MtronPTIS Oscillator Jitter Basics HAT IS IT TE  OW IS JITTER M ASU  EC TI ON  IT TE R EF IN IT ION An i eal clock signal at

MtronPTIS Oscillator Jitter Basics HAT IS IT TE OW IS JITTER M ASU EC TI ON IT TE R EF IN IT ION An i eal clock signal at - Description

52 M z will have a pe riod of 6430 pic sec onds for one c lete cycle Successi ve c cles of a no is free wa ve form will measure exactly 6430 picos ec onds The noise elements that will cause the cloc k pe riod to va ry from 6430 p S are kn own as jitt ID: 30075 Download Pdf

137K - views

MtronPTIS Oscillator Jitter Basics HAT IS IT TE OW IS JITTER M ASU EC TI ON IT TE R EF IN IT ION An i eal clock signal at

52 M z will have a pe riod of 6430 pic sec onds for one c lete cycle Successi ve c cles of a no is free wa ve form will measure exactly 6430 picos ec onds The noise elements that will cause the cloc k pe riod to va ry from 6430 p S are kn own as jitt

Similar presentations


Tags : will have
Download Pdf

MtronPTIS Oscillator Jitter Basics HAT IS IT TE OW IS JITTER M ASU EC TI ON IT TE R EF IN IT ION An i eal clock signal at




Download Pdf - The PPT/PDF document "MtronPTIS Oscillator Jitter Basics HAT I..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.



Presentation on theme: "MtronPTIS Oscillator Jitter Basics HAT IS IT TE OW IS JITTER M ASU EC TI ON IT TE R EF IN IT ION An i eal clock signal at"— Presentation transcript:


Page 1
MtronPTIS Oscillator Jitter Basics HAT IS IT TE ? OW IS JITTER M ASU ? EC TI ON 1 IT TE R EF IN IT ION An i eal clock signal at 155.52 M z will have a pe riod of 6430 pic -sec onds for one c lete cycle. Successi ve c cles of a no is free wa ve form will measure exactly 6430 pico-s ec onds. The noise elements that will cause the cloc k pe riod to va ry from 6430 p S are kn own as jitter. Jitter is c ed both d term in isti c (fr sp ecific cau ses lik e po wer su pp ly no ise), and ran co en t. Th e rando tion o Jitter can b ch aracterized u ing Gau ssian stri bu tio

statistics. For exam ple, th e hun dr ed su ccessiv measu em en ts th e riod of a 1 52 MHz clo o cillato r th at con ains o y Ran dom Jit er ele ts wou appear as a Gau ssi n di st ri ( o e peak ). Pl us or m nus one st an dar vi at on oul co nt ai n 68 .2 o al pe ri m easurem ent dat p +/ - 2 si gm a wo ul d c nt ai n % o al m asurem ent t ke n. +/ - 3 si gm a wo ul d c nt ai n . +/ - 4 si gm a wo ul d c nt ai n +/ - 7 si gm a wo ul d c nt ai n (1 1e % of al m easurem ent t ken A Ga ussi di st ri but n wi a l rger st rd de vi at on wo ul d ha ve wi der t , m re peri od m easurem ent furt he r

fr om t e an (b ot h sm al r and l rge T e di st ce f t e sm al st peri od easurem ent to the largest period measurem ent is known as the Peak-to Peak Jitter lev l in p co -secon ds. Th e eak-p eak valu e is d en t on th e samp le size. Larger sam le em en ts of the sam u it u nder test will yield larg er ak -p eak v es. Th e e of St an rd Dev atio n (1 -si ) is on ly v lid in pu re Gau ssia n stribu tio ns. If an y Determ istic Jitter (h as sp ecific cau es as ou tlin ed low) ex ists i th e stribu ti , th e e of si gm a based on th e en tire Jitter h sto ram fo r th e esti o ility ccurren ce

is in lid . If Determin istic Ji er ex ists in th e sig l, th an th e stog ram o period m easu emen ts will com rise ltip p ak s. It is li ly th at a domin an t con ri bu tor to Jitter such as we s ppl ripp le is m latin th e clo to bring abou t m ltip le p eak s. Or, th e cillato r si al co co ain lo ts su b-h rm c er. To tal Jitter is a co in atio n of Rando Jitte r in teracti g wi th Determ in isti c Jitter. EC TI ON 2 OURC ES O ITT Ran Jit er co s fro man sou ces. Th erm l v at io sem co tor crystal stru cture cau se ility to v ry d ng pon th e in st an tan s te mp erat e of the terial.

