#704 – Applied Embedded Electronics with Jerry Twomey

Chris Gammell: This is The Amp Hour Podcast. Released October 2nd, 2025. Episode 704. Applied Embedded Electronics with Jerry Toomey. Welcome to the Amp Hour. I'm Chris Gammell of Contextual Electronics. Hi, and I'm Jerry Toomey from Effective Electrons. Welcome, Jerry. How are you doing? Doing well today. Doing well. Great. Well, I first heard you on the Embedded Podcast a little bit more than a year ago. I think it was June of 2024. And I immediately thought, man, I wish we would have gotten him on the Amp Hour first. But I enjoyed the heck out of that interview. So nice job on there. And we're going to extend on that today.

Dave Jones: Thanks. What kind of questions can I answer for you today?

Chris Gammell: How do I make better products, Jerry? I think that's the thing. Like your book, Applied Embedded Electronics. And, you know, it's about making better embedded electronics. How do I do that? How do I move things forward so that I don't have random field failures? How do I improve as an engineer? You know, I'm kind of at that point in my career where I'm capable, but I'm, you know, I'm not perfect.

Dave Jones: Well, I don't know if we can ever achieve perfect, especially in the fast-moving field of electronics and the methods changing all the time. How long have you been in the business, Chris?

Chris Gammell: 21 years, depending on how you count it. 2006 was really my first non-co-op job. So maybe pushing 20 years is probably a better way to say it, but yeah.

Dave Jones: Okay. I'm feeling a little bit old because I've been doing this for over 40 now.

Chris Gammell: I'm chasing you, but yeah. And I do kind of wonder as well, like from 20 to 40, you know, how do things migrate? How do you keep filling out the chapters in a book like this, I suppose?

Dave Jones: It's an ongoing process with the methods changing, but there's a number of things that covered in the books that are pretty time static that don't change. And one of the things that when I was developing this book, I said, I did not want to do something that was going to be outdated in six months. You know, here's an applications manual for a ST microelectronics microcontroller and how to use it, because that's going to be something that's historically significant within a year or two. I was dealing more with how to do robust systems. And a lot of what's in the book is motivated by a problem solving I was involved with for many, many clients where they'd come to me with a project that had been developed and they either hit problems with performance or noise or regulatory testing or something that just had them stymied. I'd get dragged into it. Usually ended up fixing a multitude of things that became the motivation for first large number of articles in electronic design magazine. I've written 30 odd articles for them and for EDN and a few other places as well on various technical topics. It was basically a situation of saying, okay, these are the things people have problems with. So the name of the game is let's put something together in an organized manner that'll take me out of business fixing other people's problems. I can say, read the book, read the book.

Chris Gammell: You've been consulting for a long time, I suppose. So what were the kind of the subject areas? You mentioned EMI style things. I saw you're in San Diego where there's a lot of RF stuff. What were the industries and verticals you were often working in?

Dave Jones: The list is long. A little bit of a quick history. I did college and grad school in Massachusetts. I worked in the electronics industry around the Boston area for a while. Got recruited to go out to San Jose, Silicon Valley and spent about 10 years there. And then I ended up going down into San Diego where the weather's nicer and there's people with lives and things other than work. Silicon Valley can be a little bit work and fast.

Chris Gammell: Much better tacos and stuff too, I think, right? Yeah, for sure. Yeah. For sure. I think San Diego, I think the military electronics, I think about Qualcomm's down there. There's a bunch of RF stuff. One of our longtime guests of the show was a RF designer out there. I often kind of think military RF, high-end chip stuff.

Dave Jones: Yeah, there's a lot of that down here for sure. There's also a lot of life science stuff. I've also migrated in the past 10 years or so to doing a lot of medical electronics. Before that, I've had times when I've done consumer stuff, aerospace stuff, defense projects. I've done DARPA research things. And the list kind of goes on. And I'm a little bit of an anomaly because I consider myself a DC to daylight kind of analog person. But a lot of what I've ended up doing has been for the digital world. Very high-speed data systems. They're basically a communications channel problem. They're not really a digital problem. A lot of things with robust design, you ask the question about, you know, how do you become a better designer? One of the things I did with the book is I broke it down into a bunch of topics that people frequently have problems with and covered each of those in depth as much as I could. I mean, the book's 600 pages, so consequently, there's a lot there. It was a situation where one of the first things I had to do was teach people about the fact that what they learned at university about the nature of signals and systems is frequently not the real world things. Academic stuff teaches you about ideal digital systems, for sure. They're going to teach you about ideal signal systems, but they're not going to go into things like noise, immunity, and second-order problems with things like inductance and cross-coupling and a multitude of other things as well.

Chris Gammell: Got it. So they call you up and they say, Jerry, Jerry, look at this logic analyzer. It's got all this garbage on here. And then you pull out the scope. You pull out the eye diagram. You pull out the VNA, that sort of thing.

