Daniel Beeker is one of the top experts in the field of EMC and signal integrity. We met him during PCB West 2018 and asked him about his design techniques.
0:10 You’re an expert in EMC and signal integrity design techniques. What kind of guidelines would you suggest to designers?
There are some really simple rules that you follow, and you can be successful in developing a power supply system. The difference that I take, than other people, is my approach is based on the actual amount of energy that needs to be moved. And my perspective is that energy moves through plumbing that’s created by the conductors of the printed circuit board, and the dielectric. The dielectric being the actual medium through which the energy moves. So I look at the amount of energy that is required, and I make sure that the capacity of the plumbing to move the energy, matches the requirement. And that’s a little bit different than what everybody else looks at.
And my power supply designs are successful. Let’s say that.
1:04 Let’s say you want to move ten volts of power at 3.3 volts…
Then what I do is I design a single connection in my simulation, and determine what the impedance of that connection is. If the impedance is 50 ohms, then I only have the capability for moving a tenth of an amp at each wave cycle, because the energy travels in waves. So I have two choices, either to make the capacity larger by lowering the impedance, or I add more pipes to increase the amount of energy each wave cycle. So I would go to the best solution which seems to be on adding more pipes, close to each other, so you can have a concentration of field moving in a small space, but at a much higher rate, than if I use a single dielectric.
1:45 Is the presence of one ounce of copper on the power plane or ground plane really effective?
The weight of the copper is really not as important as people would think. It’s important if you use it for dissipation of heat. Then you get better thermal transfer with the one ounce copper. But if you design your system properly, half ounce copper is more than adequate for any type of system, including moving high energy.
The reason people have trouble, and want to use higher weights, is because they haven’t considered the capacity of the transmission line to move the energy. So, what happens is you end up with more of the energy being lost in the movement as it’s interacting with the molecules of the conductor, and causing them to vibrate, and that starts to heat up the copper. And you lose energy.
If you properly design the transmission lines so they are capable of moving the energy, then you can use half ounce copper because you aren’t going to have the losses you do if you under-design the plumbing.
A lot of people think they need to use one ounce copper for their power and ground planes. This is useful if you need that material to dissipate the heat from an IC. They typically use it in power planes, because I believe it’s a result of inadequately designing the transmission lines that you use to move the energy.
More or less copper?
What happens in improperly designed power systems is a lot of the energy is converted from electromagnetic field, into kinetic energy, heating up the board conductor material. And so people want to have more copper because it’s capable of withstanding more of this thermal interaction.
A well designed power distribution system has very little interaction with the copper. You’re not moving so much energy through that space as you are in a poorly designed one. So I don’t specify one ounce copper for my power distribution systems. I design them based on the requirement of the power supply and the capability of the transmission lines. If I need more than one transmission line, I add multiple transmission lines, and design for the energy movement. And I never even consider IR drop, or any of the traditional parameters. That doesn’t come into the picture if you have the proper plumbing in place.
3:16 Do you use any special tools, or simulation software, while designing?
Unfortunately, the simulation tools that are available today really don’t evaluate the printed circuit board design from the perspective of the basic rules of electromagnetic field management, which is to ensure that there is one dielectric between each of the signal conductors, and the ground. They don’t look at that three dimensional structure and see if the plumbing is intact.
So all of my design analysis is done by just simply looking through the board stack. And looking at the traces, and helping to ensure that from the load to the switch I can see an unbroken dielectric that’s bounded by an unbroken conductor, ground, and the switched conductor, the signal control or the power regulator interface. There has to be an unbroken path through all three dimensions for the energy to move through. If that’s not true, then you have signal integrity issues. No tool evaluates the design from that perspective.
And unfortunately, no company who does simulation tools has a plan to provide such a tool. It basically needs the ability to do a finite element analysis of the design. And if that were available, the number of boards and systems that fail EMC would be reduced dramatically.
4:53 How is this related to EMC?
