#474 – An Interview with Nash Reilly

Chris Gammell: This is The Amp Hour Podcast. Released January 12th, 2020. Episode 474. An interview with Nash Riley. Welcome to the Amp Hour. I'm Chris Gammell of Contextual Electronics.

Nash Reilly: And I'm Nash Riley. I'm an electrical engineer at Sonos. Hey Nash, how you doing? I'm still kind of shocked I managed to find my way on here, but I'm doing great.

Chris Gammell: Well, you know, that's how serendipity goes with meeting at events and having a cool background doesn't hurt either. So we met at the Open Hardware Summit back in 2018, right?

Nash Reilly: Yeah, I was actually just starting to scheme how I could figure out how to make it to Oshawa 2020, which I think is New York City this year?

Chris Gammell: New York City in March, yeah.

Nash Reilly: Awesome. Yeah, no, I think I might have to take a day or two off to go head down to that. It was a lot of fun the last time I went. I mean, you could take a train down there even if you wanted to. Oh, I love the train. I love the train in the Northeast Corridor. I grew up in the middle of nowhere in Montana where there were no trains. Well, sometimes some parts have trained, but yeah.

Chris Gammell: Not the kind of train you want to ride. Yeah, right, right. Yeah, okay. Well, so you work at Sonos, but how did you find yourself there at Sonos?

Nash Reilly: I went to school at Montana State University. Go Bobcats. I grew up in Montana. Yeah, for all of the Bobcats fans listening to this podcast, they're few and far between.

Chris Gammell: I hope there are. I mean, Montana probably has a couple of electrical engineers, right? There are a few. We might get one or two of them, yeah.

Nash Reilly: I sure hope so because, I mean, I would love to go and be an electrical engineer in Montana again. Montana is a great place, but that's a subject for another time. I could talk about that all day. But yeah, I basically found out when I got out of college that there wasn't a whole lot of opportunity to do the electrical engineering subset that I specialized in in school. I mostly focused on embedded systems and high-speed digital design. And I got out of school and I went to work for Micron, which I usually have to explain. It's in the Pacific Northwest at least, right? It's up in Oregon, right? Yeah, it's in Boise, Idaho. Oh, Boise. That's right. Yeah, yeah, yeah. Yeah, and that was interesting. I wasn't doing much electronics work there, though. I was working doing a lot of data analysis for, like, returned DRAM modules. So, like, the DRAM modules that go into your computer or actually a server are in this form factor called SODIM. And generally, when a big web customer, like a Google or an Amazon or something like that, has a problem with one of these SODIM modules, they return it. And they don't just return that one module, they return the whole batch of them because they're like, this is too much of a threat to our reliability. We want a refund. Whoa. Yeah. It's good to be king. It's good to be king. And that actually, like, had huge effects on Micron's bottom line because it was millions of dollars worth of, like, their most cutting-edge chip technology. Yeah. And these are the sort of things that people have written about for years, like cosmic ray strikes can cause bit flips in DRAM. And we're not 100% sure why that happens. People have built really fruitful PhD careers about trying to figure out why that happens. But anyway, long story short, I was basically a data analyst trying to find similarities between these chips that were getting returned by these big vendors and piece together what was happening. And I got tired of that after about a year just because I didn't feel like I was designing electronics. And I was just, like, I wasn't happy. I was writing Perl scripts all day.

Chris Gammell: Well, you're adjacent to electronics, and that's something, but it's not quite what you wanted.

Nash Reilly: That's a really good way to put it. It was, I didn't have a lot of lab time. I didn't really have a chance to use, like, scopes. There's, there's not a lot of soldering that you do because, like, when you're actually looking at a DRAM chip, you're looking at a decapsulated die. And I know there are people who love that, and, like, they're, like, great test fixtures and all that sort of stuff. But, like, it wasn't quite hands-on enough for me. I mean, it was, it was pretty cool when I needed to take a measurement on a DRAM chip of, like, going to the guy who's got the laser deposition machine and the vapor deposition to me and to put, to put, like, a tiny little test point that's, like, smaller than a period on something that I have to probe with a probe that's finer than a human hair through a microscope. But, like, I did that, like, twice in a year. I wasn't doing much actual electrical design. I wasn't doing much simulation. I wasn't doing any layout. I was mostly just, like, doing data analysis stuff. And that, that just, like, didn't make me happy.

Chris Gammell: Well, and Mike Crunch, so, like, I went to a chip company after school, similar kind of thing. And it's, like, they just want talent that's moldable, right? They don't really care. Because it's, like, most of the stuff they're doing, it's, like, where are you going to find someone who's done laser deposition and analyzing that kind of thing, right? I mean, it's just, like, you have to train people for it.