Dave Jones: That's basically been part of it many times. I'll give you a very typical example. Medical devices, company can remain nameless, but I got called in on a project and it was a situation where they had gone in for regulatory testing. The regulatory testing associated with medical devices is pretty strenuous. You've got applied EMI issues, applied ESD issues, magnetic field issues, and things like that where the system has to keep working throughout all of that stuff. And you can't have the blue screen of death on ESD pulse because it's a medical device and that's not going to be well-received in the operating room, to say the least. So it's a little bit more strenuous in the testing. It's kind of like an aerospace way of thinking about it where reliability is first and foremost cost is much less of an issue. But getting back to the topic at hand, this particular client, medical device, and they went into regulatory testing and they found all kinds of problems. The guy that was in charge of the project was much more of a digital person and very little experience on the analog or the signal integrity side of things. I found things like no ESD protection and no running cables off of boards where there was data streams, where there was no error correction. There was no detection of errors. There was no common mode rejection of noise. It was ground reference signal going off of a board, which is a suicide jump when you're trying to have something work in a hostile environment. Early on in the book, the first one of the early things I do is I talk about situations with the fact that a lot of what you get in academia is ideal. Ideal signals, ideal data streams, and no noise, no corruption of signals, things like that. You might see a little bit of that in grad school, but that's kind of about it. I go into that. I go into basic components, and certain components are non-ideal relative to capacitor variances. Everybody treats capacitors until they get some experience. They treat them as, oh, I put down a 0.1 microfarad capacitor. The reality is something very different depending upon what the actual capacitor is. That's very simple stuff. Right, right. But a lot of it never gets covered in academia.

Chris Gammell: What were the levels of engineers you were working with as well? I mean, I guess it's going to depend company to company, but when are you getting called for these kind of engagements as well?

Dave Jones: I've gotten it from people at all levels. Case in point, a few years back, I was developing high-speed data for IBM mainframes, and that got done because of the fact I had done something similar a year before that was with Cray, again, doing high-speed data over copper within the computer systems. And IBM wanted a similar system. So I ended up doing high-speed data for groups of people that were extremely experienced when it came to developing computer systems. But this niche area of high-speed data signal processing was a little bit outside their forte.

Chris Gammell: Yeah, those names are not around much anymore, Jerry. In the same context, at least. Now it's like IBM's like, I'd like to sell you some SaaS software solutions.

Dave Jones: Ah, but IBM still does mainframes, and Cray is still out there doing NSA computers. No way. Okay, I learned today. I had no idea. Cray still does massive multiprocessor systems, whatever the acronym is for that. Lots of parallel computing processing. Those projects were over 10 years ago, but I think they serve as a good example where you're going to get extremely experienced people that are lacking knowledge in a particular area.

Chris Gammell: Right. And I can imagine, too, as systems scale, like you said, the interconnect, the things that are, you're going to get at the edges, you're going to have problems at the edges, versus someone who's like a complex VLSI, like silicon designer, and thinking about like, they're not thinking about the wire, they're thinking about the silicon, that sort of thing. So like having that cross-knowledge capabilities do seem like you need to kind of bring a bunch of skills in there.

Dave Jones: Yeah, it becomes more of a communications channel problem at that point, because of the fact that you're trying to put data across wires, where it's not ones and zeros going across the wires, it's actually multiple standing wave transitions. So you have multiple bits on the same wire at the same time, just doing its transmission line going across the wire. So you start to get down to your bits underneath a tenth of a nanosecond and worse, and you basically start running into all kinds of interesting analog problems. But, you know, that's just one particular area.

Chris Gammell: That's actually probably way past the kind of the threshold of what I would actually be doing as well. So like when I think about, you know, improving my own skills, I mean, I guess there are nanosecond transitions, but they're, you know, they're clock edges that need to be straight, but it's not, it's not something that I'm like clocking quite that fast.

Dave Jones: Well, it's an interesting thing that you mentioned that because of the fact that when you have a data stream going across wires, what are the things that limit your performance? And a lot of people will say data rate, but I think it comes down to two things. And I go into this in the book quite a bit in the chapter on reliable data communication. And that is rise and fall times and distance traveled. If you have extremely fast rise fall times, you have a situation where you have a transmission line problem. Even when the distance traveled is, you know, under a centimeter, you need to understand those two things and how they play off of each other.

Chris Gammell: All right. Well, maybe we could do some example quasi design activities here, just as like, you know, gaming it out. Thinking about when I call Jerry, when I, when I read the book, that sort of thing, maybe we could talk through it. You have some stuff in there about choosing power supplies, about choosing microcontrollers, about defensive coding kind of really does cover the gamut of runs, the gamut rather of different thought processes you need to have different considerations needed to make. But if you were brought in at the right at the beginning of a project, let's talk about like a glucose monitor. That's a Bluetooth glucose monitor talking, you know, to a, to a gateway. Is there, are there any special considerations you start with in those kind of application spaces or, or is it more, or maybe that's too pedestrian even for a lot of these considerations? Well, it's funny.

Dave Jones: You mentioned glucose monitors because I've done those for Dexcom. Oh, okay. Yeah. Okay. There we go. I helped them get through three of their early generation products. Two of their on-body monitor systems, and then a combined insulin pump and glucose monitor for hospital use.

Chris Gammell: All right. So we're right, right in the wheelhouse then. That sounds great. But I think medical devices these days, I think a lot of that stuff, it's kind of the time is right. There's a bunch of stuff around it.