Well, signal integrity and EMC are directly a result of the design of the transmission lines that are used for moving the signals.
A quick definition of a transmission line is two conductors separated by a dielectric.
Some people jump to the idea of a controlled impedance transmission line. Well, that is a transmission line but it’s a special use case.
The EMC and signal integrity are created, or destroyed, by the wave front. When you turn on a switch it starts a movement of field energy into the new space. So it’s not turning on a switch and current flowing in the conductors. It’s turning on a switch and allowing the field that’s behind the switch to move into a new space. It does so sequentially, like water moving into an empty hose. The front of that wave is where all of the action happens. The displacement current creates all of the noise. And depending on the switching speed of the device, and how well the geometry of the transmission line matches that wave front, determines whether or not you will have any radiated emissions issues, or signal integrity issues.
The ideal transmission line
If you have one nanosecond switching event then you need to make sure that you have a transmission line that is structured to accept that wave. Any EMC or crosstalk occurs during the wave transition. The ideal transmission line allows for a single wave, after the switch is turned on, to completely fill the space with the new field density, the voltage.
If there is a mismatch then you will start to get reflections, because the energy will not be able to be absorbed by the load, and that energy piles up and reflects back adding another wave. So you get more waves happening, and that starts the ringing so that the energy hits the receiver, returns back to the driver, can’t go back to the driver, returns back to the receiver. So you get this succession of waves in a system that you don’t want waves in, and that energy will continue to ring back and forth until it’s either converted into heat as it interacts with the molecules of the dielectric, and the molecules of the conductor. Some of it, finally, is absorbed by the receiver, but a great deal of it is radiated into the dielectric in the surrounding circuit crosstalk.
7:46 What are the other challenges in signal integrity and EMC in automotive?
The greater challenges posed by the automotive designs are a few. One is the extreme temperature range that these systems must operate in. We’re looking at systems that must function properly between minus 40 as high as 125 or 135 Celsius degrees. So that makes for a challenge itself.
The radiator emissions specifications are extremely tight because of the continued addition of radio receivers in the vehicles both the traditional AM/FM bands, the Sirius networks, and now with Bluetooth, then you’ve got cellphones in the vehicles, so you have to have a very low radiator emissions. The other side is they have a much higher susceptibility requirement because they put transmitters all over the car now. That challenge is what really my focus is on a daily basis. The focus in automotive is always low cost because the volume of the modules is very high. Millions and millions of each of these types of controllers are designed so that they want to reduce the cost.
I work with them and one of the classes I teach is called Effective PCB Design and the idea is to take the materials that have been selected for their cost model that we can use them effectively to achieve the goal of full EMC compliance and the performance they require for the application.
9:26 Are there any special PCB materials that are recommended for automotive applications?
The PCB materials that we use in automotive are primarily the standard FR4 type materials. I don’t know very many special use cases where a different material is used. There are some sub modules because now we have radar all over the car but they use this traditional materials and traditional manufacturing technologies they can, they’re not really excited about HDI or blind and buried vias and using any special materials such as the high Dk materials. There are starting to become a few higher performance modules. Now has been a crossover point where connectors have come down in price and they will do a motherboard with four, six layer traditional manufacturing and a couple of connectors to a CPU board with a high performance quad core gigahertz processor and external memory. Because size is important they might in those cases use HDI or blind vias.
10:32 How do you keep automotive electronic boards cool in that harsh environment?
Managing the temperature changes in a circuit board are extremely difficult and they vary rarely will do anything special to cool the electronics. The best I see is in the audio realm where the power amplifiers will be mounted against the metal case of the radio and they will use that as the heat dissipation engine. In most of the controllers they may or may not take the power driver transistors for the ignition system, for example, and mount them against the metal case as well.
But as a general rule your electronics must just live in the environment. Which is why 135 Celsius degrees is pretty high. There is a lot of dream in their future. They would like to embed the electronics into the machines. So they would like to put the transmission controller in the transmission. They would like to put the engine controller in the engine. So they’re always pushing the semiconductor and other companies to come up with solutions that can live at higher and higher temperatures.
