What Every PCB Designer Should Know - Return Current Path (with Eric Bogatin)

In this video we are going to speak about return currents.

And...

I wanted to create a video on this topic because I can see that many engineers, when they are doing a layout; they keep thinking about how to route the tracks but they completely forget about how the signals will...

How the return currents for these signals, what they are routing; How these return currents are going to flow on their PCBs.

And if you don't know what i'm talking about then maybe you may want to consider to watch this video.

If you know something about the return currents, this video still can be very interesting because: I'm going to talk to Eric about this topic.

I have recorded our call with Eric, where we were discussing how these return currents flow for simple signals or for simple tracks...

In the call or in this video You will see also our discussion about return currents for differential pairs and also about some other topics.

I really would like to say that this call with Eric, it was one of the best calls what I ever had.

And... It's the kind of call, or it's the kind of discussion what I imagine like when two engineers are talking about some really interesting topic.

You will see, I really hope...

You will enjoy this call same as I did and...

Don't forget leave comments okay, let me know what do you think about this kind of videos, created during my calls with different people.

Okay, so that's everything from me.

Now let's play the video from my call with Eric.

Here it is.

What does it mean return current?

I... I have some...

Image in my mind, but I know many people, they imagine return current the wrong way.

Because, I will tell you; When I was studying electronics, and when people were saying like return current, I was always thinking like you know, current flows one way and it goes the other way back and that's what is called a return current.

But for higher speed signals, that's not really how people should imagine return currents, correct?

Eric: You're right, you're right And this is one of the most important concepts in signal integrity.

In other words, so many effects in signal...

in signal...

In fact maybe every problem in signal integrity is tied to or related to that return path.

And... When we look at signals, you know even, even low frequency signals in the audio range.

You can't think of just the signal path, and design the signal path.

You have to think of the signal and it's return path.

And... I, you know it's a...

It's a concept that you know...

What you described of okay, You know, you have like a battery and a switch and a light bulb, and you throw the switch and the current goes you know, from the battery and it travels through the switch and it goes and lights the light bulb and then it comes back and comes around again.

That's... Robert: It's what they teach you at school Eric: That's what we all learned, exactly and...

And this is the concept that we all learned in elementary school and high school and that was the basis of how we thought about currents in circuits.

And that's okay for DC.

Everything that we do with DC, that circuit model is just fine, really easy to understand.

It's when we try to extrapolate that simple model into anything above, you know, 10 kilohertz; 100 kilohertz.

That's where it breaks down.

And... And so anything related to noise due to the current paths is...

is affected by where exactly is the signal of the return and when is it... when is it flowing.

And so when I teach signal integrity that's one of the first concepts that I...

That I talk about is this idea of the return current...

I have a...

I teach an advanced PCB class at CU...

Graduate students you know they've been...

They're electrical engineering graduates, they've been doing this forever, and many of them come into the class with the same concept.

They think when they see that ground symbol in a... schematic,

they think of it as an infinite sink of current.

And wherever that ground symbol is, okay that ground current is going to go in there and it's going to just you know...

Do whatever it's going to do and it's going to disappear.

You know, I gave a talk at...

Altium live keynote, I don't know, a couple years ago.

And I talked about...

The... It was about Signal integrity lives in the white space on a schematic.

That the schematic tells you nothing about signal integrity which is so much based in in the return currents.

And so this whole concept is...

You never... You never encounter it if you're doing schematics because it doesn't tell you about where the currents are flowing.

And when you translate the schematic into a layout, if you don't know about this concept of where return currents flow, you would never know how to...

Why to worry about controlling the return currents as carefully as the signal currents.

And that's how...

Robert: So how people can imagine these return currents?

Robert: So how people can imagine these return currents?

Eric: Here are two ways. So...

Robert: How you imagine it?

Eric: Right... So I'm going to give you...

A short answer and then a long answer.

The short answer is... I've got a bunch of webinars that I've done over the years about how to think about...

You know, I call them the four most important principles of signal integrity.

About a transmission line, about the fact that their...

Dyna... Signals are dynamic.

And... This other idea of where return currents flow.

And how they flow and...

So if you go to... my bethesignal.com website on the left-hand side there are a whole bunch of webinars that are... they're all free, they're all ones that I've done over the years.

that are... they're all free, they're all ones that I've done over the years.

And there's one about...

What every oscilloscope user needs to know about transmission lines.

Because that is where I think, if you're not familiar with signal integrity and you use an oscilloscope, you have to know about transmission lines and return currents.

