開關(guān)型調(diào)節(jié)器的電路板布局技術(shù)(完整版)實(shí)用資料

開關(guān)型調(diào)節(jié)器的電路板布局技術(shù)（完整版）實(shí)用資料(可以直接使用，可編輯完整版實(shí)用資料，歡迎下載）

開關(guān)型調(diào)節(jié)器的電路板布局技術(shù)開關(guān)型調(diào)節(jié)器的電路板布局技術(shù)（完整版）實(shí)用資料(可以直接使用，可編輯完整版實(shí)用資料，歡迎下載）

摘要：本文介紹了開關(guān)型調(diào)節(jié)器的電路板布局的基本原則。盡管本文集中分析的是升壓型開關(guān)型調(diào)節(jié)器，但它所包含的原理同樣適合其它類型的開關(guān)調(diào)節(jié)器。本文討論了接地方法、元器件布局、降低噪聲輻射以及減少寄生電容和電感的重要性。

當(dāng)考慮怎樣才能最好地為開關(guān)電源設(shè)計電路板時，最好首先考慮一下它的最終目的，即提供一個特定數(shù)值的穩(wěn)定電壓。有經(jīng)驗(yàn)的設(shè)計人員會謹(jǐn)慎考慮電路的接地方法，從而獲得穩(wěn)定的電壓。他們知道很難獲得完美的接地方案—因?yàn)檫@不僅僅是接地問題，任何接地工作都會直接影響到電路的性能，設(shè)計人員還要特別注意各種穩(wěn)壓元件的位置。

接地讓沒有經(jīng)驗(yàn)的工程師簡單地畫三條短線表示接地可能是一個誤區(qū)，這個符號會給初學(xué)者一種錯覺，簡單認(rèn)為接地是一種理想情況。如果用一條較長的引線把電路的各種元器件連接到電源或電池的負(fù)端，您可能從直覺上意識到這條地線并非理想的接地。這條引線表明電流通過地層或地線的電阻、電感流回電源，在這個過程中會產(chǎn)生相應(yīng)的壓降。因此，接地回路不會穩(wěn)定在一個理想的穩(wěn)壓值—即通常所說的0V。

圖1所示boost轉(zhuǎn)換器說明了考慮接地的重要性，該調(diào)節(jié)器依靠控制器IC內(nèi)部的基準(zhǔn)電壓和兩個反饋電阻產(chǎn)生特定的輸出電壓。為了獲得正確的反饋從而得到正確的輸出電壓，電壓基準(zhǔn)、電阻分壓器以及輸出電容必須處于同一電位。確切地說，控制器的模擬地引腳(電壓基準(zhǔn)的地)和電阻分壓器的地電位必須與輸出電容的地電位相等。輸出電容接地端的電壓至關(guān)重要，因?yàn)橐蠓€(wěn)壓器提供精確電壓的負(fù)載通常緊靠著輸出電容安裝—這部分地是反饋電壓的參考端。

圖1.升壓型開關(guān)轉(zhuǎn)換器的電路板布局設(shè)計原則同樣適用于其它拓?fù)涞拈_關(guān)型調(diào)節(jié)器

另一原因是控制器需要精確的電壓反饋，為了實(shí)現(xiàn)無抖動的開關(guān)操作，控制器需要在輸出電壓出現(xiàn)任何交流干擾時能夠產(chǎn)生一個準(zhǔn)確的取樣，而這個精確的取樣是通過反饋網(wǎng)絡(luò)得到的。

元件布局除了接地方案，合理的布局穩(wěn)壓元件也很重要。例如，控制器內(nèi)部的電壓基準(zhǔn)必須通過緊靠REF引腳安裝的電容旁路；基準(zhǔn)電壓的噪聲會直接影響輸出電壓。同樣，該旁路電容的地端必須連接到低噪聲的參考地(與控制器的模擬地以及電阻分壓器的地端相連)，遠(yuǎn)離嘈雜的功率地。這個低噪聲參考地和嘈雜的功率地之間的隔離至關(guān)重要。

既然嘈雜的功率地和低噪聲參考地最終還是要連接到一起，為何還必須將二者隔離呢?為了防止較大的開關(guān)電流通過模擬小信號的地回路進(jìn)入電池或電源，這樣的隔離是必需的。一旦出現(xiàn)這種狀況，敏感信號的地回路遭到干擾；較大的開關(guān)電流流經(jīng)地回路的電阻、電感，使地回路的電平沿著路徑發(fā)生變化。

了解嘈雜的功率電路有助于找尋將它與其余電路隔離開的最佳方法。圖2描述了調(diào)節(jié)器的功率電路，包括兩條電流路徑：當(dāng)MOSFET導(dǎo)通時，電流流過輸入回路；當(dāng)MOSFET斷開時，電流流過輸出回路。將這兩個環(huán)路的元件相互靠近布局，可以把大電流限制在調(diào)節(jié)器的功率電路部分(遠(yuǎn)離低噪聲元件的地回路)。CIN、L1和Q1必須相互靠近放置，CIN、L1、D1和COUT也必須相互靠近。圖2特別指明了這兩個環(huán)路以及需要靠近安裝的元件。

