ddjpp = 34.3117 ps picoseconds voltage pulse response pre-cursor cursor post-cursor tim hollis, micron senior project proposal micron engineering clinicfall 08 spring 09 re

30、sults calculated performance vs given performance -1-0.500.51 x 10 -10 0 0.2 0.4 0.6 0.8 1 micron engineering clinicfall 08 spring 09 conclusion worst case analysis is beneficial computation is pulse based analysis user can define and add any distortion as desired results of worst case analysis matc

31、h results of given test case micron engineering clinicfall 08 spring 09 questions? micron engineering clinicfall 08 spring 09 statistical analysis of electrical signaling jordan kemp technical lead electrical engineering micron engineering clinicfall 08 spring 09 introduction how why what summary mi

32、cron engineering clinicfall 08 spring 09 introduction worst case eye good for pass/fail mask, but doesnt give details need for probability of error, rather than rigid “pass/fail” pass/fail maskpass/fail mask micron engineering clinicfall 08 spring 09 use channel impulse response, p(t), and transmitt

33、er symbol response, c(t) introduction ( )( )( )y tc tp t find pdf (probability density function) & cdf (cumulative distribution function) of the channel output micron engineering clinicfall 08 spring 09 introduction shows ber of transmitted data given timing uncertainty (data jitter, clock jitter) a

34、nd voltage uncertainty (vref, rx sensitivity, isi) plot ber eye-diagram as a function of sample time, sample voltage, and probability of error micron engineering clinicfall 08 spring 09 why trends increase: speed, capacity decrease: form-factor, power, cost all above decrease signal integrity theore

35、tically impossible to send error- free data # errors # transmitted bits bit error rate (ber) = ( )0110101001101000 ( )0110101 011010001 x t y t certain number of errors per number of bits sent specified by user/system usually specified below 1 1 trillion would require a 1 trillion bit simulation! mi

36、cron engineering clinicfall 08 spring 09 what probabilistic data eye using channel impulse response, p(t), and transmitter symbol response, c(t) to find the pdf & cdf of the channel output what is a pdf? - probability density function - shows the probability that a specific value is likely to happen

37、 - integrates to 1 what is a cdf? - cumulative distribution function - shows the probability is less than or equal to a specific value - integral of the pdf micron engineering clinicfall 08 spring 09 how (1) 1st way (from *casper paper): recursively convolve 1ui sample recursively convolve 1ui sampl

38、e terms assuming equal probability terms assuming equal probability of a transmitted 0 or 1of a transmitted 0 or 1 0 -0.01 0 -0.01 0 0.59 0 0.59 0 -0.070 -0.07 0 0.0150 0.015 0 0.0550 0.055 0 0.20 0.2 0 0.50 0.5 0 00 0 0 0.70 0.7(each step scaled by to account for p(0) = p(1) )(each step scaled by t

39、o account for p(0) = p(1) ) 0.590.59 0.00.0 1ui1ui y(t) micron engineering clinicfall 08 spring 09 how (1) very hardware intensive (must compute multiple convolutions) must maintain certain amount of resolution, slowing computations down even more very quickly run out of memory performing calculatio

40、ns problems: 0 -0.01 0 -0.01 0 0.59 0 0.59 0 -0.07 0 -0.07 0 0.015 0 0.015 0 0.055 0 0.055 0 0.2 0 0.2 0 0.5 0 0.5 0 0 0 0 . . . . . . . . 1.452 1 (no decimal resolution) 1.452 1.4 (one decimal resolution) 1.452 1.45 (two decimal resolution) delta function 1: 0 1 0 0 1 2 . . . delta function 1.4: 0

41、0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 1.1 1.2 1.3 1.4 1.5 . . . micron engineering clinicfall 08 spring 09 how (2) implemented method (modified from previous): 0 -0.01 + 0 0.590 -0.01 + 0 0.59 + + 0 -0.070 -0.07 + + 0 0.0150 0.015 + + 0 0.0550 0.055 + + 0 0.20 0.2 + + 0 0.50 0

42、.5 + + 0 00 0 + + 0 0.70 0.7 same points taken as before, but add instead of convolve keeps track of locations of delta functions keeps track of heights of each delta plot locations versus heights micron engineering clinicfall 08 spring 09 how (2) -.01 0 height = 0.5 0.59 height = 0.5 height = 0.25

43、-.01 0 .58 .59 0 -0.01 0 0.59 = 0 -.01 (0+.59) (-.01+.59) = -.01 0 .58 .59 instead of convolving, proposed method adds & concatenates micron engineering clinicfall 08 spring 09 advantages: very quick easy to implement infinite precision (in theory) how (2) micron engineering clinicfall 08 spring 09