And so I... have some nice animation there to show it.

In fact you know what... Can I...

I have a... So I have a buddy of mine who's our application engineer in Japan Yoshi, and he's an expert at flash animation.

And he created this really cool...

Flash animation which you can download, It's free... Freely available on my website,

It's free... Freely available on my website, that shows dynamically how the return currents flow.

Can I show that? Can I show that to you? Robert: Oh, yeah.

Eric: Okay so... I have to...

Robert: You can share your screen, I can stop my Eric: Yeah uh Eric: And let me go find it here...

And I apologize I should have had it lined up here but i didn't think about it.

but i didn't think about it. Robert: No it's fine.

Eric: Let's see Gosh you know I can't remember...

I have so much stuff floating around I can never remember where I left... Okay here it is.

Yeah okay.

So I'm gonna... I'm gonna share my screen Okay so here is a simple transmission line, a microstrip transmission line.

And... can you see that on your screen?

Robert: Yes, I can Eric: Okay...

Eric: Okay...

And so we're going to launch a signal at the beginning here.

And it's just going to get a (little voltage sort of a little step voltage.)

It's got some source impedance, you know 50 ohms or something.

And we're going to increase the voltage between the signal and the return path over here at the beginning.

And that...

So the first principle of signal integrity is all interconnects are transmission lines.

I happen to draw this one as a microstrip, nice and uniform, but it's any signal and the return path.

And so, you know, you usually...

What we learn in elementary school is when we turn on the...

The voltage, we'll put a voltage from the signal return...

If this is a 50 ohm input impedance line...

And and we put one volt in, then you know one volt into 50 ohms, that's 20 milliamps.

So we get 20 milliamps going down this transmission line.

And we think, okay, 20 milliamps goes down the line it goes down the line, goes down the line and then it hits the end and then it's going to return into the return path, and then it's going to come out and come out and come out.

And that's... that's how we learn it in elementary school.

It goes down one conductor and then returns in the other conductor.

That's not how.

That's kind of what... What it looks like in DC but that's not how it gets there.

That's not what an...

A signal is going to look like as it's propagating down Instead, when we launch 20 milliamps into the transmission line, it doesn't have to go all the way down and then come back to come out.

Because, what happens if this is an open over here?

How does the return current get from the signal to the return?

What if it's a 50 ohm, or what if it's some other resistance?

How does that return current get to the bottom conductor and come out?

And you think, wait I have this insulating dielectric between the signal and the return.

How could you possibly get current flowing between the signal and the return conductor if it doesn't go all the way down?

And the answer, the secret to understanding return current is displacement current.

And displacement current is a concept that James Clerk Maxwell introduced...

In the 1860s, and that was one of the most important innovations that he... contributed.

And he said you know, Yeah there's... There's some conduction current,

Yeah there's... There's some conduction current, there's the motion of charges in the conductor.

That's a current and... you have currents flowing through conductors...

And they generate magnetic fields and you change the current that changes magnetic field and all that kind of stuff.

But he said that's not the only kind of current.

That when I have an electric field, That electric field is generated because of, or it generates a voltage between the two conductors.

So if I put an electric field between the first conductor and the second conductor, I put an electric field there, I have a voltage difference between them.

If it's static, not changing, there's no...

Current flowing between the signal and the return.

But if that electric field changes, if it increases for example.

Wherever I have electric field lines, I'm going to have a new kind of current.

And he called that displacement current.

It is... just as real as the motion of charges flowing through the conductor.

And where do you have a changing electric field, or, because it's between two conductors a changing voltage between them, wherever you have that change in electric field, that's where you have the current flowing.

The displacement current.

And so when we launch a signal in the front and that voltage turns on, and when it turns on... So I'm going to launch that signal...

I'm going to turn it on.

We're going to get that electric field changing right at the beginning, and right where it's beginning, that's where we're going to see that current flowing between the signal and return.

And as that edge propagates, wherever that signal is changing, that's where we get the displacement current flowing between the signal and return.

So let's turn that on, let's watch that.

I'm trying to... Let me see if I can freeze it.

I can't freeze it very well So we'll just watch it dynamically.

So here's that edge, here's that voltage wave propagating down.

It's only in this region where the dI dT is, the changing voltages.

It's only in that region where I have the changing electric field and that's where I have the displacement current.

And so as that signal propagates down, we have this current wavefront, coincident with the voltage wavefront, propagating down.

And when we talk about the impedance that signal sees as it's propagating down, it is literally the ratio of that voltage that is the signal divided by the current that is the displacement current.