Nash Reilly: Yeah, absolutely. And, like, I don't want to throw people off the semiconductor industry because mileage may vary intensely. Like, Micron works as a very different business than, like, an NXP would work or an analog devices would work or an on-semiconductor would work because they're building memory devices. That's a whole different subset that's optimized for a whole different thing than most chip companies do. Like, they're just one player in a vast market. But, yeah, at the end of the day, not really for me. So, I called up my old boss who was my boss at my internship, which was at Sonos, which at the time was a very small company. I interned there when it was about 100 people. That was between my junior and senior year of college. That was 2011, summer of 2011. And if you're familiar with Sonos' product line, that was right when we were getting ready to launch the Play 3. And the Play 3 was a big deal for us because it was the first product that we launched that was under $300, which was a big deal for us because it incorporated a lot of speakers. And it was a little bit more affordable. The only product that we'd had that was an all-in-one speaker and amplifier package that was wireless was $500 before that. So, it was a pretty big step down in price point and a lot more affordable for a lot more people.

Chris Gammell: Right. And Sonos is known for being kind of high-end stuff anyways, right?

Nash Reilly: I mean, it's not high-end sound and things like that. Definitely. Definitely. It's definitely a premium product. And we definitely position ourselves that way. And we pride ourselves on delivering a really premium experience too. We promise rock-solid Wi-Fi. We promise integrating with any streaming music service that you possibly could want. We promise earth-shaking sound in really small packages. Like, we went on from the Play 3 to deliver the Play 1, which was sub $200, which we've since – the price point is now at about $150. We've added a voice assistant to it. A lot has changed. I've gotten to see the company go through a really, really big time in the time I've been there.

Chris Gammell: Yeah, and it sounds like it moves from, you know, like a high-end space moving into a more broader market. You move into more consumer level grade. Your part selection probably changes. Your volume goes up. And you're changing decision types, I imagine, as well. Oh, yes.

Nash Reilly: Oh, yes, certainly. Yeah. Like, that's something I've been on the front lines of being able to watch is kind of going from a company where it was really hard to get certain vendors of components to call you back to having them banging on your door and, like, saying, use our newest stuff. Like, we'll give you a really sweet price on it. Like, you're moving huge volumes of this thing and, like, seeing just how much volume you have to get to see component prices really slide. And it's interesting because your business relationships with your component vendors change. Because, like, when you're still pretty small, you have to work with people who are kind of scrappy like you are. And I've seen a lot of stuff on, like, open source hardware blogs about how, like, smaller chip vendors will maybe work a little bit more closely with smaller companies because they have nothing to lose and everything to gain by it. But if you contrast that to, like, a truly giant chip maker, like, for example, Qualcomm. Like, Qualcomm is enormous. Like, if you're not moving five million units, like, good luck getting anybody to pick up the phone because, like, Qualcomm is too busy keeping their relationship with Apple. They won't sell you a chip, right? They just... Oh, no. Like, there was a time where we were using Qualcomm chips and we were never actually interfacing with Qualcomm. We were just interfacing with people who sold radio cards that featured Qualcomm chips. Like, we couldn't even get Qualcomm to pick up the phone. We were relying on a lot of open source drivers for Qualcomm PCIe interfaces and that sort of stuff. Yeah.

Chris Gammell: Yeah. Right. So then, okay, so it's 2011. You call your old boss and you're like, hey, I'm sick of data stuff. I want to do electronic stuff. Sounds like he went for it.

Nash Reilly: Yeah, he did went for it. He did went for it. And I done went for it myself.

Chris Gammell: Uh-huh. Uh-huh.

Nash Reilly: Wow, a little Montana coming out of there, huh? Yeah, a little bit, yeah. People tell me I have an accent and then some people tell me I don't have an accent. I haven't decided which camp I'm in.

Chris Gammell: I mean, maybe it's like one of those things where like as you relax around people or if you have a drink or two, you know, like that accent comes out more. I don't know. Do you sound different when you're around like friends from back home? Because that's usually a big test too.

Nash Reilly: I don't feel like it. I've been told I sound different on the phone. I'm sure if somebody comes up to me after the podcast and is like, hey, weren't you on the amp hour? They're going to be like, oh, you sound different in person. I'm sure. I mean, can you turn up twang though like consciously? I hate to break it to you, but we don't have much of a twang in Montana. There's – I don't know.

Chris Gammell: I don't – I mean, I think you may be the first person I met from Montana. Montana is a big place with not many people.

Nash Reilly: That's true.

Chris Gammell: It's true. It's worth a visit if you ever get out there.

Nash Reilly: I've heard it's gorgeous out there. Yeah. I mean – It's beautiful. I was just there two weeks ago visiting my parents for the holidays. Oh, that's nice. Went skiing. Oh, yeah. It is truly one of the last best places. Nice. No engineering jobs, sadly.

