Embedded - 216: Bavarian Folk Metal

Episode Date: September 22, 2017

Carmen Parisi (@FakeEEQuips) joined us to talk about electronics and podcasts. Carmen works on switching regulators. If you want to know more, he sent along some very basic application notes: How to Apply DC-DC Step Down Regulators (Analog Devices) and Switching Regulator Fundamentals (TI). The digital communication method with these switchers is the I2C-like PMBus. If all those make sense, dive a little deeper with chapter 9 of the online and free Linear Circuit Design Handbook. Carmen says the whole book is excellent for analog information. Also, the free chapter of the Art of Electronics is on power. If all that still makes sense, you may be Carmen if you can also write an app note like this one: Multiphase Buck Design From Start to Finish (Part 1). Carmen is a host on The Engineering Commons (@TEC_Podcast). Some episodes you might enjoy are 93: Capacitors with James Lewis of KEMET (aka BaldEngineer) and 77: Remote Host Toast with Elecia White. Some suggested books from Carmen: The Art and Science of Analog Circuit Design by Jim Williams Analog Circuit Design: Art, Science and Personalities by Jim Williams An Engineer's Guide to Solving Problems by Bob Schmidt Elecia mentioned How to Diagnose and Fix Everything Electronic by Michael Jay Geier and promised a PID image from her book Making Embedded Systems.   

Welcome to Embedded. I am Alicia White, and I'm here alongside Christopher White. Our guest this week is Carmen Parisi, co-host of the Engineering Commons podcast and an electrical engineer. Hi, Carmen. It's nice to talk to you today. Yeah, nice talking with you guys, too. It's weird being on this side of the microphone for a change. So you are a host of the Engineering Commons podcast, which I was on a while back.

It's been about a year and a half now, hasn't it? I think longer. It could be two. All the episodes blend together after a while. They do, a little bit. But how do you introduce yourself if you're like a technical conference usually i like to dim the ceiling you know the dim the house lights come in through a cloud of fog strobe lights what's your what's your entrance yeah what's the music uh i think i'll have to go with the classic final countdown all right do you give them technical qualifications so you just

figure that's it and start i let my name speak for itself it's like the godfather people seek me out uh no i mean uh you know long time podcast host it's been about four years now long time listener of this one by the the way. Big fan. Thank you. You're welcome. So yeah, I've been an applications engineer in the semiconductor industry since I graduated school back in 2011. I actually went to school with former guest Alvaro. Oh, yeah. Yes, he mentioned that. I haven't seen him in a while, but I ran into him last time I was out in the Bay Area a few years ago.

He's very busy making cheese. Yeah. Hi, Alvar ago. He's very busy making cheese. Yeah. Hi, Alvaro. Casey's listening. Yeah, and since I started, I've worked at Intersil and Texas Instruments as an applications engineer, as I said, and I've been doing power electronics. Unlike a lot of your guests, I'm a terrible coder. But you're a good electrical engineer.

I'd like to think so. Okay. I haven't gotten fired yet. That works. So you listen so you know what lightning round is. Oh, yeah, I'm prepared. So we're just going to go into it. All right.

Preferred voltage. 60 volts or less. Favorite flux. Favorite flux? I like the kind that comes in the little syringes as opposed to the flux pen so i can switch out the tips and get different sizes favorite fictional robot i'm gonna have to get some nerd points with this one uh his name is vape he's an astromech droid just like r2d2 but he was the uh the droid for Face, who was part of Wraith Squadron. And he's my favorite because he was modified to shoot out beer if you said the secret command.

I think I just got out Star Wars'd. For extra nerd points, it was eblo beer. Favorite non-fictional robot? favorite non-fictional robot favorite non-fictional um I have to go with curiosity just because the landing was pretty cool

yeah it was yeah most hated electrical component most hated I don't know I'm gonna have to go with the processors

because you know they make my job difficult well this has been a lovely time talking to you yeah it's this is going to be one of with the processors because they make my job difficult. Well, this has been a lovely time talking to you. Yeah, this is where I'm going to be on the shortest podcast. I got to push that big red button you guys asked me to push before the hang-up one. Now I don't know whether I should go with favorite microcontroller now that I know that they're all awful. Well, you can have a favorite of the worst things.

Yeah. That's true. Tell you my least favorite, I guess I shouldn't say, but the MSP430 from TI, just because it brings back memories of school. That was my intro to microcontrollers class in assembly. And yeah, I had a lot of late nights and long days with that one. Wait a minute. Where do you work now? I work at TI, but I'm in the analog section. We can talk about those power guys. No, from what I hear, it's a good part. But, you know, I don't want to.

I want to code them. I like digging around with electrons. That sounds fair. And you. So what do you do with these electrons? I mean, when I do things with electrons, it's to make lights blink and robots move and all of that. But your electrons do different things. Yeah, I use the bigger, heavy electrons for power.

So, you know, that encompasses, I'm in a pretty niche aspect of power. So I do what's called V-core power. So if you have a big Intel processor or, you know, some kind of ASIC from a non-Intel vendor, they have a wide range of voltage rails. And I supply those voltage rails. And then specifically the core voltage, the main voltage the processor runs on, it has a laundry list of specs you have to hit, like tight DC tolerances and big slugs of current. Some of these things can pull hundreds of amps. That's a lot. Yes. So at Interstell, I worked more on the low end.