And as we move the bottom conductor around, change the spacing, change it's shape, anything like that.

When we change the return conductor we influence the displacement current, we change the electric fields, we influence the displacement current, and that changes the impedance.

And when... So that says that right at this very instant when i'm beginning, just when it launches, look at the beginning.

I don't know where that edge is, all I know is I have signal going in and I have signal coming back.

All I know is, there's 50 ohms looking in for as long as that signal is propagating.

That's why a transmission line, when we look at the instantaneous impedance that signal sees, it doesn't care what's happening at the end until it gets there. And...

Robert: I like the picture down because you can very nicely see, that basically because there is the edge, there are higher frequencies.

And for higher frequencies the capacitor will be basically short circuit.

So that's where you get these short circuits going down, that's...

That could be also one of the explanations or I'm wrong?

Eric: Well it's a little more complicated only because... The...

The current, the displacement current is flowing between the electric field, between the signal and the return, and that electric field is through the capacitor, so that's one way of thinking about it.

But the inductor also plays a little bit of a role in the propagation of the signal.

But yeah, it's the... I hesitate to say the short circuit, but it is the changing electric field in the capacitance between the signal and return that the current is flowing through.

So those electric field lines... and so

It's the... You know the...

I show this... model underneath, the LC model.

You know only to...

Kind of illustrate the two different ways of thinking about a transmission line.

This is not necessarily a great way of thinking about how a signal propagates on a transmission line.

It's an approximation of the transmission line but you know what.

This, the actual physical structure and the idea of the propagating changing electric field is a much better way of thinking about how that signal interacts with the transmission line.

Because this is really how it does.

It's all about the fields, it's not about the approximation of the L and the C.

Robert: Okay, now when there is... for example cut Eric: Yeah... Robert: In... In

Eric: Yeah... Robert: In... In

the plane so what will happen then?

How the return current will look?

Eric: Right I apologize for interrupting the call.

I just would like to be sure everyone understand what kind of cut I mean, so I created this very simple picture.

Here you can see the cut, okay?

So when there is a cut in this plane, then the return current, it cannot flow directly under the track.

It has to go somehow around, and that's what we are going to talk about next.

Let's go back to our call and let's continue.

Eric: So when there's a cut in the return plane, the current can't flow through that...

through that path underneath, it has to snake around.

And in snaking around, it's taking a longer path, and now I have a little bit more inductance in that path, a little impedance discontinuity in that path.

And that means that return current that's flowing from one side to the other, going through that impedance discontinuity in the middle, is now going to have a voltage drop across that path.

That means this side of the board will be a different voltage than this side and when you have that transient current, it's going to excite that...

Gap and that higher inductance, and we're going to see a voltage noise peak.

Robert: So, but the return current will go like around the gap, or it will cross the gap?

Eric: So...

Two ways of thinking about it.

If I want to kind of approximate, an easy way of thinking about what that current does is, I think about it as going around that gap In reality what it does is, it excites a... What's called a slot wave mode in that gap.

And... we get current going transverse, and that current going transverse creates the voltage drop from one side to the other.

So they both give you the right answer, the slot wave mode is the... You know, a little bit more correct way of thinking about it, but it's also really hard to conceptualize.

And so it's easier to approximate it by...

Yeah the current goes around and then going around there's a longer path, there's tighter concentration of the current so there's higher inductance, and so there's an inductive discontinuity...

We approximate that gap as an inductive discontinuity in the return path.

And if you have a dI dT through an inductor what do you get?

If you have... If you have an inductor and you have a fast changing current through that inductor you get a voltage drop from one side to the other.

And that's what happens in the plane, you get a voltage drop from this side to this side because of that inductive gap.

And... here's the thing about the planes and why we use planes.

You know this doesn't show where the current is actually flowing in the plane.

What we really need to show is, well where is that current flowing? And...

I have another slide... So I just gave a... I'm doing a presentation at...

PCB west coming up in... next week.

And I'm talking about... Differential pairs with no return path And... there, it's really important to kind of think about...

And... there, it's really important to kind of think about...

The return path, and what happens... Where is that return current flowing. And...

And so I have an example of that...

Let me see where I put that...

Let's see, okay I have it in here And...

And...

Here we go.

And I'm just going to pull up this other slide.

And just to show where it... Where does that current flow I have...

I have...

There are other...

So when we have a gap in the ??, when we have a discontinuity of any sort so that return current is not flowing in a nice wide plane, we call that kind of noise we get, from one side to the other, we call that kind of noise... ground bounce.