圖2.在為本圖所示的兩個電流環(huán)路的元件布板時，需要特別謹(jǐn)慎。使用短且寬的引線實(shí)現(xiàn)如此密集的布線，可以提高效率，減小振鈴，并可避免干擾低噪電路。

實(shí)際的電路板布局需要一些折中考慮，特別是在為上述兩個大電流環(huán)路布局時。如果需要決定將哪些需要就近安裝的元件真正地實(shí)現(xiàn)就近安裝，擇須確定每個環(huán)路中的哪些元件有不連續(xù)的電流流過。就近安裝元件可以最大限度地減少寄生電感，而這些具有不連續(xù)電流的元件位置對于減少寄生電感非常重要。請參考將寄生電容和寄生電感減至最小。

其它考慮事項(xiàng)不管是采用電池還是電源為升壓型開關(guān)調(diào)節(jié)器供電，電源阻抗都不為零。這意味著當(dāng)調(diào)節(jié)器從電源汲取快速變化的電流時，電源的電壓將發(fā)生變化。為了改善這種效應(yīng)，電路設(shè)計人員在靠近上述兩個功率環(huán)路的位置安裝了輸入旁路電容(有時使用兩個電容：一個陶瓷電容與一個有極性電容并聯(lián))。這一舉措并非為了保持功率電路的電源穩(wěn)定—即使電源電壓發(fā)生變化，功率電路也能很好工作。然而，將旁路電容靠近功率電路安裝可以限制大電流注入功率電路，避免對低噪電路的干擾。

干擾是如何產(chǎn)生的呢?有三個途徑：首先，如上所述，如果功率電路的接地返回電流流經(jīng)調(diào)節(jié)器模擬電路的部分地回路或全部地回路，由于地回路的寄生電阻、電感，該電流將在這部分地通道上產(chǎn)生開關(guān)噪聲。地回路的噪聲會降低穩(wěn)壓輸出精度，這個電流還可能干擾同一電路板的其它敏感電路。其次，與地回路類似，電池或電源正端的開關(guān)噪聲還可能耦合至用同一電源供電的其它元件。包括控制芯片，使基準(zhǔn)電壓發(fā)生抖動。若輸入旁路電容兩端的電壓不穩(wěn)定，在控制器的電源引腳前加一級R/C濾波器有助于穩(wěn)定其供電電壓。最后，交流電流流經(jīng)的環(huán)路面積越大，所產(chǎn)生的磁場也越強(qiáng)，產(chǎn)生干擾的幾率也大大增加。將輸入旁路電容緊靠功率電路安裝可以縮小環(huán)路面積，從而降低產(chǎn)生干擾的可能性。

如果輸出端的兩個分壓電阻布局不合理，噪聲也會引發(fā)其它問題。將這兩個電阻靠近控制器的FB引腳放置，可以保證得到一個對噪聲相對不敏感的電壓反饋控制環(huán)路。這種布局可以使電阻分壓器中點(diǎn)至開關(guān)調(diào)節(jié)器的FB引腳的引線最短。這是非常必要的，因?yàn)殡娮璺謮浩髦悬c(diǎn)和控制器FB引腳的內(nèi)部比較器輸入都為高阻抗，連接二者的引線易于耦合(主要通過容性耦合)開關(guān)調(diào)節(jié)器的噪聲。當(dāng)然，必要的話，可以考慮延長電阻分壓器與輸出端相連的引線，以及電阻分壓器與輸出電容地端相連的引線，開關(guān)型調(diào)節(jié)器的低輸出阻抗可抑制這些引線上的耦合噪聲。

將寄生電容和寄生電感減至最小找出圖1電路中電壓發(fā)生快速變化的節(jié)點(diǎn)，也就找出了需要將寄生電容減至最小的位置，這是因?yàn)殡娙輧啥说碾妷翰荒苘S變。在該電路中僅有一個這樣的節(jié)點(diǎn)，即由功率電感、二極管和MOSFET連接形成的節(jié)點(diǎn)。開關(guān)導(dǎo)通時，該節(jié)點(diǎn)的電壓接近地電位；開關(guān)關(guān)閉時，該節(jié)點(diǎn)電壓攀升至比輸出電壓高出一個二極管壓降的電平。須確保電路板的走線使該節(jié)點(diǎn)的寄生電容最小，若寄生電容減緩了該節(jié)點(diǎn)的電壓瞬變，調(diào)節(jié)器的效率將受到一定損失。保持該節(jié)點(diǎn)較小的尺寸不但有助于減小寄生電容，還可降低EMI輻射。不能犧牲布線寬度來縮小該節(jié)點(diǎn)的尺寸，相反，應(yīng)該采用短而寬的走線。

找出具有快速變化電流的分支，也就找到了需要將寄生電感減至最小的支路。電感電流不能發(fā)生躍變，當(dāng)電感電流快速變化時，電感兩端的電壓將產(chǎn)生毛刺和振鈴，從而導(dǎo)致潛在的EMI問題。而且，該振鈴電壓的幅度有可能非常高，以至于損壞電路元件。

