What Every PCB Designer Should Know - Crosstalk Explained (with Eric Bogatin)

Summary

In this engaging discussion between Robert Feranec and Eric Bogatin, the complexities of crosstalk in PCB design are elucidated. Crosstalk arises when unintended electrical noise interferes with signals on a circuit board, often due to improper design choices. Eric explains the phenomenon using the interaction of electric and magnetic fields and demonstrates the concept through animations. The conversation highlights the intricacies of capacitive and inductive coupling and their effects on signal integrity, especially relevant in two-layer PCBs, where maintaining a solid ground plane is challenging.

Highlights

• Robert Feranec shares personal crosstalk experiences in PCB design. 🤔
• Eric Bogatin uses animations to demystify the dynamics of crosstalk. 🔍
• The fundamental causes of crosstalk are explained through electric and magnetic fields. ⚡
• Strategies to reduce crosstalk, like moving return paths closer, are discussed. 💡
• Eric provides invaluable insights into capacitive and inductive couplings' impact on PCB performance. 📊

Key Takeaways

• Crosstalk in PCBs can lead to unexpected behavior in circuits, such as untriggered interrupts. 🚨
• Two-layer PCBs are particularly prone to crosstalk issues due to difficulty in maintaining a solid ground plane. 🛠️
• Crosstalk is caused by fringe electric and magnetic fields between aggressor and victim lines. 📡
• Understanding the roles of capacitive and inductive coupling can help manage crosstalk effectively. 🧠
• Animations can greatly aid in visualizing and understanding crosstalk dynamics. 🎥

Overview

Robert Feranec dives into a common yet perplexing problem in PCB design: crosstalk. He recounts his initial struggles with unexpected behaviors in his circuits, attributed to the sneaky nature of crosstalk. To tackle this, he teams up with the expert Eric Bogatin, who brings clarity to the subject with engaging explanations and animations.

Eric Bogatin takes the stage by breaking down the science of crosstalk into understandable bits. He explains how fringe electric and magnetic fields cause crosstalk and demonstrates the concept using both capacitive and inductive coupling models. His vivid animation helps viewers visualize how these fields interact and influence each other on a PCB.

The conversation capstones with practical tips on managing crosstalk, especially in the challenging two-layer PCB design. Eric reveals key strategies, such as altering return path distances to mitigate crosstalk effects, and Robert enjoys a lightbulb moment, underscoring the importance of visual tools in mastering intricate design concepts.

What Every PCB Designer Should Know - Crosstalk Explained (with Eric Bogatin) Transcription

• 00:00 - 00:30 so have you ever had problems with crosstalk when i was starting with hardware design my very first problem which i had no clue what is happening on my board was crosstalk basically the microcontroller on the board was receiving uh some interrupt requests which were never happening and then i designed another board
• 00:30 - 01:00 where i had a lot of problems with crosstalk it was a two-layer pcb where i routed a lot of tracks close to each other and the board was sometimes freezing and crashing and resetting and i had no idea what is happening and i have seen so many pcbs which are routed this way and i still see people routing two layer pcbs with tracks which
• 01:00 - 01:30 are just every single track is close to the other one and i believe it's like super important to understand at least a little bit about crosstalk so i decided to create a video about this topic and i was talking to eric bogatin and he has like a really cool animation which explains when crosstalk happened and what exactly
• 01:30 - 02:00 it is and when you will see this animation you will be like oh so this is how crosstalk appears on my boards and that's basically what we are going to talk about so in this video i'm going to talk to eric bogatin we had a call which i recorded
• 02:00 - 02:30 in this call eric is going to initially explain a little bit what is behind the theory of crosstalk and then how also he will play this animation and explain actually what crosstalk is my name is robert feranek i'm from federal academy i really hope you will find this video interesting and useful here it is here is the video from my call
• 02:30 - 03:00 with eric bogatin so so fundamentally crosstalk is all about electric and magnetic fields and when we talk about crosstalk there's always one or multiple signal and return paths that are the aggressor that are causing the the noise and there's a victim on which we see that noise uh and uh that the fundamental root
• 03:00 - 03:30 cause for all crosstalk is always a combination of fringe electric fields between the signal and the uh the aggressor and the victim and fringe magnetic fields and if if all the you know if you put a signal on the signal and the return if all the electric fields were confined in the vicinity of the signal in return so none of them spread out there would be no crosstalk it's and we call these fields that spread out we call them fringe fields and so
• 03:30 - 04:00 crosstalk is all about fringe electric and fringe magnetic fields you need a lot of math to understand electric and magnetic fields and in fact if you want to do a practical problem gosh it's really hard to you know get a paper and pencil and match those equations and write down equations solve the differential equations get the fields and do something with them and so we approximate the behavior of electric and magnetic fields with circuit elements but keep in mind that the real way the world works the the
• 04:00 - 04:30 real effects that are going on is related to the fields we're only approximating that behavior with circuit elements and how do we approximate the electric field coupling do you know what the circuit element that approximates how electric fields couple on its capacitors yes exactly right yeah you have that very good intuition and likewise if we have um fringe magnetic fields from one signal return path over here to a