Ano r source o Rand Jitter co s fro th e i rfection du e to semico tor p cess v riatio su ch as non -un form d opin ity. Mu ltip le Ran Jitter so ces add in an R S fash ion bu t a eak-to-p ak valu e is eed ed wh en add ng R Jitter to Determ in istic Jitter to ob tain To tal Jitter, ak to p ak . Determ in istic Jitter is created id en tifiable in terferen e sig ls. It is always bou nded in am litu and s sp ecific cau se s (no random . Determ inistic Jitter can be ge nera ted by cros s talk betwee n adjacent signal traces. This happe s when inc em ental inductance from one conduc or converts

induce m gnetic field from a adjacent signal line into induce d curre nt . This induce cu rren t i creases or ecreases th e ltage, t cau ing Jitter. EMI Rad ation can cau e term in ist c Jitter. A sen itiv e si l th can be affected th e m tic field fro an EMI sou ce . E I sou ces i cl po we r su s, AC po we r l es, a nd R -si gnal s urces . Li ke cr oss t a noi se c rre i i nduce on e t g si gn al th , wh ich th ereb y m du lates th e tim in sign al vo ltag level. Determ in istic Jitter can b g rated b n se in p wer layers o a m lt layer su bstrat e. Th is n se can ch an thresho vo tag s i lo

c tes. Or, a ch ang i und referen e at th e thresh ltag will resu lt in ch an th e req red vo ltag to switch th e g te. Determ in istic J itter can app ar wh en m ltip g tes switch to th e sam lo c state sim an eou ly . Curren sp es can ind ed o po we r a g pl ane , c ng a not her o ni fo r t res ol d ge l vel shi .
Page 2
EC TI ON 3 OW TRON PTI M EAS RES IT TE ron TI uses t easure ent sy st em s from W vec est p ase noi se st em s, and di gi scope s t easure bot h t -dom ai n an freq y-do main Jitter. Wh il e m easu en t tech will b ie fly d scu ssed in th is tu to rial, Wav crest h

s written a nu er of ap lication no s on th e unda als o Jit er m easu ts u g th eir eq en t. Their w ite is lo cated at ww.w ecr est. com. Two rticu ar app licatio notes to start with in clud e: DTS ME AS UREMENT T CHNIQUE (Getting Starte d) JITTER ANALYSIS (Getting Sta ted). The DT S m eas ures the tim e b tween two events . Inte rnall ny tim measurem ents (s ples) are com iled into a histogram (Figur e 1 directly below). Basic statistics such as s le size, an, eak -t pe ak, a nd 1-si gm a can be vi o t e f panel T is d ta th e o cillato r ou tpu si al can liv ered ov er GPIB to Wav ecrest

Virtu In stru s so ft ware, API so ft ware t ers cust om soft wa re. Dat fr om hu re ds or t ous an ds of hi st og ram can be com d i pl ot s, h hi ghl ht m re i rm at on ab out out put signal quality. Specific values of Jitter accum lation suc as Jitter freque cy and power, and magnitude of De terminist and Ran Jitter can sp lay in th ese p s. FIGURE 1 Th e DTS m es tim e b een two ev en ts, wh ich are th resh cr ssin . A ital sam lin o cillo sco m eas es th e vo ltages with re spect t tim relative to a tri gge r. Event base d m easurem ents allow the DT S to determ ine actual edge placement

to withi 800 fem o-secon s (ha e res on o D S) . Sam ng sc ope s i rp ol at e dat bet ee sam d poi nt s t det rm e e t of t he threshol d crossing. Because scope s depe nd upon the use of a trigger si gnal, any Jitter o this wave form could pote ally sk out a itter con ribu r. On e p rt icu ar d ital sco fro an in du stry-lead ng supp lier listed 1.5 as th e rm s Jitter erro r for th eir easurem ent set . The D S u es asy no us ra nd sam li ng of ev en ts to estab lish a v lid statistica d strib tion of ev ent ti s. It g th ers sam les ev ery 2 to 25 uS in en t o th e DUT o ratin frequ en cy.

