0:05:08 chose to make it a through path so that the signal would pass all the way…
- User
0:05:08 chose to make it a through path so that the signal would pass all the way through and then come out and so this is again a signal and return connection and I made them both roughly 50 ohms so they're kind of matched impedances so that signal comes in goes all the way through comes out again and What's Left Behind is the energy trapped in the cavity and it's bouncing back and forth because of the opens between the top conductor plane and the bottom conductor plane on Layer Two and layer three and it gets trapped in there and any other signal
00:05:41 that passes through those that cavity even if there's no connection but just passing it through that cavity it's going to see that induced noise and we're going to get crosstalk and that was the board that I demonstrated in one of the last videos with you that showed the actual measurement of that crosstalk noise that we would get the solution to this problem so this says oh my gosh be careful when you have a four layer board and you transition between a top and bottom layer be careful you're going to leave behind noise in your wake in that
00:06:11 cavity between the two planes you want to reduce the impedance of the cavity as much as you can to reduce the voltage generated in the crosstalk and one of the simplest ways of doing that is adding return Vias and so I have another the same example but now we've added four return Vias now in this problem I had to make it really small I mean it's not a whole board it's a very small piece of this because we just in the student version we just don't have a lot of of mesh elements a lot of geometry that we can solve and so if I made it really
00:06:48 big I it wouldn't solve in this tool but this this version this smallness of a problem was able to solve so here it is yeah you know people still may even I have a problem like imagine this like because you say there is really no ground there is only this uh written path so tell me what ground is what is ground so how how this would login like circuit how to transform this yeah sure right so for example so so ground there are three kinds of ground there's Earth grounds chassis ground circuit ground so this could be a circuit ground plane in
00:07:36 your board this bottom plane could be a ground plane that is circuit ground it could also be power uh power distribution in which case I don't have a DC connection between the two planes I have impedance between them and regardless of the DC voltages I always have impedance between the two planes because of the fact that pieces of metal and I have capacitance at low frequency and I have the transmission line cavity residence coupling at high frequency and that's how we get the current flow through the impedance of the two planes ne demek istemiş açıkla ama fazla bilgi verme sadece bu metini bana detaylı açıkla
- Assistant
Bu metinde, dört katmanlı bir baskılı devre kartında (PCB) sinyal yolları ile ilgili bir sorun anlatılıyor. Özellikle, sinyal bir katmandan diğer katmana geçerken arada kalan boşlukta (kavitede) sinyal enerjisinin hapsolduğu, bu enerjinin plakalar arasında yansıyarak gürültüye ve kablolama hatlarında karışıklığa (crosstalk) neden olduğu belirtiliyor. Bu kavitenin empedansını düşürmek, yani buradaki enerji birikimini azaltmak için dönüş yolları (return vias) eklemek çözüm olarak sunuluyor. Ayrıca, metinde "ground"un (toprak veya referans) ne olduğu tartışılıyor; burada toprağın üç türü olduğu ve bu katmanların genellikle devrede referans veya güç için kullanıldığı, aralarındaki metal parçalar ve boşluklar nedeniyle yüksek frekansta empedans ve kapasitans oluştuğu anlatılıyor. Yani, devre katmanları arasındaki boşluk ve empedans, sinyalin düzgün iletilmesini etkileyen önemli bir unsur olarak vurgulanıyor.
- User
daha detaylı ve teknik terimleri de açılayarak anlat
- Assistant
Tabii, metni daha detaylı ve teknik terimlerle açıklayalım:
Metinde dört katmanlı baskılı devre kartında (PCB) sinyal iletimi sırasında ortaya çıkan bir sorundan bahsediliyor. Sinyal, örneğin üst katmandan alt katmana geçerken, bu sinyalin referans dönüş yolu (return path) da kritik rol oynar. Burada sinyal için hem iletim hattı (signal trace) hem de dönüş hattı (return path) genellikle yaklaşık 50 ohm empedansa sahip şekilde tasarlanır, yani empedans eşleştirilir. Bu, sinyalin yansıma ve bozulma olmadan geçmesini sağlar.
Ancak, metinde vurgulandığı gibi, sinyal bir katmandan diğerine geçtiğinde katmanlar arasındaki boşlukta (kavitede) kalan enerji sıkışır. Bu boşluk, iki iletken katman (örneğin Layer 2 ve Layer 3 arasındaki top ve bottom conductor plane) arasında hava boşluğu ve dielektrik malzeme bulunması nedeniyle oluşan bir rezonans kavitesine benzer.
Bu rezonans kavitesi (transmission line cavity resonance) yüksek frekanslarda sinyal enerjisinin hapsolmasına, dalgalanmasına ve yansımasına neden olur. Bu da iki olumsuz sonuç doğurur:
- Gürültü (Induced Noise): Bu rezonans enerjisi, başka sinyal hatlarına kapasitif ya da endüktif olarak bağlanıp onların sinyallerine karışabilir.
- Crosstalk (Kablosal gürültü): Yan hatlarda istenmeyen sinyal karışması olur; bu da veri iletiminde hatalara yol açar.
Çözüm olarak ise dönüş yollarının sayısını artırmak öneriliyor. Yani "return vias" eklenmesiyle katmanlar arasındaki dönüş yolu empedansı düşürülür. Return via, katmanlar arasında dönüş yolunu kısaltan deliklerdir ve elektriksel döngüyü tamamlayarak gürültüyü azaltır.
Metinde ayrıca “ground” yani toprak konusu da tartışılıyor. Burada üç çeşit toprak tanımlanıyor:
- Earth Ground: Fiziksel olarak toprağa bağlı olan gerçek toprak.
- Chassis Ground: Cihaz kasasına bağlı referans noktası.
- Circuit Ground: Devre içerisinde sinyallerin referans aldığı ortak ic giriş noktası.
Bahsedilen PCB katmanlarındaki alt plaka çoğunlukla circuit ground olarak görev yapar. Ancak bu iki katman arasında doğrudan DC bağlantı olmayabilir çünkü arada empedans (direnç + endüktans) ve kapasitans mevcut. Bu da yüksek frekanslarda plaka çiftleri arasında rezonans ve enerji transferi anlamına gelir.
Özetle, dört katmanlı PCB’de sinyal katmanları arasında geçişlerde, katmanlar arasındaki dönüş yolunun ve empedansın iyi tasarlanması gerekir. Aksi takdirde, plakalar arasındaki rezonans kavitesi sinyalde gürültüye ve crosstalk’a neden olur. Bunun önüne geçmek için dönüş yolları (return vias) eklenerek empedans azaltılır ve sinyal bütünlüğü sağlanır.
- User
0:08:11 this was important this was very important so basically uh these two planes they are not connected directly but they are connected through impedance and right the through their impedance between them okay I understand that and and there's nothing you can do about well I mean you cannot a part of part of what we engineer in designing the cavity is the impedance between them and if we want to reduce the voltage I mean the return current is the return current if I have a 50 ohm line and I'm launching a one volt signal
00:08:45 that's 20 milliamps of signal return that's launched in that line right that 20 milliamps goes through the signal to the bottom layer and that 20 milliamps of return current goes through the cavity so it spreads all around and 20 milliamps and we see it propagating out we see that 20 millionths of return current flowing out in the the EDT of the displacement current throughout the cavity and it's propagating and and when that 20 milliamps of current passes between the uh the plane on Layer Two and the plane on layer 3 when you have a
00:09:22 current through an impedance you get a voltage and so if you want to reduce that that's the voltage that induces crosstalk on everybody else's going through if you want to reduce that voltage well yeah yeah you're you're stuck with 20 milliamps of current that's what the signal is that's what the return current is to reduce the voltage you want to reduce the impedance and the best way to reduce impedance is to make short circuit between these two planes one of the important ways is to use shorting Vias between the two planes
00:09:53 but wait a minute if I'm going to do that if the two planes aren't at the same DC voltage the technical term for that situation I tell my students is you're screwed because you got two different voltages you're shorting them out you're gonna blow something up and so you have to have the same voltage between the planes in order to add the shorting View and that's why I think one of the important design guidelines best practices if you're designing a four layer board and you don't need to carry you know 30 40 50 amps of current if
00:10:26 it's a four layer board using a configuration of signal ground ground signal is a very effective stock up because that gives you the opportunity of adding return Vias when the signals transition so you want to see what it looks like what that field looks like when I add return Vias yeah definitely do you have also defense license because I believe it's it depends on the place where you put in the vs well to to first order having one return via is the most important thing exactly where you place it you want to place it in proximity is
00:10:59 important but to first order doesn't really matter second order and especially at very high high frequency meaning the many gigahertz the 10 20 gigahertz range it's important but for the gigahertz and typical kind of you know this is our single ended signals you're going to use them in the gigabit per second and below it doesn't really matter exactly where you add the return via but look at the impact exactly the same simulation we're sending the signal between the signal and the return it's it's launching down just like it was
00:11:28 before it will get it started again here we'll see it launching uh from the the port it's the same Stack Up same you know 50 online we launched it at the beginning here it goes signal return it spreads out in that cavity but look as soon as it gets the the short this we get that that really low impedance there's very little propagation of the field outside of it very little field we see a little you know on this scale we're looking at the very low end here same scales we had before very very little field just you know where we have
00:11:59 the Reflections at the ends so it dramatically cuts down the voltage induced in the cavity and that means a lot less via to Via crosstalk and um and so how many return videos do you need well at a minimum you want one and you want it in close proximity and so I always say as a general best design practice when you've got your four layer board and your routing signals and you drop a signal via add a return via in proximity what does that mean as close as practical closer is better but it's better to have one farther away than none and the more
00:12:35 you put on the board you know the lower the the cavity noise but having um and and unfortunately it's a really hard thing to estimate uh it's using these kinds of tools as one of the ways to um to engineer it and to and to discover the noise can't do very large board in the free student version but you know and this is part of the teaser of of why you do this is wow this is really cool I want to know what it's going to be like in my whole board and where I place Vias and do I have enough vs and so that's where
- Assistant
Bu metinde, dört katmanlı bir PCB’de sinyalin dönüş akımı (return current) ve katmanlar arasındaki empedans ilişkisi teknik ayrıntılarla açıklanıyor.