• 04:30 - 05:00 victim line over here it's the magnetic field this is the inductor it's inductor right mutual inductance right and so when we describe crosstalk we will sometimes approximate the electric field coupling with a capacitive coupling and the magnetic field coupling with inductive coupling so those are the words that connect those fields but just looking at the field lines if you imagine and so the other important principle is how how how how do you get crosstalk
• 05:00 - 05:30 through electric fields i mean here here's this electric field from the signal return over here and it's going to the victim line how does the electric field couple anything over to the victim line well if you think of it capacitively then how do you get current if there's if we approximately electric field is a little capacitor how do you get current through a capacitor it's the only way there are different voltages uh more than just different i mean and the different voltage gives you feels but even if you have a
• 05:30 - 06:00 constant voltage across the capacitor there's no current flowing through the capacitor what does it take in order to get current through get energy flowing through a capacitor change a change right and so it's the changing electric fields that if it's modeled as a capacitor the changing voltage across the capacitor that you get capacity coupled currents as a form of crossover i feel like i have the exam it is an oral exam but you know what you
• 06:00 - 06:30 have very good intuition about this and so i have high confidence you're going to get everything correct the and here's how to think about and this is one of the most important concepts in the electric field perspective for crosstalk i have these electric field lines that i've drawn between the aggressor and the victim line they are we approximate them by capacitance but as electric field lines when i have a changing voltage it's the voltage difference that gives me the feelings but if that voltage difference changes
• 06:30 - 07:00 the electric field changes and a changing electric field is equivalent to a current we're used to thinking about current as little moving charges in a conductor that's conduction current that's the electrons moving in in in copper but there's another kind of current as well and that's a changing electric field and that's how you get current through a capacitor because in a capacitor remember they're two conductors insulating dielectric
• 07:00 - 07:30 you don't have little charges moving through the insulating dielectric so how do you get current through a capacitor is because of the changing voltage which means a changing electric field and we call ashes james clark maxwell that that that um named that new kind of current that flows through the electric field lines when those it's called yeah is it the displacement current exactly you see you are on a roll
• 07:30 - 08:00 yeah it is displacement current and so here's what you need to do you need to visualize when you look at the electric field lines you need to visualize those electric field lines because when they change when the voltage changes you're going to have it's not really charges but it's still a current flowing through those field lines from the aggressor to the victim and we call that displacement current and it is just as real and behaves exactly the same way as conduction current even though it's through space literally through space or through the dielectric so that's so that so
• 08:00 - 08:30 whenever you see electric field lines you're going to get displacement current when the field lines change when the voltage changes and likewise with magnetic fields you can have static magnetic fields when there's a constant current when do we get induced voltage on the victim line from the magnetic field lines only when it is changing only when the current the magnetic fields change and they change because of a changing current exactly right and so it's the changing rings of magnetic field lines the changing number
• 08:30 - 09:00 of rings of field lines that surround the signal path the the victim line it's when those rings of feelings change that we get induced voltage then induced voltage drives the current down the line so it's the magnetic fields and how many of these rings of magnetic fields go around the victim line is a measure of how much inductive coupling how much magnetic field coupling and so just thinking about it in terms of the fields if we want to reduce the crosstalk we want to reduce
• 09:00 - 09:30 the number of field lines that go between the victi the aggressor and the victim electric field and magnetic fields and there are two ways we can engineer the interconnects to reduce the number of electric field lines and magnetic field lines between the aggressor and the victim what is one of those ways it away exactly right so we can pull the aggressor the victim away from the aggressor reduces the number of fringe electric
• 09:30 - 10:00 fields reduces the number of fringe magnetic fields and that reduces crosstalk there's another way as well uh we could move the reference plane closer and what would that do you're right move the reference plane closer and it will reduce the fields make them smaller what we do by adjusting the return plane is we sculpt those fringe field lines by bringing the return plane closer we can find the fringe electric feelings closer to the vicinity of the aggressor
• 10:00 - 10:30 they don't spread out as much and same thing with magnetic field lines if we bring the return path closer we constrict those rings the felines fewer of them go around the victim line and we reduce the magnetic field coupling or the inductive coupling i'm going to interrupt eric because there is something super important what i would like to point out notice in these examples
• 10:30 - 11:00 we always have this reference plane what does it mean a reference plane for example in your pcb what can be usually a really good reference plane nice and solid ground plane so for example when you are routing a pcb then you would like to have tracks on one layer and then on the other layer maybe you would like to have this nice