So here is...

See if this works. Here we go.

I'm going to interrupt the call for a moment, and it's because when I saw this picture I was a little bit confused.

You will see it in the call And before we start talking about this picture, I would like to make it clear.

So this is how it looks when I redrawn the picture.

On the top, what we can see here, this is basically the copper of the track which is routed here on the top.

And this down here, this is the copper of the plane.

When we have a look back on the original picture, this is the copper of the track and this is the copper of the plane.

Okay, now we can continue with the call.

Eric: Okay so here is a plot of the current distribution in that same microstrip that we were just looking at.

Can you see that that new plot there?

Robert: Yes I can Eric: Okay so here's the signal on the top here's the return plane, and I'm sending current in at like... 10 megahertz or so.

Not real high.

And we'll talk about that transition in a second.

So I'm turning in current... sending in current at... 10 megahertz or so.

Here's what the... The colors are the current density in the... conductors.

Blue means no current and red means a lot of current.

And you can see that in the return... So you know, there's current in the signal line and there's a little bit of ?? effect that's causing the current to go to the outer surface here.

But it's, you know, it's in this... It's confined in the signal path.

In the return path that current can flow anywhere.

Where... Where is it flowing?

Well, at frequencies above about 100 kilohertz, and I'll show you where that number comes from in a second, at frequencies above 100 kilohertz, that's the high audio range, the current in the return path is going to flow underneath the signal line.

It's not going to spread out everywhere, it's only going to be under the signal line.

And so... And it's not using the whole plane.

It's only using this narrow region of the plane.

And if I...

If I were to...

Robert: That's the place where the edge is...

Eric: Well this is... This so we're looking at a side...

Eric: Well this is... This

so we're looking at a side... Robert: Ah okay, so this is like this way Eric: Yeah right this is the cross section of it, so the signal is going into the board into the plane.

Okay, and this is showing the lateral extent of the return current.

Robert: And why this... picture which is above the line, why it is above...

I would expect it between the line and the plane.

Or I'm imagining this the wrong way...

Eric: So let me see if I can describe this.

So here's the signal line.

The... This region in here that's white, that is the insulating dielectric. Like the...

Robert: Ahh okay so the signal line is like whole copper.

Eric: Yes this is the copper trace Eric: Yes this is the copper trace. Robert: Okay okay.

Eric: I made it extra thick just so we could see the (current).

Robert: I understand now, I understand now Eric: Right, and I made the return plane extra thick so we could see the current in the return plane.

Robert: Ahh okay so you can very nicely see the red areas, it means the current basically flows...

on the surface of the... track.

Eric: And as we go up to higher frequency, the current is going to flow more to the outer surface in the signal line.

But it's confined, you know, it's...

It's got walls so you can't... The conduction current can't flow outside.

And so we're looking at conduction current here.

There's... If we have this changing frequency...

A changing voltage, then of course we'll have...

A displacement current between the signal and return as well.

But we're just looking at the conduction current in the conductors.

And so here's the signal line, and at this frequency, you know, a lot of the current's on the outside surface.

Here is the return plane, and... look there's no current over here.

There's no current over here and in fact, if you go, you know, a couple line widths away, this is roughly the dimensions for a 50 ohm line, if you go a couple line widths away, you could cut this copper away and not affect the return current. Right?

And so we say that this is the path the return current wants to take naturally, this is where it's going to flow and...

And as long as the return path is wide enough to accommodate that return current, then it looks like an infinite plane.

I could literally cut out this copper over here and this copper over here, not affect the signal quality or the impedance the return current sees.

Robert: It... It means that when you need to route two tracks parallel, and if you go beyond the green area then you will be absolutely safe and the...

tracks will not influence each other or... Eric: Uh... Not quite

Eric: So now you're... talking about crosstalk.

And crosstalk is about...

When I have a nice wide plane, crosstalk is about if I had another victim line over here...

Crosstalk would be about... electric and magnetic field coupling.

Or... we approximate it by...

Eric: Capacitive and inductive coupling. Robert: Ahh okay Eric: So there will still be crosstalk, this is long range. I mean you don't have to have overlapping return current to have inductive crosstalk.

Eric: This is going to give us... Yeah

Robert: I understand now, I understand what you mean.

Eric: But...

If we get it really close, and we have some...

And you have to get really... extremely close in order to have that that extra crosstalk due to the overlap of the return currents.

But if you pull them apart you're still going to have crosstalk.