I was doing power products for like tablets and ultra books. And I dabbled in desktop and some battery charger stuff when those guys needed help. But that was my bread and butter was doing tablets and ultra book. Power always strikes me as something almost as black magic-y as RF. Yeah, I was going to ask about that. Am I mistaken or am I just too afraid of it? No, I can see that. You know, if RF is the black magic,

then I don't know what I am. I'm mugging you in an alleyway somewhere. Yeah, it's not the sexy aspect of electrical engineering everyone talks about, you know, like when you go through school, they tend to focus on signal processing Yeah, it's not the sexy aspect of electrical engineering. Everyone talks about, you know, like when you go through school, they tend to focus on signal processing and EM fields. And that does come into play, but there's not too many schools with a good power program. What would make up a good power program? Well, we may as well go to the model. If you look at the CEPAS lab, Center for Power Electronic Systems at Virginia Tech, world-class lab. And they've literally written the book on power electronics. They only started, I didn't go to Virginia Tech, but I did go there for a week this summer for some training put on by the professors who run that lab. And they do everything from low voltage DC to utility scale power. So as a software engineer, as an embedded software

engineer, what do I need to know? I guess how to, you know, if I can have one goal for this podcast is to get you guys comfortable, at least with the lingo of power. So if you're trying to spec something, you know what you're doing. And you won't ask the question you see on Reddit all the time is, you know, I have to run this controller. I only have nine volts. I need to get five. Can I just use a resistor? The answer is no. Are you sure? I mean, because V equals IR and voltage drop. I should just be able to figure that out if it was that simple

i'd be out of a job yeah no the the very short answer to to know i guess the slightly shorter answer is uh you know it's not regulated so as you pull different amounts of currents especially if you're doing 300 amps at one volt uh you know you to get some voltage drop. That I gets pretty big, and so you have no voltage left to work with if you do just that. And so I can calculate I need to drop four volts at a constant current, but as soon as my processor starts drawing a little more current, I'm no longer dropping four volts. Is that what you're saying? Yes. Yeah, you could dropping uh yeah four volts to 100 volts if you're just using a simple resistor it all depends on the load that v equals ir so power electronics is the way to pro i don't know you know make that resistor

fancy i guess in simple terms so what's inside a power switcher i mean now now i have this idea of a bunch of resistors that now you i have this idea of a bunch of resistors that now you're switching based on current yeah and they're in little tracks yeah and the trains yeah oh yeah yeah pretty much that's it we have little trains inside the chips uh no so i guess before we jump from resistors right to switchers we should talk about uh linear regulators first um and instead of using that, you know, there's two basic components to that is there's a pass transistor, typically a MOSFET these days, because there's lower dropout voltage, you can go from a, you know, five volts down to 4.5 if you had to. Whereas if you're using

an older BJT one, you typically need more space, headroom. So you might need six volts to get down to 4.5 instead of going down to five and you have an op amp for some feedback that you connect to the output a couple resistors and a reference voltage and you use that transistor basically as a variable resistor and you use that to control your your voltage drop across the device and you go from 12 volts to 5 volts but that gets very inefficient as you start pulling a lot of current or have big voltage drops. So that's where switchers come into play. Before switchers, there was one thing you said in there

that always baffled me when I looked at setting up the whole power subsystem. And that was the idea that I needed more voltage in than out. I mean, if I only need five volts out why do i have a minimum of six in that doesn't make any sense why is there so much loss um well that's that's the fun analog world i play in instead of you digital guys my transistors aren't just ideal switches um so there is some uh resistance with a transistor. And as we talked about before, that resistance, you know, will drop some voltage as you pass current through it. So as you pull, you know, one amp from your linear regulator, you'll have some voltage drop

associated with the transistor. And that specs how low you can go before the transistor uh you know drops out of its linear read engine and just shuts off okay that makes sense now now i assume i'm ready to design switchers wait i have more questions oh okay go ahead fire away because i've got all these acronyms in my head that fly around at work and now i want to ask them what they are and explain them to me is it going to be ldo it was exactly going to be ldo low dropout yes i know that slow dropout i don't know what it means yeah so that that means kind of what we've been talking about so uh if you use an older part uh what's that one that everyone specs like the 7805 or something like that i don't remember part numbers very well i always have to look

these things up um you know the minimum on that might be eight volts to get you know you need at least two volts of headroom that means uh so that means whatever you want your output voltage to be you have to have an input voltage of at least two volts higher than that in order to properly regulate so low dropout means you can get with some of these ldos you can get down to a couple hundred milliamps so So if you only need 3.3 volts out, you might be able to get away with 3.5. And that comes from using different control topologies, but mostly from switching to newer and newer transistors that don't have as much parasitic resistance.

Okay. That makes sense. It's a thing. It's a thing. I get that. You have certain properties. Well, I mean, the resistor part part parasitic resistance it's it's yeah how much is it using for itself is kind of how i think about it yeah okay switchers now are you ready well i was gonna ask about dc to dc converters but i'm not sure they're they're in the same ballpark

well that's oh no they're still dc to dc um but switchers can do both but we won't focus on going from dc to ac because i don't know anything about that very well okay that's not my area um no so both switchers and linear regulators are dc to dc converters but how they go about doing that is very very different um a linear regulator is very linear obviously it's in the name uh switchers are a student's worst nightmare i guess you could say um do you guys remember from your signal uh signal course back in school how you had a linear time invariant system yes so the output was proportional to the input and it was always the same no matter what time you looked at it and everything was perfect uh yes switchers are non-linear time variant systems so it's the exact opposite of

all that stuff it's it's wonderful why would i want to do that um well one of the problems associated with the ldo is the dropout voltage um so if you want to go tighter you know then the dropout voltage or you're trying to do a big step down like uh there was just actually i'm working on some of this these parts there's an episode on the amp hour recently about data centers and they were talking about going from their 48 volt bus directly down to a core voltage of about one volt in one stage conversion um that sounds scary yeah it's a big step the industry is trying to feel that out right now but um if you try to do that with a linear regulator, you'll probably just desolder it off the board.

Because if you're passing two amps and you have a dropout voltage of two volts, that's four watts dissipation right there in your switcher. So things get very hot with linear regulators regulators but switchers are much more efficient so you don't have that power loss associated with them and if it's a properly designed switcher you get well over 90 efficiency and things run much cooler that way okay but how do they work i mean i assume that something in there is being switched yes good guess yes so uh we'll start with the buck regulator i guess because that's what i know best it's most of my day-to-day that's when i have to put in like three volts and i want like five out or that's a boost converter yeah buck only goes down okay so a buck is when i want

when i have nine volts in and I want three volts out. Exactly. So you guys are familiar with PWMs, right? Yes. Yes, every microcontroller these days has at least one or two PWM outputs. So in simplest terms, a buck regulator is you're creating a PWM waveform, so your output voltage is dependent on the duty cycle of that PWM, and you're then passing that through a low pass filter to get a dc voltage out okay so is there that sounds great yes is there so is there

a cleanliness aspect of that i mean you're not getting a pure dc signal are you uh well it's an lc low pass filter uh inductor and a capacitor. So you're getting, um, you know, quite a bit of filtering. Okay. It's two stages. Like as if you had just two resistor and capacitor low pass filters stacked together.