• 11:00 - 11:30 and solid ground plane because that's the situation what we can see here because this nice and solid ground plane can help us for example control these fields if we move the ground plane further from the tracks then the fields are going to be bigger and when we move this ground plane closer under our tracks then these fields are
• 11:30 - 12:00 going to be smaller now the question is what about two layer pcb if you are like very lucky you may be able to route all the tracks on the top layer of two layer pcb and you may be able to create nice solid ground plane on the bottom layer of two layer pcb
• 12:00 - 12:30 however the distance between top layer and bottom layer onto layer pcb is huge and even if you would have this nice solid ground plane on two-layer pcb these fields are going to be very large in reality there is almost never a way
• 12:30 - 13:00 to create nice solid ground plane in two layer pcb in reality usually there is no nice uh reference plane on two layer pcb so if we remove this reference plane from this picture what is going to happen with these fields they are going to spread all around these fields are going to be picked up with all the other traces on your pcb
• 13:00 - 13:30 so that's one of the reasons why two layer pcbs are so bad because you cannot have this nice solid ground plane which can help you for example to to control these fields okay i wanted to point out this because uh when we talk about these pictures we use always this reference plane and uh
• 13:30 - 14:00 i wanted everyone to be aware that this what we are talking about it applies only in these situations when there is this reference plane if there is no reference plane everything is going to be much much worse okay let's go back to our talk with eric and let's listen what eric is going to say next here it is here's an example of
• 14:00 - 14:30 dynamically how that crosstalk looks so here's here's what i've got and i've kind of exaggerated the drawing a little bit but here's a trace on a board aggressor here's a trace on a board a victim line here's the return plane and i've kind of you know hand drawn in here's the capacitance between the signal and return and here's that coupling capacitance you've got to imagine that the fringe fields between the aggressor and the victim they're creating this distributed capacitive coupling all the way down the line we're just looking at
• 14:30 - 15:00 you know one little region over here and the capacitive coupling between it right so when here here's that edge and i've drawn this purposely in an example showing what i would call an electrically long interconnect electrically long meaning that the rise time is short compared to the physical length of the interconnect so example i get a nanosecond edge then electrically long is like six inches because the edge is moving at six inches a nanosecond
• 15:00 - 15:30 nanosecond long its spatial extend is about six inches in extent so if it's shorter rise time then it'll be a smaller spatial extent and and that just makes it easy to think about the this crosstalk from an aggressor to a victim and so here's this this edge moving by when do i get the noise flowing between the aggressor and the victim through the coupling capacitance when is the only instant of time that i'm going to have noise flowing from an aggressive victim through the coupling capacitance
• 15:30 - 16:00 when the change is happening when the change happens exactly right so at this instant of time right now here's where the edge is here's the capacitor i'm not getting anything flowing through here it's only when this edge passes by this region that i'm going to get that capacity coupled current flowing from the aggressor to the victim and as soon as i get that that dv dt across the changing voltage across the capacitor and i get that burst of displacement current through the capacitor and it comes over here and
• 16:00 - 16:30 it's now on the victim line i get capacity between the signal and return of the victim line that current flowing over is going to build up some charge here i'm going to get a voltage now i've got a a voltage pulse a voltage signal appearing on the the signal and return of the victim line what's that voltage pulse going to do it will travel with the edge it will travel it will propagate exactly and which way will it go will it go in the forward direction the same direction
• 16:30 - 17:00 the signal is moving or we'll go in the backward direction the opposite direction it looks like for capacitance it will go both direction yes it will split because it's just a voltage pulse it's like you you added a a little voltage source between the signal and return and and so it's going to go in both directions and and that voltage from the capacitor of the current is going to travel signal return in one direction signal return in the other direction and so we're going to
• 17:00 - 17:30 get it at this end and just to distinguish the direction so so here's this case here i should have i should have gotten my directions right so i've got this pulse it's moving down this way down there yeah yeah i know yeah and so here's where we get the di dt or the dv dt we get the current in here we get some of the the signal return current going in this direction some in this direction to describe this direction or that direction we refer to the source where the signal is coming from on the aggressor we call that the source that becomes the
• 17:30 - 18:00 forward direction and this is the forward direction of the current this is the backward direction or because this is also where the source is this direction is toward the near end of the source and this direction on the victim line is the far end far from the source so we use those terms interchangeably backward or near forward or far so the capacity coupled current is going to flow between the signal and return it's a voltage pulse