Chris Gammell: Well, that is how it goes with more remote areas. So unless you do your own thing, you know, you can work anywhere. Oh, that's the hope.

Nash Reilly: There is like kind of a small tech sector bubbling up in Bozeman, the town that I grew up in and the town I went to college in. There was a laser diode company there, ILX Lightwave, for a really, really long time. And I think they were just bought by Newport Scientific a couple years ago. There was a different company that made LiDAR units for autonomous vehicles, Blackmore sensors, and they were just acquired by a bigger self-driving car startup, I think, called Aurora Sciences.

Chris Gammell: Yeah. I mean, those things are like catnip to, you know, big autonomous driving things. They're like, oh, we'll just buy the whole technology stack.

Nash Reilly: Okay. It creates some interesting options if you cannot hire talent and have a couple million dollars lying around. I think Patrick McKenzie.

Chris Gammell: Which they unfortunately do.

Nash Reilly: Yeah. Yeah. Patrick McKenzie wrote something on his blog about that once. It's like, oh, Google can just hire somebody by just buying their whole company if they don't want to change jobs. That's right. That's right. And then they lock them in for four years or whatever.

Chris Gammell: Yeah. They get some work out of them. Yeah. That's fun.

Nash Reilly: It sounds like you're speaking from personal experience there.

Chris Gammell: No. I mean, I have not been bought out yet. Don't worry about me.

Nash Reilly: Well, I hope that comes along for you at some point if that's what you want. It seems like a pretty good way to get rich quick, but, you know, I wouldn't want the amp hour to go away.

Chris Gammell: Well, I don't think that's going to be – that's not a risk right now, so don't worry about that. Thank God. Yeah. Okay. So, you're now – you're at Sonos, and what, you just started diving – I mean, like, you took a break there then, too, so you had to kind of, like, readjust, or how did you get into it then?

Nash Reilly: So, I finished school. I went to work for Micron. I called my old boss, and he said, we're hiring in the sustaining and manufacturing engineering group. He talked to all of my former coworkers in Boston, and they were like, yeah, he doesn't need to interview. He can, like, hire him, but I was working with a group that was mostly centered on the West Coast. So, I flew out to interview, met all of them. I think I passed the interview when I fixed a test bed for one of the people who was interviewing me when he was like, I'm having a problem with this test bed. That's a great interview tactic, by the way. I was like, oh, I remember how that test bed worked. It really is. I remember how that test bed worked for my internship, so I got it fixed for him, and he got it done. And conveniently, that was actually when the hiring manager walked up, and I was like, oh, hi. And that worked out. So, you hear that, folks?

Chris Gammell: Just find something broken, or, you know, another option is break something when you go to an interview, and then be like, I know how to fix this.

Nash Reilly: Just disconnect one cable on one breadboard on the way in, and you're good. Right, right. You didn't hear that from me. So, I worked in that group for about three years, and sustaining engineering in larger companies is, if that's a term that nobody's familiar with, is basically your first-line engineering support for when pre-designed products have something going wrong with them. So, sustaining engineering is like when you have a field return issue where a lot of customers are returning products. It's your job to figure out why and to track trends. And if that trend has some sort of root cause, to find that root cause and to design it out of the product and make a change to get that failure mode out of the product. It also includes a lot more work like making sure your supply chain is stable. Like, hence all these watching our supplier relations change over time. Like, as we got to the point where we were selling more and more products, we were able to get better and better deals out of a lot of our chip vendors and better and better support. And actually, like, start being able to work with different vendors who were willing to meet the demands that we had for chips. And that was a lot of travel to China. I spent probably – I was in China probably one week out of every quarter. There were some challenging times in that. I once had to leave a vacation and fly directly to China to deal with a line down issue. So, the worst thing that you can – the worst thing that can happen when you're running a manufacturing line with a contract manufacturer is the line going down. And I'm not 100% sure of the fiduciary responsibilities that your company has to the contract manufacturer. That's a long way outside of my department. But basically, every day your line is sitting idle, it's costing you money and it's not making you any money. Yeah, double whammy. It's a bad situation to be in. Nobody likes it. Everybody is very, very upset because they want the line to start running so everybody is making money again. So, I got a call on vacation about two years after I started for my boss saying that the line had gone down in China and they need somebody to go out and troubleshoot it. And fortunately or unfortunately for me, I was actually on vacation in Hawaii. So, I was halfway there. Oh, halfway there. Exactly. Yeah. And I had to leave my vacation. I was able to see my vacation through to the end, which was cool. And then I got on what was actually the most turbulent flight I've ever been on and got to China and got the line back running after about four nail-biting days. Oh, man. That's crazy. Yeah. It was very, very exciting.