Um, but there's no loss associated with the inductor ideally because it's a lossless component. I'm like the resistor. So you do get some pretty good DC filtering. And if you make the capacitor big enough, you'll have very low ripple and a pretty quiet output. And then there's some other tricks you can play to get lower EMI, but we don't need to dive into those dirty details just yet. Why would anybody use a linear

regulator? That sounds easy. Yes, it is easy. And actually, we've come a long way in switching technology. So you don't have to be a power expert to use them, although is easy. And actually, we've come a long way in switching technology, so you don't have to be a power expert to use them, although it helps. There are three terminal switching regulators you can get nowadays. I think the industry term form is power modules, but that may change as you go from company to company and they put their own marketing spin on it. So are these things that we'd see in small devices or switchers generally for

larger things like computers or things that are taking mains power or or would you see this in a wearable going you would see these everywhere because they're so high efficient um i i work on some that you would see on like computer motherboards or server motherboards okay but they're everywhere they're in your cell phone they are um they're connected to your solar panels on your house you would use a dc to ac converter though and that's a switching topology and you would go from the dc voltage of your batteries to the mains voltage to power your house uh they're they're everywhere they're they're ubiquitous that's sort of the opposite of my question,

which was why does anybody bother with linear? If these are so efficient and so ubiquitous, are they more expensive? Potentially, depending on how high performance they have to be. Also, just the efficiency is so much better. Like I said, you have that dropout voltage associated with a linear regulator. you know power is voltage times current if you're passing

10 amps at two volts dropout that's 20 watts you have to dissipate um yeah 10 times two don't do math on air that's a podcasting tip i should know it's definitely a podcasting tip so you got 20 watts you got to dissipate in that linear regulator. That thing would be the size of a tabletop. You know, you'd need a heat sink as tall as I am. Okay. So you would never be able to fit that into a cell phone. And, you know, your cell phone processor is probably pulling about 20 amps at peak current.

Based on the little bit I've dabbled into cell phone switchers. And so are they a lot more expensive? No, you can get the controller and the FATs. We'll get into the three basic parts. But you can get the stuff for pretty cheap. It's just it takes a lot more design knowledge. We try to make it as easy as possible, but they are more complicated.

But it's definitely not uncrackable. You can definitely come up to speed pretty quickly. Okay. What else do I need to know then in order to use it? definitely not uncrackable you can definitely come up to speed pretty quickly okay what else do i need to know then in order to use it uh how much you know about capacitors and inductors more about capacitors than inductors inductors are some sort of magnetic marvel that involves coils inductors just turns a wire basically um you've got to know a little bit more about capacitors and inductors because some LDOs will spec minimum ESR on a capacitor.

They wrote it in a data sheet, and it says use 10 microfarads at 1 milliohm. 10 milliohms, they'll say to spec it in a range. That also becomes more important in switchers because they play a role in stability and how well your converter responds to load transients on the output. So as your current jumps from one amp to 10 amps or 10 amps to 100 amps, the capacitors and the inductors become real important. And so with a linear one, I might be able to just slap whatever I have on in the garage, even though it says it wants some external parts of some range. I'll just put on whatever I have. It'll be close enough.

But with a switcher, I need to be a little more careful. Yeah, it's a lot more forgiving when you design with a linear regulator. Okay. You may have some stability issues if you have a transient. You may see some ringing. You're all familiar with the basic control systems where you over-damped and under-damped systems. We are, but go ahead and define them. Sure.

So in an over-damped system, you get pretty close to an RC time constant charging, where it's just nice and slow and there's no overshoot. So if you're going from one volt to five volts or one amp to five amps, basically, you just get a nice slow, slow ramp up to that exponential curve. If you're over or under damped, you would snap up real fast, but there'd be a lot of ringing. You'd overshoot, come back down and eventually settle out or not settle out at all if it's a really poorly designed system. And you may see some of that in a linear regulator, but again, it's a lot more forgiving. It's a lot simpler control loop. It's basically just an op-amp. And there's a little more to a switching regulator.

So you got to know a little bit of everything, control systems, sample data theory, passive components. That's if you're diving deep like I am. But a lot of these guys have simple switchers out there where you just slap them down, follow the data sheet, and you're on your way. If the oversampled and undersampled doesn't sound friendly to you listeners, I'm putting a picture, and it's with PID control, but the theory is the same with over and under. Yes, yeah, exactly the same concepts, just apply it in a different manner. So look at the picture and it will make more sense.

I hope. Okay, so there are some simple ones. It's a lot like PID controllers with switchers. It's a lot more efficient with switchers. I need to be careful about the components I put on the board with my switcher. What else do I need to know? I mean, the only thing I'm really worried about at this point is how fast does it come up and does it make my processor unhappy?

That depends on how fast or how complex your processor is. For something like the Arduino, they use just a simple linear regulator on that, I'm pretty sure. And that's not a very demanding power application. I'm pretty focused in a niche. So those big networking ASICs have quite a lot of requirements, like how fast you come up, how tight that DC tolerance is. Are you exactly 9 volts or 5 volts or 1 volt or anywhere in between? And then they have a lot of digital components, actually.

So every now and then I get to dabble in the embedded world because these things get pretty complex these days and there's all sorts of fun little digital features on there. Like what? Well, these days there's... I forget when exactly it came around, but it's on revision revision three of the spec there's an industry standard open spec called uh power management bus or pm bus oh yes yes it's uh

it's pretty similar to i squared c but there's some differences i guess in how they format the the commands and the the data and so what kinds of things can you control with that just voltage level well like if you are if you have a wearable and you have like four power rails okay and you you know you have a 1.8 you have 3.3 and you have five volts for something and and your processor runs off the 1.8 you can turn off the other ones to increase the efficiency oh Oh, so this is the power management IC. Yeah. Is that a separate part or is that built in with the switcher? Well, the switcher has to have some digital stuff on its end

to interpret the PMBus commands. But yeah, typically on a board that complex, you'd have something like that. There would be a system controller that would talk to the processor or sometimes the processor itself would send out those commands and turn on and off regulators. So the spec's kind of wide open. There's a few registers that are agreed upon by, you know, all the companies in the industry that come up with the spec and they, you know, they say register one is the manufacturer ID. So, you know, you're talking to a TI part or an intercell part or an analog devices part or whoever makes it. And then so on and so forth. And then, you know, they set a block

of registers, you know, are marked off as manufacturing registers. And you can do whatever you want with those. The sky's the limit as long as you document it in your data sheet. So, some regulators will put in protection features. So, you can say, hey, shut off at 20 amps or 25 amps. Don't go higher than that. Some will let you change compensation. They also let you read back voltages and currents as well. So if you need some telemetry, you have to know exactly what your processor is pulling for fan management or anything like that. PMBus will let you do that too.