Dave Jones: Oh, for sure. I helped develop a wearable EEG monitor as well. A few years back, that was kind of a nice project that I enjoyed because of the fact that before, if you were trying to monitor EEGs on a patient, you had to have them in the hospital in a bed with, you know, all kinds of stuff wired up to them. This became a manifestation that looked like a deck of cards kind of thing that you wore. And it linked out, but it also recorded and kept the memory so that you could just hand it back in, and it was all there. So even if the RF data links were lost, you still had the whole EEG recording. But I would start with something of that nature. If it was a new project of architecting the system, architecting a system depends upon, you know, what is it you're trying to do? What's coming in? What does the user need to see or do? Defining your needs, defining your interfaces, and things of that nature. And that can be simple from a standalone, bare metal, microcontroller kind of setup at the heart of the whole thing. Or it can be a multiprocessor distributed system. Industrial systems use that a lot. You can have two extremes there. And then you have to scope out the system for what's needed in the way of communication between devices. What are the devices that are needed in the system? And before you even go into the guts of things, you're looking at it as a black box. What the application end user's person sees and figuring out from there going one level down into the electronics. That's typically a starting point.

Chris Gammell: Got it. So if you're advising on like these kind of architecture level decisions, usually in medical space, it's not super cost constrained. So that's not driving a lot of it. But are there thoughts around, you know, just generic, like how often is it like jelly bean, like, you know, five volt to three, three buck converter type of things that you worry about versus the more kind of esoteric, weird parts that might be needed? Like how are you thinking about those sort of things?

Dave Jones: Once you get a basic architecture done, you've got a situation where you're determining your peripherals and your interconnects. You're starting to develop a little bit more at the block level of diagram of a system. And usually what happens with stuff like that is things start to very quickly rear the heads going. This is going to be the bottleneck of the system. Sometimes it's getting power consumption down to virtually nothing. Medical implantables, that's a big issue. So you want to get the power consumption down to, you know, like virtually nothing. And that could take up 90 percent of your effort. And the rest of it's straightforward. Or I worked on a multi-camera endoscope. And the thing that turned out to be the gating problem with that system was the miniaturization of the distal head on the endoscope. In this particular case, we were packing five video cameras into an area that was 11 millimeters in diameter and 13 millimeters long. And we had five video cameras in there. Wow.

Chris Gammell: And that's the actual like sensor down in the endoscope, not like fiber coming back out to a sensor up top.

Dave Jones: Correct. The fiber sensor method was done many years ago and has become obsolete. The cameras are in the distal heads of the endoscopes these days. But for that particular situation, we developed an entire endoscope system front to back. The desktop box, it basically runs and controls everything, all the computer systems, et cetera, et cetera, et cetera. And it was a situation where the gating issue that took the most amount of engineering time was the miniaturization of that distal head.

Chris Gammell: Got it. That's good, though, because like in both those cases, you're kind of talking about kind of figuring out the hardest tasks. It's not saying, oh, the Bluetooth part will be easy or, oh, the power will be easy. But it's just like, where are we going to focus our attention? Do you have any methods for kind of intuition? Is it experience? Like, how do you figure out what will be the hardest part of those black boxes?

Dave Jones: I think it's an experience question largely. Once you define the architecture of a system, certain things will stand out as straightforward to experienced designers. You know, you've been in the business for 20 odd years. So I'm sure you've looked at a system and gone, this is the beast. The rest of it's straightforward. And consequently, those types of things largely gate on experience. I've always said with a good engineering team, you can have 10 fresh graduates out of college. But if you got one or two experienced people to give them guidance, you can get something productive out of all of them.

Chris Gammell: It's tough to have that experience sometimes, too. Like, I mean, it sounds like in this scenario, too, we're also talking about companies that are going external to hire your experience or similar kind of things as well. Right. And so then you'd be coming in, you'd be guiding them. You'd be like kind of pointing out like that. Oh, there'd be there'd be dragons, that sort of thing.

Dave Jones: There'd be dragons with big headaches. Yeah. Make sure you have a large bottle of Tylenol.

Chris Gammell: Yeah. And it does kind of seem like from your, you know, just going through the various kind of high level chapters as well, the power systems, battery power, EMI, external driving peripheral stuff like that. It sounds like each of these might be definitely areas of focus for those junior teams that you're advising.

Dave Jones: Sure. But it's also a situation where if you look at the nature of the book, it's very much a situation where the things that I focus on are the things that screw people up in their design cycle. And those things predominantly are analog in nature. You may be dealing with a digital system end to end, but the problems tend to be analog. Well, think about it. Digital communication is all about data corruption. Short from low speed and short proximity connections, you know, like an I2C situation, it's easy on a board because you're going maybe a megahertz, a few megahertz in data rate, a couple of inches across a common ground board. That's easy. But you start to do high speed data, it starts to get messy. It starts to turn into an analog problem and a signal corruption problem. Power systems are similar. What screws up in power systems? The things that fail are usually due to noise. You get a situation with switching supplies. You got a couple of buck converters in a system. They're all generating all kinds of noise and they can get into your analog front ends or in the way of switching noise, or they can just make you fail FCC section 15 on EMI, radiated EMI. Again, it's analog stuff that has problems. The power systems, one of the things in the power systems chapter in the book, I go into the basics that are available to you. And I will usually suggest that if you need an AC to DC converter, buy that commercially. You could develop it yourself, but the amount of time and effort involved with that, when there's thousands of options out there from highly qualified companies and they're cutthroat competitive on pricing, it's usually more time and money effective to just go buy the AC to DC converter off the shelf. Then you get like, let's say 12 volts or 5 volts or 24 volts or whatever's needed into your circuit boards. And at that point, you've probably got to go down to 5. You've got to go down to 3.3. You've probably got some stuff at 1.8. You've probably got some stuff at 1.2. And, you know, the different voltages of the various CMOS families. So, you've got a situation where those buck converters are going to be something that is generating EMI or classic case of digital thinking in an analog scenario was when somebody had a very sensitive analog front end for signal processing and the power supply that they put on the thing was coming from switching converter. And basically, the switching converter was swapping the analog front end with noise. So, again, it's the analog stuff that comes back and bites you.