Chris Gammell: That's a great word to use there, but it's probably not one I would use. I mean, stressful, nerve-wracking. I don't know about exciting. Exciting is a good word to use there, though, right? I'm sure that you learned a ton just in those four days.

Nash Reilly: Oh, yeah. Oh, man. Those were a wild and crazy four days. My fiance has a word for this kind of situation. She calls it type two fun, which is it's not any fun when you're doing it, but it's really fun to talk about afterwards.

Chris Gammell: Yeah.

Nash Reilly: That's good. That's good. I find that certain kind of people actually kind of like these situations. It's high stress and it's demanding, but it really brings out a side of people that you don't see any other way. And you can actually see a lot of really great teamwork out of people in times like that.

Chris Gammell: Yeah, I think the people usually don't say that they like it specifically, but there's certain people that will keep going back to it. And they put themselves in those situations over and over again. They stay in that kind of job or they are the go-to person. They like that identity as well. And I think that that definitely speaks to having that kind of desire to help, of course, but then I think being able to handle it, wanting to rush, that kind of thing. Oh, yeah.

Nash Reilly: I totally, totally agree with that. Related to that, I work with a guy now at Sonos who is one of my mentors. He's been an electrical engineer since before I was born. He has worked for tons of little startups, tons of big companies, designed digital systems that are like mind-bogglingly complex, done a lot of cool stuff. But anyway, he was telling me about one of his bring-up efforts for his prior company was on this big X86 Intel behemoth chip that they were one of the first customers for. And this was a huge project. I don't actually recall the customer. And this is a secondhand story, so I don't feel so bad in getting some of the details wrong. But he was talking to me about how it took him six months to get this chip bring-up done because it was stepping through all of the power sequences, like all of the rails. Like this is an X86 chip, like it's not uncommon for these to have 20 distinct rails with very, very tight power sequence requirements for each voltage rail. Like really, really intense stuff. And he did this for about a month himself and felt like he'd exhausted some of these solutions. And actually, at that point, he managed to leverage through his employer a contact with Intel. And they sent him one of their FAEs to work with him for about a week, and he couldn't figure it out. And that actually got him to the top tier of Intel FAEs. So this is crazy. This is a group of people that Intel keeps on call. And I don't know if this is still the case, so somebody keep me honest if this is no longer the case. But this is a group of people at Intel. It's a small team that are capable of chartering Intel's corporate jet, dropping everything they're doing, and going on-site with the client. Like paratroopers, basically. Exactly. They're the smoke jumpers for Intel. They can fly in at a moment's notice and help get your chip running and solve your problems. So he actually, after a couple more months, got one of those guys, the paratroopers, to come in and work with him for a week and made some progress. And I think after about two weeks after that, they got the chip up and running. Everything was satisfactory. But that was six months of his life.

Chris Gammell: That is, yeah, that is determination, you know, like just to keep going with that, too. You know, like that's killer.

Nash Reilly: I've had some brutal chip bring-ups of my own, but nothing that approaches six months on the same piece of silicon.

Chris Gammell: Yeah, I mean, that's the point where you're like staring at this thing. You've read the data sheet five times through, and you're starting your six, and you're like, could I have possibly? I mean, granted, an Intel, an x86 data sheet has got to be, you know, thousands and thousands of pages long. But at that point, like, you know, you just start asking yourself, like, have I missed something? Like, I just must be missing something, or, you know, there must be a bug or something here. And it's just, you know, you start running yourself in circles. I don't know about you, but I've been running, not running into that lately, but like I've just been making sure I document. And so, because I don't have anyone keeping me honest anymore, you know, other than like the clock and the real world. Yeah, no, that's super important. Because you can have like, you can have like someone to check your work and make your, make you, keep you honest, I suppose.

Nash Reilly: Yeah, no, I definitely, I definitely try to use writing and note-taking as kind of a second brain where I can keep track of all these things. Like, I'm, I am a really obsessive digital note taker. I, I've talked to a lot of people about this, like Evernote, Microsoft OneNote, Google Keep. I know a lot of people who use some combination or one of those platforms. And like, I, like, I find it's, it's super essential to me. It definitely helps me be more effective in that sense. And like, and especially in those situations where you find that the error and the thing not coming up is just like a zero ohm resistor that you thought, you know, loaded, but you loaded it. Or like this one bit and this one EEPROM that you forgot to set. And that was actually the bit that it relied upon for a whole other chain of events. And that's kind of its own special nightmare.