And for current monitoring too, some of them. Yes, exactly. Yeah, that helps set the protection features as well. Well, from a software perspective, knowing how much current you're pulling at different times can help you figure out where the system is going wrong. Or optimize your power usage. Yeah, optimize your power usage.

Exactly. The power usage one is big and uh my new job here at ti do a lot of stuff for data centers and you know if you can save a watt times a million servers that adds up pretty quickly yeah so they're they use the pm bus telemetry to really fine-tune what their what their system is doing and always run at that optimum point where they get performance and a low electric bill we're jumping all over the switcher chart right here we're we're every which way is there anything else i should know before i go out and put it on a board uh yeah i've probably made it sound terrifying and you know you should always consult with an expert

but it's really not um you know to to hit the ground running if you read some app notes if you read the data sheet i'm the same as you guys i don't want to work harder than i have to so writing a good data sheet and some app notes um you know answers a lot of questions and it you shouldn't be afraid to just dive in so So one of the things that people might encounter as they get deeper into embedded and not necessarily design boards, but get contact with more complicated systems is that oftentimes there's multiple power domains, right?

There's maybe a 1.something and maybe a 1.8 and a 3.3 and a 5. And then you might have different peripherals that are all running off of those things and it's all well and good to all plug that together but uh well you know you assuming you have a proper power design yes but when things get more complicated and you start adding radios or storage or things that take transient power but high power for short bursts of time, what do people need to know about kind of a brownout situation? Or, oh, I turned my BLE on and now my CPU doesn't work anymore. Or every time I write to my storage, I, you know, something faults

and it just gets, it doesn't make any sense. Is that something people can learn about easily or is it just kind of a trial by error sort of thing? That's a good question. It's something I'm trying to learn every day. Being in the IC industry, I'm kind of insulated with that. So I work on reference designs for a specific part and I get to see them integrated

when I help debug on a customer setup. And that really kind of blows my mind when I see it sometimes too, whether it's a motherboard for a new laptop that's the size of a deck of cards or something that's the size of my desk here that's going inside of a server. It's definitely an art. I don't know if you could just pick that up easily with an app note. Certainly after it's happened a few times, you start to be very suspicious. Yes, you definitely dig through.

I can say you definitely should dig through and find your worst case scenario. If by chance everything turns on at once and your processor is really cranking away on something and you know you're you're blasting music out your speakers so your audio amplifier needs all its power that it's pulling you turn on your bluetooth radio definitely know what those corner cases are um so you can you can handle that make sure you know whatever is providing your main bus whether that's 12 volts that's going across or 24 volts, whatever it may be, make sure that can provide all the power you need. Yeah, because just blindly coming out, I look at my battery,

my battery's fully charged. What's the problem? Yeah. I've got checks left. Yeah, exactly. Make sure, especially I mentioned you can set different protection features like overcurrent and everything. Make sure those aren't set too low. And switchers also, you have to wonder what happens at your light loads as well, because they behave very different at light loads versus heavy loads.

So they have what's called continuous and discontinuous conduction mode. And that depends on the current in the inductor. So now that's scary. Yeah, now we're back to scary land yeah yeah that'll depend on each design um and you have to worry about you know stability issues in that range uh the light load and then also for battery life too you know you can you know making sure you count nano amps on your processor in a sleep state doesn't matter if your power stage is still drawing, you know, hundreds of milliamps, you'll still kill your battery life. And I've spent many weeks debugging sleep modes

and switchers as well. Well, and then that goes back to the PM bus and being able to set things digitally. It's always nice when you, nice, quote, nice uh when you have worked really hard to put everything to sleep that needs to go to sleep before the processor goes down and then you wake up and you don't know what was asleep and what wasn't so now you try to send things out ble the reason it doesn't work is because it doesn't have power and it some of these these chips are really complicated and it gets that way on the uh you know the the power side as well i was working on a part at intersil and i had multiple outputs associated with one uh switching regulator there were three separate ones in one package and you could put the part into a sleep state because it was meant for, you know, a laptop and you always want the most battery life out of those that you can get.

And there was a race condition when it woke up that we missed in simulation and the initial validation. And that came up down the road when it got designed in. And, you know, you test as well as you can, but in production, things come up that you don't plan for was that more of a an electrical bug or more of a software bug uh if i remember right it was an electrical bug there was something you know crossing crossing digital domains you know from the outside world into whatever our part was trying to do you had to i don't know all the digital stuff that went on in there, but they had to switch clock domains. You guys could probably lecture me on that.

And there was, yes, there's something where a clock cycle was getting dropped and that was causing the race condition. And it would work well, like 92 times out of a hundred, but then those other eight times you would see a failure. I was a nightmare. But now there are essentially processors inside these power regulator switchers yes how they used to be almost entirely analog except for that pwm clock didn't they uh yeah and actually that pwm clock is is pretty analog as well

um so they've always been sort of mixed domain, I guess, because you are banging around that PWM signal. So that's how you get your time variant system, because you have to know what state PWM is in in order to analyze the circuit. If you connect the MOSFET between your switch node and the input voltage, you have one circuit. And if the other MOSFET between your switch node and the input voltage, you have one circuit. And if the other MOSFET's on between the switch node and ground, you have another circuit to analyze. So there's your time and variance system. And then because you're banging the MOSFETs hard on and hard off between VN and ground, that's your nonlinear system. So that's how you get those. So it's always been a little bit digital because you have that signal and then

we we don't have to dive into all these weeds as well but there's different control methodologies and it's like obscure metal genres like bavarian folk metal and screamo finish control and you know so the two big ones are voltage mode and current mode and current mode is a sample data system so it's always been mixed signal as well and um you've had to know a little bit everything to do a current mode regulator design but they seem to be getting more drop-in no not the more i mean yes yes the more drop-in, more intelligent, more processor, more digital, really. Yeah, so you always can still go out and find the simple drop-in, easy to design with parts that are analog, that you only need 500 milliamps of current for.