Chris Gammell: And when you come in and you like kind of parachute in for these sort of things as well, is it like you are often working directly with embedded engineers and you're kind of the analog guru who's kind of helping to wave the wand and make the problems go away?

Dave Jones: That's been largely it. I wish it was simple as waving wands.

Chris Gammell: Well, of course, RFI wands, you know, little current loop wands, right? Okay. If you're sniffing out signals, for sure. It's interesting. I do wonder about like kind of effective communication with those teams as well. Like, I guess even in a design context, too, like what does it look like when you're interacting with an embedded team?

Dave Jones: Well, I like to basically work with teams in a manner that doesn't alienate people. And as we all know, in a group of engineers, you've got a group of egos that you've got to deal with as well. And the name of the game is let's have a conversation about what's going on. Let's understand it together and help guide them in a self-discovery manner to boot. I think we've all seen the big pretentious person that comes into the room. Seagull manager is a lovely, often known as, you know, fly in, dumps on everything, flies away.

Chris Gammell: I've never heard that term, actually.

Dave Jones: That's pretty good. Seagull management? Yeah, I've never heard of that. Yeah, that's good. The word is usually something a little bit different, but you get the idea. Got it. Got it. OK. You don't want to do that because then the people aren't going to listen to you the next time they see you. It's a situation where, you know, you get in a lab with a person, you start looking at things on a scope and you say, well, you see this and here's the ground over on this board and here's the ground over on that board. And let's see what happens when we turn the system on to the two grounds relative to each other. Then what happens is you get a situation where the person, where people learn about the fact there is no such thing as an ideal connection between things. There's always some, you know, RLC, inductance, resistance in the connections. And then you say, OK, now typically what gets done to deal with this variation in ground between the two is the signals ascent differentially. And by doing that, you can usually compensate for a volt or two of common mode noise. That works very well at, you know, two adjacent circuit boards, the reasonably good chassis ground underneath them. But, oh, now we're going to extend that out to 100 feet. Now we have to take a different approach. And again, it's a matter of having a conversation with the involved parties about what's the proper way to deal with it. But let's have a self-discovery process where you learn about the nature of the problem in a hands-on manner, because if you got the signals in front of you, you can't argue with the nature of the signals.

Chris Gammell: Got it. Yes. So it does sound like a lot of like bench time, a lot of kind of empirical measurement, things like that, you know, kind of discovery, discovery on the bench, that kind of piece.

Dave Jones: That's a good part of it, because when I'm doing medical devices, usually what happens is it's not so much a discovery on the bench as in the testing shop came back with all the regulatory testing results. And because of the failures to pass on all of that stuff, that is your lab data that you then have to sit down and address together.

Chris Gammell: And then you strategize through that sort of thing. Talk about what board turns you might be needing to make, that sort of remedial actions. I'm sure that's also one that when the lab report comes back, that's also when your phone starts ringing in the general case anyway. It happens that way. Who could fix this? Jerry could fix this. That sort of thing.

Dave Jones: Yeah. And sometimes when I've developed relationships with clients, I've had situations where they get me in earlier in the game rather than later in the game so they don't make the same mistakes. And that can go well. And I've had situations where the design engineers have said, can we get Jerry Toomey in here to coach us on this a little bit because we've never done this before? And sometimes what happens as well as if I've worked with some of these guys or ladies in the past, I'll get a call informally and I'll try to help them out as much as I can off the record. They come to me with questions and stuff because they've worked with me in the past.

Chris Gammell: Got it. Yeah. It's interesting because I mean, you do have a chapter on coding as well, but all the things you're talking about, it sounds like a lot of this stuff really does get solved. And the title of the book too, Applied Embedded Electronics. When I first approached it, when I first heard you on Embedded FM actually as well, I kind of assumed that it was going to be more kind of embedded methods that were going to fix things. But like you keep saying, you kind of keep coming back to the analog side of things. So it does seem like a lot of this is happening at the board layout stage, at the block diagram stage, how you're choosing architecture, that sort of thing.

Dave Jones: It's a situation where things like architecture decisions. One of the things I always like to say is I never have noise problems on anything that I design. And the reason I can say that usually with pretty good authority is because of the fact that when I build a system up, design it myself from scratch, I include all of the contingencies and all of the proactive solutions to noise and EMI and ESD issues from the get-go.

Chris Gammell: Here's a man that likes the zero-ohm resistors and your extra placeholders for capacitors and that sort of thing.

Dave Jones: I have no problems with doing do-not stuffs on first-generation boards. Oh yeah, big time.

Chris Gammell: It does kind of contrast a little bit with size constraints, but I feel like it's like you just say, the first one's not going to be the smallest design. You're going to move your way down the stack and make smaller and smaller things. Over time, right? Sure.

Dave Jones: You would have gotten a kick out of, it wasn't a project I was involved in, but a friend of mine was involved with this one back in my San Jose days. The very first generation DVD player, this would be early 90s, I think it was.