Chris Gammell: I, I have one that I tweeted about today, actually, I was bringing up a DAC and I'm looking at this board and I'm getting these weird ass values out of this DAC. And I just think, you know, I think it's my code, right? I try all these different things and I'm like, this is not making any sense. And then finally, I'm like, I better check the reference on this thing. And I'm like, oh man, the reference is like slightly off. Why is, why is it off? You know, references aren't this bad. And then I go look at the schematic or the layout rather. And A ground, which is a separate ground, it's connected to three different caps. But guess where it's not connected? Oh, yeah. Anywhere else.

Nash Reilly: Oh, that's fun. I've, I've, I've, I've seen, I've seen the other half of that before. Like, I don't know. Yeah. It's like, is it bringing up religion on this podcast to talk about like separate A ground, separate D ground or unified A ground D ground?

Chris Gammell: Well, you know, we've had Hank Zumblin on before and he's got that fantastic app note from ADI about, I remember, but I've referenced it so many times. I never remember the name of it though. Um, but it's because of that app note and ADCs, um, I asked him to come on the show.

Nash Reilly: Oh yes. I need this app note in my life. That's, uh, let me see if I can find it. Oh yeah. Hank Zumblin, staying well grounded. I like it.

Chris Gammell: That's the one. Yep. Standing, staying well grounded. And, uh, and we've referenced that and he's been on the show and we asked him about it. And like, it's, it's basically, you know, he's saying like, there's a lot of ways you can skin the cat, but like, you know, having some kind of point where you at least in theory have, you know, your grounds coming together, even if they are one contiguous ground, that's, that's the thing to do, you know? Yeah. No. At least you need to know, know where things are flowing, you know?

Nash Reilly: Yeah, definitely. I, one of the, probably one of the best, the best weeks of learning in my life, uh, at least in my electrical engineering career so far was taking Lee Hill's silent solutions training class, which is looking at the same sort of thing, but from the angle of, uh, of emissions compliance, looking at radiated emissions and, uh, and, and, um, radiated immunity. That's, that's a little bit of lesser importance to me. Radiated immunity is more like a, a defense industry thing and a medical industry thing just to show that some huge field isn't going to cause some stray current, which is going to cause something to break. But emissions is basically like you're trying to design something. So you have as few unintentional radios designed into your product as possible.

Chris Gammell: Big old loops in your, uh, DC to DC or similar kind of things.

Nash Reilly: Oh man, the loops in the DC to DC is like. Uh, that I, I have never been personally bitten for the things that always amazed me are just ground return planes. Like the thing that you're talking about Dax, the thing that I always have show up on radiated emission scans, I2S clocks. It's always I2S master clock. It's always the master clock. And you know, it's the master clock because the master clock has this nice fast edge rate to meet some rise time spec at the DAC or the ADC. And it's just at a higher enough rate of repetitiveness that you're going to start to see the generally not the third or fifth overtone showing up on an EMI scan, but like the ninth, 10th or ninth, 11th, 13th overtone, like right in like a hundred megahertz where all of your cable interconnects inside of the product function is really good antennas. That's always where the problem is. I see that interest. I feel like I've been bitten by that so many times and that's its own little adventure. I actually wrote a blog post about that. I've spent so many weeks in windowless rooms that I want to save other people all those weeks in windowless rooms and compliance chambers.

Chris Gammell: Well, how do you, how do you usually end up fixing it too? I mean, is it just like well-placed ferrites or something or what are you usually doing?

Nash Reilly: So well-placed ferrites, I've actually had a, had a hard time with ferrites. Like they can help in certain places, but like when I, when I find like a narrow band radiator, like a digital radiator, I find the thing that always ends up being most helpful and it seems super stupid to say this, but like, I want to help other people who have to struggle against this. Uh, so IO drive strength, like changing the output impedance of your driver in, in your IO. Like I'm talking about in the context of like an I2S bus, like you're, you're driving a bunch of single-ended IO across a connector. Like if you can keep that on board and not go across a connector, like a flat flex cable or something like that, you're typically in good shape. Um, but if you forget a stitching via or return via that can start to radiate that way. So generally what you want to do is you want to turn down your current drive strength or increase your drive impedance so that you're not sinking as much current into it. Um, the next thing that you want to do after that is put a really, really small external RC filter on the line that functions as a little serious termination and then a little cap to start rolling off the frequency past a certain, uh, past a certain knee frequency. You're like just basically setting a three dB frequency. Um, and I was actually, uh, there's actually a pretty easy way to calculate this. I was, I was looking at, um, I was actually looking at the, uh, list of folks that you've had on the podcast before. And I noticed that you had Howard Johnson on here. Yep. Yeah. A long time ago now. Oh man. He's like, I wish I could take a picture of this because I'm actually using my copy of high speed digital design as a mic stand right now. Um, Oh, nice. Yeah. Howard Johnson, like big hero of mine, uh, learned a lot from his book, still learn a lot more from it. But, uh, the first 20 pages of that book are incredible for digital design, but they're also incredible for emissions because it tells you everything you need to know about getting radiated emissions for like narrow band digital signals out of, out of your product, because it basically tells you like, here's the effective bandwidth given the rising edge. And if you basically set your three dB frequency of that external RC filter to that knee frequency, you're not going to degrade your digital SI at all, but you're going to get rid of those higher harmonics. It's sweet.