Pick this inductor from the chart in the data sheet, pick these caps, and you're good to go. But yeah, with PMBus, you are getting a lot more digital functionality into these primarily, you know, back in the day, analog parts. Does that change your outlook on being a hardware engineer? Uh, yeah, I definitely have to know a lot about buses. You know, there's the industry PMBus that we talked about that's like I squared C. Intel has its own SVID bus, it's called, which is like I squared C, but on steroids, it's a much faster frequency and a much lower voltage. And, you know, as the processors iterate, what they expect out of the controllers changes quite a bit as well. And it's a, you know, a toss up-up you know every ic company goes about it differently how they provide these parts and what digital functionality they enable let me ask you sort

of a weird question if you're designing something for mass production there's always trade-offs like knobs you can turn to say well i'm going to accept a little bit lower quality of this to get my my cost down or maybe i need higher quality here but that's going to impact so little bit lower quality of this to get my cost down, or maybe I need higher quality here, but that's going to impact. So when you're picking power parts, what are the kinds of trade-offs that you might be making when designing? I mean, you do have to know what you need to provide to the board and your inputs and that kind of thing, but eventually you're going to come down to a choice of a few parts i hope

and so how do you yes how do you decide between them if maybe the big difference is price uh so you you know much like you guys with the microcontroller it's how much do you want to integrate onto one chip so just talking from a silicon standpoint you know if we'll ignore the inductors and capacitors for a second there are three parts to a switching regulator there's your controller your mosfet driver and the mosfets themselves and depending on what your budget is you can have some combination of integration there between them so you can get a separate controller IC and gate driver IC and a MOS, you know, you can do MOSFETs on their own. And that's usually your cheapest solution because you're getting three different ICs and they can be as big or as small as they need to be.

And you get a lot of design flexibility that way, but you have to be a better designer in order to get the maximum performance out of that. Then you can integrate some combination of, you know, the controller and driver together or the driver and MOSFETs together. And you get a somewhat more integrated design that doesn't take as much knowledge on your part, the designer, but that's a little more expensive. And then you can integrate everything onto one IC, you know, the controller, the driver, and the FETss and that's the most expensive but takes the least amount of design time so it you got to work your budget and your design schedule out there works the same with code the more the more expert you are the more you you know how to tailor things and maybe fit things in smaller parts that would be cheaper so yeah that makes sense

yeah exactly and you know on a cell phone where, you know, every millimeter of space is critical, you'd want to integrate as much as you can. But, you know, on something like a desktop motherboard, there's a lot more space because you're trying to fit into, you know, the ITX motherboard or the ATX power supply or whatever it is. And, you know, it's a fixed form factor, basically. So you can work within those constraints and save a little cost. I think that's as much of switcher information as I can take right now. Well, I hope I didn't confuse everybody too much. It's pretty complicated. And I can give you some good app notes to send to people if they want to read more about it.

Yeah, we'll have some for the show notes. So I want to switch gears a bit and talk about podcasting. Sure. Actually, before that. Okay. Just strip the gear. I know. No clutch at all.

How long have you been out of school? That's a more polite question than how old are you, right? Yes. So I don't mind. I'm 29 years old. I don't have to lie about my age. And I've been out of school for a little over six years now. I graduated in 2011.

And how long have you been podcasting? Four, five years now. Four and a half, I want to say. So you were pretty early in your career path when you started podcasting. Yeah, I was. The amp hour started up. Don't tell Chris and Dave, they'll get a big head. The amp hour started up my senior year in school, and I've been listening since the get-go. I forget how I stumbled across that podcast, but I was like, oh, cool, an electronics podcast. So I got hooked into podcasts through them, and then Chris started the Engineering Commons. And when he left that to do other things, I threw my hat in the ring when Jeff wanted co-hosts.

And you said, well, if Chris and Dave can do it, I totally can. No, exactly. How hard can it be? You speak to yourself in a room with a microphone instead of just doing it when no one can hear. How much of an impact has it had on your career? It's hard to say because I'm still getting shaped, but I can't say it's had none at all. Just hearing all the different stories from the engineers we've interviewed, I've kind of learned where I wanted to focus my career. I hear stories, as funny as they are, about engineers on the road

who live out of a suitcase for 20 weeks out of the year. I know that's not the life for me. And then hearing from guys who've gone into academia, and Jeff himself, we hear from him all the time about getting a PhD and teaching. And there's a certain allure to that, but I don't think I'm ready for that right now. I like being in industry and, you know, doing a little bit of R and D work. And it's always nice when the experts come and talk to you and you don't, you know, it's kind of like they have the, the meetups after work, depending on where you, where you are. You can often catch some expert talking about their, their area.

And, and that's all nice, but as a podcast host, you just invite them to come talk to you. Yeah, and it's amazing what kind of people want to come on podcasts. I mean, we got the CEO of iFixit to come on back earlier this year, and all I did was fire off an email into their suggestion box saying, hey, if you're interested. So it's been a crazy four and a half years but i've talked to a lot of cool people yourself included i i don't i fix it thing i have to admit i saw that and i was kind of annoyed do you know how many times i asked them to be on our show we got a little bit of a rivalry going here well Well, and we have our sworn enemies, so we have to stick with those. Exactly.