44、summary use channel impulse response, p(t), and transmitter symbol response, c(t) ( )( )( )y tc tp t shows ber of transmitted data given timing uncertainty (data jitter, clock jitter) and voltage uncertainty (vref, rx sensitivity, isi) find pdf (probability density function) & cdf (cumulative distri

45、bution function) of the channel plot ber eye-diagram as a function of sample time, sample voltage, and probability of error micron engineering clinicfall 08 spring 09 questions? micron engineering clinicfall 08 spring 09 modeling jitter in chip-to-chip communication m. lucas loero technical lead ele

46、ctrical engineer l micron engineering clinicfall 08 spring 09 presentation objectives defining jitter problems caused by jitter modeling jitter receiver jitter transmitter jitter total jitter micron engineering clinicfall 08 spring 09 defining jitter time voltage ideal edge locatio

47、n edge location shifted ideal edge location jitter micron engineering clinicfall 08 spring 09 problems caused by jitter power supply and environment noise causes jitter. jitter can lead to: time uncertainty suboptimal sampling time reduce noise margin micron engineering clinicfall 08 spring 09 probl

48、ems caused by jitter micron engineering clinicfall 08 spring 09 model the effects of jitter in high-speed serial links serial links are used for high-speed chip-to-chip communications modeling jitter micron engineering clinicfall 08 spring 09 modeling jitter serial links transmitter generates a trai

49、n of pulses transmitter clock sampler decision circuit micron engineering clinicfall 08 spring 09 modeling jitter traditional approach to modeling jitter there is two main problems with this approached first, simulation time second, difficulty simulating serial links micron engineering clinicfall 08

50、 spring 09 receiver jitter independent and identically distributed transmitted bits step response channel impulse response jitter sequence modeling jitter micron engineering clinicfall 08 spring 09 calculated eye diagramsimulated eye diagram modeling jitter micron engineering clinicfall 08 spring 09

51、 worst-case receiver jitter modeling jitter micron engineering clinicfall 08 spring 09 worst-case isiworst-case receiver jitter modeling jitter worst-case receiver eye time voltage voltage time micron engineering clinicfall 08 spring 09 modeling jitter worst-case receiver eye micron engineering clin

52、icfall 08 spring 09 transmitted bits step responsechannel impulse response jitter sequence modeling jitter transmitter jitter micron engineering clinicfall 08 spring 09 modeling jitter worst-case transmitter jitter micron engineering clinicfall 08 spring 09 worst-case isiworst-case transmitter jitte

53、r modeling jitter worst-case transmitter eye time time voltagevoltage micron engineering clinicfall 08 spring 09 modeling jitter worst-case transmitter eye micron engineering clinicfall 08 spring 09 modeling jitter total jitter transmitter jitterreceiver jitter micron engineering clinicfall 08 sprin

54、g 09 modeling jitter worst-case total jitter micron engineering clinicfall 08 spring 09 worst-case isiworst-case total jitter modeling jitter worst-case total eye time time voltage voltage micron engineering clinicfall 08 spring 09 modeling jitter worst-case total eye micron engineering clinicfall 0

55、8 spring 09 conclusion voltage time micron engineering clinicfall 08 spring 09 questions? micron engineering clinicfall 08 spring 09 project software engineering, development, and results ben meakin team lead computer engineer micron engineering clinic spring 09 micron engineering

56、clinicfall 08 spring 09 introduction open eye : a formal verification tool for high speed electrical signaling what is open eye? provides software infrastructure required to deliver a useful verification platform based on existing state-of-the-art methods presentation outline requirements and object

57、ives software architecture graphical user interface (gui) software infrastructure use case summary and conclusions micron engineering clinicfall 08 spring 09 product requirements cross platform (windows, linux, etc) graphical user interface no proprietary software license (i.e. matlab license) corre

58、ctly implement worst-case and statistical link analysis pass fail mask and data eye visualization selectable transmitter and receiver jitter multiple sources of co channel interference spice data file input micron engineering clinicfall 08 spring 09 product release status featurephasestatus worst-ca

59、se analysisrelease statistical analysistesting25% static analysisrelease basic plottingrelease advanced plottingdevelopment50% jitter modelingdebug75% co-channel interferencetesting75% file inputrelease file outputdevelopment75% graphical user interfacetesting50% multi-threadingtesting75% documentationdevelopment50% micron engineering clinicfall 08 spring 09 design goals portability extensibility usability view micron engineering clinicfall 08 spring 09 object oriented design model data io signal analysis parameters static data statistical data worst-case data range math utility signal modify