Chris Gammell: I just remember them being very, very expensive. Like, wow, that stuff was high end at the time.

Dave Jones: The first generation DVD players out were, there were no IC system on a chip for those devices at that time. And it was a big, messy circuit board scenario for the first couple of builds. And they got it down to something like a single board that 10 by 12 inches, something pretty big like that. And that's what went out as the first generation DVD players. And this is ancient history, but the name of the game is the thing hasn't changed in the fact that they got something out there. There was a large amount of discrete circuitry in it, not specialized circuitry, not system on a chip mentality. For that first generation product, the product went for over a thousand dollars. And this was back in early 1990s money.

Chris Gammell: Of course. Yeah. I look at the, you know, I was at Keithley Instruments and the stuff that was out in the nineties. I looked at that in the 2010s when I was working on it. And I was like, oh, this stuff looks so old. But at the time it was like, no, that was, that was what was available. And that's what the chips were at. And it was not state of the art at the time, but it was, it was just the normal electronics at the time. And it's what you had available. So you worked with it.

Dave Jones: As you go back, historically, you've got a situation, you go back to a lot of single transistor level designs, stuff that was more, to me, that's 1970s, 1980s kind of thinking where there were tons and tons of discreet in design. Because you're going back into an era where the definition of an integrated circuit back then was 555 timers, 741 op amp and 7400 series logic. That was it. Beyond that, you went to discrete transistors.

Chris Gammell: Right. The era of looking at Jim Williams schematics and similar, right?

Dave Jones: Yeah. Dealing with, be it Barry Gilbert or Bob Pease. And, you know, this is all going to be analog. And I'm sorry, this is, it may be an analog world, but the way, the way you interface to it is still very digital, but you do have all kinds of analog considerations making that real.

Chris Gammell: You know, thinking about your book and kind of your role as you, when you come into companies too, like how are the youngins going to learn this stuff? That's, you know, one thing I think about is like, you know, I'm sure LLMs will get better, but there's no Jerry Toomey, LLM, nor is, would it be able to act in the same manner? You know, there's design rules you can put into place in CAD programs and similar, but it does feel like when stuff hits the fan, at some point people are going to need to have this knowledge and kind of the intuition there. How are the newest engineers supposed to gather this intuition?

Dave Jones: Frequently it's by hard, painful learning experiences, but it was that way for us when we were starting out too. One of the reasons for the book was I said, I don't want people to have these learning experiences. They follow the guidelines that I've given them in the book for, you know, everything from how to set up a battery system to how to do their digital system to how to do the power system, EMI, their driving peripheral devices, sensing things, all of that. If they follow the guidelines that are here, they are going to eliminate a lot of the problems that they're going to run into if they were on a learning curve. Because first time design in any particular area, you're not aware of a lot of things. You're not aware of noise. You're not aware of the special needs of the system. And consequently, you may end up doing it more than once.

Chris Gammell: You don't know what you don't know. And then you eventually you do know what you don't know, right?

Dave Jones: You put it together and then you go fight the fires, put the fires out. And it's funny because about half of my career has been in board and system level work, but about half of my career has been in integrated circuits. I've done a lot of analog mixed signal RF design at the chip level. Nowadays, modern engineers, they go, oh, we're going to simulate this. We're going to simulate this. We're going to simulate this. And if I'm designing a chip, I'm going to be living on a simulator. No discussion on that. That's a given. But if I'm designing a board, my personal approach on that is I will rarely simulate anything. I will tend to architect it out by hand. I will tend to build something up and then I will experimentally work with it and improve it as on an as needed basis.

Chris Gammell: What kind of chip level stuff?

Dave Jones: Are you still doing that or is that more in the past? More in the past. I've been called in as a consultant a couple of times recently on stuff, but it's basically been people with special needs. They didn't understand the issues and stuff that I solved in a day or two kind of thinking. The last big, big chip projects I did were probably RF power amplifiers. I used to work with a company down here in Southern California, Axiom Micro Devices. They bought out and gone. Broadcom bought their product line. But I was, I think, employee number six in the door there as a startup. We were doing RF power amplifiers, but we were using CMOS foundry processes to do that. And if you start putting together the voltages and currents and all the information associated with an RF power amplifier and you go, wait a second, this does not work on a CMOS process because of the voltages involved. You have to go to gallium arsenide or, you know, some other friendly high frequency, high voltage type of process. We were doing electromagnetic methods, doing things down at lower voltages at the CMOS get handling it and then going out through an RF transformer to create an actual RF power amplifier, but doing it on a cheap CMOS process. So it's been a couple of years since I've been doing IC design, except for, like I said, doing some firefighting for various associates with specific problems.

Chris Gammell: Yeah, I never gotten into that space. I mean, I can imagine that that really brings home, especially with the simulation kind of aspect of it, but like the base level nature of understanding, you know, more integrated circuit you might be pulling off the shelf and using in a design at a board level kind of thing.

Dave Jones: Different world. And even when you're pulling any chip off the shelf and using it, the nature of the challenges are going to be very associated with the problem you're trying to solve and the signals you're trying to process. Because sometimes it's fairly straightforward stuff because it's low in frequency and you don't require crazy resolution to do it. But other times it's really high frequency stuff or really small signals that you have to be extremely sensitive to how things get done.

Chris Gammell: Ground planes and things like that on silicon, are there kind of gotchas in that space as well?