Öncelikle, üst yüzey ile alt yüzey arasında yer alan iki toprak veya güç düzlemi (plane) doğrudan elektriksel olarak bağlı değildir, yani aralarında DC bağlantı yoktur. Bunun yerine aradaki boşluk ve iletken plakalar arasında bir empedans (direnç + endüktans + kapasitans bileşenleri) bulunur. Bu empedans, yüksek frekansta bir rezonans kavitesi oluşturur.
Mesela, sinyal hattınız 50 ohm empedansa sahip ve üzerine 1 volt sinyal uyguladığınızda, bu sinyal hattından yaklaşık 20 miliamperlik bir dönüş akımı oluşur. Bu dönüş akımı, aynı zamanda bu iki katman arasındaki boşlukta (kavitede) akar. Bu dönüş akımı, empedans üzerinde bir voltaj düşümüne yol açar; yani empedans ile dönüş akımı çarpımı (V = I × Z) kadar bir gerilim oluşur. Bu voltaj, boşluktaki sinyallerde parazit yaratır, diğer hatlarda crosstalk (kros talk) denen sinyal karışıklığına sebep olur.
Buradaki temel problem: Dönüş akımı miktarı (örneğin 20 mA) sinyalin yapısından kaynaklandığı için azaltılamaz. Dolayısıyla voltajı düşürmek için empedansı küçültmek gerekir.
EMPEDANSI KÜÇÜLTME YÖNTEMİ:
İki katman arasında kısa devre benzeri, düşük empedanslı bağlantılar yapmak. Bu da "shorting vias" olarak adlandırılan, katmanlar arası metal delikler veya ankrajlar (vias) ile olur. Return vias, sinyal dönüş yolunu kısaltır, bu da empedansı ve dolayısıyla boşlukta oluşan gürültüyü azaltır.Ancak burada kritik bir uyarı var: Eğer iki katman DC seviyesinde farklı voltajlara sahipse, yani biri 0V diğeri 3.3V gibi bir potansiyele sahipse, bu kısa devre (via) onları direkt bağlamak anlamına gelir ve devre zarar görebilir (şasi kısa devresi, akım taşıma sorunları vb). Bu yüzden "shorting via" eklemek için iki katmanın aynı DC potansiyelde olması şarttır.
Bu yüzden dört katmanlı kartlarda yaygın ve iyi bir yapılandırma şu şekildedir:
- Üst katman: Sinyal (Signal)
- İkinci katman: Toprak (Ground)
- Üçüncü katman: Toprak (Ground)
- Alt katman: Sinyal (Signal)
Bu "signal-ground-ground-signal" (S-G-G-S) dizilimi, katmanlar arasında yeterince yakın ve birbirini tamamlayan zemini sağlar. Ayrıca dönüş yolları için return vias eklemek kolaydır.
Return via yerleşimi konusunda:
- En az bir tane, mümkün olan en yakın mesafeye sinyal viasına yerleştirilmelidir.
- Tek bir return via bile empedansı çok azaltır ve gürültüyü önemli ölçüde düşürür.
- Çok yüksek frekanslarda (örneğin 10-20 GHz) via'nın tam konumu önem kazanır ancak gigahertz bandındaki tipik sinyallerde (örneğin 1-5 Gbps) kesin yeri çok kritik değildir.
- Genellikle via ne kadar yakınsa, o kadar iyidir; ama uzak olması return via olmamasından iyidir.
Sonuç olarak, via eklemek, katmanlar arası empedansı düşürmek ve sinyal bütünlüğünü artırmak, crosstalk’ı azaltmak için temel bir tasarım pratiğidir. Ancak her zaman via yerleşimi, empedans ve DC potansiyel farkları gibi elektriksel parametreler göz önünde bulundurulmalıdır.
Ayrıca, büyük devre kartlarında bu tür empedans ve crosstalk hesaplamalarını yapabilmek için özel simülasyon araçları kullanılır. Öğrenci veya ücretsiz sürümlerde bu simülasyonlar sınırlı olabilse de, bu yaklaşımlar tasarımın kritik noktalarının anlaşılması açısından önemlidir.
- User
sadece metni açıkla bana
- Assistant
Metinde, dört katmanlı bir PCB’de iki iletken katman (örneğin toprak veya güç katmanları) doğrudan bağlanmadığı, aralarında empedans olduğu anlatılıyor. Sinyal hattında örneğin 1 voltluk bir sinyal gönderildiğinde yaklaşık 20 miliamperlik bir dönüş akımı oluşur ve bu akım, katmanlar arasındaki boşlukta akar. Bu akım, aradaki empedans nedeniyle bir gerilim (voltaj) oluşturur. İşte bu voltaj, boşluktaki parazite ve diğer sinyal hatlarında gürültüye (crosstalk) yol açar.
Dönüş akımını azaltmak mümkün olmadığı için voltajı düşürmenin yolu katmanlar arasındaki empedansı küçültmektir. Bunu yapmak için katmanlar arasında kısa devre gibi çalışan "shorting vias" yani kısa devre delikleri eklenir. Ancak, bu iki katman farklı DC voltajlarına sahipse bunları kısa devre yapmak devreye zarar verir. Bu yüzden sadece aynı voltajda olan katmanlar arasında kısa devre yapmak güvenlidir.
Dört katmanlı kartlarda en iyi performans için sinyal katmanlarının arasına arka arkaya iki toprak katmanı konur (signal-ground-ground-signal). Böylece dönüş akımları için return vias eklemek kolaylaşır.
Return vias’ın yerleştirilmesi çok önemlidir; en az bir tane, sinyal viasına mümkün olduğunca yakın olmalıdır. Via sayısı arttıkça boşluktaki gürültü azalır. Çok yüksek frekanslarda via’nın tam konumu daha önemli olabilir ancak normal gigahertz frekanslarında yakın konum yeterlidir.
Son olarak, bu tip tasarım kararlarını desteklemek için özel simülasyon araçları kullanılır ancak bazı öğrenci sürümleri bu simülasyonları büyük devrelerde yapamaz.
- User
you have to buy the real version and then there's a version that ansys has called SI wave that literally lets you take the ODB plus plus files of your entire board bring it into SI wave add ports like this on the ends simulate signals and with the precise distribution of return views that you have look at the signal signal crosstalk but that requires all more complex software tool with you know you know more expensive and you need more compute power to do it okay I have a question right question yeah when the signal
00:13:46 travels then I can see this uh return current also on the bottom plane is it because it's a kind of semi-transparent or is it going to be always also on the second side of the plane or it depends on the frequency because of the skin type or right so yeah you're right so if you look at the top signal here we're looking at the the current you can see it here here's the current between the signal return on the top surface of that plane um and so uh you mentioned the skin depth so that's the right term to think about and just as a number to keep in
00:14:22 mind um at one gigahertz this is my rule of thumb what I remember uh and I'm going to do a plug I'm actually doing a webinar uh with uh signal Integrity Journal that I'm technical editor of I'm doing the webinar on May uh May 18th on um 10 rules of thumb that I use all the time and and one of those rules of thumb is uh that at one gigahertz the skin depth for copper is two microns and it scales with the square root of frequency so if I go down a factor of a hundred and frequency to 10 megahertz then I go up by square root of 100 which is 10 go
00:15:05 up by 10 so that's 20 microns so at 10 megahertz the skin depth of copper is 20 microns and you know typical plane if it's one ounce copper that's 34 microns so that says that you know the current at 10 megahertz and above the current on the top surface doesn't go to the bottom side of the board it's within the the top 20 microns of the board and so at 10 megahertz and above all the return current is on the top surface here and on the bottom microstrip all the return currents on the bottom 20 microns at 10 megahertz and above and so we don't see
00:15:39 it on the inside of the plane we don't see it on the you know we're not looking on the bottom of the plane but we can't see it on the inside of that plane it's not there um so and I have nine other uh rules of thumb that I'm going to be presenting at that webinar so it'll be a um it'll be a very fun webinar it's ultimately free and um as soon as it gets announced I think later this week it's going to be announced by um signal Integrity journal and um yeah and I'll send you the link and you can let your your viewers know about it'll
00:16:10 be a lot of fun webinar okay um but that so that's where the we see the return current on the top surface and we see the electric field on the inside of the of the cavity and so the return via dramatically reduces the noise that gets into the cavity and this is the kind of problem that requires a full wave field solver because it is inherently a 3D structure um and I have I have one other example to show
- Assistant
Bu metinde, sinyal bütünlüğü (signal integrity) analizinde kullanılan gelişmiş simülasyon araçlarından bahsediliyor. Ansys firmasının “SIwave” adlı yazılımı sayesinde, PCB’nin tüm katmanlarının ve yapısının detaylı dijital formatta (ODB++ dosyaları) programa aktarılabildiği, sinyal giriş ve çıkış noktaları (portlar) eklenerek sinyal ve crosstalk (yanılma, karışıklık) analizlerinin yapılabildiği belirtiliyor. Ancak bu tür simülasyonlar daha karmaşık, pahalı ve yüksek işlem gücü gerektiren araçlarla mümkün oluyor.