By rand izin th e acq sition ti th ere is no ch an ce m sk in g ou t a Jitter si gn al th at cou match a sam li rate.
Page 3
Si C cl e t Cy cl e d f 4 Os ci ur ed usi ng cr D S- 20 75 20 40 60 80 10 12 15 22 29 36 43 50 57 64 71 78 85 92 99 106 113 120 127 134 141 148 155 162 169 176 183 190 197 204 211 218 225 232 239 246 253 of cy cl es tte ( 155 52 M P L V nd or X 1 s gm a = 75 pS 55. M P L V ndo 1 si gm a = pS 4 M fu nd am 5. 7 pS 15 5. 52 A 1si 9. 2 pS FIGURE 2 Th e DTS m eas em en t system in clu s m software t s to enh ce o s u rstan o wav form ch aracteristics. Th e Jitter

an alysis to al lo ws t e er to ew Jitter m du latio n. For ex am e, th e er can m eas e hu ndreds or tho sand s of stog ram each con ng of 30 0 sam easurem ent o a wa vef rm s peri od . The st anda rd devi at on f om al of t e sam le gr ou ps ca n be plotted to show accum late d Jitter over a designate d time period (Fig ure 2 de picted directly above) . Other critical waveform ram ters su ch as rise tim e, fall ti , p op ag atio n d la ys, and freque ncy can be am in ed in th is fash io n. The t e dom ai n dat can b con ect ed t t e freq cy dom ai n by usi ng a FFT t det rm e t e fre que

ncy an d am pl ude of t Jitter com pone nts. Adva nce s ftwa re tools beyond t e scope of this discussion are a so available from W vecrest. The Tail-fi TM al go ri enabl s t e use anal ze odal di st ri but ns (m ore t an o e pea In no n- ussi an di st ri , Ga ussi an assu tio ap y to th e tai s (furth st left an righ t regio ) wh en it is p ssi e to calcu late th e eq len 1-sigma o th se tail regi ons . I t way Tai TM llo ws for th e calcu lation prob ab ilistic o ccurren e o ou tlyin g measu em en ts fo r no n-Gau sian str bu tio ns. Thu , a pred ictio of eak-to-p eak Jitter for a g n wav fo

rm can app ox im ed (with v ry h co nfid en ce) a llio cycles aft wer-u witho waitin g fo r on e illio si gn al p riod s t co lete. Ano r adv nced software featu e i vo es th e calcu lation in teg ated j itter wh en th e u er sp ecifies a frequ y b wid s uc h a e SO NET ra nge of 12 z t 20 M z. D pen g u t e num ber o sa le easurem ents specified in th e set-up , th e in teg ated itter test wou tak b een 0 and m es p r it. EC TI ON 4 YP ES OF JI TT ER SP EC IED F R OS CI LLA TO RS Measurem ents in the tim e domain Cycle to cycle Jitter m easu es th e fference in riod between ad acen cycles.

Peak t eak cy cle Jitter is the wi of the distribution of peri od m easure ents from s llest value to largest val Peak-to-pe is obtaine d from a histogra of success ive edge m easure ents. The more sam le asurem ents tha are taken, t e larg er t e di ffe rence betwe n the sm alles and the large t peri od dat . The di st ri but n sam peri od m easurem ent has a R S or st an dar de vi at on (1 -si ) asso ci at ed wi it. The 1-si gm a l vel o a di st ri on pe ri easurem ent can be use to calculate bit-error rate. T e expecte BER is define d as t e ob ab ility o cau sing a b it error wh en t e To

tal Jitter (Rand p Determ in ist c), p ak to p ak , ex ceed t e Jitter bu dg et.
Page 4
Listed b w is a tab e Rand Jitter (pk to -p k) rsu BER: PROBAB ILI Y OF D TA RROR PEA PE AK = N*RM 10 -10 Pk -Pk = .7*RMS 10 -11 Pk -Pk = .4*RMS 10 -12 Pk -Pk = .1*RMS ( /- si gma) 10 -13 Pk -Pk = .7*RMS 10 -14 Pk -Pk = .3*RMS Measurem ents in the fre quenc dom ain In teg ated Jitter ov er a sp eci fied b nd wi can b ob tain ed fro Mtro TIs DTS ti easu em en t syste . Fo r ex am e, SONET stan rds req ire j itter p rfo man ce with in t e frequ y ban 12 KHz to 20 MHz offset fro th e carrier frequ