Dave Jones: Ground planes and silicon are an exercise in suffering the fact that you will have high frequency inductance issues and you will have the ground plane twist like a chip due to current injection into the ground plane if you don't know what you're doing. And you will have a similar problem with boards, as I'm sure you know. You've probably, I'm sure you've seen variation in ground across a board dynamically. So that's definitely an issue or problem. There are ways to deal with it. When you're inside of a chip, absolutely without doubt or question, it will always be a fully differential design, no matter what it is. And that is because you are never going to get a good hard ground reference system on a chip. It doesn't happen.

Chris Gammell: So kind of like a rule of thumb for you is just to think differentially on almost any ground impacted system, sounds like.

Dave Jones: If it's inside of a chip, it's always differential. If it's on a board and if it's a five volt swing signal and I need to detect, you know, one volt, three volt and four volts with 100 millivolts of variance on it, they don't care. That can be ground referenced.

Chris Gammell: Well, and you've written articles about grounding in the past as well. I mean, usually when I start thinking about grounding, I think about like, you know, star ground versus common ground versus, you know, split grounds. Where do you kind of fall on a lot of these things?

Dave Jones: Well, that's to me more of a chassis and multi-board thinking. Typically grounding, if it's on a singular circuit board, common uncut ground is the way to go. Dedicated ground plane is the way to go. And sometimes if it's a big complex board, there'll even be multiple ground planes, multiple ground plane layers, but stitched together with vias. Yeah, tied together. Yeah, totally. Yeah. Yeah.

Chris Gammell: And I feel like the prevalence of like low cost multi-layer PCBs have really kind of changed the starting point for me. Whereas I may have like suffered a two-layer board in the past. I'm like, why am I even bothering, you know, just like started four and don't even be shy about going to six these days.

Dave Jones: Totally agree with you. A hundred percent. The thing is you have to then teach people going, Hey, why have you cut the ground plane up here? Oh, I needed to get a signal across. And then you go, okay, you get out the soldering iron and you patch the signal out through a discrete thing and you put a copper patch over the thing and you go back to a solid ground plane. And then you look at the spectrum analyzer on EMI and you say, see, you just changed the signal radiating from the board by about seven decibels. Right.

Chris Gammell: Got rid of your, your slot antenna there.

Dave Jones: There's a lot of people that they, they come out of college and they don't understand this stuff, especially if their focus is digital.

Chris Gammell: You're talking to one of them, man. I mean, that's the thing. Like, where do you learn this sort of thing? It's like you learn it from sitting down next to a engineer such as yourself, or, you know, maybe someone who's at your company, but it's, I've written in the past about like the solo engineer and I've written about it lovingly. Like, I think it is possible to do so much more as a standalone engineer, either at a small business or just on your own if you want to build your own stuff. But the loss there, and often it's driven by monetary factors, is just like the fact that there's not anyone to learn from, you know, it's like I can learn from the internet maybe, but that's not quite the same as being on a bench next to you, Jerry. Well, that's also separating the wheat from the chaff when it comes to the internet.

Dave Jones: I don't know if that's getting better though. I think it's getting worse. And probably so. We've all had situations where people get on Reddit or EDA board or one of the other places that host engineering discussion groups. And some people love to have their opinion, but unfortunately they don't have the experience or knowledge to back up what they're proselytizing.

Chris Gammell: Sure. Sure. But on the internet, nobody knows your dog.

Dave Jones: Yeah, this is true. Even if you do get a blue check mark.

Chris Gammell: Right.

Dave Jones: You've got to pay for that.

Chris Gammell: You've just got to pay for it. That's right. Honestly, that's what I worry about with the, you know, the LLMs kind of being in the mix these days too, is that like, you know, I've fallen subject to it myself of just like, oh, well, I lovingly call it the dummy box because it makes me into a dummy. But like I asked the dummy box for a question and it confidently tells me an answer. But like, if it doesn't know anything about signal grounding and, you know, best practices, I'm, you know, if I'm taking it on faith and not looking at references and not looking at multiple sources and not looking at, you know, sources of truth, like books and primary sources, it's like, okay, well, I might just be listening to a blue check mark or, you know, a dog on the internet, but it's actually just in the form of a LLM, you know, that sort of thing.

Dave Jones: Yeah. And the thing is the AI tools that are out there, they're in their infancy, but they're growing and they're improving. Most of those have lots of room for improvement. But again, I cautioned that, you know, that was the case when the Osborne portable computer came out 40 years ago and it was a huge box with a huge handle. And this is a portable computer back in that era and look where we are today. So I think you're going to see that improve. But there's room for improvement. And the thing I tell people that feel threatened by AI, as I say, go ask AI about the topic that you know more about than anybody else in the world, or you've worked with extensively, and you'll actually find out how stupid AI is right now. It'll get in better. It will get better.

Chris Gammell: I'm not concerned about it getting being not being good enough. Now I'm worried about my own kind of reliance on it, you know, kind of that sneaking reliance on it where I ask it a question. Oh, okay. You know, try a thing. It's just try a thing. It's just that sort of thing. But then more broadly, I worry about kind of the whole world being in that in that direction as well. And it's like, yeah, I'm sure it's going to get better. But the true intuition and you know, when the rubber hits the road, when someone's not passing EMI testing, and they're on a bench, and the LLM or the dummy box is no longer suggesting things that make sense, then what, right? Like when there's truly no one who knows anymore, you know, like, what do we do then?