If however I do anything to screw up this return path so that the return current is not in this region, like I... take all of this conductor and I move it over here

like I... take all of this conductor and I move it over here and I make it a pin and a connector.

Or a (lead) and a package.

Or I put a gap underneath here, and I force this return current to flow over here.

I have pulled that return current far away, I've created more inductance in that return path.

Anything I do, other than this region for the return current, I am going to increase the inductance in the return path and...

that is going to cause voltage noise in the... return path.

And if that same return path is shared by another conductor, like I have another conductor here and I...

cut this return plane so that I make the return plane over here and I have another conductor over here and his return current is in this region and his return current is in this region, if they share... so if I screw up the return path so I increase the inductance and I share that same return path, then this poor little guy sitting here is going to see all the extra voltage noise

from the aggressor's dI dT through that extra inductance, and he's going to see that as part of his voltage noise.

And we can call... Robert: This was exactly the question what I wanted to ask because if you make gap in the solid ground plane and for example multiple return paths from multiple different...

signals will go through same area, and I wanted to know what do you think if...

If like going multiple return currents through same area if they are going to influence each other.

and if they are going to make the signal worse Eric: Right Very good, absolutely right.

So if you have the same inductance, same return path... We'll make the return path over here.

And in fact I don't know if maybe I can draw that here.

Let me see if I can do that.

So I'm going to... Grab a...

I'm going... to pretend that my return path is really only over here.

And... so all the return current for this signal line is over here and it's really narrow and so all that return current flowing through this region here sees that higher inductance in the return path.

And now I've got another signal line over here, his return current is going to be in here as well, maybe I can make that a different color.

So his return current, we'll try to make it green here, is also in here.

The return current from the aggressor over here, his return current generates voltage noise in this path because that is...

has the high inductance and the the poor victim line over here, his return path is part of this higher inductance path, and I'm going to get the dI dT from this guy in here as part of his signal.

And now if I have not this guy but I have another one over here, and another one over here, and another one over here and they're all sharing the same path over here as their return path, I have not one dI dT but...

This guy's... two dI dTs, three dI dTs, four dI dTs.

If they all switch simultaneously I'm going to have four times the voltage noise in this common return path and this poor victim line is going to see it all.

And it gets worse and worse.

And we call this behavior, we call this noise source when I have a screwed up return path, so it's not...

It's anything other than a wide line.

A screwed up return path and multiple signals sharing that same screwed up return path, we call that ground bounce.

That's what ground bounce is.

It is a type of crosstalk that is dominated by the shared return path.

And that says, how do you... fix ground bounce.

How do you eliminate ground bounce.

It says number one, don't screw up the return path and number two, don't share return paths.

That's the secret of solving ground bounce problems. That's why you want to do everything you can to avoid gaps in the return path.

And that's why when you watch the other guys show how to route traces on a board and they route all their traces first and they use the top and the bottom layer to route the traces and then they throw copper pour and they say, okay this is now my ground connection... connected the ground net.

Worst thing you can possibly do because you have automatically engineered ground bounce in your design.

Worst thing you can do.

I'm going to interrupt our call again because I would like to...

Point out, that...

In my last video, I created this simulation where I routed three different tracks, three different signals over a big gap in a ground plane.

And in this simulation you can very nicely see the return currents of these three tracks, they go around this gap which is here.

So all these red color here, these are the return currents of these tracks which are routed here.

And as you can see, they go around the gap and they are all mixed together.

So this is one of the reasons why you really don't want to route your signals over a gap in...

In ground planes.

This is one of the reasons why you really would like to be sure that you have nice and continuous return current path under your tracks.

Okay let's go back into our call and let's learn some other new awesome things.

Robert: Another question. You started talking about return currents or return path for differential pairs.

If I understand right, differential pairs are usually just single-ended pins but with opposite signals.

So the...

Basically what I think about is, even for differential pairs, the... return path, for these signals may also go through the ground back.

Is it correct?

Eric: Yep.

So... this is exactly the topic of the presentation I'm doing at PCB west...

Is about, well, how do you think about the return path of differential pairs and...

Does a differential pair need a return path and...

What happens when you don't have a plane for your differential pair?

Where... is the return path?

You know, if you have a region... if you have a one layer board for example, a differential... where's the return path?

a differential... where's the return path?

Or if you have unshielded twisted pair, where's the return path?

And so I'm going to show one other slide here.

This is...

Let's see if I can get this on here, maybe I'll do it this way.

Here we go.

Okay this is three different situations of...