So the silicon sapphire and some of the other things may become attractive at some point in time.
11:46 There are a lot of cameras, a lot of transmitters, a lot of high speed communications, wireless going on between one device and another device in automotive…
As the vehicles start to add more and more data collection systems both the cameras, the sonars, the radars, the lidars, the big challenge is how do you move that data from the sensor node into an environment where they can process that data and do their object recognition and pursuit of the autonomous vehicle world. They are pursuing high speed ethernets in the vehicles. We’re working with our customers to create ethernet gateway systems that will bring the data. There’s an effort now to do some of the re-processing by putting high power processors at the sensor node so they can reduce the amount of data that needs to be transmitted back to the central processor where the final decisions on driving will be made.
12:47 A kind of distributive process?
Yes. There was the trend in the past years to bring the processing into a central system but now they’re going the other direction. It’s becoming a distributed high performance segregated system where each different function is relegated to a different network so the vehicle control system is separate from the driver, enhanced driver applications to a separate network. And then the infotainment will be on a separate network. Those will all go to a central gateway and then that gateway will connect them to the communications devices, the radios, the other ways for the vehicle to talk to the network and the goal is that most of the cars will all talk to each other and as one car learns not to run over something they all learn not run over that same item. And shared intelligence.
13:49 Aren’t there any specific devices which help in reducing EMC radiation?
Ferrite beads and inductors used in control of EMC issues, that’s one of my favorite subjects. Early in my career, we worked with a consultant on a noise problem for a Chrysler vehicle. After a couple of months of analysis, they came up with a long list of ferrites and inductors. They wanted to add these to the controller at an added cost of 10 or 15 dollars. Well, that was absolutely unacceptable in the automotive world. You don’t ask them to add cost you show them ways to reduce cost. I never use ferrites and inductors in my designs. I recently did my first network control board, which is an ethernet gateway. It has a high powered network processor, 4 gigabytes of DDR-4.
It also has an ethernet controller, ethernet switches, ethernet phys, CAN and LIN control interfaces. The reference design that I started with had 28 ferrites and inductors in all of the inputs to the power supply for the main CPU. I took them all of the board and I designed it managing the plumbing properly. Then, I put the energy storage network and the board passed EMC first pass.
15:18 Your insight seems to be pretty new and non-traditional in that sense. Have you written down your guidelines so that designers can take advantage of it?
I’ve been working hard to put this perspective into a useful easy to understand format for a number of years. I teach this in all of my classes when I go around the world. Where I can, I teach at industry conferences to make this knowledge available. I teach it directly at my customers sites when I get the opportunity. And all of the NXP technology events, I teach as well. I go around the world to do this.
I’ve learned the basic rules because I was fortunate enough to meet Ralph Morrison. He is currently the oldest living EMC engineer. Ralph is 93 years-old. Wiley asked him to publish a new book last year which is now in print. And Ralph decided he was going to make me his apprentice. He was going to teach me electromagnetic field physics or it was going to kill me. And so I’m still alive and I actually met with Mr. Morrison on Monday. He made me feel very good because he said, ‘Dan I think you understand it now.’
My wife is always asking me to publish my work and I will someday. Well, Mr. Morrison wrote the first book on EMC. It was called Grounding and Shielding back in 1967. All of the subsequent EMC engineers always say that they have used that book as a reference. What Ralph has told me is that nobody ever really got the right perspective from what he was trying to tell people, which is very simple.
The energy moves in the space between the conductors. They don’t teach it that way in school. The methodology that we used for most of my career did not accept that idea. It was connecting wires because electrons carry the energy. The energy transfer doesn’t involve electrons. It’s an electro-magnetic field, and that doesn’t travel in conductors. It travels in the space. We use the conductors to create the boundaries for the field to move in.