Chris Gammell: Right. Yeah. You're still going to, you're still going to get the, the, the fundamental, right. And you might round off the corners a little bit. So I don't know if that, does that end up impacting like your rise time requirements or is that okay usually still, or is it a balance?

Nash Reilly: Uh, typically funnily enough, most of the, of the DACs and ADCs that I've worked with that are I2C compliant. Um, there's not a lot that these days that are super finicky about rise time anymore. Um, and, and I think it's just because they've moved to processes that I guess, like I'm guessing like for one thing, they're small enough that it's not such a big deal to just put a Schmidt input on all of those digital inputs. Um, the other half of that is that, um, I think a lot of these chip vendors have realized that by requiring a master clock, like most traditional I2S buses do, uh, they're actually creating more radiated problems than they solve for customers. So a lot of people are integrating onboard PLLs under those chips and just regenerating the master clock from the bit clock. Cause it's gotta be a integer multiple anyway, if that makes sense.

Chris Gammell: I can't say I know anything about I2S. Uh, I, I've seen it. I know it's on chips and I'm like, yeah, I just kind of like, well, I don't really do sound. And then that's something I should bring up here too is like, we don't usually have people that are doing audio stuff on here too. So that's kind of an interesting insight into the whole. Oh yeah. Yours is an interesting insight into the whole world.

Nash Reilly: Yeah. It's, it's, uh, it's, it's super easy. If you've done spy, you can do I2S. I like I2S is just spy. It's just spy in one direction. In fact, a lot of microcontroller, uh, spy peripherals generally can do I2S if you find the right bit for it. I know a lot of microchips, uh, chip lines can do I2S with fairly minimal mucking about.

Chris Gammell: And so that's what you're saying about having a master clock because it's sending a clock signal by itself instead of it's like an I2C where it's, you know, clocking just on the data byte. Yeah.

Nash Reilly: Something like that. Well, so you need a master clock in a lot of these instances just because a lot, uh, so audio is one of the few arenas where Sigma Delta ADCs and DACs really rule the roost. Um, and if you're not familiar with Sigma Delta Dax, there's a, like an incredible amount of resources on analog devices website about them. There's all sorts of fun stuff about it. It's a super cool nexus of digital and analog technology. And you said Sigma Delta Dax cause I've used ADCs before, but never Dax. Oh yeah. It goes both ways. It's great. Um, yeah, no, it, and, uh, it's, it's the same process just in reverse. Um, so the whole principle about your, your oversampling, uh, but you can oversample and modulate in the reverse direction by feeding like an I2S word into a DAC and then having an up sampler, uh, output basically a single bit, bit stream. Um, uh, like the, the actual mechanics of this are way beyond my pay grade. I've just bought chips that do it professionally, but, um, but we're talking like THD total harmonic distortion at like negative 110 dB, which if, if I put that into noise terms, that's like about three microvolts RMS across the audio bandwidth. Uh, and that's like, um, yeah, again, I don't have a reference point here.

Chris Gammell: So like, uh, what's good, what's bad. Okay. Uh, like what's a, um, what's like a cheapo Bluetooth speaker I might get from AliExpress versus, you know, like a nice Sonos speaker, you know, like in terms of like noise, noise levels and, and, and harmonic distortion.

Nash Reilly: Yeah. Okay. That's a great question. Um, so a cheapo Bluetooth speaker, it really depends. Um, the, the cheapest, junkiest stuff I would imagine to be like 60 dB of harmonic distortion, which means that like, and total harmonic distortion is, is, is kind of a niche measurement for, uh, for the audio field. It's kind of like signal to noise ratio, but it's also adding distortion because that's basically the point where your signal chain is going to start adding imperfections to your analog signal. So it's, it's this cumulative analog imperfections of your chip plus the noise and the bandwidth of interest.

Chris Gammell: Hmm. I remember like, uh, is something with like a, it's like how close a sine wave is to a sine wave. Is that, is that kind of like a THD measure?

Nash Reilly: Yeah. Yeah. That's, that's a, that's a great way to describe it. Um, uh, the, one of the companies that we leverage really heavily for audio test equipment is called audio precision. Uh, and they have a great guide for THD and THD plus N and, and a much better explanation than me trying to do it. Um, uh, with, with lots of nice graphs and nice pictures and that sort of stuff.