en cy. Dep nd in g pon th e nu er sam les measu ed, th e Wav crest syst will calcu late th is v e in to 90 second s r un it. In teg ated Jitter can also ob tain ed fro a p se no ise m easu em en t syste o r a ndwid th. Dep nd ing pon th e nu er of avera es withi eac h fre que ncy offset ba nd, t e m easure ent coul d t ke as l a f m nut es per u Ea ch m easurem ent echni e ha i advant es and di sa nt ages . Sup nt al di scussi s wi ron TI t hni cal su pp can revi ew t ese easurem ent t echni que di ffe nces i m re det On a rticu ar 1 52 MHz, PECL o cillator produ ct, th e In tegrat ed Jitter

was i th e freq en cy b nd of KHz to M . Thi sam DU T ha d a cy cl e-t -cy 1-si gm a l vel of 5 pS a m easure i t t -d om ai n. T e pe ak t peak i t e t e dom ai n was 45.8 pS. Ev en tho ugh th e In tegrated Ph ase Jitter loo tter ecau e it is a s al ler n r, t e ti d in measu em en ts co mm cate t he sam in fo rm ati n abou t th e cillato r wav form n se sign at e. SECTION 5 CON ERS OF PH ASE NO ISE D TO I TE GR ATE P ASE JITT ER: Th ere are m p rs th at describ how to d riv eq tion to exp ess RMS Jitter in t rm s o Ph ase No ise. at we will sh ow y ou is a practical application tha was de

veloped for a Microsoft E cel sp reads eet . T i a t ool t can be use by any e ngi ne er in terested in an alyzin Ph ase No ise ta. We start ou t with th e well-estab lish equ tio t calculate RMS Jitte r in seco nd s ro m pha se noi se. Eq.1) Eq.2) Eq.3) fo = The ca rrier f eque ncy in Hz 1, 2 = The offset fre quency range of i teres in Hz ) = Phas e noi se i radi / Hz dBc = Phase noise power relativ e to th e carrier (f) = ectra l Den ity at a rticular of fset fre que ncy f in Hz, (see Eq .2 .) is related p se n se power Eq .3 . Sub titu tin g we ar ri ve at ou r fi nal defi ni on f r S (f )

i E . Eq.4) To a pr oxi te t e i gral we use t e s on o t oi ds Eq .5 . Eq.5) Phase Noise test syste s plot frequency al on g a logarithm c scale, at least 3 points should be picked i each decade. For exam pl e t get eve s ced poi nt s we cho e t use a 10 , 20 , 50 , 10 2 et c. m . M re e ual be r cor el at
Page 5
All that is left is to take the square ro ot of the summation and the di vide by (2*PI* Fo). We ha ve desc ribed all the pieces nec essary to mak th is a si le task to im en t in to a spreadsh eet or co th is in a BASIC C og ram Here are resu lts listed b w co arin g th is

m th with t e Ae ro flex PN900 0 Ph ase No ise Test Set ing = 16 po in ts fro 10 Hz t 1 M z. 46 .7 20 MHz VCXO 161 .0 00 MHz clo m ltip lier Ran ge: rm s noise (p s) Meas % Diff. 10Hz-1M Hz 2.93 100 Hz-1M z 0.59 1KHz-1M z 0.22 10KHz-1MHz 0.20 Ran ge: rm s noise (p s) Meas % Diff. 10Hz-1M Hz 15.16 100 Hz-1M z 12.65 1KHz-1M z 12.53 10KHz-1MHz 12.52 EC TI ON 6 ORR LATION OF SCILLATO R IT TE R F OM UPPL IE R T USTOM Reg rd less wh at typ of Ji tter is sp ecified , t e m o cil ato Jitter lev l th at wo cau e th e en d produ ct to fail it s syste requ irem en ts sh ou ld d termin ed . In m st cases

th e o cillato r Jitter c ribu tio to rall syste Jitter is 2 o less. To he lp cillato r co mp an ies sim lat op eration in th e cu sto ers bo ard informatio abo noise-ripp le in th e cillato r su pp ly sh ou ld be to th e su pp lier. Th en, th e Jitter asu em en ts can b p rfo with and wi th ou t su pp ly-lin e no ise in ect io so th at the cust om er bet r un der s t e de vi ce per wi som on- ar d si gnal i rf erence . Mtro nPTI recommen to its cu st er b se th at: 1) . Ti -dom ai n 1 si gm a Ji er a pea -t o- peak l vel be speci fi ed u 5 peri od cy cl es. . In tegrated Ph ase Jitter

over th e SONET ndwid th (1 KHz to 20 MHz) sh ou ld b specified as well. . Ph ase No ise p rform ce at d fferen freq y offsets sho specified to o cillato r su pp liers a co te p ctu e of t cillato r si l n se profile.