圖3顯示了電路的三個支路電流波形，電流I1不會產(chǎn)生問題，因?yàn)樗韵鄬ζ骄彽姆绞阶兓?，另外，該支路已?jīng)具備了一個大電感，也就是L1。與MOSFET串聯(lián)的寄生電感則會產(chǎn)生問題，因?yàn)殡娏鱅3有突變。該串聯(lián)電感包括I3至CIN地端返回路徑的任何感抗：Q1引腳的寄生電感以及地回路自身的電感。注意流經(jīng)CIN的電流并未躍變，而是和電感電流(I1)的交流部分相等(電池提供其直流部分)。MOSFET關(guān)閉時，環(huán)路的一部分同樣有快速變化的電流流過。該電流I2流過D1和COUT以及地回路的覆銅部分，因此，這些元件和地回路的寄生電感必須減至最小。

圖3.開關(guān)調(diào)節(jié)器各個支路的電流波形，由此可以看出哪些部分需要將寄生電容減至最小。電流快速變化(也就是I2和I3)的支路要求感抗最小。

當(dāng)考慮負(fù)載通路上的感抗是否會造成問題時，應(yīng)注意到輸出電容具有較大容值，而且具有很低的ESR，電容兩端的電壓保持相對穩(wěn)定。這意味著流過負(fù)載的電流不會變化太大，因此其等效串聯(lián)電感并不重要—除非負(fù)載本身動態(tài)變化。

創(chuàng)建切實(shí)可行的電路板布局有很多種方法可以處理開關(guān)電源的接地，一種方法是為所有的接地電路提供一個單獨(dú)的地層—這種方法可能不會運(yùn)行在很好的狀態(tài)下。采用這種方法時，電路的功率地電流可能流經(jīng)電阻分壓器、控制器特定引腳的旁路電容以及控制器的模擬地或是這三者的地回路，從而造成它們的地電位抖動。

也許最好的方法是創(chuàng)建兩個單獨(dú)的地層—一個用于功率電路，另一個用于調(diào)節(jié)器的低噪聲模擬電路。參考圖4a，功率地包括輸入和輸出電容的地端以及MOSFET的源極，這些連線必須采用短而寬的引線，確保功率電路的地線最寬、最短，可以降低感抗、提高效率。

模擬地部分為控制器的模擬地引腳、電阻分壓器的地端和控制器任何特定引腳的旁路電容(輸入旁路電容CIN除外)的地端。該模擬地不必是一個平面，可以使用較寬的長引線，因?yàn)槠潆娏鞣浅Ｎ⑷醪⑶蚁鄬Ψ€(wěn)定，引線電阻和電感不再是重要因素。

按照圖4a所示連接控制器的AGND引腳和PGND引腳，在這些引腳之間連接兩個地可以確保模擬地內(nèi)沒有開關(guān)電流，AGND、PGND之間的連線可以相對較窄，幾乎沒有電流流過該路徑。盡管理想情況下AGND可以直接連接到COUT的地端，多數(shù)控制器仍然要求兩個地引腳(AGND和PGND)直接連接(這是因?yàn)镃OUT的地和PGND之間總會存在一定的阻抗，若AGND和COUT的地直接相連，負(fù)載電流在該阻抗上產(chǎn)生的壓降會達(dá)到足以讓AGND和PGND之間的二極管導(dǎo)通電壓，造成嚴(yán)重后果)。在PGND和COUT之間使用短而寬的引線，可以使反饋電阻和控制器內(nèi)部基準(zhǔn)共用相同的地電位，與調(diào)節(jié)器的輸出端的參考地相同。這一點(diǎn)非常重要，因?yàn)檩敵鲭妷菏怯蛇@些元件設(shè)置控制的。

圖4使用不同的模擬和功率地隔離較大的功率地電流與低噪聲模擬地電流，從而保護(hù)低噪電流路徑。

圖4.采用隔離的模擬和功率地隔離較大的功率地電流與低噪聲模擬地電流，從而保護(hù)低噪電流回路。

有時控制器的某些旁路電容既不能連接至模擬地也不能連接至功率地，其中一個例子是使用R/C濾波器旁路升壓開關(guān)調(diào)節(jié)器的V+引腳(如上所述)。這種情況下，該電容接地引腳對于模擬地來說太嘈雜；同時，對于該電容來說功率地的噪聲也太大。必須將這樣的電容地直接返回至AGND和PGND引腳之間的連線(若控制器僅提供一個接地引腳，直接連接至該引腳)。

最后，電路板的層數(shù)在PCB布局中也是一個關(guān)鍵因素。在多層板上，可以使用一個中間層作為屏蔽。屏蔽層允許在電路板的底層放置元件，從而降低干擾的機(jī)會。配合使用屏蔽層時，將功率元件的地穿越屏蔽層連接并非一個好的方法。相反，應(yīng)該將它們連接在一個隔離的、受限制區(qū)域，可以清晰地分辨出這些電流的流向以及它們的影響。

