Ah, that's how it works! - AltiumLive 2022

The important "Aha Moments" from the life of hardware design engineers. The moments when you start looking different ways at things such tracks, stackup, vias, components, signals, power distribution, and other elements in electronics circuits and PCB layout.

Keynote Highlights:

• Importance of understanding return currents 
• Understanding impedance
• Factors that determine high-speed design
• Higher frequencies and the problems they cause
• Understanding how signals travel through components or PCBs

Additional Resources:

• Altium 365: Where the World Designs Electronics
• Connect with Robert Feranec on LinkedIn
• Visit FEDEVEL Academy website here 
• Differential Impedance Deep Dive
• Characteristic Impedance Deep Dive
• Six Basic Design Rules for Any PCB Design - Altium Whiteboard Videos
   

Transcript:

Robert Feranec:

Hello. I would like to thank you very much for deciding to spend your time by watching this presentation. I really hope you will find it interesting and useful. Ah, that's how it works. Why this topic? In every hardware design engineer's life, there are these moments when they realize that what they thought is happening in their circuits and boards was actually wrong or different from what they imagine. And in this presentation, we are going to talk about some of these moments because sometimes it may take years to get into these moments and there is no reason to wait. Sooner engineers find out about these special moments sooner it'll help them to maybe change or correct their view on circuit design and PCB layout. And it will help them to design better bots.

Now, before we start, I would like to say many of the topics, what we are going to talk about, are so big and difficult to understand that become an expert in just one of these areas may take years of work and study. The goal of this presentation is not to be too specific, but to point out the topics that may be important to know about. If you would like to learn about these specific topics, then you will need to find an expert. Okay? So let's do it. Let's have a closer look at the first of these important moments that I included in this presentation.

The first important moment I would like to talk about is the moment when a hardware design engineer find out that current doesn't always flow the shortest way. What? Now this may look like a very basic topic by now almost every engineer knows that the return currents are traveling under signals, especially for high speed. However, I still would like to talk about it because it's super important. So what is it all about? Here is an example. On this first picture, we can see three tracks. It's the green color routed about a big solid ground plane, that is the brown color. On the second picture, there is the same board, but now there are 500 hertz signals running through the trucks. And we can see the return current for these signals are somehow trying to flow on the ground plane in the area under the track.

The blue color means there are no currents. The red color means a lot of current. When we increase the frequency to one 1 megahertz, that is what we can see on the third picture, the return current for 1 megahertz signals, they are very nicely flowing on the ground plane exactly under the tracks. Now, what does it mean return current? It's cool we are usually told that current flows, for example, from battery through a switch to a light, and then it somehow returns back to the battery and many engineers keep thinking this way, even after they start designing boards professionally. I'm going to be honest, when I started designing boards I was thinking exactly the same way, current goes there and return current goes back like this.

And then there was this moment when I realized, oh, that's not how I should imagine currents high speed circuits. The better way to imagine return currents in high speed signal is that these return currents travel together with the signal edge like this. Now, why is it important to understand return currents? Because suddenly it will help us to understand other things. For example, why we may need to use stitching vias. In this animation, a signal is routed from layer one to layer four, and we are using a ground via, or it also called a stitching via to make it easier for the return current to flow together with the signal. And because we are aware of the return current, now we can be like, "Oh wait, but what will happen if there is no stitching via?"

This is why it is important to understand return currents, because you suddenly start thinking about your layout differently. By the way, on these other pictures, you can see what will happen in different situations. For example, if there is no stitching via, you can see the return current will spread all over the whole ground plane. If there is a stitching via but placed far away from the place where the signal is changing layers, some currents will find a way to go through that via, but there still will be a lot of current flowing around the board. Much better it is to place this stitching via close to the place where signal is changing layers or even better, it is to place more vias. Now, why it can be useful to control the path of the currents flowing on your big solid ground plane? Several reasons but one very practical is that you can have, for example, analog or sensitive or digital circuits all placed on one single PCB and routed the way that they are all going to work perfectly fine on one single board.

On the last picture, there is a lot of space with the blue collar. This is the space with no currents flowing around and this blue space would be a good place for other circuits because the currents from our new circuit would not be mixing together with the currents from our track which is already routed on the board. However, when we have a look at the first picture, where would be the good place to put a new analog or sensitive circuit. There are already currents flowing everywhere, there would be no good place for the new circuit. There is no good place for analog or sensitive circuit because the currents of the existing track are flowing all around the board and they would be mixing with the currents from this new circle and they could be influencing each other.