Dave Jones: And that's why they write books. And that's why even today, people read books from 3000 years ago that were written in Greek, or written in Latin, or whatever. And they're still able to find philosophical knowledge from those things. It's not like they're depublishing your book and others. So yeah, I guess that is a good point. Goodness. Now that would be a badge of honor having my book banned. Yeah. But we've got a situation where one of the many other things that gets done here is the organization of the book, as we started to talk about before, it's based upon topics. It's based upon, okay, you got to do a battery system. So there's a chapter on how to do battery systems. And what are the options there? What are the battery chemistries there? What's viable right now? And I always have a caveat emptor on that one of the fact that, oh, battery technology is changing and changing a lot. It's growing by leaps and bounds. It's getting progressively better, but it's a slow progression. There haven't been any radical breakthroughs. It's been more of an evolution than radical steps taking forward. Yeah. I guess no new chemistries that a lot

Chris Gammell: of people are going to have access to. One thing that I've noticed in the battery space has just been maybe, and it may just be where I'm focused personally, but like, it's just like the increasingly small battery size, you know, just because of size constraint for devices and the fact that they can make, you know, 20 million amp hour batteries and just these ridiculously small form factors. And then it's down to the engineer to like, just make designs that can sip incredibly small currents and still wake up and do something useful with that. And, and, you know, persist for hours, days,

Dave Jones: weeks, whatever. Bluetooth is a great example of that. Bluetooth systems, Bluetooth, low power systems do two things. One, they work with a extremely small duty cycles. So they're very briefly on do their thing. And then they, they're, they're basically off for 95% of the time or even more. There's a background digital controller, but the, the actual RF circuits are live for extremely short periods of time. And the other thing too, is they do try to power optimize what's there. So reduce the overall power, reduce the duty cycle. And as you know, things like Bluetooth or distributed network systems nowadays, um, internet of things are all. I should probably should have known that. That's the,

Chris Gammell: that's the space I work in. I should probably should have been what you were talking about,

Dave Jones: but yeah. Yeah. And IOT stuff, a lot of it, they require two hours of battery life as to pass the standard. I'm sure you've run into that already.

Chris Gammell: Uh, actually I, I don't know what you mean by that. Like a Bluetooth based device that is IOT centric or is there some kind of regulation that I, maybe I don't know.

Dave Jones: There's a few standards that are out there that are for IOT where they've got a distributed mesh network in a house and to meet the standard of the product line that's doing it. They mandate that the battery life must be two years or a year or whatever the heck it is, because they're, they're doing a mesh network throughout a house. Now, do you really want to go back and change out all the batteries in the 35 things you've got around the house every two months? No, you don't. Of course not.

Chris Gammell: Got it. I heard you say two hours. That's where, that's where my confusion was. So did two years makes more sense? Okay. Yeah. Yeah. So got it.

Dave Jones: Two years. Yeah. IOT does a lot of that because they're, they're pushing wireless and they're pushing basically wireless communication and no hardwired power whenever possible. I've got mixed feelings about that. I mean, I've kind of, of the opinion, I prefer things that plug into the wall just for reliability, get the batteries out of the equation. Consequently, in a lot of applications, I'll say, do you really need this thing to be battery powered?

Chris Gammell: And that's like a system level design too, right? That's like an architectural type of thing. It's like, is it something that's driving product decision versus just a expectation from a product

Dave Jones: maker? Sure. As we all know, ease of use is important. Reliability without having to muss and fuss with it is important for something to be marketable. There were, there were a bunch of products out there probably back in the nineties through the 2010, 2015 region. There were a lot of things out there where the battery life was not that good. Cell phones come to mind. You probably remember Nokia phones being extremely popular because they got the battery life thing down to something low aggravation where you constantly weren't on a charger.

Chris Gammell: Whenever I think of like nineties battery problems, I was thinking of the Sega Game Gear. My friend would like load six double A's into there and it'd get like four hours of battery life. And then he'd like frustratingly like rip them out and replace them. And it would just be a pile of dead batteries if he was going on a road trip somewhere. And it's just like, oh, that's kind of wasteful. At the time they had like, they were like linear regulators and some of the handhelds that were out there too. And there were full color VGA, you know, style displays. And it's just, no, that's not going to work out great in the timeframe it was in. Well, as long as they're not having to power up a cathode ray tube, right? Yeah, right, right. Exactly.

Dave Jones: You know, a lot of this is stuff that you basically, you can get experience and you can learn about the stuff. But a lot of times what happens is you learn from your mistakes and I'm trying to minimize the number of mistakes. You're going to make it work with the chapters that are in the book, because we're dealing with how to do things and how to do things based upon what's needed to be applied in the real world.

Chris Gammell: Well, looking forward, I mean, like you've had a long storied career. What are you excited about the electronic space in the near future? Are you trying out new projects? Are you writing new books, that sort of thing?