We're way better than that other electronics podcast. And you usually have a number of hosts. Yes. Unlike a lot of podcasts, we got four hosts instead of just one or two. Yeah. And then you usually have one or two guests is that right i think we only ever had two guests like once or twice it's rare to have more you don't have a lot of room if you're starting with four hosts yeah exactly exactly um and you can tell as you go through the episodes

you know depending on who's who's our guest some of us will take a bit of a back burner to the conversation you know if if someone's done a lot more structural stuff or civil, Adam will take over and, you know, he'll lead the charge. And if they're talking about academia, Jeff definitely has a lot of opinions coming from a teacher's perspective. But we're all students, so we'll chime in as we need to. And, you know, as I brought in James from Kemet, I took the lead on that one. So we rotate out. We try not to be too congested. Keep the signal, the noise high. Having been a guest, it's a little intimidating to try to keep track of. I mean, I was taking notes on who everybody was and trying to answer questions

like, well, a civil engineer, gosh, I have no idea how to answer civil engineering questions. Maybe I should study. Oh, yeah. And this is kind of how I feel answering questions coming from you guys. I'm like, oh, man, I've never had to explain it to non-power people before. I've been so far in the industry. I got to back up a second and remember how it was. But why? I mean, why do you continue doing?

Actually, no, before that. I know. I'm really just doing this Actually, no, before that. I know. I'm really just doing this again, aren't I? Sorry. That's okay. So, for hosts and a guest, who edits your podcast? Well, that falls onto Jeff. He's the champion of the podcast. He does all the editing.

I mean, we try to help him out as the years have gone on, and we realized, okay, we're not just screwing around here. We really like it. We've bought better equipment to help him out as the years have gone on and we realized, okay, we're not just screwing around here. We really like it. We've bought better equipment to give him a better starting point. But it's Jeff doing the heavy lifting. And then why do you like it? I just like swapping stories. It's getting together with three guys who I consider myself friends with, even though I've only met Brian.

And actually, just that week I was in Virginia Tech, he happened to be taking the same class. You know, we've been talking for four and a half years. We've, you know, I've seen people have kids. I've seen people get new jobs, people, you know, change jobs and it's cool. And then our guests come on and you hear from a guy who's worked on diesel engines and, you know, for mining equipment. And you're like, you just hear all these stories. And as cool as everybody is, it's fun to find those little threads where everybody intersects, you know, where, oh my God, I had to deal with the bureaucracy at this company. And then, you know, all these meetings. And then I finally got to sit down and get some work done.

Everybody's different, but they're still the same. Yeah. I like the threads. I like hearing similar things from different people. It's just amazing how some things are so totally different than I expect and some things are just the same. Yeah, the war stories that you swap. You guys may sit and debug lines of code only to find out you're missing a semicolon somewhere and i can you know spend a bunch of time wondering why

nothing's working right only to discover uh you know everything looks like hell because the ground lead to my oscilloscope fell off and that's why i see a bunch of noise on the screen on the oscilloscope screen i did that a few weeks ago for yeah almost an hour so yeah yeah yeah it all yeah. It all just depends on where you want to bang your head against the wall. I can't do it with code, but I can do it with, you know, oh, man, the wrong resistor value got soldered down. And it's supposed to be 1K, not 10K. Where is that? Drives other people mad. At least with code, the letters haven't been getting progressively smaller.

That's true. Given my vision, it seems like they might be back at back at intercell i i you know doing this stuff for tablets and ultrabooks now it's great everybody's got a little space in the server side of things so things aren't quite so small but you're trying to debug your part on some of these little tiny motherboards that's another fun thing i've done is surgery on, you know, tablet motherboards. And, you know, they're using O2O1 components. The amount of caffeine, you know, it's not a joke. It would make a difference. I'd sit down under the microscope with my finest tweezers and my tiniest soldering iron tip. And you could just, you couldn't get your hand straight. You have to

go take a walk in a glass of water and come back later because those boards are very fragile. You'd lift a trace or what have you. It's crazy. All right, so let's shift gears again. Okay. This time maybe I'll stick with one plan. How many gears does this transmission have? So many gears.

So many. Probably more than my Honda Fit. I've been really interested in semi-trucks lately, so I'm all about lots of gears. Uh-huh. We have an episode on that if you want to download it. Ooh. Will it teach me how to drive a semi-truck?

No. Because I kind of want to learn. But if you want to learn about gears, we got that going for us. I'm not sure audio is the way to learn to drive or operate heavy machinery. Just like podcasts, and how hard could that be? Okay, we have a listener question sure as someone who has recently gotten into embedded systems as a hobby i wonder if you know of any equivalent advice on being a grown-up hardware engineer the listener goes on to say he knows just enough to get his boards to the point of works for me, but there's always this feeling of missing a resistor or capacitor or something that really should be there in order to have it be more reliable.

Are there books or tips or websites on how to learn how to be more professional professional um so coming from my perspective uh you know i've probably spewed out a bunch of incoherent stuff on switchers hoping i make sense um i i started six years ago out of school i had no clue what i was doing with power i took a very basic course that you know covered what a switcher is, but not much more than that. And I had to learn, you know, I was coming from a complete newbie perspective as well. I had to learn everything. I was scared of inductors as well. What worked for me was reading anything I could get my hands on. You know, no matter how crummy the app note was, I could always find something in it to learn and not, afraid to experiment, whether it was

getting my hands on an eval board or, you know, well, I had to design my own, it was my job. Uh, but just going through that feedback cycle and learning to take criticism, like, Hey, you want to route the traces this way. Uh, here's why asking good questions of people. Um, you can also listen to the engineering comments comments because we talk about this quite a bit. A little plug for myself there. But there's some good books that are a little bit older now. There's one, Analog Circuit Design, Art, Science, and Personalities. And then the Art and Science of Analog Circuit Design. They're edited by Jim Williams, but they're just stories by all these guys who were related to the semiconductor industry some way, whether they're designers or apps engineers.