Another example, why it is useful to think about return currents. Imagine we have situation like this, a big ground plane with a gap in the middle, because we know that return currents like to flow under the tracks. We will immediately understand, oh, there may be a problem. What kind of problem? The return currents can't flow in this area because there is no ground under the tracks. So the return currents will have to find a different way. They will have to go around the gap. And this may cause number of issues. For example, the return currents from all these three tracks are going to mix together or this may create different voltages on the left and right part of this ground point which can cause, for example, EMC problems.

Understanding return currents can help you, for example, understand how important it can be, have a solid ground plan in your PCB stackup. It will help you to understand why two layer PCB may be difficult to design. And if you have to design a two layered PCB, you know what you need to think about, for example, how will return currents flow in your two layered PCB.

So we were talking about currents, but there may be even better way how to imagine your signals. When I ask other engineers, what were their biggest aha moments? This was the most often mentioned answer, instead of voltages and currents, try to think about energy and fields. I will not go into details, you can find many great presentation on this topic, but basically the point is, better way to imagine signals traveling in your PCB is to think about them in terms of fields and energy. So not just like something traveling in your tracks or on ground planes, but as the fields between parts of your circuit. Suddenly you will notice it's not only about the copper, what is important in your PCB, but it suddenly includes everything. What is in the space where these fields are?

If you go back to our previous examples, the fields in our simple PCB would look for example like this, but what will happen with the fields when your track is changing layer? What will happen with the fields when you have a gap in the middle of the ground plane? How these fields are going to look? How large or small these fields are going to be? What everything is going to be inside of these fields? What everything these fields will influence? This will completely change the view how you see the signals traveling in your PCB? The point is to show you that when looking at your circuit, maybe you don't want to think only about this one single track, which is there, but also about all the fields around it. And it's only about fields around the tracks or PCB. You may need to think about fields around components, capacitors, inductors, they can influence each other. Fields around virus and cables, which are connected to your board or to your system or to your product they will be fields everywhere and these fields can cause for example, EMI or EMC problems. Yeah?

Next topic. The next aha moment may happen when you realize how important is impedance in circuits? We learn about impedance at school. We know it exists. We get these questions in our test to calculate impedance, but at school often they don't teach what impedance actually means for real circuits. So I would like to go through couple of examples which could help us to understand importance of impedance. First example. This is a layout with a PCB antenna. Output of a chip is connected to this point, and this big area is ground.

Hmm. We are connecting output of the chip directly to ground. Why would we do that? This is short circuit or is it? When I was starting with hardware design, I didn't think about impedance. I only used to think about resistance simply to say, I only could see the resistance, which can be measured by a simple multimeter. When I measured very high resistance, it meant to me that there was no connection. The circuit was disconnected. And if I would measure zero ohm, it means there is short circuit. If I put a multi between this point and this, it would measure zero ohm, short circuit. However, this is not how we should be looking at our circuit. This is not very accurate view because we can have the same physical circuit or connection but it can behave differently for different frequencies. And that's why we should not only think about resistance but also impedance, and very important property of impedance is that the same circuit it may look differently for different frequencies.

For example, the circuit on our picture may be a short circuit for DC or very low frequency, but for 2.5 gigahertz, this is actually a circuit built from capacitors, inductors and resistors. So for 2.5 gigahertz, the connection between chip output and ground is not zero ohm. And that is very important to realize. Signal is not just one signal, it may be a mix of signals of many different frequencies. In your circuit there are often all the kind of different frequencies and each of these frequencies is getting to see or is going to see your circle differently. For example, each frequency can see a different impedance.

Here is another example. Let's talk a little bit about power delivery and something called PD and impedance. PD and power distribution network. Okay? On this picture, we can see PD and independence graph. Now, how do we get this graph and what does it show? You can measure or simulate the PD and impedance graph for a specific power pin in your circuit and the graph will show you how much impedance this specific pin can see for different frequencies. How this can be useful. Imagine that you have a board with a micro controller or processor or FPGA and your chip needs one [inaudible 00:16:56] to power it up. So what do you do? You directly connect a output of the power supply with the power pin on your chip. And if you are thinking in terms of resistance, you basically directly connected these two pins together. So the resistance of this connection is going to be very low. For example, 0.2 ohm.