• 18:00 - 18:30 that's signal return signal return in the back direction signal return signal return in the forward direction that's the capacity coupled current and so we'll see it eventually when it gets to the near end and as it propagates in the forward direction the aggressor is moving in the forward direction and so they're kind of coincident with each other and so in the backward direction i get this little bit of capacity couple current that's going to dribble back to me the aggressor is going to move forward we're going to get capacity public current it's going to dribble back moves forward get a little bit more
• 18:30 - 19:00 dribbling back and if i sit over here at the near end i'm going to see this capacity couple current kind of dribbling back dribbling drop back dribbling back now it's a little hard to visualize it's it's a very much a dynamic kind of thing really hard to visualize on this flat screen and just moving my mouse and so i've got this really cool animation that shows how this capacity couple current behaves but this is the principle you get displacement current going to the quiet line that builds up some charge makes a voltage that voltage signal propagates
• 19:00 - 19:30 you get current signal return current loop going the backward direction signal turn in the forward direction so here is the the animation so this was done by uh so uh one of my buddies at telangana yoshi tetsuy he is an ace at writing flash animation and he developed this flash animation to describe crosstalk and so here's what we've got here's a aggressor line here's a victim line
• 19:30 - 20:00 and and here is the region in which they're coupled and i can change that region make it big and we can make it short so make it really small just a small region in which they're getting close together normally they're far apart they get close together here and they move apart so they're only coupled in this little region here and so i have this capacit we're going to look at capacity coupled current first i have this capacitive coupling between the aggressor and the victim line here and i'm going to i'm going to send a signal down we look at the edge of that signal and this is going to
• 20:00 - 20:30 be a plot of the voltage versus position as we send it down and it's in this region that i'm going to get capacity coupled uh noise and once it gets over here it's going to propagate in the backward direction and the forward direction and we'll look at that voltage that propagates okay so that's the setup and and i'm going to use a a strip line so the capacitor and dr kepler are the same so let's run that animation oh i'm sorry i need to do that here it is so here's that edge it hits the coupled region and when we
• 20:30 - 21:00 hit that coupled region we get a little bit of capacity couple voltage moving that way a little bit moving that way okay so let me see if i can slow it down a little bit so here's that it's only look we're not getting anything nothing nothing only where the edges and after the edges pass by nothing nothing nothing we get a burst of crosstalk moving in the backward direction of the four direction equal amount they're both positive because there's that voltage that um that drives the
• 21:00 - 21:30 signal return in both directions same amount if i increase the coupled region then i'm going to have that same burst generated where i have that little bit of coupling capacitance and then another region another region another region every place where that edge encounters that capacitor is going to generate a little bit of backward going noise and forward growing noise but as the backward going noise propagates in backward direction the aggressor is moving forward it's encountering a new region and we need another amount of backward going so it will increase it
• 21:30 - 22:00 will it will dribble back in the forward direction as i get my little burst of capacity coupled current in the forward direction it moves forward the signal moves forward and that's where it's going to snowball so let's watch it now so i've gotten a little bit more capacitive coupling here let's look and see so here's this region so the edge encounters the region i get a little bit a little bit a little bit i get it but they're all on top of each other in the forward direction
• 22:00 - 22:30 and so the signature of the noise we see on the victim line looks different when we look at the noise propagating in the backward direction we see over here at the near end versus in the forward direction that we see over here at the far end so let's do that again so you can see the backward going noise it's it here it is it's dribbling back dribbling back dribbling back it looks like a constant steady amount in the forward direction in the forward direction
• 22:30 - 23:00 they're being generated coincident with each other on top of each other as the aggressor moves it gets bigger and bigger and bigger the more the coupled region until it all comes out we get a big amount let's watch and i think for people we can say the coupling region is basically the region which uh we consider that when the tracks influence each other right they're far apart they come together because they're necking down
• 23:00 - 23:30 and then they're they're pulling apart so this is the region when they're close together and now i've made it really long and now we look at the capacitor coupled current and we're going to get a steady amount of current generator voltage signal going back dribbling back drill back and look it's this constant steady amount even though we're increasing the coupled region we're not increasing the amount the magnitude of the near end crosstalk we're increasing how long it lasts but not the magnitude whereas the forward direction look what happens the longer the coupled
• 23:30 - 24:00 region the more opportunity we have for that far noise to get bigger and bigger and bigger and it's just growing we get a large pulse coming out very different signature in the backward direction than the forward direction but is it going to grow forever no but it will grow so that the the magnitude of the far end noise could be comparable to half of the signal and long before it gets that big the technical term for that situation is