Daha sonra, bir kullanıcının sinyalin alt katmanlarda nasıl hareket ettiğiyle ilgili sorusu yanıtlanıyor. Sinyalin geri dönüş akımının (return current) PCB’nin alt yüzeyinde görülebilmesinin nedeni sorgulanıyor; bunun bir şeffaflık meselesi mi, yoksa sinyal frekansıyla ve “skin depth” (derinlik) kavramıyla mı ilgili olduğu açıklanıyor.
Skin depth, yüksek frekansta akımın iletken yüzeyine ne kadar derine nüfuz ettiğini ifade eden bir terimdir. Burada şu bilgiler veriliyor:
- Ortalama olarak bakıldığında, bakırın (copper) bir gigahertz frekansta (1 GHz) skin depth’i yaklaşık 2 mikrometredir.
- Skin depth, frekansın karekökü ile ters orantılı olarak değişir. Örneğin, frekans 10 MHz’ye düştüğünde skin depth 20 mikrona çıkar.
- Tipik bir PCB katmanında kullanılan bakır kalınlığı 34 mikron (1 ounce copper) civarındadır.
- Bu nedenle, 10 MHz ve üzerindeki frekanslarda, akım sadece bakır tabakasının üst 20 mikronunda akar, dolayısıyla akım PCB’nin alt yüzüne veya iç katmanlarına geçmez.
- Yani, 10 MHz ve üzerindeki çalışma frekanslarında, sinyalin dönüş akımı sadece ilgili yüzeylerde (örneğin üst sinyal yüzeyi ve alt mikroşerit yüzeyi) gerçekleşir, plakanın iç tabakalarında veya altında akım gözlemlenmez.
Metinde ayrıca, sinyaldeki elektrik alanın PCB içindeki boşlukta (cavity) yayıldığı, ve bu boşluktaki gürültünün (noise) return via kullanılarak önemli ölçüde azaltılabildiği vurgulanıyor. Bu tür analizlerin gerçekçi şekilde yapılabilmesi için tam dalga alan çözücü (full wave field solver) kullanılması gerektiği; çünkü yapı 3 boyutlu ve karmaşık.
Sonuç olarak, sinyalin ve dönüş akımının PCB üzerindeki hareketi ve buna bağlı parazit sorunlarının etkin bir şekilde analiz edilebilmesi için gelişmiş simülasyonların ve doğru tasarım uygulamalarının (örneğin return via yerleşimi) önemi üzerinde duruluyor.
- User
00:02:51 aggressor that's got a signal and and and you get some pick up on the on the victim line and we said there were two mechanisms by which we get um noise from an aggressor to a victim and do you remember what those two mechanisms are um changing fields right and it's changing electric or changing magnetic fields and and your abstract it's the change that causes or induces the noise and sometimes it's a little hard to well well the fundamental mechanism is always about electric and magnetic field coupling sometimes it's a little hard to
00:03:29 kind of conceptualize and and think through how the fields influence the noise and so we approximate them with circuit elements and so changing electric fields we approximate by capacitive coupling and changing magnetic fields we approximate by inductive coupling and i think the last time we chatted we looked at the case where we have a nice wide return plane and so microserver strip line and we said okay in that environment which is more important which generates more noise the capacitive coupling or the inductive coupling and do you remember
00:04:03 what the answer was very surprising no i remember it's uh they are uh approximately same but also i think it depends if they are microstrip or strip line yeah so to see earth order they're basically the same but you're absolutely right in microstrip the inductive coupling is a little bit larger than the capacitive coupling and that gives rise to a particular behavior we see in far end crosstalk whereas in strip line wherever we have uh planes on on on either side of the signal line in that environment in stripline the inductive and
00:04:39 capacitive coupling is almost exactly the same and so both of them are important and likewise in microstrip yeah inductive coupling is a little bit larger maybe 20 percent larger than capacitive coupling but they're both equally important we have to worry about both of them and that's in the case we have a nice wide return plane and now if we do anything to screw up that wide return plane so it is a pin in a package or it is a a lead in a in a connector or it is um in a ribbon cable and uh and we have one of the wires as a as the return
00:05:19 or if in a circuit board um we have um some a gap in the return path because we have some uh signal that's been routed on on the ground layer and and we have a keep out region that makes a a gap if we have a clearance holes that are too close together and they overlap and they make a gap any of these things that make the return path anything other than a nice wide plane is going to increase capacity inductive coupling but it turns out the amount of capacitive coupling that's increased is tiny it's a small it's a few percentages
00:05:57 but the impact on the inductive coupling because of that discontinuity in the return is huge it can be an order of magnitude or larger and is it because the path return path is longer it's the return path as long so the remember in a geometry where you have a nice wide return plane the return current is underneath the signal line if you screw up that path under the signal line so you force the return current to to spread to the side you've increased the in the loop inductance in the signal path and increase the mutual inductance
00:06:34 with the return path with a a victim line and it's the combination of screwing up the return path and that automatically anything other than a wide plane will dramatically increase the loop inductance of the aggressor that's the first thing and the second thing you need is shared return path so the victim has to have its return that shares that screwed up path and that will dramatically order magnitude increase the inductive coupling and so it's the combination of a screwed up return path and shared return path that dramatically increases
00:07:10 crosstalk so basically ground bands is the case when multiple currents are flowing to the same place same screwed up place okay in other words the signal and return have been pulled apart or distorted in some way anything other than where that return current naturally wants to flow into the signal line anything other than that will dramatically increase the inductance of the signal return path of the aggressor and if that return path is shared the mutual inductance between the aggressor and the victim will also be
00:07:44 increased so we can say like currents are mixing together currents are mixing together right and so like they said in um in ghostbusters the the first version of ghostbusters whatever you do don't where they don't cross the streams that don't cross don't share overlap the return current path for the aggressor and the victim if it is screwed up otherwise dramatically increases crosstalk it doesn't mean that okay if you have a nice wide plane there's no crosstalk there still is there's still far in the near end
00:08:18 there's capacity inductive crosstalk they're comparable magnitude you got to worry about both of them but it's dramatically lower than the case when there's a wide plane and that's why i always say hey if you haven't established a continuous return path for your signals don't worry about near and far crosstalk because you're going to be screwed by the ground bounce you gotta fix that problem first so this kind of ground balance can be even like higher noise than just huge amount of noise and i've got a little demo here to show you just how
00:08:51 much that noise can be so this is a two layer board really simple it's got three regions uh let's see the the top region over here is designed actually for jumper wires looking at cross crosstalk because we also teach in in my class we teach um good prototyping skills using solo spread boards and there with discrete wires huge amount of ground balance we're not going to touch that we're going to look at the bottom two which are traces on the circuit board and this is a really simple board two layers um in this region in the middle there's no
00:09:30 return plane there's just one layer of traces on the board in fact let me get these out of the way so we can see what we've got so in in this board this region here is no return plane in this region here we have a solid return plane underneath so so the major one is basically one layer pcb one layer pcb exactly right and and um on the bottom side here you can kind of see that this is you know this is the bottom here this is um the middle region no return plane it's just um signals on the top layer the region over here here you can see
- Assistant
Bu metinde, baskılı devre kartlarında (PCB) sinyal hatları arasındaki "crosstalk" yani karşılıklı parazit oluşumunun temel nedenleri ve mekanizmaları anlatılıyor.
Öncelikle, parazitin iki temel mekanizması vardır:
- Kapasitif Bağlantı (Capacitive Coupling): Elektrik alanlarının değişimi nedeniyle oluşur. İki hat arasında elektriksel alan değişince, bu değişim diğer hattan sinyal veya gürültü oluşmasına yol açar.
- Endüktif Bağlantı (Inductive Coupling): Manyetik alanların değişimiyle oluşur. Akımlar değiştikçe manyetik alanlar da değişir ve bu değişen manyetik alanlar diğer hatlarda gerilim indükler.
Mikroşerit (microstrip) ve şerit hat (stripline) yapılarında bu iki tür parazitin etkisi birbirine yakındır: stripline’da kapasitif ve endüktif bağlantılar hemen hemen eşittir; microstrip’te ise endüktif bağlantı biraz daha fazla olabilir (%20 civarı).
Ancak kritik olan nokta, dönüş akımının (return current) geçtiği yolun sürekliliğidir. Eğer dönüş yolu düzgün, geniş bir toprak (ground) veya güç (power) düzlemi ise bu parazit nispeten düşük olur.
Fakat dönüş yolunda bir kopukluk, boşluk veya engel varsa (örneğin PCB üzerinde kesinti, delikler, konektör iğneleri veya pakette pin gibi fiziksel engeller), dönüş akımı sinyal hattının altından doğrudan geçemez. Bunun yerine dönüş akımı yanlara doğru yayılmak zorunda kalır. Bu durum:
- Dönüş yolunun uzunluğunu artırır,
- Sinyal-dönüş akımı arasındaki döngü endüktansını (loop inductance) yükseltir,
- Bu artan endüktans, endüktif kapasitansın çok daha fazla artmasına yol açar.