However, if you start thinking about this connection in terms of impedance, then it's not so simple. If the PD and impedance graph for the power pin would look for example like this, it would mean that yes, for DC, the pin see 0.2 ohm, however, if your chip is working at frequency, for example, 30 kilohertz, then suddenly it is not going to be 0.2 ohm, but 2 ohms. And that is a big difference. What does it mean? It means that if your chip, for example, needs 100 milliamps at 30 kilohertz, then if we multiply current 100 milliamps with 2 ohms, we will get 200 millivolts drop from our 1 volt power and the pin would only get 0.8 volts instead of 1 volt. Wow.

And that is not everything. These peaks are points where your power delivery circuit is resonating. It can make a lot of noise and possibly problems, not only with power delivery, but also with EMC. I know the part for power delivery or PD and Impedance so if you would like to know more, have a look into this topic. This was just to give you an idea that thinking about impedance may completely change your look at circuit.

Next example, we are talking about direct connection from power supply to our power pin. But when you are thinking about terms of impedance, it'll change your view also on unconnected parts of your system, for example, metal enclosure or shields. Some parts of your board or circuit, they may look like they are not connected together. However, when you think your metal enclosure, maybe just like one part of a capacitor, second part of this capacitor maybe the traces on your PCB and for high frequencies, these unconnected parts of those circuits are suddenly connected together through a capacitor. There is some impedance between these parts, between your circuit and metal enclosure or shield, and this Impedance maybe lower or higher for different frequencies and in different places. So even if some parts of your circles may look, they are not connected together, they may be connected for some frequencies. So conclusion is we have to understand that connections between some elements in our circuits are not simply what we would measure by our multimeter.

So now we should understand that if we have a signal in our circuit, often this may not just be one signal, often it may be mix of number of different signals with different frequencies. Now, we also should understand the direct connection may not really be a direct connection. It may be, for example, a circuit built from capacitors, inductors and resistors. Yeah. Remember the antenna example and the same it is with components. Often they are not really a one component, but they may be a group of different components connected together.

For example, a real capacitor is not just a capacitor. I guess everyone knows this. I knew this at university, but I did not realize for a long time why this could be important. That is why I included this capacitor or sometimes inductor topic in this presentation. Everyone has seen this graph at least once. What took me a long time to realize is that this graph shows that a capacitor is actually a capacitor only on this left part of the graph on the right part of the graph, our capacitor is becoming an inductor and having a look at what frequency these 100 microfarad capacitor is becoming an Inductor just for 500 kilohertz. If you think you place a capacitor into your circuit, are you sure?

In today's circuits often we have signals faster than 1 gigahertz, and we may need to carefully choose and place decoupling capacitors to power up these circuits. Now, I don't want to go into extremes because of the very high frequencies need it to power up your fast processor may not even reach your PCB. They may need to be handled on the Silicon level or on the chip level, but I still wanted to point out that when we are working with higher frequencies, the circuit, what we intended to design may look as a completely different circuit for higher frequencies. For example, we wanted to design a nice circuit full of decoupling capacitors but instead the higher frequencies may see this circuit as a bunch of inductors.

Here is a practical example. You can see it very nicely for 500 kilohertz. The impedance was 2 milliamps, but for 100 megahertz, impedance is suddenly 200 milliamps. And what could be also another complication is that if we have another capacitor in the circuit, this inductive part of this capacitor can create a circuit together with capacitive part of the other capacitor. And it may create resonance circuit, which can make some resonance peaks and noise and other problems in our circuit. So don't forget for high frequencies components may behave completely differently as you may expect it.

When we are talking about impedance, we have to talk also about inductance of tracks and vias. For a very long time I was like, "These are super small numbers. I don't really need to care about them." However, once I have seen how quickly the impedance for higher frequencies can go up because of the inductive part of the capacitors, and I started looking into this topic more, I found out every small inductance is increasing impedance for higher frequencies, even a small inductance as 1 nanohenry is going to increase impedance for higher frequencies and this increase can be high enough to influence for example, power delivery. And do you know what is the inductance of a track? You can calculate it, but just to give you an idea, we are talking about nanohenry, one or five or 10 nanohenry, okay? It may look like a very small number, but if you put it into an impedance graph for high frequencies, this small inductance can move the inductive part of the graph by 0.1 or 0.5 or even 1 ohm up and 1 ohm for power delivery can be a big number.

Again, these are not exact numbers, but they are just to give you an idea that even these small inductances actually can make difference. And do you know what is inductance of a via. Very similar as for the tracks. It depends on the length of the via, for example, a via going from top to bottom side of the PCB can be, I don't know, one or 5 nanohenry. Just to give you an idea. Okay? And often decoupling capacitors are placed on the opposite side of the PCB. So this inductance of a via is going to influence the impedance graph for high frequencies.