• 24:00 - 24:30 you're screwed because as that as that far end noise grows it's going to get here it is it's going to get bigger and bigger and bigger and at some point that's going to be big enough to cause real problems and it will get big enough to cause problems way before it saturates and so that's and that's why foreign crosstalk when it happens can be much more important than near-end crosstalk because it can be so much larger so you say half of the
• 24:30 - 25:00 signal yeah they can gradually large is half of the signal right okay and and and but we're only looking at half of the foreign crosstalk because we only looked at the capacitively coupled current i said that when you have the nice wide plane the relative amount of capacitive and inductively coupled ground are comparable they're both important we have to worry about both but here's where the subtlety i would like to ask why is it how
• 25:00 - 25:30 so um imagine what's so going back to this drawing here as the signal is moving forward the energy that's coupling from the aggressor to the victim and is building up and getting bigger and bigger in the victim line where's that coming from where's that energy coming from this that's moving down the the the victim line ah okay it will be half because it kind of splits
• 25:30 - 26:00 the between these two yeah so it's gonna well it's gonna it's gonna come from the aggressor so that means we're gonna distort the edge of the aggressor if we get so much far end crosstalk that it becomes you know close to half the magnitude of the signalling over here that means we have taken so much energy out of the the aggressor line that we've distorted this edge and now that same edge that is half the voltage over here that's the same magnitude that i have
• 26:00 - 26:30 flowing this way it's going to act like an aggressor and it's going to go over and so they equilibrate roughly to like a half roughly but long before that happens you're screwed because you have so much fire and crosstalk here it's going to look like a bit error but that's only half of the crosstalk it's only capacitive copy now we look at inductive coupling so here's this edge and here's some mutual inductance between an aggressor and a victim so this is where it's really gosh it's really uh
• 26:30 - 27:00 hard to conceptualize what goes on here this is where understanding electric and magnetic fields is really valuable because we have a loop here signal return loop for the aggressor and we have a loop here a signal return for the victim when i if i have a constant current there's no change in magnetic field there's nothing induced when i have zero current there's no change in magnetic field there's nothing induced it's only when i have that changing that edge i have that changing current in the rise
• 27:00 - 27:30 time on the aggressor where i have a di dt over here a change in magnetic field and that generates the changing voltage over here and that's generates a changing a a current over here and here's where the very very very subtle part comes in suppose that my current increases this way in the counterclockwise direction because here's this edge the current gets bigger and bigger and bigger increasing in the counterclockwise direction right over here that's going to generate a changing
• 27:30 - 28:00 voltage over here and that's going to in this victim loop and that's going to generate an induced current and now comes the very subtle question if i have a di dt increasing the clockwise i'm sorry this is the counterclockwise direction in the aggressor what's the direction of the induced current loop in the victim is it going to be increasing in the same direction or in the opposite direction so again an aggressor loop increasing the counterclockwise direction over here generating change in magnetic fields
• 28:00 - 28:30 generates changing voltage generates changing current is the induced current going to be going in the same direction circulating in the same direction or circulating in the opposite direction i guess it's the arrow there huh is it the arrow there it it it is i kind of gave you the answer okay so here's this edge moving down you get a aggressor current increasing the circulating in the counterclockwise direction it's still propagating in the forward direction
• 28:30 - 29:00 just circulating in the counterclockwise direction and now over here on the poor victim line the edge goes by we get mutual inductance in this region we're going to get a clockwise induced circulating current loop over here and now comes the second really subtle point what direction is that induced current loop in the quiet line what direction is that loop going to propagate it's going to propagate this way and be a signal return signal return
• 29:00 - 29:30 signal return or is it going to propagate this way and be a return signal return signal return signal which way will it propagate this loop of current that was just excited is going to propagate in the backward direction in the forward direction and when it goes in the backward direction it's going to be a signal return signal return signal just like the capacitor coupled current you're going to be exactly there's going to be a little positive burst of voltage moving in the backward direction from the inductively coupled current
• 29:30 - 30:00 in the forward direction it's going to be a return signal a return signal a return signal return signal and if if signal return is a positive voltage what kind of voltage signature is a return signal direction of circulation it's going to be a negative voltage and so in the forward direction we have a positive voltage burst that's moving back and in the four direction we have a negative voltage pulse moving in the forward direction from the inductively coupled
• 30:00 - 30:30 current and again really hard to visualize let's see the animation the animation comes in right so i'm going to go back i've got lost yeah it's it's unless you have look this is where thinking about magnetic field lines and how they're the induced current is really valuable so so here's that very short coupling region and and here's that edge that's going to come by we have just a very narrow region which they're coupled and we're going to look instead of the