Let me see if I can get my pointer going here.

Here we go, okay.

So three different configurations of cross sections...

And one of my students...

He's finishing up his...

Master's degree and...

He's really an ace when it comes to a lot of these simulation tools and so he put this this...

Calculation together, it's using HFSS.

Looking at the current distribution, and I think it was about 100 megahertz or so at that frequency.

A convenient frequency and it shows the...

And we adjust the scale to show where most of the current is flowing.

And so this is... and we set it up with... a differential pair.

So here's a signal line called the p line, here's the n line.

And this is set up as...

They're tightly coupled, the spacing, so it's a 5 mil wide line, 5 mil space...

It's designed for roughly 100 ohm differential impedance but even when they're tightly spaced, right, this is as tight as you're going to route a differential pair on a board, even at this spacing when you look at the current distribution...

I mean it's one of those things that, if there are no, you know, I showed an approximation just a little while ago to give you an idea of where the current distribution is, but when you combine two of them and you get them close together, it's really hard to calculate where's that current.

That's where...

A field solver...

In this case it's a full wave field solver, is really important to help us visualize where those currents are located.

That's what the example you had from Heidi of the...

Keysight's SI pro, no PI pro is really valuable, because it helps you visualize where those return currents are.

And... we set up this problem as a really really simple problem just to help us kind of build that engineering intuition about how do the currents flow.

So here's the first case, where it's a typical two-layer board with...

Well it's, or I ?? but we're only looking at two layers.

The signal line and the return.

And you know typical kind of thickness and...

line widths for like a four or six layer board.

Uh... Tightly spaced, so it's tightly coupled.

And look where the return current so...

We see the signal current in the... p line.

Red means it's going into the board, blue means it's coming out of the board.

And so we have a 0 to 1 volt signal into the p line and we have a 0 to -1 volt signal into the n line.

So it's a true differential signal.

Well you got the current going into the board in the signal line and look it's blue in the bottom, it's coming out in the return plane.

If we didn't have this here, if we moved him far away we would see exactly that current distribution that I showed you earlier.

And there would be some return current over here for the p line.

Now we come along with the n line and we do the same and we put a 0 to -1 volt signal so the current comes out of the board of the signal line and goes into the board in the return path.

Um well wait a minute, in the return path I have this guy's, the p line's, return current coming out at me, and I have the return current of the n line going in where they overlap they're going to cancel out.

But where are they overlapping? It's just in this little region here right between them.

Most of the return current for this signal line is over here.

Most of his return is over here.

And so...

It is not true that the return current of one line is carried by the other line.

That in fact... Robert: That's I think what many people...

Robert: That's I think what many people...

Eric: Yes, I see that all the time.

In fact...

You know I'm going to add this to my list so...

I'm doing a... a talk at Altium live in October.

It's... I take my talk from uh...

I quote Yoda in my title.

I say you must unlearn what you have learned.

And... now as we're talking here I have a bunch of examples I'm going to show, but I think I will use this one as an example of one of the things to unlearn.

That it is not true that the return current of one line is carried by the other.

I'm going to show you in a minute that I'm actually wrong and that's why the correct answer is of...

Does the return current of one line get carried by the other line?

The correct answer is it depends and I'll show you the difference but...

In this case, when I have the two signal lines on the top layer and the return close proximity, so it's a 100 ohm differential pair as we would design most surface traces, it's perfectly clear that in the best case and the tightest coupling, the return current of one line is... clear and succinct and the return current on the other line is clear and succinct in the return plane there's only a little bit of overlap.

And if you actually do the calculation, it's about 10 percent of the return currents are overlapping and cancel out.

Ninety percent of this guy's return current is in the plane, ninety percent is in the plane here.

That says if we do anything to screw up that return path so it's not under here, we're going to create an inductive discontinuity and we will increase the impedance of the differential pair.

Now...

The impact of screwing up the return path is to change the differential impedance and it will only increase.

But let's look at... Let's continue this... process here and let's look at what happens to the return current in the plane when i pull the plane farther apart.

Because I... In this example...

Let me see if I can get to that example.

In this example here, that we're looking at...

I didn't really say it completely but I wanted to mention that when I pull... so this is the return current distribution for this single transmission line.

If I pull the signal in away or I pull the return path farther away that return current spreads out.

And... So here's the return current distribution when the signal line is pretty close to 50 ohm line.

When I pull the signal in away, or I pull the return path farther away, increase that separation, of course the impedance increases.

But look what happens to the current distribution. It spreads out a lot.