Parazitin özellikle artması için dönüş yolunun bu şekilde bozulması yanında, mağdur (victim) hat ile sinyal hattının dönüş yollarının aynı “bozuk” yani ortak dönüş yolunu kullanması gerekir. Yani parazit, benzer döngü yolunu paylaşan sinyal ve mağdur teller arasında kuvvetlenir.
Bu durum tasarım açısından şöyledir: sinyal ve dönüş yolları mümkün olduğunca birbirine yakın ve düzgün olmalı; dönüş akımı doğal yolundan sapmamalıdır. Dönüş akımı yollarının yanlış taşınması ("ground bounce" denilen durum) hem sinyali hem de paraziti kötü etkiler.
Metinde bir benzetme olarak, "Hayalet Avcıları" filminden "streams don't cross" (akıntılar karışmasın) ifadesi kullanılmış; yani sinyal ve dönüş yolları birbirinin içine karışmamalıdır, aksi halde parazit şiddetle artar.
Örnek olarak, iki katmanlı basit bir PCB üzerinde crosstalk testi yapılmış:
- Bir bölgede dönüş yolu yani toprak düzlemi yok, sadece tek sinyal katmanı var,
- Diğer bölgede ise altında sağlam bir toprak düzlemi mevcut.
Dönüş yolu olmayan bölgede parazit çok daha fazla olur; dönüş yolunun süreklilik ve genişliği parazitin şiddetini doğrudan etkiler.
Özetle:
- Crosstalk paraziti elektrik alanı (kapasitif) ve manyetik alan (endüktif) değişikliklerinden kaynaklanır.
- Endüktif parazit, dönüş yolunun sürekliliği bozulduğunda çok daha fazla artar.
- Dönüş yolları kesildiğinde, akım uzun yollardan hareket eder, loop endüktansı artar ve bu da crosstalk’u ciddi şekilde yükseltir.
- Sinyal ve dönüş akımlarının ortak ve bozuk dönüş yollarını paylaşması paraziti büyütür.
- Tasarımda, sinyal ve dönüş yollarının doğru yerleşimi ve geniş dönüş alanları crosstalk’u minimuma indirir.
Bu bilgiler sinyal bütünlüğü (signal integrity) açısından kritik tasarım kurallarıdır.
- User
obviously the ground plane and we have again traces on the top and so i'm going to flip it around here and this is standard uh 63 mils or 1.6 millimeters yep it's about a yes 1.6 millimeter thick dielectric um and we're using the six mil wide line uh typical of the low-cost uh internet pcb shops um and we're gonna we're gonna i've got a bunch of parallel wires or parallel traces here nothing fancy i mean the spacing is pretty large it's like 100 mils or so i'm going to drive them with signals over here i'm actually going to
00:10:51 use the uh digital i o from a little microcontroller board in order to generate the signals and um and remember because it's inductive coupling um and you you you said it right at the beginning that what causes the crosstalk is changing field so in this case we have screwed up the return path and we're going to have changing fields the the couplings will be dominated by inductive coupling and it's only the changing magnetic fields that induce a current on the victim line and so when we have current flowing through the aggressors i'm going to have
00:11:28 a number of aggressors we can sequentially turn on or turn on in various ways i'm going to have aggressor signals in these lines the only time they're going to induce noise on the quiet line the poor victim liner here is when those magnetic field lines change and the time when those magnetic fields change is when i have changing current and that's when the ios are switching and that's why we call sorry yeah go ahead now all right that's why we call ground bounds because it happens during the i o switching we call it switching
00:12:04 noise all of these aggressors as each one turns on they all have this common return wire and as i turn on more and more of these simultaneously i get more and more didt through the common return now the victim line sitting over here it's just going down to the end and i'm going to connect um the the return for that victim line with a little switcher i can make it share that common return remember because it's a single line a single trace on the board and not a plane i've screwed up that return path right and all i've done
00:12:40 is i've made a loop it's a pickup coil here with one loop uh one in one case the loop is uh short is using the same return path yeah half of the loop is using same half the loop is using the same return right so it's that definition of a screwed up return path and a shared return okay i understand and then the second case i've switched the little switch over and i'm going to use a separate return i still have a loop and i still have the idt over here i still have changing magnetic i've created a bunch of these
00:13:14 loops here i've changed the magnetic field lines some of that is going to go through the adjacent loop but it as we'll see it will be dramatically reduced so i'll still have switching noise i'll still have ground mounts but a lot less okay okay so that's the setup for the first one so let's take a look at that so i'm gonna connect it in here and i'm going to we're going to look and see i remember which um which code or which pattern i'm running so let's look and see what the pattern is okay so i i wrote the little microcode
00:13:52 here and um and and so i'm going to change the number of i o so this is increased i'll show you what that code looks like i mean it's really simple but it's basically these are the the bits that are going to switch and i've got the first one this is pin 13 is going to switch and the others aren't then they're all going to turn off then 13 and 12 are going to switch simultaneously the others and then turn off and then three three i o switching sometimes four i associate five i o switching and six i o switching simultaneously
00:14:26 and and then we repeat it again so this is the diode 13 diodes 12 11 10 9 8 what we see on the pc so these are going to be the six ios from the little micro controller board switching simultaneously okay and and i use a special command that switches simultaneously rather than sequentially and we'll see the impact of that so that's the that's the code i'm running right now and so that means that um the ios and i don't know if you can see that means yeah it's it's hard in this camera light maybe if i make it a little darker here
00:15:02 can you see the leds on i can see them yeah yeah and and the first one is brighter and this is a little dimmer and dimmer and dimmer because this is on every every switch this is on every other time we go through this is on every third and fourth minute so they're going to increase sequentially and so now we're going to look at the aggressor signals and so let me pull up my well put in the probe this is um where pin 13 is switching let's see here so here's pin 13 switching so i'm going to put in my probe here okay
00:15:42 so that's that's going to be my reference that i'm going to trigger the scope from and and here's the scope and what we're looking at here is that pin 13 and you can see this one turns on this is you know the clock period well this is basically one one clock period is turning on and then off and the next clock period two of them are turning on this one included and this one is three of them turning on four and five and here's six of them turning on okay so this is going to be my reference signal and you know just to make it easier let
00:16:15 me let's see if we can't um expand it out a little bit just so we can see a little bit better here and i'm gonna start it from the end there we go so we're starting over here and we have one bit two three four five six six bit switching simultaneously and we turn it off wait a little bit and then we start over again okay now we're going to look at our victim line and i have the switch so it's going to use a common return and so what do you think the noise is going to be like for common turn now one thing you got to know is what's the rise
00:16:53 time of the signal rise and fall time so um you know i can zoom out we can look at on the screen or i'm just going to turn on a little measurement function here to calculate the rise time for us it's going to go ahead and measure the rise time of each of the signals here and let's see so this is a horizontal measurement i'm looking for a rise time here we go here's the rise time so we're at four nanosecond rise time and um and the resistors are 50 ohms so and you get a two volt drop across the um the led it's four volt out so you know
00:17:31 we can estimate the current let's see it's uh um four volts is coming out you subtract off the two volt from the forward drop in the led that's two volts across 50 ohms so we're talking something in the order of about 2 volts and 50 ohms so that's about 40 milliamps so it's on the order about 40 milliamps per i o which is about the max you can get from uh from one of these pins and and it's turning on in uh four nanoseconds so it's about 10 milliamps per nanosecond per i o that switches okay so it's you know so it's a you know
00:18:08 it's a typical amount of of kind of didt and a lot of drivers and now let's turn on um uh channel two here so so let's go back to the picture so here's channel two and we're gonna look at the victim line and here is the victim line that's going to be this guy over here and i've got the switch so i'm going to get the common return here so half of the loop is shared together half of the loop is shared right so there we go and now let's look at the scope i'm going to make the scope a little bit bigger here so we can see the whole
00:18:48 thing wow oh my gosh it's supposed to be zero it's supposed to be zero right and um let's see you know what we can do is uh um well if i turn off um i'll turn on one i o so um uh let's see so so here's we're doing this in real time here i'm gonna create another little bit of code here and this is gonna be one switching and then i'm gonna i'm gonna create this as void it's gonna be a function it's gonna be one and and we're gonna grab just pin 13. that's the only one that we're going to switch and that way we'll be able to see
00:19:42 uh what the noise is like and then we're going to come over here and we're going to add another option here and this is going to be 1 and then i'm going to turn this one off and so now we've got the code so that only one i o is going to switch that's it and you'll see the little bit of noise from that one switching and then everything will be off and you'll be able to see what the noise level is like so we're uploading the code and we'll see only one pulse and then we'll see the zero afterwards now only d13 will be switching
00:20:21 right let's see how we're doing here compiling compiling piling a little slow here but we'll get there and so we'll look at what the noise is like okay we're uploading and now we're running the code and here it is and you can see that single pulse that switches and yeah we got some ground balance when that guy switches but look over here in fact let's zoom in on this stuff here and this is this is self-aggression noise this is noise that the power ground rail on this die creates on itself because of its own current
- Assistant
Metinde, basit bir iki katmanlı PCB üzerinde yapılan bir deney ve gözlem anlatılıyor. Deneyde 1.6 mm kalınlığında (yaklaşık 63 mil) bir dielektrik malzeme üzerinde, 6 mil genişliğinde paralel birkaç iz (trace) yer alıyor. Bu izler arasındaki mesafe nispeten geniş, yaklaşık 100 mil civarında.