Now, to be honest, when I design PCBs often, there is so little space that there are not really many options where to place for example, decoupling capacitors or how to connect them. And my boards always work perfectly fine. So even if you don't think about impedance of these tracks or via too much, you can design perfectly working boards. However, I wanted to talk about this topic because it may become important soon. And especially if you are developing or designing something completely new, or if you are working with higher frequencies. If you are optimizing power delivery or cost of your board, you really may want to be aware of this.

You don't have to separate grounds. I still see some people to worry about splitting or not splitting grounds, and I'm not an expert for super sensitive circuits. So I can only talk about my own experience. When I was starting with hardware design, I used to follow the design guidelines, which recommended to split analog in digital grounds. For example, when I was placing an analog audio circuit on my digital boards. Actually once I read a book where they recommended not only separate analog and digital grounds, but they actually recommended to make visible gaps in all layers for every single subcircuit on the board. The thing is I was following these design guidelines and on one board, we still had problems with ESD, which was damaging our audio chip. And on the other board, I had EMC issues caused by noise coming out from the analog ground through the audio cables.

And I was super worried when a friend of mine told me that in the next board, I should try, do not separate the grounds. I was worried that the results are going to be even worse. So there was this first time when I only use one nice solid ground, plane and surprise, audio work with no problems, no noise. Since then, I have always used solid ground plane. And I have never had any issues with audio circuits or ad circuits or power supply circuits. I'm not saying that this is a general rule and it can be applied for every situation. But I wanted to mention my experience. So if someone is still not sure about this topic, that for me, it really worked. And I had this moment when I found out that I don't need to separate grounds.

The next on my list is, high speed is not about frequency. Okay. Almost always there is someone who says, "Yes, it is about frequency" but you know what I mean. Yeah. I mean, it's not about the clock frequency or about frequency of switching output, pin, or bas frequency. Okay? Because I get this question a lot where people ask me when they should consider the design to be a high speed design. I would like to talk about this. So sometimes people think when they're designing a micro controller board, that they don't need to care about high speed design rules, because their micro controller is running at 10 megahertz and their interfaces are running at frequency just in hundreds of kilo Hertz. But when designing high speed boards, there was this moment when I realized that everything, what is happening in high speed can actually happen also on these slow boards.

I believe by now, almost everyone knows that high speed is a lot about signal edge. Sharper the edge is higher, higher frequencies will be running through your circle because sharper the edge is, it has to include higher frequencies. These higher frequencies will make the edge sharp or fast. Okay? So even your circuit itself is not really using these high frequencies. The high frequencies will be automatically, or maybe automatically in your circuit, if your signals have fast or sharp edges. But why should I care? Once there are these high frequencies included in your circuit, you may run into number of problems. I don't want to go too much into details, but one of the biggest problems, what I have seen caused by not following high speed rules for these slow boards is reliability. Boards are randomly resetting, crushing, or freezing. This is often caused by crosstalk or by noisy power.

And these problems are often caused by higher frequencies. We have already learned that higher frequencies see or seek differently. For them, every track is an inductor and makes it harder to deliver power or suddenly two tracks with a space between them become a capacitor. And for hard frequencies, capacitors are not problems. They can travel through them and cause higher cross talk between your signals. I keep saying if, for example, you route an output pin, close to a reset signal and this output pin has a very sharp edge, sometimes changes on this output pin can create so high crosstalk that it can resend your microcontroller processor and that output pin doesn't have to run at clock 1 gigahertz. It can run at 1 Hz or change only occasionally it still can reset your chip, if the tracks are routed too close to each other and the edge of the signal is [inaudible 00:34:41] enough. So yeah, don't wait to use high speed design techniques until you start designing high speed boards because later you realize that even you were designing slow board, you still should follow some good high speed rules and techniques

Hands up, who has ever created a tool layer board with a weird behavior? The kind of board which was usually working fine and sometimes something happen and you could not figure out why your board suddenly behave so weird. Me, I have designed this kind of board. By the way this is the picture of my two layer board where I learn a lot. This is the board in the good light. And this is the reality. When I was starting with hardware design, I was asked to route one complicated design on two layer board to save money and make the board as cheap as possible. And I was like, "Sure, I will try to connect all the pins on this small board." And I was proud of myself when I connected everything together. Of course the board had all the kind of issues, especially random occasional freezing.