• 30:30 - 31:00 capacity coupled current we're going to look at the inductively coupled current and so what we expect to see is nothing nothing nothing until the edge gets over here we have a signal return circulating in in this orientation here we're propagating this way it's circulating in the clockwise direction the induced current inductively coupled current in the quiet line is going to be going the opposite direction that's lenses law and we're going to get half of it moving in the forward direction half of it moving in the backward i'm
• 31:00 - 31:30 sorry half of it moving in the back direction half of it moving in the forward direction and so that's going to be a signal return signal return signal return moving backwards that's a positive voltage and it's going to be a return signal return signal return signal return signal negative voltage in the forward direction so let's watch that so here's that edge nothing nothing nothing here we get that little positive foreign
• 31:30 - 32:00 positive inductively coupled current moving in the back direction and we get that little negative burst of of uh voltage moving in the forward direction and so once again we see a very different signature in the backward direction is the forward direction for inductively coupled current remember what it was for capacitive coupling it was both positive in both directions right but for inductive coupling
• 32:00 - 32:30 it's positive in the backward negative in the forward direction and if we increase the coupled region the behavior the signature is going to be exactly what we saw for capacitive coupling let's make this a little bit bigger so let's look at that inductive coupling so it's going to be we get that little positive burst and it's going to be dribbling back dribbling back dribbling back dribbling back the forward direction it gets bigger and bigger it's snowballing it gets bigger more negative and if i make it really
• 32:30 - 33:00 big again we're only looking at the inductively coupled noise if we make it really big we're going to see in the backward direction in the near end when we're sitting over here a constant steady amount lasting for some amount of time the round trip time constant steady amount independent of the coupling length but in the forward direction oh my gosh it got bigger and bigger and bigger bigger and we get this large burst coming out coincident with when the edge but now i'm going to be really
• 33:00 - 33:30 confused because it looks like it's going to eliminate each other what an astute observation you are exactly right if the capacitive coupled noise and the inductively coupled noise are of the same magnitude then they're going to cancel out in the forward direction let's take a look let's do one section now we're going to look at both so we looked at capacity we looked at inductive now we're going to look at the total
• 33:30 - 34:00 and so we're going to have the capacity coupled noise in the forward direction here it is in red and the green is the inductively coupled noise both the same magnitude and we get twice the amount in the forward direction wait a minute what happened what happened in the forward direction we get a little bit of positive capacity coupled current we get a little bit of negative inductively coupled current they're exactly cancelling out there's nothing in the forward direction and in fact if we make it
• 34:00 - 34:30 really long coupling region then the capacity and inductively coupling will dribble back dribble back dribble back to black but the capacitor inductive couple in the four direction get bigger and bigger but they're exactly the same they cancel out there is no far end crosstalk okay and now you know now there is this question then why we are talking about crosstalk ah because well number one we have near-end crosstalk and near-end crosstalk is a problem and
• 34:30 - 35:00 this is in the special case of a strip line when i have two planes and i have nice homogeneous material in between so in the stack up i got the same dielectric you know above and below same material above and below in that environment the inductive and capacitive coupling are the same they're opposite sign they cancel out no foreign crosstalk okay we still have ideally you would like to route all the traces between two grounds yeah if we care about far end crosstalk
• 35:00 - 35:30 and if you want to eliminate foreign crosstalk you are exactly right you want to use strip line geometry but they know the distance needs to be kind of same from the point now the only thing that's important is you get the same dielectric material above and below which is normally you know you have a core and you have pre-peg they're close they're very close and so you'll mostly eliminate foreign crosstalk if we
• 35:30 - 36:00 don't have that if we have a micro strip surf because you always have surface traces and if you look at a lot of especially four layer boards you'll find long parallel buses routed in the microscope traces that are really close together so here's the case doing the same analysis with microstrip the only difference is the relative amount of capacitive inductive coupling when we i'm going to make our coupling region really short
• 36:00 - 36:30 so in this environment there's going to be a little bit more inductive coupling than capacitive coupling and so here's the capacitive coupling and we get in a little positive burst a little positive earth just like we saw before here's the inductive coupling we're going to get a little positive burst going in the back direction negative in the forward direction but these the capacitive and inductive coupled noise not quite the same there's a little bit more inductive coupling
• 36:30 - 37:00 so when we look at the total they don't cancel out they add in the forward direction so i get more they don't quite cancel out in the forward direction and that means that when i have a long coupled region i'm going to have in microstrip surface traces where i have more inductive coupling the indefinite coupling is larger i have a net negative voltage in the far end so in microstrip far end