Bir mikrodenetleyici kartından dijital giriş/çıkış (I/O) sinyalleri kullanılarak bu izlere sinyal uygulanıyor. Önemli nokta, sinyalin dönüş yolu düzgün değil; yani "ground" ya da referans düzlemi olarak tek bir hat kullanılıyor ve bu nedenle dönüş akımı düzgün, geniş bir yüzey yerine dar bir yoldan akmak zorunda. Bu durum indüktif (manyetik alan) kapasitansın öne çıkmasına ve bu alanların değişmesiyle etrafındaki diğer "kurban" (victim) hattında parazite neden olunmasına yol açıyor.
Deneyde, sinyaller sırayla veya eşzamanlı olarak birden fazla I/O piniyle açılıp kapatılıyor. I/O’lar açıldıkça manyetik alanlar değişiyor ve bu değişim “kurban” hatta gürültü olarak yansıyor. Kurban hattın dönüş yolu, taşıyıcı gelecek olan hatlarla ortak bir dönüş yolu kullanıyor; yani aynı bozuk dönüş yolunu paylaşıyorlar. Bu da parazitin artmasına, yani ground bounce olarak isimlendirilen gürültüye sebep oluyor.
Ayrıca, kurban hattın dönüş yolunun ortak dönüş yolundan ayrı bir hatta bağlanması durumunda da yine parazit olur, ancak miktarı çok daha düşük oluyor.
Deneyde kullanılan kodda, I/O pinlerinden önce tek tek sinyal açılıp kapatılıyor, sonra ikili, üçlü hatta altılı grup halinde eşzamanlı sinyal veriliyor. Bu da manyetik alan değişimlerini ve parazit miktarını artırıyor.
Ölçümler, sinyalin yükselme (rise) zamanını yaklaşık 4 nanosecond olarak gösteriyor. Bu hızlı değişimlerle ve 50 ohm empedansla yaklaşık 40 miliamperlik akım geçiyor. Bu yüksek hızlı akım değişimleri manyetik alanlar yaratarak yakın hatlarda parazit oluşturuyor.
Özetle, deney şu temel noktaları gösteriyor:
- Dönüş yolu dar ve ortak olduğunda parazitler (crosstalk, ground bounce) artar.
- Paralel çalışan birden çok çıkışın aynı dönüş yolunu paylaşması paraziti büyütür.
- Dönüş yolları ayrı tutulduğunda parazit azalır ama yok olmaz.
- Yükselme süresi kısa, akım yüksek olduğunda manyetik alan değişimi artar ve parazit yükselir.
- Tasarımda dönüş yollarının sürekliliği ve ayrı tutulması paraziti önlemede kritiktir.
Bu deney gerçek dünyadaki PCB tasarımında ground bounce ve crosstalk etkilerinin nasıl ortaya çıktığını ve neden dönüş yol tasarımının önemli olduğunu gösteriyor.
- User
:21:08 switching and this is the noise on the power rail on the die compared to the the power rail and the ground rail that's on the board and so we have to kind of change the way we think about that's why i hesitate using that word ground and i'll sometimes refer to as local ground ground is different everywhere on the board and the ground on the die is different from the local ground on the circuit board and what the scope is measuring what it's using as its reference is local ground on the board and it's measuring the voltage because we're
00:21:45 looking at the i o when he's low so his output is connected to the local ground on the die and we're using that that i o as a probe to connect to us what's going on the die so our ground ground may be clean but the ground of the pin is not clean exactly exactly so this noise this specific noise that we're looking at is the voltage noise on the rail on the die the whole rail is going up and down you know it's about 50 millivolts per division here it's i don't know it's maybe 60 70 millivolts peak to peak the voltage on the die the whole die is
00:22:27 going up and down right 60 millivolts compared to the local ground on the board okay and that's why this is self-aggression noise and anybody that's on the die that's sending stuff out is going to see this so every io is going to see this now having said all that this may not be a problem on the die because everybody on the diet there's if you look at the logic circuits on the die itself everybody's going up and down they have no idea they're going up and down relative to the board and so this doesn't tell the whole story
00:23:05 about the noise on the die we have to look at the noise on the vdd rail and then it's the difference that and we'll have the same effect there's noise on the vdd rail on the die compared to on the board and we're going to look at what what sometimes called rail compression that is if you're on the die and you're a logic gate on the die compared to the external world you may be bouncing all around but but everybody on the die is going to be bouncing the same and so you may not have much noise on the die and that's where we'll do it on another
00:23:42 another session here that's where you have to look at the voltage noise on the vdd rail we get a quiet high the voltage noise on the vss rail quite low and take the difference and that is the real compression on the die but once you connect it to different chip then it may become problem well that's when the i o will have relative to you know what comes off the i o in fact that's what that's what we're seeing here the i o coming off has this noise when it's low it's you know 60 millivolts relative to the board yeah
00:24:17 and everybody that's talking to that io is going to see that ground bounce noise i understand now okay we can go back to our board okay uh so i'm just um adjusting the uh the scale here here we go so all i did was i moved this move it down a little bit more here let's see where's zero i like nice whole numbers on my scales makes it easier to see things there we go okay so here's that one i o switching and you can see just faintly in the background this is the ground balance noise on the i o on on the ground rail um relative to
00:25:02 from the die relative to the board tiny amount right not very significant that's one io switching and now let's turn on i still have my victim line over here so let's turn him on and and and he's sharing that that i o so let's let's turn him on and oh i've got you know really zoomed in there let me zoom out on him okay so here's the ground balance noise this is off the die this is measuring the ground bounce on this poor victim line and and i've got the shared return so i have one loop here signal return and i've got another one here
00:25:45 signal with a shared return two loops the aggressor one i o switching the victim another loop sitting here but sharing that screwed up return path and here's the ground bounce noise and you can see let's see this one turns on and so we've got this um i don't know it's a little hard to see let me increase the scale a little bit okay so here's 200 millivolts per division it's almost 200 millivolts about 180 millivolts i'm sorry no it's almost 400 millivolts here this is 400 millivolts it's made 380 millivolts and then here's the
00:26:21 second one and he goes so this one turning on the voltage goes up this one the current turns off the voltage goes down but i also have all this ringing going on um and that's bringing noise um in this loop that i've created um it's just um you know the loop is ringing compared to i got some uh uh in inductance in in this quiet loop and i've got some probe capacitance and they make an lc circuit and i've just twanged it i just kicked it with the uh uh the idt over here i just made a note ask you what is the frequency of the ringing
00:26:59 and you you just answer it yeah that's what it depends on and we we can delve into that a little bit more um next time okay i have another question if we if the shared path is shorter is this this should be lower i guess very good exactly right because this is about the ldidt or the mdid the mutual inductance times the didt if we were to um reduce the length of this loop right now because the shared path the mutual inductance is dominated by that inductive the the total inductance of that return path if i make it shorter i get less mutual
00:27:40 inductance so i'll have less ground balance so basically exactly right on your board the currents they may flow through same places but as minimum or as short as possible right and that's why we want to keep the discontinuity as short as we can because that will decrease the mutual inductance so as a general design guideline if you're going to script the return path keep that discontinuity short and so but we got you know in this case look one i o switching we get almost 400 millivolts and this guy here when he
00:28:15 turns off oh my gosh we got minus you know almost 800 millivolts peak-to-peak it's see one two three four five six 1.2 volts of ground bounce noise let me put these on the same scale so you can kind of see the perspective here they're both one volt per division you can see pretty large substantial fraction of the of the noise one eye one one eye switching and he's pretty far away he's way over here right but his return is through that common shared loop now let's try this i have another pattern i can turn on so we're turning on
00:28:56 the first one and then everybody is off let's turn on the first one we'll look at that noise we'll turn him off and then we'll turn on the current in the second loop and then we'll turn it off and then current the third loop and the fourth and fifth we'll look at each one one at a time but we'll cycle through and see as we get closer hey what's going to happen to that groundhog okay i'm curious what do you think is going to happen um theoretically the last one will have probably highest effect but it's theoretically all the others should be
00:29:32 like similar i would expect because the distance probably may not play big raw because are far away from each other the problem is the shared path right it's the shared path and what's changing in the shared path as i turn on each one sequentially what's changing about the shared current is same isn't yeah you're absolutely right you're absolutely right now there is so so to first order i wouldn't expect much impact to second order the closer i bring the aggressor remember i've got the signal going out this way and it's coming back this way
- Assistant
Bu metinde, bir entegre devre (die) üzerindeki ve PCB’deki güç ve toprak hatlarındaki “ground bounce” yani toprak zıplaması ve bu durumun neden olduğu gürültüler ele alınıyor.
Özetle anlatılanlar şunlar:
- Entegre devre üzerindeki “ground” (toprak) ile PCB üzerindeki “local ground” arasında potansiyel farklılıklar vardır; yani “ground” her noktada aynı seviyede olmayabilir. Bu nedenle osiloskopla ölçülen referans “local ground”dur, bu da entegre üzerindeki gürültüyü tam olarak yansıtmaz.