確保功率元件的地位于頂層，這種連接與電路板的層數(shù)無關(guān)；這樣處理可以將其電流限制在已知的路徑內(nèi)，不會干擾其它地回路。若無法實(shí)現(xiàn)這一點(diǎn)，可以通過使用其它電路板層的隔離覆銅區(qū)域和過孔進(jìn)行連接。對于每個接地點(diǎn)，應(yīng)使用多個過孔并聯(lián)以降低電阻和電感。本文來源自《電子產(chǎn)品世界》論壇SIMPLESWITCHER?PCBLayoutGuidelinesINTRODUCTIONOneproblemwithwritinganApplicationNoteonPCBlayoutisthatthepeoplewhoreaditareusuallynottheoneswhoaregoingtouseit.Evenifthedesignerhasstruggledthroughelectromagneticfields,EMC,EMI,boardparasitics,transmissionlineeffects,grounding,etc.,hewillinallprob-abilitythengoonwithhisprimarydesigntask,leavingthelayouttotheCAD/layoutperson.Unfortunately,especiallywhenitcomestoswitchingregulators,itisnotenoughtobeconcernedwithjustbasicrouting/connectivityandmechani-calissues.BoththedesignerandtheCADpersonneedtobeawarethatthedesignofaswitchingpowercon-verterisonlyasgoodasitslayout.Whichprobablyex-plainswhyagreatmanyofcustomercallsreceived,con-cerningswitcherapplications,areultimatelytracedtopoorlayoutpractices.Sadly,thesecouldandshouldhavebeenavoidedontheveryfirstprototypeboard,savingtimeandmoneyonallsides.TheoverallsubjectofPCBdesignisanextremelywideone,embracingseveraltest/mechanical/productionissuesandalsoinsomecasescompliance/regulatoryissues.Thereisalsoacertainamountofphysics/electromagneticsinvolved,ifaclearerunderstandingissought.ButthepurposeofthisApplicationNoteistoreachtheaudiencemostlikelytouseit.Thoughthereisenoughdesigninformationforthemoreexperienceddesigner/CADperson,theApplicationNotein-cludesaquick-setofclearandconcisebasicrulesthatshouldbescrupulouslyfollowedtoavoidamajorityofprob-lems.Inparticular,wehaveprovidedrecommendedstart-ingpointsforlayoutwhenusingthepopularLM267x,LM259xandLM257xfamilies(Figure2Thefocusisonthestep-down(BuckSimpleSwitcherICsfromNational,butthesameprinciplesholdforanytopologyandswitchingpowerapplication.Quick-SetofRulesforSIMPLESWITCHERPCBLayout(BuckaPlacethecatchdiodeandinputcapacitorasshowninFigure2.cConnectviastoaGroundplaneifavailable(optional,marked‘X’inFigure2dIfviasfallundertabofSMTpowerdevice,theseareconsidered‘thermalvias’.Usecorrectdimensionsasdiscussedtoavoidproductionissues.Orplacetheviasclosetobutnotdirectlyunderthetab.eRoutefeedbacktracecorrectlyasdiscussed,awayfromnoisesourcessuchastheinductorandthediode.fDonotincreasewidthofcopperonswitchingnodeinjudiciously.gIfverylargeheatsinkareaisrequiredforcatchdiode(havingestimatedtheheatsinkrequirementcorrectlyuseisolationasdiscussed.hForhigherpowerSMTapplications,use2ozboardforbetterthermalmanagementwithlesscopperarea.SIMPLESWITCHERisaRegisteredTrademarkofNationalSemiconductor.NationalSemiconductorApplicationNote1229SanjayaManiktalaJuly2002SIMPLESWITCHERPCBLayoutGuidelinesAN-1229INTRODUCTION(ContinuedTHEACANDDCCURRENTPATHSReferringtoFigure1a,theboldlinesrepresentsthemain(powercurrentflowintheconverterduringthetimetheswitchisON.AstheswitchturnsON,theedgeoftheofthecurrentwaveformisprovidedlargelybyCBYPASS,there-maindercomingmainlyfromCIN.Someslowercurrentcom-ponentscomefromtheinputDCpowersupply(notshownandalsorefreshtheseinputcaps.Figure1brepresentsthesituationwhentheswitchisOFF.Wecanthereforeseethatincertaintracesections,currenthastostartflowingsud-denlyduringtheinstantofswitchturn-offandinsomesec-tionsitneedstostopflowingequallysuddenly.Figure1c20042601FIGURE1.AN-12292THEACANDDCCURRENTPATHS(Continuedcurrentdoesnotchangeinasteppedfashionandsotheharmoniccontentislower.ItisalsonosurprisethattheDCsectionsarethoseinserieswiththemaininductor,becauseitisknownthatinductorshavethepropertyofpreventingsuddenchangesincurrent(thisisanalogoustoacapacitorwhich‘resists’suddenchangesinvoltage.Summingup:Inswitchingregulatorlayout,itistheACpathsthatareconsideredcritical,whereastheDCpathsarenot.Thatistheonlybasicruletobekeptinmind,andfromwhichalltheothersfollow.Thisisalsotrueforanytopology.PerformananalysisofthecurrentflowforanytopologyinthesamemanneraswedidfortheBuck,tofindthe’differencetraces’:andthesetracesarebydefintionthe’