Chris Gammell: We do like graphs.

Nash Reilly: We like pictures. That's good.

Chris Gammell: Yep.

Nash Reilly: I love a good graph. What's your, what's the best graph you've seen lately? Hmm.

Chris Gammell: Um, I don't have a good answer for the best graph I've seen lately, but I do have a party game that's based on graphs.

Nash Reilly: Oh, tell me, I need this. I need this in my life.

Chris Gammell: It's basically like apples to apples where like one person like has a chart and it gives you the X axis and then everybody plays a card where you play the Y axis. And so it might be like, uh, you know, it's going like down into the, it's a, it's going down as you go left to right. And then it'll give you like the X axis and it's like, you know, years of your life and it's like, you know, college and midlife and end of life and stuff like that. And then, you know, you would submit cards that are like, I don't know, usefulness of your knees or, you know, that would be like a really boring answer, but there's a lot of funny answers too. Yeah. I can't remember the name of it now. I'll look it up while you tell me about your favorite graph.

Nash Reilly: My favorite graph lately, uh, I've subscribed to analog dialogue, uh, and they had a great, uh, a great article in the December issue about designing seismographs and earthquake sensor networks. And they show the different kinds of waves that travel through the earth and propagate through the ground, uh, when earthquakes occur. And the graph of that and the visualization of that was super cool. Um, highly recommend that. Um, anyway, uh, that was a large tangent. Sigma Delta. Oh yes.

Chris Gammell: Wait, I found it.

Nash Reilly: Charty party.

Chris Gammell: It's called a charty party. Charty party. I'm writing that down. Kickstarter apparently. Yeah. I'll, I'll link it in too, but I actually have the, uh, the actual version. It, you know, it, it, it, it's tough to get it first, but it was pretty fun. Oh, I need this in my life. Apples to apples.

Nash Reilly: That fiance I was talking about is a, is a math teacher. So this will, this will, this is going to rule.

Chris Gammell: Oh yeah.

Nash Reilly: You're going to, you're going to, uh, make someone's day. So that's great. Absolutely. Um, so yeah, Sigma Delta is, is definitely the big thing in audio, um, because you can get very high signal to noise ratios, uh, and all you're trading for it is latency. Um, which the human ear is usually okay with, right?

Chris Gammell: I mean, like if you're, oh yes, you know, you don't care that your Bluetooth you're on your phone starts and then a hundred milliseconds later that just comes out of the speaker. You don't really have to care about that. Cause there's, you know, there's, it's not like a thing.

Nash Reilly: Yeah. So here, here's a, here's a fun question for you. How much, uh, how much time delay do you think humans can recognize and in terms like what, what do you think has to be the time interval for a human to recognize events is instantaneous? I thought it was 30 milliseconds. That's the number I always remember hearing. Yeah. That's about the rule of thumb for a lot of, uh, instantaneousness ratings. Yeah. So, so the part two of that, how much, uh, how much phase detection does the human ear have in terms of time? None. It's, it's substantially more sensitive to that. And then in actual research, actually research, uh, is, is somewhat debated about how, how much phase delta between two signals you can detect with your ear, but like your ears are very, very sensitive and it's in the neighborhood of one millisecond in some people to about five milliseconds in most of the population. Is that because of, because we're binaural, like that it starts to give you like beats or something like that? I would have to assume so, but I have to tell you that we pay people much smarter than me, much more money to know that at Sonos. And I, I, I'm afraid I don't have a hardcore answer on you, but, um, on that for you here.

Chris Gammell: That's okay. I mean, that's, that, that is very interesting though. So how does, how does that end up playing out then in your designs though?

Nash Reilly: That ends up playing out because, um, the nice thing about Sigma Delta ADCs is that they do have high latency, but since they're largely digital, that latency is constant and you can anticipate it and you can plan for it. Um, like in isolation, if, if, if you don't need to synchronize anything, uh, a Sigma Delta ADC or Sigma Delta DAC makes a lot of sense in a lot of systems where you have a relatively, uh, low bandwidth of interest. Like that's why they're so great for audio because they function great from DC to 20 kilohertz or so, but you have to oversample at such a higher rate that it rapidly becomes impractical. I mean, like, uh, like a by 64 oversampling rate is not unusual for a Sigma Delta ADC or a by two 56, um, oversampling rate is not unusual either. So then your clock from your I2S would have to be like megahertz, many, many megahertz or what? So I think I'm trying to do the math in my head. So a 50 megahertz sampling rate, uh, would require a gigahertz clock at a 256 sample.

Chris Gammell: Oh, I think you said it was a, Oh, okay. I was thinking it was a sampling of the audio bandwidth, like of like 20 to 25 kilohertz. You're saying it's at the, the sampling of the, the clock, the I2S clock itself.