And I chose them because they were near and dear to my heart. That's what I went to school for was IC design. I wound up being apps because it's a lot more fun than driving a simulator all day. But stuff like that, it just gives you the human side of how they approached a problem. And you learn like, oh, you know, everybody struggles from this stuff. Everyone's wondering if they're missing that resistor and capacitor. So that helps you get over that imposter syndrome because we all feel it. Are you familiar with Michael J. Geyer's

How to Diagnose and Fix Everything Electronic? It sounds familiar, but I can't say I've read Gay years, how to diagnose and fix everything electronic. It sounds familiar, but I can't say I've read it myself. No, is it in similar vein? It's a little focused on diagnostics. You're given a broken thing, what do you do next? But there's a fair amount of personality in it, of all the things he's had to fix over his lifetime there's stories like that in these books too and yeah stuff like that is great uh another one a guest on the engineering commons has come on a few times he's got a book uh an engineer's guide to

solving problems and again it walks you through you know how to troubleshoot how to approach where to start um from personal experience i can say say, you know, keep it simple. Don't just immediately jump to, oh my God, everything's broken, or there's some crazy second order effect I never thought about causing everything to go wrong. A lot of times, I forgot to turn the stupid input supply on. So, that's why I'm not starting up. Or, you know, I thought I plugged the cable in, but I didn i didn't or you know my my scope probe is hooked up to the wrong node and just the number of silly mistakes i've made that way and still make to this day uh happens all the time and you learn okay let's start simple is everything actually plugged in have you turned it off and on again really is a good troubleshooting tip

because usually discover you never turned it on in the first place yes exactly yes yeah what's that uh i've seen it a couple times online you know all the it support guys say oh sometimes dust gets on the cable can you can you blow on it for me and plug it back in and it's just to have them check and see if it's actually plugged in you know so it's to start simple and just you know work your way down and don't be afraid afraid to blow stuff up. I've done it on co-ops. I've done it on internships. I've done it on the job. I've accidentally broken customer boards. And it happens.

It's part of being an engineer is making mistakes like that. Well, I think what you're saying about stories is a good point because we do pick up a lot of institutional knowledge that we don't get in school. And we don't get working on our own when we have coworkers, and particularly good coworkers. And you have to be really lucky to get connected with good mentors or people who are willing to show you their tricks

and the mistakes they made and that kind of thing. So getting a jumpstart by reading, finding accounts and things like that. Yeah. And going back and rereading as well. I've paged through those books quite a bit and they talk about circuits I never understood. But every couple of years, I'll pull them back out and page through and I'm always picking up on new things because I'm at a different point in my career. I get things now. I'll find something I missed before. And yeah, learning to ask good questions and knowing when to ask a question, you know, don't

quit two minutes in, quit two hours in, quit two days in, depending on your, you know, what kind of timeframe you're working on and then go to an engineer, you know, above you and say, hey, here's what I've tried. This is where I'm struggling. And, you know, the advice they give you will click so much better if you've struggled through it yourself a little bit. Yes. Oh, I have found that so many times that if I spend some time on my own trying to fix it, that when I finally find somebody who can help me, it all works so much better. Yes. Yes. I've been, you know, I'll plug that class of Virginia Tech I took. It's definitely not a beginner course, but, you know, I took it six years into my career. And throughout those six years, I've been reading app notes and doing board designs myself and picking up nuggets here and there.

And then all of a sudden, I was able to ask the right questions during this class. And so many things just clicked for me. And it's because you put in that legwork, that groundwork to struggle yourself. But it's hard to figure out. I mean, because there are times when you shouldn't be struggling for three days when you could have asked a question

that would have solved it in five minutes. Yes, exactly. That's an experience thing, I think. Yeah, yeah. I've definitely gone to guys two minutes in and asked for a sanity check just to, you know, make sure I wasn't doing anything dumb. And then, yeah, then I go beat my head against the problem. And yeah, it's experience. Like you said, if, if this listener still in school, you know, finding good co-ops and internships is key. I was very lucky to have, you know, three excellent ones for many different reasons, all for different reasons, but they taught me quite a bit. And you mentioned blowing things up. Yes.

I'm in favor of this. Not just the recreational type, but the trying it, and then it fails and it's expensive and you look around wondering how you can dustbin this board that used to exist so that nobody knows without pissing off the tracking system uh not not that i would know in detail uh-huh uh-huh no never uh have you ever gotten in trouble for that not really no i've gotten some good-n, never. Have you ever gotten in trouble for that? Not really, no. I've gotten some good-natured ribbing, but I'm never in trouble. You know, sometimes a customer board does break, and that's what happens.

We were trying to run a cold test on a customer unit, and a frost formed on the board, and it shorted something. Oops. You know, we didn't have a way to control for humidity as well as we needed to. And thankfully it was towards the end of our testing when we realized we should be testing for cold. So we could extrapolate around what we data we could get, but it happens. And, you know, at least doing low voltage DC stuff for power, I can fail a lot and not die as opposed to working on, you know, main stuff. I just wanted to point that out that you probably hadn't gotten into any trouble. I didn't get into any trouble. It's, it's the price of education. I mean,

that's a big lesson you just learned and they should be a little happy that you don't have to learn it again. You're not going to fall under that spell again. Yeah, exactly. If you're an intern, as a student, you're expected to fail. And when I was on one of my co-ops, I was soldering an image sensor under the microscope. And I don't know what the heck kind of package this was, but there were some almost paper mache-like things I had to solder down, these little tiny flexible leads.

And yeah, doing the first two or three of them, I would rip a lead and we'd have to start over. And things probably cost as much as I made in a week as an intern, but it was part of the job. They wanted me to learn. They were okay with it. Yeah, they probably gave you broken months to start with. Possibly. But I mean, that's part of having an intern and it's part of teaching a junior engineer is knowing that they're going to fail and that failure is okay. That's how you learn.

Failure is always an option. Exactly. And being in the semiconductor industry and working on the validation side of things, to fail is my job because it's better if the board blows up or the chip explodes while I'm working on it during validation than when it's on your board powering your $10,000 high-end FPGA.

Oh, I feel that way a lot. I much rather have the bug on my desk than have a customer ever see it. Yeah, yeah. I mean, I got to take proper precautions, obviously, when I'm using a customer board. But when I got a board back on my own that I've designed and I'm testing a new part, I don't care about ESD. I don't care about making sure what the current limit of my supply is exactly. I want to know how robust the part is.

Things in your industry and power have already been changing so fast because it is going to more mixed signal and it's going more digital and processors get cheaper. What technologies in general, in power or in general, do you think will be interesting in 10 years? Just the kind of configurability parts will have for one. Various companies I've worked with and talked about, there's a lot more processing going on for the power. So they're looking at things like inside the IC, looking how the MOSFET ages and is the resistance increasing with time and changing how we operate based on that to keep providing good power and stuff like that.