Dave Jones: I've been debating doing a couple of other books, but I don't think they'd be electronic centric. I have a pet peeve of mine. You're familiar with Ray Kierswell? He's a futurist that likes to claim the singularity is coming. And he writes books about the singularity approaching and things like that. And I've read a couple of his books and I look at some of the things he predicts and he's just so detached from what can be realized. You know, magical things. We've reached limitations on CMOS scaling. It's kind of like they talk about the two nanometer CMOS node, but the truth be known is you really can't go below. I think it's 10 or 11 nanometers due to physics. And consequently, they're sort of lying to you when they say they, oh, this is the two nanometer mode. And I go, okay, where's the two nanometer features in this thing? And they don't exist. So we've been running Moore's law, hot and heavy for 40 years. Yes. Right. And then in the right here and now, the way people are improving things is they're not improving things with smaller transistors. They're more improving things with more parallel processing. And that is an issue that I haven't seen any breakthroughs. I, I read a little bit about quantum computing here and there, but I haven't seen anything yet that I would consider is going to get the same kind of widespread application that you would actually have,

Chris Gammell: you know, with a typical CMOS ASIC. Yeah. I wonder about in, in a lot of these cases too, of like they were driving up efficiency by driving down power rails in the past as well. So parallelization, lower voltages, but we're just like, we've approached diode voltages for, you know, like it does seem like kind of the, the gas is out of the tank. And so then where, where do you go from there? Software also does improve a lot of these things, you know, just writing better algorithms, that sort of thing.

Dave Jones: There's an article I did for electronic design. It's still up on my website and electronics design still has it on their website as well. And the title of the article, I may, I'm paraphrasing here was,

Chris Gammell: um, tiny transistors, giant molecules, Moore's law crashes into the laws of physics. Yeah. I'm

Dave Jones: looking at it right now. That's it. That's a good one. The predictions that I had in that sort of, you know, they rang true, but the way we did things even back 15 years ago is we're going to more and more parallel processing. And if you look at the clocking of PCs microprocessor, uh, the PC microprocessor clocking rate hasn't really significantly changed in the last five or 10 years. You know, it's incrementally gone up a little bit, but it wasn't like way back when the PC was a new thing and, Oh, 20 megahertz, uh, and six months later, 80 megahertz and six months later, 200 megahertz and 133 megahertz with a turbo button. Right. And what happened is, you know, we kind of got up into the three gigahertz clocking range on the internals on a microprocessor. And I think the fastest thing out there right now in common CMOS microprocessors is on the order of about a five gigahertz internal clock. That is a situation where the reason they're getting beyond that is parasitics. They've got a situation with distributed capacitance, distributed inductances of interconnects. I can get those transistors to go at 20 gigahertz and I've done it. But once you go out into a distributed logic system, you know, with auto routing and auto layout and all that good stuff, you're back down at, you know, the five gigahertz are even worse, worse than that, just to keep the thing surviving in the environment that's got distributed capacitance and inductance associated with the signals.

Chris Gammell: Yeah. Like making timing across the chip, that sort of thing. And all the edges you need to make, all of the pipeline you need to do that sort of thing. Sure. Exactly. Yeah. It is interesting, like looking at the spec, you know, I'm on a Mac M4, whatever, and it's just like the number of cores and GPUs. And like you said, it's just more and more of the same, but put towards different tasks internally with software, that sort of thing.

Dave Jones: It's basically more parallel processing. And we're also dealing with a situation where aren't they still, isn't Intel still trying to do x86 compliance devices?

Chris Gammell: I don't think anyone knows what Intel is doing these days, Jerry. Yeah, this is true. If anyone's listening from Intel, good luck to you, sir or madam.

Dave Jones: Yeah, because let's face it, the microprocessor has gone to risk architecture devices. And there's a real good reason for that because you eliminate a ton of useless architecture issues and transistors and things like that, that were created to support a rather complex instruction set. You didn't need that. They went to risk processors, done much better with that. As you know, things like ARM architectures are all risk processor architectures. They don't support a complex instruction set like an Intel x86 compliant device does. It's kind of, how do I put it this way? Legacy devices, unfortunately, trying to make things backwards compatible. You pay prices.

Chris Gammell: Yeah, totally. Yeah, you're hauling a lot of baggage with you on the way forward, huh? Oh, yeah. Yeah, that actually could probably be Intel's current tagline, hauling a lot of baggage.

Dave Jones: Well, I haven't worked with Intel in a long time. I'd be curious to know what's going on in there relative to foundry issues and what they're doing there. Yeah, totally. And what they're doing looking at next generation architectures because, yeah, they've kind of fallen by the wayside to a certain extent.

Chris Gammell: Yeah. Yep. Well, Jerry, where can people find you? Where can they find your book? Where can they reach out if they're interested?

Dave Jones: Okay. The book is published by O'Reilly Publishing. They're huge. A lot of people in the software world, tons of software books put out by O'Reilly. Title of the book is Applied Embedded Electronics, and the subtitle is Design Essentials for Robust Systems. It's available online on Amazon, Barnes & Noble. There's actually a specialty copy that's done for the India market that's coming out of a publisher in India as well. And they can find me at effectiveelectrons.com. On that site, you'll find a list of my publications, information about the book and what's in the book. Also, I personally can be found on LinkedIn. If anybody in the world of hardware engineering electronics asks me for a connection, I'll give them a connection on LinkedIn just because of the fact that the hardware guys I keep close to me and the marketing and salespeople I kind of ignore.

Chris Gammell: Got it. Yeah. All right. That's great. Well, Jerry, thank you for being here. I really appreciate it. Thanks for writing this book. I'm sure a lot of people are going to get a lot out of it. So thanks

Dave Jones: for being here. Thanks for having me, Chris. Enjoy doing it. Please reach out if you've got any questions on things because I'd be happy to help you out. Thanks so much. Okay. Have a good day now. Bye.