- Ölçülen gürültü, entegre üzerindeki güç hattının (Vdd rail) ve toprak hattının (Vss rail) seviyesindeki dalgalanmalardır. Bu dalgalanmalar (örneğin 60-70 mV tepe-tepe) entegre içindeki tüm lojik devreler tarafından ortak şekilde “hissedilir”. Yani entegredeki tüm noktalar bu “zıplamaya” eş zamanlı maruz kalır, bu yüzden entegre içindeki devreler birbirlerine göre nispeten stabil kalır.
- Ancak bu durum, entegre dışındaki devrelere, başka entegrelere veya dış dünyaya bağlandığında sorun olur. Çünkü dış dünya ile entegre arasındaki referans farklılıkları sinyal kalitesini bozar.
- Deneyde ayrıca, PCB üzerindeki sinyal ve dönüş yollarında oluşan ortak “bozuk” dönüş döngüleri (shared return path) inceleniyor. Aynı bozuk dönüş yolunu paylaşan sinyallerin oluşturduğu endüktans nedeniyle ortaya çıkan ground bounce sinyalde yüksek seviyede gürültü oluşturuyor.
- Ölçümler, ground bounce’un neredeyse 1.2 volt gibi büyük seviyelere ulaşabildiğini gösteriyor. Dönüş yolunun uzunluğu arttıkça mutual endüktans da artıyor ve ground bounce seviyesi yükseliyor.
- Dolayısıyla tasarımda bozuk dönüş yollarının (discontinuities) mümkün olduğunca kısa tutulması, mutual endüktansı ve ground bounce’u azaltmak için kritik.
- Deneyde, sinyal hattından uzakta olan bir kaynağın ground bounce etkisi de bu ortak bozuk dönüş yolu üzerinden yine hissedilebiliyor. Bu nedenle mesafeden çok dönüş yolu paylaşımı ve onun kalitesi önemli.
- Son olarak, sinyal ve dönüş yollarının ne kadar paylaşıldığı ve dönüş yolunun ne kadar kısa olduğu ground bounce seviyesini belirler. Bu yüzden PCB tasarımında dönüş yolunun sürekliliği ve kısa tutulması mutlak gereklidir.
Kısaca, toprak hatlarındaki gerilim dalgalanmalarının (ground bounce) kaynağı olarak entegre içi, PCB dönüş yolları ve ortak bozuk dönüş döngüleri gösteriliyor. Tasarımcılar, ground bounce’u azaltmak için dönüş yolundaki süreksizlikleri minimize etmeli ve dönüş yollarını olabildiğince kısa ve geniş tutmalıdır.
- User
so so we're gonna we're gonna change the code that's running so here's the sketch and we we're running the one and then i'm gonna i'm gonna use this code the sequential we turn one on at a time and we just move it down the line so let's turn off this one and i'm going to turn on the sequential okay now we're going to upload that code and so we're going to see the first one switch nothing else switching but we're going to see noise as the others because i'm looking at pin 13 that's just going to fire once but we'll see the noise
00:31:27 generated um and it shouldn't change much but i think it'll get a little bit bigger when we get closer but that's second order the first order should be about the same and so let's just check and see how we're doing with that it's compiling there we go yep uploading uploading uploading we're done and here's the result wow not don't see a whole lot of impact about the same but the last one is you're right it it looks a little smaller um and you know that's kind of what we expected but i'm going to increase the scale so we can see a little bit better
00:32:09 okay you know what maybe you can see it so i'm not sure why it kind of peaks here maybe because of the you know the the the uh all the i o may not have exactly the same rise time but you can see the closer we get you know that peak value is dropping off a little bit and it's because the fields are cancelling each other right right we're getting the the uh aggressor closer to its the common return path and that's decreasing the fields a little bit second order not a big impact but in this case we're able to see it yeah so we get field lines
00:32:43 from the aggressor are going to be around the return in the opposite direction from the returns current and that's going to help decrease the number of fuel lines in there so another time we can do about field magnetic field lines around conductors because there's a lot of what's the term i i see a lot of people talking about inductance and magnetic field lines and boy is there a lot of confusion out there about the correct way of thinking about it and so um i'll plug my signal integrity book i have another book out
00:33:16 it's been a while signal and power integrity simplified and i tried to be very careful in that book to establish a really good foundation in how to think about uh loops and magnetic field lines and inductance so if anybody wants a preview you can grab a copy of um signal and power integrity simplified um but we can talk about that uh no it's uh uh you get it through amazon it's college by prentice hall um and it's basically you know if you've taken a a sophomore e m class uh even then they don't teach inductance very well
00:33:53 but i tried to describe it clearly in my in my textbook but we'll spend time on that another time but this is the case let's look at that noise this is the case when i've got a common return one i o switching getting closer and closer common return while we're here um let's let's try should we change the return and make it um uh uh a different return um no let's do all of them at once all of them okay let's turn them all on okay so here's what we're gonna do so i'm gonna turn off this and now this is going to turn off turn
00:34:34 on sequentially one after the other so we'll have a lot of i o switching six altogether right um and and so here's what the code looks like and so it's one switching two switching three four five six simultaneously and you have to be careful when you write the code because not all um depending on what device you're using you have to make sure they're switching simultaneously okay so we're compiling so now and we're still looking at a common return that's going to be the worst case we're ever going to see we should see the biggest noise oh my
00:35:11 gosh oh my gosh i'm changing the scale here on the on the scope wow so they're on the same scale look one volt per division for both here's with one of them switching two three four five six look at that i have i have two volts almost four volt peak to peak can we move it round down they have same ground uh okay the purple yeah okay yeah so they're both centered at ground and and now you can see that huge amount of ground bounds it's two volts peak of ground miles peak to peak we're talking uh you know more than three volts of ground
00:35:58 bounce noise as we increase the number of io the didt increases the uh mutual inductance is about the same for all of them and we get that larger and larger amount of ground bounce when you simulate your circuit can you simulate also this oh absolutely but you have to have a good model for the mutual inductance so that's the that's the challenge is you gotta have a good model for the mutual inductance so you need to include some element into your simulation circuit which will uh show the shared pod or something yeah exactly shared
00:36:37 inductance total inductance and it's shared now the best tool out there to take a in this case um uh is you know traces on the board there's also the package leads the best tool out there to take a package or a connector and generate that spice model that has all the shared inductances so you can decide given the path you take for the signal and returns how much ground mouse best tool out there is ansys um q3d it's built into their electronic desktop because it will calculate the inductance matrix and the spice model to allow you
00:37:15 to take into account these factors who is the manufacturer of the software uh ansys they make hmss okay i think they're the world's largest simulation company right now wow okay still this is really important to know because this means when someone for example design two layer pcb and they have let's say some logic gates they may think they have clean signals or ground and this is what may be happening and this is crosstalk on the board this is the crosstalk between the aggressors that are switching here on the board and their shared return and
00:37:57 with the victim line the shared return for the those signal lines right and and in a minute here we're going to switch to this region here and you'll see the dramatic difference when we have um nice wide return path on the board first let's let's try the the unshared path okay let's we'll switch this over now let's you know tell you what let's save this so we can see it so i'm going to go over here and i'm going to say let's save the waveform like two walls this is really big yeah if you have like a reset or something
00:38:34 then it may even like oh absolutely it will absolutely trigger an io yeah um uh okay so there it's saved so we can compare it uh now let's i'm just going to move this switch so that the return is now a separate return and what do you think is going to be the impact this is i call it rule number nine thinking about anticipating what you expect to see so i'm going to change the return what's going to happen to the crosstalk i hope so i'm going to have a separate return right a separate return so this the victim line is going to be signal
00:39:09 return it's going to be a separate loop i'm still going to have all these guys switching they're still going to contribute the idt but it's going to be a lot less because it's not shared okay there we go and now let's see how much oh my gosh i'm going to put this over here and this is just influence of layout layout exactly right and this is just because we have a um separate return right now while we're here let me just save this as well i'm going to save this in m2 here okay so now we have a record of that so this is basically prove
00:39:54 that designing boards is not only about connecting pins right you have to pay just as much attention to the return as you do the signal to minimize ground bounce and that's why i say you know there's you design for connectivity which is just about connecting pins but once you've done that everything about design is all about noise control so you have connectivity and that's what most most layout folks pay attention to this connectivity really important but once you've got connectivity everything else is about noise and ground bounce is so
- Assistant
Bu metinde, bir mikrodenetleyici kartındaki birkaç I/O pini ile yapılan crosstalk ve ground bounce (toprak zıplaması) deneyleri anlatılıyor. Özetle süreç şöyle:
-
Öncelikle, I/O pinleri tek tek ve sonra artan sayıda eşzamanlı olarak açılıp kapatılıyor. Bu, sinyallerin ortak dönüş yolu (common return) üzerinden oluşturduğu manyetik alan değişimlerini ve dolayısıyla paraziti (ground bounce) inceliyor.
-
Deney sırasında, sinyalin paylaştığı dönüş yolunun alanı daraltıldığında, manyetik alanlar birbirini kısmen iptal ediyor. Bu da parazitin biraz azalmasına neden oluyor ama bu etki ikinci dereceden ve büyük değil.
-
I/O’lar eşzamanlı açıldığında ground bounce seviyesi ciddi oranda yükseliyor; osiloskop ekranında 2 voltluk, hatta 4 volt seviyesine varan pik-pik gerilim dalgalanmaları gözlemleniyor. Bu, dönüş yolundaki manyetik alan değişiminin şiddetli olduğunu gösteriyor.