And so the way... Robert: But that's, what is the emission then, no?

Robert: But that's, what is the emission then, no?

Eric: Not necessarily because we... will have near field Robert: Ahh okay Eric: And so the signal and the return and so we get more fringe field but that fringe field drops off pretty quickly.

It doesn't necessarily radiate, it doesn't contribute to far-field emissions.

Robert: But it can mix with other currents Eric: Yes, yes, very good, absolutely right.

When we pull that return plane farther away, we increase the characteristic impedance of this line, and we increase the extent of those fringe field lines that spread out even more and that's going to contribute to more crosstalk.

So we... when we... that's why one of the important ways of reducing crosstalk is pull that return in close proximity to the signal line.

When we pull it farther away consequence number one is we increase the impedance.

Consequence number two is we increase the extent of the fringe field lines that contribute to crosstalk.

And consequence number three is, we spread the return current out more in the plane.

It spreads out more, and so the way to think about it...

You know, the simple... you know, intuitive engineering way of thinking about it...

If here's our signal line here's our return path, think about it as kind of like a a 45 degree right triangle.

That's the extent of the return current.

When it's... When the dielectric thickness is thin, so it's close to 50 ohms, then the return current spreads out a little bit on either side.

As I pull that plane farther away, the return current spreads out more and more and more and more.

That means that it's available to interact with other signals more, we get more fringe field, we get more crosstalk.

Robert: So that's why... that's the reason or...

one of the reasons why, when you have a...

solid ground plane close to the signal layer then you can basically route tracks a little bit closer, and there will be no crosstalk.

Eric: Right, well I... never say no.

Eric: There will be less crosstalk. Robert: Okay so, Robert: it will be still fine.

Eric: Well I hesitate saying fine, but it will be less Robert: Okay Eric: The only way we can answer those kinds of questions about...

Is it good enough, is it acceptable, is it okay?

Robert: Yes, that's what I mean Eric: The only way we can answer this, we have to put in the numbers.

And... so it's the...

And the only way... and you know that question about crosstalk, between two adjacent lines, the only way we can answer that it's all about the fringe electric and magnetic fields.

The only way we can answer that question of how much crosstalk is there, is using a 2D field solver.

Because this is the only way to calculate those fringe field lines and that's a very simple straightforward calculation.

Robert: But there are these like kind of...

rules what you can read in some documents which will say like 5H or 3H maybe, kind of like enough for...

routing the tracks.

Eric: Right, they're based on two assumptions.

They're based on... And I have the same ones, I... publish those...

rules of thumb as well.

But they're based on two assumptions.

Assumption one is how much crosstalk is too much.

And it depends on the application.

If we're dealing with sensitive analog signals, I may want less than...

A tenth of a percent crosstalk.

Because I want high isolation.

But if I'm dealing with digital signals, then five percent crosstalk... may be acceptable.

So assumption number one is how much crosstalk is too much.

Assumption number two then, well, not assumption but the question number two is, well, as I move the traces far apart, how much crosstalk do I have and how far apart do I have to get them so that I'm below that threshold that I say is too much.

That allocated noise budget for crosstalk.

And typical digital systems for microstrip and strip line, if I keep the spacing between...

so that's assuming 50 ohm lines, if I keep the spacing between the two lines greater than twice the line width, that defines a dielectric, or spacing, of about four times the dielectric thickness for microstrip traces or in strip line.

If I keep the spacing twice the line width, the crosstalk between an aggressor here and a victim over here is less than two percent.

And so if I have that worst case of here's my victim line and he's got an aggressor over here and he's got an aggressor over here and they're both switching the same time, I'm going to have noise from...

The first aggressive noise...

I'm at two percent from this guy, two percent from this guy That's four percent. That's less than my five percent criteria.

So that's where that rule of thumb comes from.

Making those... the assumption how much is too much and then calculating with a field solver how far away you go to get that less than five percent.

Robert: What is your opinion if...

all these tracks are from the same bus.

For example uh...

Let's say they are all from data bank zero and they will be all switching at the same time or very similar time, so can you route them a little bit closer, because even if there is a crosstalk it doesn't matter, because it... the crosstalk will happen at the time when the bus is not read or written,

because it... the crosstalk will happen at the time when the bus is not read or written, so what do you think about this?

Eric: So there are a lot of folks that say, you know, yeah I'm going to get... so ground bounce is a form of switching noise...

That...

Not all of crosstalk happens at the edge.