Accept the rules
It took me more than 8 years before Ralph would stop yelling at me. I would say something improperly using the wrong language and it would earn me a one hour lecture on what the physics really was doing. The reason I’m successful is because I finally accepted the rules. It’s the one dielectric rule. You must have only one dielectric for each signal. You can’t put multiple signals in the same space or they become well connected, this is crosstalk.
The other person who’s responsible for my being here today as an instructor is one of the perennial instructors of PCB West and that’s Rick Hartley. My first experience with the signal integrity class was about 12 years ago. I came to PCB West for the first time and took a signal integrity class that Rick was teaching. I was a senior engineer, I designed 68,020 emulators so I was at the peak of my game.
At that time, I was in the process of designing a pre-silicon emulator. It had gate array based IP. My PC Board design team had to go get some learning. I didn’t know what I was doing. It was only about 15 minutes into Rick’s class that I came to the realization that every design I’d ever done worked by accident not by purpose. As a senior engineer that was a very humbling moment so my job is solutions.
I’m an applications engineer. I have to find out what it is. How to make it work to help enable my customers to be successful. So, that started me to pursue the knowledge. And it led to subsequently my meeting Mr Morrison. i said the right bad things to have him take mercy on me and teach me physics. I’ve been able to extract the simple rules that are behind all of the high level mathematics, you don’t need to understand that to do good PC board designs, you just need to understand that you’re a plumber now. You’re designing pipes for energy to move in.
20:13 What is the reason for failing EMC?
I want to take a moment to talk about the current state of the industry and electronics and the impact that the proper understanding can have. If you designed systems based on the actual physics involved, you can design a system that will be compliant both from a signal integrity perspective and from an EMC perspective.
The physics of electronics is quite simple if you just follow the rules. The problem is most people don’t understand what they do for a living. We design, manufacture systems that generate, manage and consume electromagnetic field energy. You can also do three things with electromagnetic field energy. You can store it, you can move it or you can convert to kinetic energy. When we start to think of things in that perseverative, it all becomes a lot simpler.
The problem is people want it to be complicated. As a result, especially driven by the changes in IC geometry and the increased difficulty of passing EMC standards as they continue to become more and more stringent across a wider band of frequencies, the status quo for the industry has become failure. Everybody expects to fail EMC. They design the module and don’t feel confident that it will behave properly.
This is well known and unfortunately accepted. The problem is nobody budgets for redoing the modules, both from a time perspective or from a cost perspective. They end up in a situation where they have to do the new design in an expedited manner. And they have to pay expedite charges for materials to fabricate the new boards to manufacture the new boards. They beg for time in the test chambers and do it over again with no confidence that it’s going pass the second or third or fourth time.
Come up with a method and a solution
I work with customers, they always come to me after they’ve failed two or three times and usually the solutions are very simple. Most of my board analysis time is about three to five minutes. Typically I get to the board stack and I’m done. If done properly, it enables you to violate the rules instantly. We can’t continue to do this, that is not engineering. That is children playing in a sandbox. Engineering is where you come up with a method and a solution.
We have to take the physics of electronics and we have to create systems that manage it properly. The status quo can’t continue to be we are going to fail. It has to start to become we are going to pass. And that comes from the knowledge that you’ve followed the rules. You actually have to accept those rules.
We have to start teaching the rules in school and we have to start getting the engineering teams around the world into the proper training so that we don’t have continued failure. This is not OK. We waste hundreds of millions of dollars every year building boards they know that will not pass compliance.
They don’t know whether they’ll pass or not?
I ask people “when you send your board back for the second test or the third test do you feel confident that it’s going to pass?” The answer is always no. Following the rules that I’ve learned working with Mr Morrison, when I send my system off for testing, I don’t think about it again. I know I’ve followed the rules and it’s going to pass EMC.
It’s very simple rules. I don’t, I’m not very smart I just figure it out, how to put my tinker toys together.