• 37:00 - 37:30 can be large it scales with length so the longer the coupling length the region the more the foreign crosstalk and it scales inversely with rise time so is it going to be again like maximum voltage yeah but again long before then you're screwed long before the foreign crosstalk is you know half the signal you you've got so much noise that now you're getting false triggering and uh why this is happening is it
• 37:30 - 38:00 because uh half of the environment is air and the other one is pcb in that environment so so one of the things you learn when you take a fields class is electric fields interact with dielectrics right there's the dielectric constant tells you how much is the electric field in the material kind of amplified a little bit by the dielectric magnetic fields on the other hand don't interact with dielectrics at all and so whether i have the in in
• 38:00 - 38:30 in strip line or microstrip regardless of what my distribution of dielectric is i don't change the magnetic fields the magnetic field coupling is the same but when i when i add uh when i take away the dielectric above the um the signal lines to go from strip line so here's the strip line i have dielectric up above if the electric and magnetic field coupling capacitance is the same in this environment
• 38:30 - 39:00 and i take away the top dielectric what's happened to the magnetic fields and the magnetic field coupling nothing they're not affected by the presence of the dielectric so in this environment i have exactly the same magnetic field coupling between these as in this environment so inductive coupling stays the same but when i had the dielectric up above here what happens to remember the capacitive coupling is all about these fringe field lines when i have the dielectric up here i
• 39:00 - 39:30 have extra coupling between these fringe field lines because of the dielectric they've kind of amplified the the the electric field a little bit and so i have a lot of electric field coupling that makes the capacitive inductive coupling comparable amount when i take the dielectric away and i look at microstrip those same field lines they don't have the dielectric here i have less coupling so here's the counter-intuitive principle i decreased
• 39:30 - 40:00 the capacitive coupling and yet i've increased the foreign crosstalk because i've made the capacitive and inductive coupling not equal anymore they're not cancelling each other out by decreasing the capacitive coupling i've actually increased the far end crosstalk because i don't have that capacitance to cancel out the inductance and that's why it's the inhomogeneity of the dielectric material the fact i have air and dielectric here that gives a different amount of
• 40:00 - 40:30 coupling capacitance and that's why we have foreign crosstalk okay and so we have these two behaviors something different at the near end versus the far end and that's why we call it near and crosstalk and far and crosstalk to distinguish those this is very interesting because when i was thinking about making this call i imagine the like standard pulls and then you know some ringing or something and i was like oh this is the crosstalk what we can see on
• 40:30 - 41:00 the other weak theme and i was not really thinking about crosstalk this way so thank you very much there are two things that complicate the simple analysis this is exactly what happens this is the coupling this is what you measure you know i do this in my class quite a bit we build a couple of transmission lines we send a sep edge in we look at the near and far and crosstalk looks like this two things that make it a little more complicated
• 41:00 - 41:30 on the victim line i've just introduced this pulse remember when we sent the signal in we we get this the signal on the this is on the aggressor we get the signal on the victim line the victim line doesn't know it's a victim line doesn't know this is crosstalk noise it thinks this is a signal and it's going to interact with the terminations at the ends exactly the same way a signal would interact if i have an open here foreign cross is going to reflect here and it's going to bounce over here and it's going to bounce back and forth
• 41:30 - 42:00 the near end crosstalk is going to bounce back and forth so the terminations on the victim line influence how that noise rattles so termination will make the ring we'll make some ringing and the termination on the aggressor will make the ringing and if i have an open over here like source series termination for example if i have an open at the far end of the aggressor the signal is going to go this way i'm going to get the near and far end crossing over here it's going to reflect and now this is
• 42:00 - 42:30 going to be the forward direction and i'm going to have far end crosstalk over here at the receiver and near and cross circle over here that may reflect and so the multiple reflections in the aggressor and the multiple reflections in the victim are going to make the signature look a little more complicated so that's why if you if your lines and output and input if they meet the required impedance it makes it much better because
• 42:30 - 43:00 you kind of stick with this one simple example simply to say but you eliminate all the other possible crosstalk which can appear from the ringing or from the reflections right and so that's why you're exactly right if we terminate the line so we're using you know controlled impedance we terminate the lines we don't have the reflections we'll have the lowest amount of also crosstalk the lowest amount
• 43:00 - 43:30 any reflections will change the signature a little bit it doesn't change the amount of crosstalk the amount of crosstalk is still about the coupling over here it's what that crosstalk appears as because of all the reflections so this the mechanism that the thing you do to reduce the crosstalk is still about the coupling but you want to pay attention to the reflections because that can only make things worse okay there's one sorry yeah go ahead go ahead