And only after a couple of years later, I have got this moment when I realized how naive i was when I used two layer board for the design. There were whole bases routed on two layer board and I can't imagine how much crosstalk and noise there must have been between the signals and on the powers. So sooner engineers realize how bad two layer boards are, sooner they will start designing more reliable boards. I'm not saying it is impossible to design two layer boards, it just maybe more complicated and simply using at least a four layer PCB can make your design much better and much more reliable.

I learned this twice. First. It was when I designed a circuit built from couple of logic gates and it didn't work. I couldn't figure out what the problem was. Schematics was okay, all the calculations were okay and my real circuit didn't work. It turned out that one of my outputs was 10 nanoseconds light because it went through an additional gate. And I was like, "Oh, when signal travels through a component, that actually is a visible delay" which has to be considered when designing circuits. Until that moment, I thought all component work with no delay. No.

The second time I learned about this was when I realized that even the signal itself is actually traveling with a speed which we need to consider. When I was designing a memory interface, again, until that moment, I was like signal travel so fast that we don't really need to care about it. Turn out the signal speed is not actually so fast, then we have to consider it. Not only when we think about signals, how they are traveling through components or PCBs, but also what materials they are traveling through. That can make big difference.

A good example is a different speed between signals traveling on the top and bottom side of our PCB, comparing to the speed of the signal, traveling inside of our PCB. They are different. We learn, we should imagine signal as fields and energy. So we can now very simply imagine that the difference in the speed is there because the fields will travel easier. For example, in the vacuum or comparing in the air, comparing to the situation when the fields have to travel through some materials inside of a PCME. In the materials, there is more resistance. It's harder to travel through them so the speed is lower. And that is why we have to do wine matching because signals are not actually so fast.

The capacitor radio may be much lower if DCBS is considered. I didn't learn this one by myself, someone told me about it and I was like, "What?" Basically on the picture, you can see a comparison of four, 1 microfarad capacitors. Now watch this. When a 20 volt DCBS is applied on these capacitors, then the real value of this blue capacitor is not 1 microfarad, but only 300 nanofarads. Wow. What does it mean? It may mean, for example, that the next time when you place a 10 microfarad on power supply output, and you think job well done, because data shit says 10 microfarad capacitor is required. Maybe the next question you may ask yourself after placing this capacitor into your circuit could be, "Are you sure it is the right capacitor?"

What are the differences between these four capacitors? Do you know why some are better and some are worse? The blue capacitor is built for 25 volts and from X5R material, the green one is the same material, but for 50 volts, the red is, you can get different material that is X7R, 25 volts. And the purple is X7R for 50 volts. Notice the big differences between them. For example, if you use the purple capacitor, then for 20 volt or DC bias voltage, the capacitor will have 700 nanofarads, which is much more, or which is more than double of the blue capacitor. Both are 1 microfarad capacitors, but they are not the same. Be careful about this.

For many years, reference designs well for me like the top of the art in electronics. I was like, "People who are designing this reference board, they have to be super clever. They have to know what they are doing. They surely run all the kind of tests and call film and validate the designs." I was very disappointed when number of very good engineers and people told me, "No, reference designs are not always like that. Some are by junior engineers. Often they are not really testing or measured or validated." and I was like, "What?" Since that moment, I'm careful about reference designs and I don't blindly believe everything, what is there. Yes, they work. So that is at least good to know, but is the design optimal or the best? Not always. So now I'm less worried when I, for example, want to modify or improve designs or when I believe some things should be maybe done differently.

When I knew my topic for this presentation on my LinkedIn, I ask you what were your aha moments? So here are some, which you mentioned. A bull capacitor behaves as a short circuit, shield of a cable is an extension of an enclosure. Many people pointed out this. Components are not perfect, we already mentioned capacitors can be sometimes inductors but this is more generic. They are not on the inductors they are sometimes also resistors and it's much more complicated. Keep in mind what is inside of components, for example, ESD protection diodes, ground and VCC plane position in the stackup may be important for good power delivery. Cable of length, Lambda slash forward is a good antenna for that frequency. The real capacitor value may be much lower when DCB is considered. Now we were talking about this. Flags may have low enough resistance to influence sensitive circuits.

Electronics can be learned even at higher age, capacitor placement and position makes difference and it is important. And the most important one, PCB layout is not just connecting pins. Thank you, everyone helping me with this. And there's everything what I wanted to say. I really hope you found some of these topics interesting. I hope maybe you learn something new and I hope you have got some new ideas what to learn about next. If you would like to contact me, you can send me an email to robert.feranec@fidel.com. And that's it. Thank you very much for watching.