It's the edge that drives it, but the signature of the noise that you see may last during the period of the... signal as it's propagating.

And... so many people say, well if I have switching noise, if I can keep that outside of the setup and hold time for the receiver then... who cares.

And generally that's mostly true, it's just that...

can you really guarantee that that switching noise that I've got...

is not going to be seen by an asynchronous bus somewhere.

If I've got a... and I see this a lot in... ?? buses where they're lower frequency than the clock, and so I'll have— I'll have switching noise due to other signals switching at at a higher data rate or higher clock frequency, and they will give me noise in...

in the ??, data or clock lines, because that's such a wide time interval for the ?? compared to the clock edge,

for the ?? compared to the clock edge, I'll have multiple clock edges switching inside there, and so if I'm looking at...

And I can absolutely guarantee my design, that I'm only going to be looking at the signal during that setup and hold time and I keep all the switching noise outside of it, I maybe get away with it...

It's just a higher risk site, but you know, sometimes you have to...

You know, do... the risks

Robert: You know why I'm asking this, because...

Very often I do like very small and high density designs and...

There is not really like...

You have to think about this kind of stuff, because there is no space.

So, you have no other options.

Eric: Yeah, one other thing you have going for you when you make a small board is, you know, for a given rise time, the shorter you make the interconnect, as long as it's short, electrically short, compared to the rising edge, if you make it electrically short, you will always reduce the crosstalk.

And so if you make your interconnect short enough then hey, you can keep them really really close and the amount of crosstalk you get is... reduced.

Robert: But I was actually very surprised. When I was doing some...

simulations, or when I read some articles, I don't know...

There are situations where you can get like the full crosstalk even like in five millimeter...

I don't know how much is it in...

Eric: Five millimeters is... about 200 mils Robert: Yeah so even like for very small...

time you route two tracks together and you get the full crosstalk.

Eric: Well let's see...

So, when you're electrically short, depending on the impedance of the two lines, that will determine whether it's dominated by capacitive coupling or inductive coupling.

If it's either one, generally the longer you make it, the larger that coupling will be.

If it's electrically short.

If it's electrically long...

Then the near-end crosstalk saturates and...

But the far end will continue to increase.

Robert: Okay I didn't know that, I learned something new.

Eric: It's... I mean all of this is all about kind of the basic pro...

electrical properties of interconnects and signal integrity.

And this... is what I teach in my... class at CU, this is what we have on the signal integrity academy website, this is what we... you know, in the last couple years I've been focusing on best...

Best measurement practices and I've developed a number of...

Test boards that illustrate these principles by measurement.

I mean simulation tools are great to show it and...

I used to do it by simulation, but I just found it was so much more visceral when you can actually have a piece of hardware in your hand and do a... send a signal and do a measurement and see the impact of the... crosstalk... in the device.

And... all of these effects that we've been talking about today and... previously,

they're all the basic fundamental principles of signal integrity that if you really have a good understanding of the principles, you'll be able to determine what direction to go longer shorter farther closer thicker thinner, if you understand the principles you can tell what direction will reduce that noise, and then armed with the tools, the simulation analysis tools,

you'll be able to estimate... is it far enough, is it thick enough is it thin enough...

And so it's those two kind of different... aspects of understanding, it's the, what are the principles, I call them the essential principles, and then to answer the... is it okay, is it good enough, you have to be able to do analysis.

And I have a variety of rules of thumb that I published, they're on my... I list them all on my faculty page they were all published in EDN magazine back in the good old days, so use rules of thumb analytical, approximations, numerical simulation tools.

If you really want to be serious about optimizing your design and having confidence, it's going to work in reducing the risk, the more analysis you... can do up front, the lower the risk that the the product is is going to break because of one of these noise sources.

That's the principle of signal integrity engineering.

And... that's everything for today's video.

Leave comments okay?

Leave comments for Eric.

Let him know if you like this... topic,

if you like his explanation about the return current path and about crosstalk and all those things what we were talking about.

Leave comments to let me know if you like this kind of videos.

If yes, then maybe also make some suggestions, maybe...

What other people you would like to see in... my videos,

what other calls I could make.

I would like to thank you very much to Eric, and I would like to thank you very much to you, because without Eric and without your views and without your likes, without your...

subscriptions...

It would not be possible to make this video.

So thank you very much.

I...

I would like to say, I really enjoyed creating this video, I really hope you liked it, if yes don't forget press like button...

Don't forget to subscribe so we can create more of these kind of videos.

And I would like to thank you very much for watching.

See you next time.

Bye!