• 43:30 - 44:00 there's one other complication that makes it a little difficult to interpret the behavior a little bit and that's we've been looking at the case when the interconnects are electrically long in other words we've got this edge that's short the spatial extent of the edge is short compared to the signal compared to the the length of the interconnects and so we can watch that edge as it moves and and its interactions along the way when the rise time is long compared to the
• 44:00 - 44:30 time delay of the interconnect then we still have the same effects but the rise time kind of smears out the behavior of the near and the far end crosstalk a little bit it's not as clear as the sink that that means that we will always when we have a longer rise time the crosstalk foreign crosstalk especially will always get less in magnitude and that's why i think i think it was in your video your tip number one didn't you say in your tip number one
• 44:30 - 45:00 use as long a rise time as you can for your signals yeah you see that it's uh yeah if you can then slow slow if you can slow the edge down you want to do it because one reason is that will always reduce foreign crosstalk and sometimes if you make the rise time long enough or the interconnect short enough it will it will decrease the amount of near-end crosstalk as well and so that's the second complication that makes it a little difficult to interpret the
• 45:00 - 45:30 voltage that you measure the ends but in if you have sub-nanosecond edges and you have interconnects that are a couple inches long you will see a different signature at the near end than the far end the dynamic nature of that signal is really important influencing the behavior of the crosstalk okay i have i don't know how much time we still have because i only have a minute or two here but we can reconvene another time okay everything i know about crosstalk i've just told you now
• 45:30 - 46:00 um how fast or well how it depends or when we exactly get the maximum or the saturation of the what what is the factor what influence this when you get the maximum okay so we're talking about foreign crosstalk yes then okay so i can show you um so so unfortunately you know you know the most common answer to all signal integrity questions it depends
• 46:00 - 46:30 the way the way you answer it depends questions is you have to put in the numbers and so i have an example here i think let me go back to the slides here i have an example here of putting in the numbers because here is how we calculate the amount of far and crosstalk this is that ratio of the voltage we see on the victim line at the far end compared to the the the voltage we see here at the far end on the victim
• 46:30 - 47:00 compared to the voltage that we're launching on the aggressor we call that ratio the foreign crosstalk coefficient so literally is a percentage and and you're asking the question when is this 50 well here's what it depends on there's a term here that's about the coupling of electric and magnetic fields that's where there's no good approximations you have to use a 2d field solver to estimate what that foreign coupling coefficient is but it scales with the coupling length
• 47:00 - 47:30 and inverse with the rise time and so i've got here so this is the far end crosstalk coefficient i've got here i did this for you i took 50 ohm 250 ohm transmission lines that defines the aspect ratio of signal to line width to dielectric thickness in in microstrip and and i'm plotting here the separation in line widths because it's easy to see when you're doing the layout but that's assuming you've got 50 ohm lines this is that coupling coefficient this
• 47:30 - 48:00 is that coefficient over here when you use the units of length in inches the rise time in nanoseconds so for example when the separation equals the line width which is you know tightly coupled this is i see this in a lot of boards when you're writing dense buses on the on the surface the coupling coefficient is 0.005 so you put in 0.005 and here's the length over the rise time if the right i put in some numbers over here
• 48:00 - 48:30 if you got a let's look at this first one if the length is 10 inches long and the rise time is ddr4 kind of numbers rise time is 0.3 nanoseconds and separation equals the line width the coupling coefficient 0.05 you put in the numbers here here's 10 inches divided by 0.3 nanoseconds times 0.05 that's 17 percent if you scale when does it saturate 17 percent is already too much you know noise uh budgets typically like
• 48:30 - 49:00 5 17 you're screwed already but if you made this longer than 10 inches if you doubled it you get 34 if you tripled it you get about 50 so 30 inches long 0.3 nanoseconds you saturate that's how you use this relation the only way to get this curve here unfortunately is with a field solver and you know a lot of tools you know polar instruments 2d field server will do that i think there maybe some approximations
• 49:00 - 49:30 out there none of them are very good but like the saturn pcb tool will calculate the foreign coupling coefficient but it's not a very good approximation so i would recommend if you want an estimate for this term here use the polar tool so that's that's how we estimate it so thank you very much eric okay gave a lot of information in a short period of time we can always you know come back and talk about details another time and uh
• 49:30 - 50:00 that's everything for today's video i need to say i really love the animation and together with eric's explanation i really hope this video will help at least some people to understand a little bit more about crosstalk i still have many questions what i wanted to ask just we didn't have enough time so if you think we should create
• 50:00 - 50:30 another video about this topic leave comments okay if you have any questions about crosstalk or if you have any questions what you would like to ask eric leave comments it would not be possible to create this video without eric's help so i would like to thank you very much to eric for helping
• 50:30 - 51:00 me to create this video and also i would like to thank you very much to you for watching this video if you like this video don't forget press like button if you would like to see my future videos don't forget to subscribe uh i really hope you found this video interesting and see you next time bye