critical’onesforlayout.Whatistheproblemwithstepcurrentchangesanyway?Inaresistorforexample,thiscausesnounexpected/unidentifiableproblem.ThevoltageisgivenbyV=IR,andsoforagivenchangeofcurrent,thevoltagewillchangepro-portionally.Forexample,a0.5cmwideCutraceofthick-ness1.4milhasaresistanceof1milliohmpercmlength(at20degC.Soitseemsthata1inchlongtracewithacurrentchangeoverof1A,wouldproduceachangeinvolt-ageofonly2.5millivoltsacrossthetrace,whichisinsignifi-cantenoughtocausethecontrolsectionsofmostICstomisbehave.Butinfacttheinducedvoltageismuchlarger.TheimportantthingtorealizeisthattracesofcopperonaPCB,thoughbarelyresistive,arealsoinductive.Now,theoft-repeatedthumb-ruleisthat‘everyinchoftracelengthhasaninductanceofabout20nH’.Likethetraceresis-tance,thattoodoesn’tseemmuchatfirstsight.ButitisthisratherminuteinductancewhichisinfactresponsibleforagreatmanycustomercallsinSIMPLESWITCHERapplica-tions!TheequationforvoltageacrossaninductanceisV=L*dl/dt,andsothevoltagedoesnotdependonthecurrentbutontherateofchangeofthecurrent.Thisfactmakesallthediffer-encewhenthe1Achangewespokeaboutoccurswithinaveryshorttime.Theinducedvoltagecanbeveryhigh,evenforsmallinductancesandcurrents,ifthedl/dtishigh.Ahighdl/dteventoccursduringtransitionfromFigure1atoFigure1b(andbackinalltheACtracesections(shownboldinFigure1c.Theinducedvoltagespikeappearsacrosseachaffectedtrace,lastingforthedurationofthecrossover.Togetabetterfeelforthenumbershere:thechangeincurrentintheACsectionsofatypicalbuckconverterisabout1.2timestheloadcurrentduringtheswitchturn-offtransitionandisabout0.8timestheloadcurrentduringtheswitchturn-ontransition(foran’optimally’designedBuckinductor,asperinductordesignguidelinesintherelevantDatasheets/SelectionSoftware.Thetransitiontimeisabout30nsforhighspeedFetswitchersliketheLM267x,andisabout75nsfortheslowerbipolarswitchersliketheLM259xseries.Thisalsoincidentallymeansthatthevoltagespikesinthehigh-speedfamiliescanbemorethantwicethatintheslowerfamilies,foracomparablelayoutandload.Thereforelayoutbecomesallthemorecriticalinhigh-speedswitchers.So,oneinchoftraceswitchingsay1Aofinstantaneouscurrentinatransitiontimeof30nsgives0.7V,ascomparedto2.5mV(thatweestimatedonthebasisofresistancealone.For3A,andtwoinchesoftrace,theinducedvoltage’tries’tobe4V!InFigure1c,thesmalltrianglesalongthesectionsindicatethedirectionofthemomentaryinducedvoltage,astheconverterchangesfromthesituationinFig-ure1atothatinFigure1b(switchturn-off.WecanseethatassumingthatthegroundpinoftheICisthereferencepoint,theswitchingnode(marked‘SW’triestogonegative(allitsseriestracesectionsaddingup.Similarlytheinputpin(marked‘VIN’goeshighthroughseriescontributionsinallitsrelatedsections.Figure1crepresentsthepictureduringtheturn-offtransition.Duringtheturn-ontransitionallthein-ducedvoltagepolaritiesshownaresimplyreversed.Inthatcase,theVINpinisdraggedlow,andtheswitchingnodepinisdraggedhighmomentarily.Theastutedesignerwillrecognizethatthiswastobeex-pectedsinceanyinductance,evenifitisparasitic,demandstobe‘reset’,whichmeansthatthevolt-secondsduringtheon-timemustequalandbeoppositeinsigntothevolt-secondsduringtheoff-time.Thedesignerwillalsoreal-izethattilltheseparasitictraceinductancesreset,theydonot’allow’thecrossovertooccur.Soforexample,traceswhichwerecarryingcurrentpriortoswitchturn-offwill’insist’oncarryingcurrenttillthevoltagespikesforcethemtodootherwise.Similarly,thetraceswhichneedtostartcarryingcurrentwill’refuse’todosotillthespikesacrossthemforcethemtodolikewise.Sinceswitchinglossesareproportionaltocrossovertime,evenifthesevoltagespikesdonotcauseanomalousbehavior,theycandegradeefficiency.Forex-ample,intransformer-basedflybackregulators,whenthetheprimarynumberofturnsismuchlargerthanthesecondaryturns,designersmaybesurprisedtolearnhowmuchthesecondarysidetraceinductancesalonecandegradeeffi-ciency.Thisisbecauseanysecondarysideuncoupled(trace/transformerleakageinductancesreflectintotheprimarysideasanequivalentparasiticinductanceinserieswiththeswitch.Thisaddsanadditionaltermtotheeffectiveleakageasseenbytheswitchthatequalsthesecondaryinductancemultipliedbysquareoftheturnsratio(turnsratiobeingNp/Ns.Thereforethedissipa-tionintheflybackclamp(zener/RCDcanincreasedramati-cally,loweringefficiency.Onelessonhereisthatthough’leakageinductance’(fromtracesorthetransformeriscon-sidered’uncoupled’,inrealityitcanmakeitspresencese-verelyfeltfromonesideofthetransformertotheother.Soitisnottotally’uncoupled’atall!Infactthishappenstobethemainreasonwhyflybackswithlowoutputvoltages(highturnsratioshowpoorerefficiencyascomparedtohigheroutputflybacks.Therefore,reducingcriticaltraceinduc-tancesisimportantforseveralreasons:efficiency,EMI,be-sidesbasicfunctionality.Themomentaryvoltagespikeswhichlastforthedurationofthetransitioncanbeveryhardtocaptureonanoscilloscope.ButtheymaybepresumedtobepresentiftheICisseentobemisbehavingforno’obviousreason’.Thesespikes,ifpresentwithsufficientlyhighamplitude,canpropagateintothecontrolsectionsoftheICcausingwhatwecallhereacontroller’upset’.Thisleadstotheobservedperformanceanomalies,andinrarecasesthiscanevencausedevicefailure.Sincenoneofthesespike-relatedproblemscanbeeasilycorrected,orband-aided,oncethelayoutisinitiallybad,theimportantthingistogetthelayout‘right’tostartwith.Thedesignermaywellask,whyisitthatthesestepcurrentchangesareaproblemwiththeparasitictraceinductances,andnotwiththemaininductoroftheBuckconverter?Thatisbecauseallinductorstrytoresistanysuddencurrentchange.Butsincethemaininductorhasamuchlargerinductance(andenergystorageascomparedtothepara-sitictraceinductances,itthereforeendsup‘dominating’.AN-12293THEACANDDCCURRENTPATHS(ContinuedFromV*dt=L*dlwecanalsoseethatifLislarge,amuchhighervoltseconds(V*dtisrequiredtocauseagivenchangeincurrent.Thetraceinductancesthereforesimply’givein’firstbeforethemaininductordoes.Buttheycertainlydon’tgodownwithoutafight...andthevoltagespikesbeartestimonytothis!Noticethatthecurrentsinthesignaltracesintheschematicarenotshown.Forexamplethoseconnectedtothecom-pensationnode(marked‘COMP’orbootstrap(marked‘BOOST’carryrelativelyminutecurrentsandthereforearenotlikelytocauseupsets.Theyarethereforenotcriticalandcanberoutedrelatively‘carelessly’.Thefeedbacktraceisanexception,andwillbediscussedlater.TheGroundpinoftheICisanotherpotentialentrypointofnoisepickup.Inex-perienceddesignersoftengrosslyunderestimatetheneedsofthispin,particularyforBuckconverters.TheyassumethatsincethemainpowerflowinaBuckconverterdoesnotpassthroughthegroundpin,the‘currentthroughthegroundpinisverylow’,andthereforethetracelengthleadinguptothispinisnotcritical.Infact,thoughtheaveragecurrentthroughthispinisverylow,thepeakcurrentoritsdl/dtisnot.ConsidertheswitchdriverasshownschematicallyinFigure1.Clearlyitneedstosupplycurrenttodrivetheswitch.InanyFetoperatedasaswitch,largepeaktopeakinstantaneouscurrentspikesareneededtochargeanddischargethegatecapacitance.ThisisessentialsoastocausetheFettoswitchfast,andthisreducestheswitching/crossoverlossesinsidetheswitchandimprovestheoverallefficiencyoftheconverter.(Actually,inapracticalIC,the‘spike’ofcurrentcomesfromthebootstrapcapacitor,andthenthebootstrapcapacitorisquicklyrefreshedbytheinternalcircuitryoftheIC----itistherefreshcurrentthatpassesthroughthegroundpin.Further,asinanyhigh-speeddigitalIC,partsoftheinternalcircuitry,clocks,gates,comparatorsetc.,canturnonandoffsuddenly,leadingtosmallbutabruptchangesinthecurrentthroughthegroundpin.Thiscancause’groundbounce’whichinturncanleadtocontrollerupsets.There-forethelengthofthetracetotheGroundpinalsoneedstobekeptassmallaspossible.Thisalsoimpliesthattheinputcapacitors,especiallythebypasscapacitor‘CBYPASS’shouldbeplacedveryclosetotheIC,evenforaBuckIC.PLACINGCOMPONENTS‘a(chǎn)cap’(ascloseaspossibleOnehasheardthisbefore:“componentXneedstobe‘a(chǎn)cap’”.Soonwearetoldthe“componentYtooneedstobe‘a(chǎn)cap’”.Then“Ztoo”.Andsoon.Whichwouldbephysicallyimpossiblebecausemattercannotoccupythesameplaceatthesametime!Sowhichonecomesfirst?Thisisthemillion-dollarpredicamentalwaysfacingswitcherlayout.ThetroublingtracelengthsarethoseindicatedFigure1c.Tokeepthemsmall,clearlytwocomponentsneedtobeacap.Thesearetheinputbypasscapacitorandalsothecatchdiode.Considertheinputcapacitorsectionfirst.Thereforeahighfrequency‘bypass’or‘decoupling’capacitorwithsmallornoleads,shownas‘CBYPASS’inFigure1,istobeplacedveryclosetotheVINandGNDpinsoftheIC.Thisisusuallya0.1μF–0.47μF(monolithicmultilayerceramic(typicallyX7Rtype,size1206orthemorerecent’inverted’terminationversionofthispopularsize,the’0612’----alsonotethatsmallersizedceramiccapsgenerallyhavehigheresr/esl,butcheckbeforeuse.Sincenowthiscom-ponentprovidesthemainpulsedcurrentwaveshape,thebulkcapacitorshownas‘CIN’,maybemovedslightlyfurtherup(aboutaninchwithoutanydeleteriouseffect.Forlighterloads,andifitispossibletoplacetheinputbulkcapacitorveryclosetotheIC,thehighfrequencybypasscapacitormaysometimesbeomitted.Butforhigh-speedswitchersAN-12294PLACINGCOMPONENTS‘a(chǎn)cap’(ascloseaspossible(ContinuedliketheLM267x,theinputceramicbypasscapacitorisconsideredalmostmandatoryforanyapplication.becauseontheschematicthereisnowaytotellthediffer-ence.However,particularlywhenthediodeisaSchottky,theprimarypurposeofsuchasnubberistoabsorbthevoltagespikesofthetraceinductances.ThereforeitspositionmustbesuchthatitprovidesbypassingofthecriticalorACtracesectionsoftheoutputsideasshowninFigure1c(righthandsideoftheswitcher----whichmeansitmustbeclosetotheIC.Ofcourse,asmentionedpreviously,itisbesttogetthelayoutrighttostartwith,ratherthanaddingsuchextracomponents.Remainingcomponentplacementscanbetakenuponlyaftertheinputbypasscapacitorandthecatchdiodearefirmlyinplaceandarebothacap.Thetracestoeitherofthesetwocomponentsshouldbeshort,fairywide,andshouldnotgopassthroughanyviasonthewaytotheIC.ForSMTboardsthisimpliesthattheinputcapacitorandcatchdiodeareonthesamelayerastheIC.InFigure2suggestedPCBstartingpointsareprovidedforseveralswitchers.Allofthemfocusonplacingthesetwocriticalcomponentscorrectly.Theselayoutsarestronglyrecom-mendedformostapplications.The’X’markssuggesttherecommendedlocationwhereviascanbeusedtocon-necttoaGroundPlane(ifpresent.Theremainingcompo-nentscanbeplacedrelativelycarelessly(thoughindoingso,theremaybeslightimpact,forexampleontheaccuracyoftheoutputvoltagerailanditsripple,butnothingcomparedtowhatcanhappeniftheinputdecouplingcapandcatchdiodeareincorrectlyplaced.Traceroutingisnowdiscussedinmoredetail.AN-12295PLACINGCOMPONENTS‘a(chǎn)cap’(ascloseaspossible(ContinuedROUTINGTHETRACESAsmentionedabove,itisnotadvisabletorouteanyofthecriticaltracesthrough‘vias’.ViasareconsideredusefulfromapurelyCADperspectivefor‘layerjumping’,butareoftenusedindiscriminatelyastheyseemaneasysolutiontocon-nectivityproblems.Buttheyalsoaddimpedance,andthatisexactlywhatwearetryingtoavoid.Theinductanceofaviaisgivenbyforsomereason,severalviasinparallelwillyieldbetterresultsthanasinglevia.Andlargerviadiameterswouldhelpfurther(unlesstheyarebeingusedas’thermalvias’---discussedlater.Itisalsosaidthat“thetracesalsoneedtobe‘wide’and‘short’”.Thenecessityofshorttracesisclearlyunderstood,usuallyintuitively,bymostengineers.Infactthethumbruleof‘20nHperinch’alsoimpliesthattraceinductanceisalmostproportionaltolength.However,acommon‘intuitive’mis-takeistoassumethatinductanceisinverselypropor-tionaltothewidthofthetrace.Sosomeengineersmis-takenly‘a(chǎn)ddcopper’lavishlytocriticaltraces(thoughtherearesomeotherreasonswhythismaybebeingdone,andthesewillbediscussedlater.Afirstapproximationfortheinductanceofaconductorhavinglength‘l’anddiameter‘d’iswherelanddareincentimeters.NotethattheequationforaPCBtraceisnotmuchdifferentfromthatofawire.20042602FIGURE2.RecommendedLayoutStartingPointsAN-1229ROUTINGTHETRACES(Continuedwhere‘w’isthewidthofthetrace.ForPCBtraces,Lhardlydependsonthethicknessofthecopper(1ozor2ozboard.BoththeaboveequationsareplottedinFigure3.Itwillbeseenthatforagivenlength,aPCBtraceofwidth‘x’hashigherinductancethanawireofdiameter‘x’.InfactthewidthofaPCBtracehastobeabout1.78timesthediameterofawireforthesameinductance.Btrace,wecanseethatthediameter/widthofawire/tracehastotypicallyincreasebyafactorof10fortheinductancetohalve.Therela-tionshipofLtodisthereforelogarithmicinnature.Thereasonforthisistheeffectsofmutualinductancebetweenparallelsections/stripsoftheconductor.