-
Ground bounce’un seviyesi, temel olarak mutual endüktansın (karşılıklı endüktans) ve sinyal değişim hızlılarının (dI/dt) çarpımıyla orantılı. Bu sebeple mutual endüktansın iyi modellenip simülasyonlara dahil edilmesi önemli.
-
En iyi simülasyon araçlarından biri Ansys firmasının Q3D yazılımı; bu yazılım PCB üzerindeki izler ve paket bağlantıları için mutual endüktans matrisini ve ona bağlı SPICE modelini çıkararak gerçekçi simülasyon yapılmasını sağlıyor.
-
Ayrıca, ortak dönüş yolu paylaşıldığında ground bounce ve crosstalk çok yüksek olurken; dönüş yolu ayrıldığında (separate return path) bu parazitler ciddi oranda azalıyor.
-
Yani tek başına sinyalleri birbirine bağlamak yeterli değil, dönüş yollarının sürekliliği, ortaklığı ve tasarımı sinyal bütünlüğü (signal integrity) ve parazitlerin minimal düzeyde olması için kritik.
-
PCB tasarımcılarının bu yüzden bağlantı (“connectivity”) kadar dönüş yolunu da dikkatli planlamaları ve özellikle “noise control” (gürültü kontrolü) konusuna önem vermeleri gerekiyor.
Özetle, deney ve açıklamalar ground bounce’un neden olduğu büyük gerilim dalgalanmalarını, bu dalgalanmaların mutual endüktans ve dönüş yolu üzerinden nasıl şekillendiğini, ve iyi bir PCB tasarımında dönüş yolunun önemini net bir şekilde ortaya koyuyor.
-
- User
daha detaylı açıkla
- Assistant
Tabii, metni daha detaylı şekilde açıklayalım:
Bu metin, mikrodenetleyici ya da dijital devrelerde birden fazla giriş/çıkış (I/O) hattının eşzamanlı veya ardışık olarak anahtarlanması (switching) sırasında ortaya çıkan elektromanyetik etkileşimler, özellikle ground bounce (toprak zıplaması) ve crosstalk (parazit) etkilerini inceliyor.
Deneyin Temel Unsurları
- PCB üzerinde dar (6 mil) paralel izler (trace) var ve bunlar üzerinde dijital sinyaller mikrodenetleyiciden üretiliyor.
- Dönüş yolu veya referans (ground) olarak farklı senaryolar kullanılıyor:
- Ortak dönüş yolu (common return path): Birden fazla hattın dönüş akımı aynı dar hatta veya bozuk temizlenmemiş bir dönüş yolunda paylaşılıyor.
- Ayrı dönüş yolu (separate return path): Hattın kendi bağımsız dönüş yolu oluyor, yani dönüş akımları karışmıyor.
Temel Kavramlar ve Olgular
-
Ground Bounce (Toprak Zıplaması):
Dijital devre sinyallerinin veya I/O pinlerinin açılıp kapanması sırasında, hatlarda ve güç/ground ağında ani akım değişimleri olur (dI/dt). Bu akımlar, PCB dönüş yollarındaki endüktans nedeniyle gerilim düşümüne yol açar. Bu düşüme "ground bounce" denir.
Yani, toprak seviyesinde ani dalgalanmalar oluşur ve sinyal kalitesini düşürür. -
Crosstalk (Parazit):
Bir sinyal hattındaki değişimlerin elektromanyetik alan yoluyla yan hattaki sinyali etkilemesi. Burada indüktif (manyetik alan) ve kapasitif (elektrik alan) etkiler temel rol oynar. -
Mutual İndüktans:
İki iletken arasındaki manyetik alanın birbirini etkileme katsayısıdır. Paylaşılan dönüş yolu ile sinyal döngüsü kapandığında, bu mutual endüktans büyür; birçok sinyal aynı dönüş yolunu paylaştığında ground bounce şiddetlenir. -
dI/dt:
Akımdaki zamanla değişim hızıdır. Dijital sinyaller hızlı yükselip alçaldığı için bu değer oldukça yüksektir, bu da endüktif etkileri artırır.
Deneyin Gidişatı ve Sonuçları
-
Tek I/O pini anahtarlandığında:
- Çok az ground bounce ve parazit gözlemlenir. Tek sinyalin dönüşü tek başına dar dönüş yolunu kullanır, manyetik alanlar sınırlıdır.
-
Birden fazla I/O pini ardışık olarak anahtarlandığında (sequential switching):
- Her sinyal açılıp kapandığında benzer küçük ground bounce oluşur.
- Parazit miktarı büyük oranda değişmez, ama ufak farklılıklar gözlemlenir; örneğin sinyal başına ground bounce biraz azalabilir, çünkü manyetik alanlar bir nebze birbirini iptal edebilir.
-
Birden fazla I/O pini eşzamanlı anahtarlandığında (simultaneous switching):
- Ground bounce seviyeleri ciddi şekilde artar.
- Ölçümlerde peak-to-peak (tepe-tepe) 2-4 volt seviyesinde değişimler görülebilir.
- Çünkü tüm sinyaller aynı dönüş yolunu ortak kullanır ve manyetik alanlar toplanarak toplam endüktansı artırır. Bu da gerilim dalgalanmasının büyümesine yol açar.
-
Dönüş yolunun ayrı ayrı kullanılması durumunda:
- Aynı sayıda sinyal anahtarlandığında ground bounce çok büyük oranda azalır.
- Çünkü sinyallerin dönüş akımları karışmaz, mutual indüktans azalır ve parazit etkisi düşer.
Teknik Detaylar ve Tasarım Önerileri
-
Dönüş Yolu Tasarımı:
Sinyal ve dönüş yolları mümkün olduğunca yakın ve kısa tutulmalı.
Dönüş yolundaki süreksizlikler (delikler, kopukluklar, ortak dar hatlar) mutual endüktansı ve dolayısıyla ground bounce’u artırır.
Bu nedenle geniş ve sürekli bir toprak veya güç düzlemi tercih edilir. -
Mutual İndüktans Modelleme:
Ground bounce’u simüle etmek ve tahmin etmek için mutual indüktansın doğru modellenmesi gerekir.
Bu modeller SPICE gibi devre simülasyonlarına dahil edilir ve tasarım öncesi analiz yapılabilir. -
İyi Simülasyon Araçları:
Ansys’in Q3D gibi gelişmiş araçları PCB üzerindeki izler, paket bağlantıları ve dönüş yolları için mutual indüktans matrislerini hesaplayarak gerçekçi simülasyon sağlar.
Bu sayede tasarımcılar hangi dönüş yolu düzenlemelerinin ground bounce’u düşüreceğini önceden görebilir. -
Yükselme ve Düşme Süreleri:
Sinyal kenar süreleri (rise/fall time) ne kadar kısa olursa dI/dt o kadar büyük olur, bu durumda ground bounce artar. Bu yüzden kenar sürelerine dikkat edilmelidir.
Sonuç
- PCB tasarımı sadece pinlerin bağlanmasından ibaret değildir; sinyal ve dönüş yollarının birlikte, sürekli, mümkün olduğunca kısa ve geniş olması gerekir.
- Dönüş yolu tasarımındaki bozukluklar, ground bounce ve crosstalk riskini önemli ölçüde artırır.
- Tasarımcıların, layout (yerleşim) aşamasında dönüş yolu sürekliliğine, mutual endüktans değerlerine ve sinyal kenar sürelerine dikkat etmesi, sinyal bütünlüğü için zorunludur.
- İyi bir dönüş yolu tasarımı ve yüksek kaliteli simülasyon, yüksek performanslı ve güvenilir elektronik devreler için şarttır.
Bu açıklama, büyük oranda yüksek hızlı dijital ve RF tasarımlarında kritik öneme sahip olan dönüş yolu ve ground bounce kavramlarının detaylarını ve deneysel gözlemlerle bağlantısını ortaya koyuyor.
- User
sadece sana verdiğim metni bana açıkla
- Assistant
Bu metinde, mikrodenetleyici kartındaki birkaç giriş/çıkış (I/O) pininin ardışık ve eşzamanlı olarak anahtarlanması (switching) sırasında ortaya çıkan ground bounce (toprak zıplaması) ve crosstalk (parazit) etkileri anlatılıyor.
Öncelikle, I/O pinlerinden biri sırayla açılıp kapatıldığında, o pinde küçük bir gürültü oluşuyor. Ardından diğer pinler sırasıyla açılıp kapatılıyor ve bu işlem devam ediyor. Bu sırada, pinler arasındaki mesafe yaklaştıkça oluşan gürültü biraz azalıyor, çünkü manyetik alanlar birbiriyle kısmen etkileşerek (iptal ederek) azalıyor.
Sonrasında, tüm pinler aynı anda anahtarlanıyor. Bu durumda ground bounce çok büyük boyutlara ulaşıyor; osiloskopta 2 volt, hatta 4 volt civarında pik-pik dalgalanmalar gözlemleniyor. Bu yüksek gürültünün nedeni, tüm sinyallerin ortak ve bozuk dönüş yolu üzerinden geçmesi ve manyetik alanların toplam endüktansı artırmasıdır.
Ayrıca, mutual endüktansın (karşılıklı endüktans) doğru şekilde modellenmesi gerektiği, bunun için de gelişmiş simülasyon araçlarının (örneğin Ansys Q3D) kullanılması gerektiği belirtiliyor.