The algorithms and the control loops you have to design get crazy. Stuff like that's really interesting. And then from my end, backing out of of the ic design just testing these kind of things becomes ridiculous as well um the next gen processors i'm looking at could pull 400 amps at 0.8 volts trying to find test equipment that can even do that you're almost outpacing the test equipment industry because the voltage drops along cables even with sense lines coming from your electronic loads becomes too much. So just seeing how we're going to get around those problems and then testing to transient specs. Not only do I have to worry about jumping from 50 amps to 150 amps or 400 amps or whatever it may be, I have to worry

about the slew rate of that load. And some of these guys can pull current at 500, a thousand amps per microsecond and trying to figure out how to test that is ridiculous. Yeah. Our processors are getting quite annoying. Yeah. Yeah. That's why they're my least favorite part. It makes my job so much harder. I'd much rather have a switcher power, uh, you know, an MSP four or an arduino or something that's going to only jump around like a couple hundred milliamps the complexity part of that i think everything is getting so much more complex just because it's yeah for some reason it's easy to be more complex uh which is kind of mind-boggling but yeah the testing exactly that's a that's a good thing to bring up because as we increase complexity all over the industry our testing methodologies

are not getting better having been playing with machine learning more i'm just like how are you gonna test that exactly yeah it's a It's a problem in and unto itself. At Intersil, and I'm sure at TI as well, but I'm a little bit more on the marketing and customer relations side here, even though I do a lot of reference designs. But you'd find all these digital engineers getting hired to do verification because, like you said before, you're basically dropping a microcontroller into some of these higher-end parts. You've got to do as much testing as you can to find all those little bugs because it's expensive to spin silicon and get new parts

out of there and then you're gonna have four more updates on your power microcontroller exactly i should order a copy of your book now and start reading it but that's probably coming sooner than i think it It probably is. I mean, if there was a software bug, they would need to update it. But goodness gracious, what if you put a virus? Wow. Security in power regulators is going to be a thing someday. Yeah, probably sooner than we realize. And yeah, I'll have to learn how to write good code because i don't like doing it

and i can write code but please don't ship it i i like python you guys didn't ask me that in the light uh lightning round but python's my preferred language python's a good language i've been doing more python i've been very happy with it yeah i've got something uh i've been updating a lot of older reports and getting them into a new format to put online. And so you just get this old Word document and you can. Here's a fun tech tip. If you save a Word document as a dot zip and go inside there, there's a media folder with all the pictures from that presentation. And you can just copy that out. So I have a Python script to go through and rename to the format I need to. I got a Python script to talk to my scope and take scope shots. I got a couple other little

ones here and there I've hacked together and there's no pressure for shipping it because only I have to use it. Having done some C++ and Python together recently, I'm surprised at how the new C++, not the C version that I learned in the 90s, but the new has containers and templates and polymorphism and all that stuff. C++. You're taking me back to some terrible flashbacks in school. I thought I wanted to do robotics for a while, and C++ gave me a lot of trouble.

It's closer to Python than it used to be. Is it? It is. I can't say I've used it too much because I tried to avoid it. But it's good to know it's useful if I ever need or easy to learn if I ever have to get back to it. It's not easy to learn. Don't listen to her. It's not easy to learn.

Just like switchers, but you can do it. One of the things that's hard about it is these libraries, except if you've been playing with Python for a while, the libraries start to look familiar. You're used to lists and sets and all of those things that are hard to implement and see. Yeah. I mean, I'm not an expert in Python by any means, but it's nice to be able to look at some example code and figure out how I want to modify it at least. All right. Well, we're pretty far off our hardware engineering topic, and I think we're just about out of right. Well, we're pretty far off our hardware engineering topic and I think we're just about out of time. Chris, do you have any questions

before we let him go? I had a question way back, but it's back on topic, so we shouldn't go all the way back on topic. It is so far from where we are now. So a million years ago when I worked at a startup, we did core routers. And one of the features of the core router was that had 90 amp dc power input something like that and

there were these giant copper bars i was always impressed by how thick the uh the power inputs were because it drew a lot of i think it was several kilowatts um yep yep that sounds about right but what i never understood was why data centers and and routers you know racks and things in in data centers uh why dc power was the way to go is it just because you don't want to be doing the ac conversion all over the place? Or why are some of these things DC powered? That's a good question. That's on almost a utility scale. I can't say I know too much about that sort of thing.

But yeah, there's got to be a reason. It must be easier to bus. In data centers, there's a 48-volt bus that goes around the whole data center. That's a couple kilowatts and they got to have their reasons for it if i had to guess a slight guesstimate i would say because you know going from ac to dc requires bigger it's it's at 60 hertz or 50 hertz depending on where in the world you are and to do magnetic components for that are pretty big and so one of the things busting around dc and you know doing it that way is you can use a switching topology to go from AC to DC as well.

That's what's in your iPhone charger and stuff. And you can use nice small transformers and high frequency. So it's got to just be from a cost standpoint if I had to pick a starting point to answer that question. I'll accept that. Sounds good. All right, then, Carmen, one more question for you. And that is, do you have any thoughts you'd like to leave us with? No, I just really want to hit home again.

Like we talked about, don't be afraid to learn. Don't be afraid to dive in feet first. It's worked out pretty well for me so far during my career. I've done the R&D side of things. I've done the app side of things. I've done the R&D side of things. I've done the app side of things. I've done some customer relations. I just got back from my first trip overseas to meet with customers.

I'd never played the salesman before, but if you fake it till you make it and you read a lot of literature, app notes, books, whatever, you'll do better than you think. That's good advice. Our guest has been Carmen Parisi, Senior Applications Engineer at Texas Instruments and co-host of the Engineering Commons podcast. Thank you for being with us.

Thanks for having me on, guys. Hopefully I didn't confuse power too much for your listeners. I was at least somewhat coherent. I understood. And that's all that matters. Exactly. You're a podcast host. It's been a pleasure.

Thanks for being such good hosts on my first time as a guest. Thank you to Christopher for producing and co-hosting. Thank you to listener Ken for his excellent question. And of course, thank you for listening. I have a quote to leave you with this by J.K. Rowling. We do not need magic to transform our world. We all carry the power we need inside ourselves already. Embedded is an independently produced radio show that focuses on the many aspects of engineering. It is a production of Logical Elegance, an embedded software consulting company in California.

If there are advertisements in the show, we did not put them there and do not receive money from them. At this time, our sponsors are Logical Elegance and listeners like you.