‘Beefingup’tracestoreducetheeffectsofparasiticinduc-tancesshouldbealastresort.Decreasingthelengthofthetraceshouldbethefirststep.Increasingthewidthofcertaintracescaninfactbecomecounterproductive.Inparticular,thetracefromtheswitchnodetothediodeis‘hot’fromanEMIpointofview.ThisisnotonlybecauseoftheAC(highfrequencycurrentitcarries,butbecauseofitsvoltage,whichisaswitchedwaveform.Anyconductorwithavaryingvoltage,irrespectiveofthecurrent,becomesanantennaifitsdimensionsarelargeenough.Radiatedemissionsfromthisantennacancauseundesirablecommon-modeinterfer-enceinitsvicinity.Thereforethiscallsfortheareaofthecopperaroundtheswitchingnodetobereduced,notin-creased.Largeplanesofswitchedvoltagealsocauseca-pacitivenoisecouplingintonearbytraces.OnatypicalSMTboard,iftheoppositesidehappenstobea’groundplane’,noisefromtheswitchingnodecancouplethroughtheFR4dielectricofthePCBintotheGroundplane.NoGroundplaneis’perfect’,andthereforethisinjectedhighfrequencynoisecanalsocausethegroundplanetonotonlyradiate,buttopassnoiseontotheICthrough’groundbounce’.Somepeoplesuggestthatacopperisland,exactlythesamesize/shapeastheswitching-nodeislandbecreatedontheoppo-sitesideofthePCB,connectedthroughseveralvias.Thisissupposedtoprevent’capacitivecross-talk’toothertracesandtoenhancethermaldissipation.Butthisobviouslyalsoleadstothebreaking-up/partioningoftheGroundplane.ThisdefeatstheverypurposeofGroundplaneasitcancausestrangeeffectsarisingduetotheoddcurrentflowpatternsinthenowdividedGroundplane.Ingeneral,theGroundplaneshouldbekeptcontinuous/unbrokenasfaraspossible,oritcouldbehavelikeaslotantenna.Fortheswitchingnodetherefore,thebestoptionistokeeptheamountofcopperaroundittotheactualminimumre-quirement.Somebasicphysicstoberemindedofhere:electricfieldsarecausedbyelectriccharge,andmagneticfieldsbycur-rents.Butifanelectricfieldvarieswithtime,itproducesacorrespondingmagneticfield.Howevermagneticfieldsareassociatedwithcurrents.ThereforeACvoltages(varyingelectricfieldsonoppositeplanesofcopperonaPCBcausea’displacementcurrent’(capacitivecouplingcurrentthroughtheFR4dielectric.Similarly,avaryingmagneticfieldcausesanelectricfield.Soforexampleinatransformer,whenwepassACcurrent(varyingmagneticfieldinawind-ing,wegetFaradayinducedvoltages(electricfield.When-evervoltageorcurrentisswitched,anelectromagneticfieldisgenerated,whichproducesEMI.AndthisEMIisinadvert-ently’helped’byantennastructures.Therefore,onaPCBlayout,theareaenclosedbyallcurrentloopscarrying’AC(switchedcurrent’mustbekeptsmall.Similarlytheareaofcopperplaneswith’AC(switchedvoltage’mustbekeptsmall.Bothcanbehaveasantennae.Inaddition,tracescarryingswitchingcurrents/voltagesmustalsobekeptawayfrom’quieter’tracestoavoidcross-coupling.Further,since’sharpedges’areknowntocauseanincreaseinfieldstrengths,two45degreebendsinatracearepreferredtoasingle90degreebend.COPPERFILLING:WHENTOSTOPAddingcopperlavishlytotracesservessomepurposeoc-casionally,sometimesnoneatall,andsometimesitevenworksagainstthedesigninanunintendedmanner.Theremaybenosimplehardandfastruleshere.Judiciousnessneedstobeapplied.Butfirstitisinstructivetoconsidersomeofthe‘reasons’whycopperislavished,andtothedegreeitisreallyrequired.Mostoftentherequirementsareactuallymuchlessthanpredictionsbasedon’gutinstinct’:Wewilltakeeachoftheseseparately:aCurrentHandlingCapability:Ifwemultiplythewidthofatracewithitsthicknesswegetthe‘crosssectionalarea’oftheconductor.Thisdeterminestheresistance(perunitlengthoftheconductorandtheconsequentself-heating.Thisleadstoanestimabletem