Son olarak, dönüş yolunun ayrı tutulması durumunda ground bounce’un önemli ölçüde azalacağı, bu yüzden good PCB tasarımında dönüş yolu sürekliliğine ve temizliğine dikkat edilmesi gerektiği vurgulanıyor.
Metin, sinyal ve dönüş yollarının iyi planlanmasının sinyal bütünlüğü ve gürültü kontrolü için ne kadar kritik olduğunu gösteriyor.
- User
so we're gonna we're gonna change the code that's running so here's the sketch and we we're running the one and then i'm gonna i'm gonna use this code the sequential we turn one on at a time and we just move it down the line so let's turn off this one and i'm going to turn on the sequential okay now we're going to upload that code and so we're going to see the first one switch nothing else switching but we're going to see noise as the others because i'm looking at pin 13 that's just going to fire once but we'll see the noise↳
00:31:27 generated um and it shouldn't change much but i think it'll get a little bit bigger when we get closer but that's second order the first order should be about the same and so let's just check and see how we're doing with that it's compiling there we go yep uploading uploading uploading we're done and here's the result wow not don't see a whole lot of impact about the same but the last one is you're right it it looks a little smaller um and you know that's kind of what we expected but i'm going to increase the scale so we can see a little bit better 00:32:09 okay you know what maybe you can see it so i'm not sure why it kind of peaks here maybe because of the you know the the the uh all the i o may not have exactly the same rise time but you can see the closer we get you know that peak value is dropping off a little bit and it's because the fields are cancelling each other right right we're getting the the uh aggressor closer to its the common return path and that's decreasing the fields a little bit second order not a big impact but in this case we're able to see it yeah so we get field lines 00:32:43 from the aggressor are going to be around the return in the opposite direction from the returns current and that's going to help decrease the number of fuel lines in there so another time we can do about field magnetic field lines around conductors because there's a lot of what's the term i i see a lot of people talking about inductance and magnetic field lines and boy is there a lot of confusion out there about the correct way of thinking about it and so um i'll plug my signal integrity book i have another book out 00:33:16 it's been a while signal and power integrity simplified and i tried to be very careful in that book to establish a really good foundation in how to think about uh loops and magnetic field lines and inductance so if anybody wants a preview you can grab a copy of um signal and power integrity simplified um but we can talk about that uh no it's uh uh you get it through amazon it's college by prentice hall um and it's basically you know if you've taken a a sophomore e m class uh even then they don't teach inductance very well 00:33:53 but i tried to describe it clearly in my in my textbook but we'll spend time on that another time but this is the case let's look at that noise this is the case when i've got a common return one i o switching getting closer and closer common return while we're here um let's let's try should we change the return and make it um uh uh a different return um no let's do all of them at once all of them okay let's turn them all on okay so here's what we're gonna do so i'm gonna turn off this and now this is going to turn off turn
00:34:34 on sequentially one after the other so we'll have a lot of i o switching six altogether right um and and so here's what the code looks like and so it's one switching two switching three four five six simultaneously and you have to be careful when you write the code because not all um depending on what device you're using you have to make sure they're switching simultaneously okay so we're compiling so now and we're still looking at a common return that's going to be the worst case we're ever going to see we should see the biggest noise oh my 00:35:11 gosh oh my gosh i'm changing the scale here on the on the scope wow so they're on the same scale look one volt per division for both here's with one of them switching two three four five six look at that i have i have two volts almost four volt peak to peak can we move it round down they have same ground uh okay the purple yeah okay yeah so they're both centered at ground and and now you can see that huge amount of ground bounds it's two volts peak of ground miles peak to peak we're talking uh you know more than three volts of ground 00:35:58 bounce noise as we increase the number of io the didt increases the uh mutual inductance is about the same for all of them and we get that larger and larger amount of ground bounce when you simulate your circuit can you simulate also this oh absolutely but you have to have a good model for the mutual inductance so that's the that's the challenge is you gotta have a good model for the mutual inductance so you need to include some element into your simulation circuit which will uh show the shared pod or something yeah exactly shared 00:36:37 inductance total inductance and it's shared now the best tool out there to take a in this case um uh is you know traces on the board there's also the package leads the best tool out there to take a package or a connector and generate that spice model that has all the shared inductances so you can decide given the path you take for the signal and returns how much ground mouse best tool out there is ansys um q3d it's built into their electronic desktop because it will calculate the inductance matrix and the spice model to allow you 00:37:15 to take into account these factors who is the manufacturer of the software uh ansys they make hmss okay i think they're the world's largest simulation company right now wow okay still this is really important to know because this means when someone for example design two layer pcb and they have let's say some logic gates they may think they have clean signals or ground and this is what may be happening and this is crosstalk on the board this is the crosstalk between the aggressors that are switching here on the board and their shared return and 00:37:57 with the victim line the shared return for the those signal lines right and and in a minute here we're going to switch to this region here and you'll see the dramatic difference when we have um nice wide return path on the board first let's let's try the the unshared path okay let's we'll switch this over now let's you know tell you what let's save this so we can see it so i'm going to go over here and i'm going to say let's save the waveform like two walls this is really big yeah if you have like a reset or something 00:38:34 then it may even like oh absolutely it will absolutely trigger an io yeah um uh okay so there it's saved so we can compare it uh now let's i'm just going to move this switch so that the return is now a separate return and what do you think is going to be the impact this is i call it rule number nine thinking about anticipating what you expect to see so i'm going to change the return what's going to happen to the crosstalk i hope so i'm going to have a separate return right a separate return so this the victim line is going to be signal 00:39:09 return it's going to be a separate loop i'm still going to have all these guys switching they're still going to contribute the idt but it's going to be a lot less because it's not shared okay there we go and now let's see how much oh my gosh i'm going to put this over here and this is just influence of layout layout exactly right and this is just because we have a um separate return right now while we're here let me just save this as well i'm going to save this in m2 here okay so now we have a record of that so this is basically prove 00:39:54 that designing boards is not only about connecting pins right you have to pay just as much attention to the return as you do the signal to minimize ground bounce and that's why i say you know there's you design for connectivity which is just about connecting pins but once you've done that everything about design is all about noise control so you have connectivity and that's what most most layout folks pay attention to this connectivity really important but once you've got connectivity everything else is about noise and ground bounce is so
- Assistant
Bu metinde, bir mikrodenetleyici kartındaki birkaç I/O pininin (giriş/çıkış pini) tek tek veya aynı anda anahtarlandığında (switching) ortaya çıkan "ground bounce" (toprak zıplaması) ve crosstalk (parazit) etkileri inceleniyor.
Öncelikle, I/O pinlerinden biri sırayla açılıp kapatıldığında, sinyaller arka arkaya tetikleniyor. Ölçümde, sadece tek I/O pini anahtarlanırken diğerleri kapalı oluyor ama gözlemlediğimiz pin (örneğin pin 13) sinyaller diğer pinlerin anahtarlanmasından dolayı hafif bir gürültü alıyor. Pinlerin yakınlaştıkça bu gürültü biraz azalıyor çünkü bazı manyetik alanlar birbirini kısmen iptal ediyor (elektromanyetik alanların etkileşimi). Ancak bu azalma ikincil (ikinci dereceden) bir etki ve büyük bir fark yaratmıyor.
Daha sonra, aynı anda tüm I/O pinleri (örneğin altı adet) eşzamanlı olarak anahtarlanıyor. Bu durumda osiloskopta pik-pik 2 volta varan (hatta 4 volta yaklaşan) çok büyük ground bounce etkisi görülüyor. Yani ortak dönüş yolunu paylaşan bu çoklu sinyallerin aynı anda açılıp kapanması, manyetik alanların birikmesine ve büyük gerilim dalgalanmalarına yol açıyor.
Ground bounce etkisinin büyüklüğü, sinyallerin akım değişim hızı (dI/dt) ile karşılıklı endüktansın (mutual inductance) çarpımı ile orantılı. Bu nedenle simülasyonlarda bu mutual inductance’in doğru modellenmesi çok önemli. Mutually endüktans etkisi simülasyonlarda yer almazsa gerçekçi sonuçlar alınamaz.
Bu modellemeyi yapmaya yarayan en iyi yazılımlardan biri olarak Ansys firmasının Q3D aracı gösteriliyor. Q3D ile PCB üzerindeki izlerin, paket bağlantılarının karşılıklı endüktansları hesaplanabiliyor ve SPICE modelleri oluşturulabiliyor. Böylece sinyal ve dönüş yolu ortaklığı, sinyal bütünlüğü ve ground bounce etkileri gerçekçi şekilde tahmin edilebiliyor.
Metin, ayrıca dönüş yolunun ortak kullanılmasının (shared return) sinyal parazitlerini artırdığını, bunun tersine sinyal dönüş yolunun ayrı tutulmasının ground bounce ve crosstalk’u ciddi miktarda azalttığını pratik örneklerle gösteriyor. Bu yüzden PCB tasarımında dönüş yolunun sürekliliğine, ayrık ve temiz olmasına özen gösterilmesi gerektiğini vurguluyor.
Sonuç olarak, devre tasarımında sadece pinleri birbirine bağlamak yeterli değil; sinyal ve dönüş yollarının doğru ve iyi planlanması sinyal kalitesi ve gürültü azaltma açısından zorunludur. Tasarımın büyük kısmı aslında gürültü kontrolüne odaklanmalıdır.