Nash Reilly: Yeah. So your clock frequency has to be an integer multiple of your sampling frequency. Okay. And a much, much higher one. So, um, so like if you're oversampling it here, I'm using the Google Chrome calculator here. So a 50 megahertz, uh, sampling bandwidth, like a, like a Nyquist frequency of 50 megahertz, uh, oversampled at 64 X is going to require a 3.2 gigahertz clock, which is what, which is what most cutting edge Intel processors are, are like trying to avoid power, power consumption.

Chris Gammell: But yeah, exactly.

Nash Reilly: And, and 3.2 gigahertz, I think is, is, is well beyond the range of what most people want to pay for, for test equipment. So that, right. Yeah. My VNA can do it, but it's a, that was not cheap. Oh no, I'd imagine not. Yeah. That's, that's a, a range of frequency that few play in and nobody is playing in the Sigma Delta ADC, in the Sigma Delta ADC camp, I'm afraid. Okay.

Chris Gammell: So what do people do instead then? So I, I'm actually a little, I'd like to take a step back and understand too, like the, the sampling frequency at 50 megahertz is higher than I would have guessed in the first place. Why, why is it that high in the first place?

Nash Reilly: So, uh, uh, this, this is one of those times where I, where I, I really wish I had a great picture. Um, so the principle of, of a Sigma Delta DAC or Sigma Delta ADC is that you take some small sampling window and you basically determine if you are over that point or under it constantly. You're doing a constant comparison. Um, uh, other great book that I've learned a bunch from, like I'm sure everybody on this podcast has, is, uh, uh, art of electronics. And the third edition has the most elegant description of a Sigma Delta ADC that I've ever seen. Um, and it's basically just showing you how a one bit DAC basically with a feedback mechanism into an integrator and a summing junction can produce a really accurate representation of the average of something. Um, this is one of those things that is really hard to describe in words, but it's really great with pictures.

Chris Gammell: Yeah. Um, but you're saying that that is, um, so even for a relatively slow signal, you're saying you can get a really precise, um, so let's say, what would it be? So like, let's say we had 50 kilohertz of audio frequency. You're saying that if you went up to 50 megahertz, so what is that? A thousand X, you know, 10,000 X, the sampling, right? So you'd get, you'd get a lot more bits per sample itself.

Nash Reilly: Does that sound right? Yeah, that sounds about right. So another way to think about it is this. So if you have a 50 kilohertz sine wave, right? Um, if you're starting to look at that over a 50 megahertz unit interval, what does that sine wave look like at, at that small amount of time? Like, you basically.

Chris Gammell: I mean, it looks flat pretty much, right? You would like for within like the 50 megahertz from one sample to the next, it would look like not much just changing. I would imagine.

Nash Reilly: Exactly. Exactly. So if you start looking at a 50 megahertz window or just the sampling interval window. So if you're looking at a 50 kilohertz sine wave, you'd have to be sampling at a hundred kilohertz to be able to reconstruct that without any aliasing. So if you look at that really small discretized portion that's sampled at 50 megahertz, it looks pretty much linear and you can feed back and integrate a comparators output across that small window to get the average value across that. Um, analog devices again, has another great visualization of this, just kind of showing like a really small windowed region and how, uh, with enough time you'll get an average value out of a Sigma Delta ADC. That's really, really accurate and really, really low distortion.

Chris Gammell: Yeah. I am looking at a, uh, Sigma Delta ADC and DAC application node and two, three, two, a three. So that's probably what we'll, I'll link that in the show notes too, but it sounds like my mind was just Googling. So I'm sure people could find other resources out there too.

Nash Reilly: For some reason, this is like showing me a bunch of like stuffed animals of woolly mammoths. Yes. First order Sigma Delta ADC. I love it.

Chris Gammell: Well, you know, Sigma Deltas are, uh, maybe it's like, you know, like in how, how stuff or how things work that, uh, that old tome. Oh, did you have the computer program?

Nash Reilly: I had the computer program for, for like an old Mac computer.

Chris Gammell: Oh, I never had that. I still have the coffee. That's actually my only coffee table book is, is the, how things work. How stuff works. How stuff works.

Nash Reilly: Right. That's great. Yeah. No, it was great. The, the, uh, animated woolly mammoth was awesome. I haven't thought about that in years.

Chris Gammell: Is that what it's called? The way, oh, the way things work. Jeez. You'd think I'd read the coffee table book once in a while. The way things work. And there's newer versions too, cause I've, I've seen a newer version.

Nash Reilly: I think we can forgive you. I mean, like you don't bring the, you don't have the coffee table book for you to read. That's for the guests to read.

Chris Gammell: That's right. That's right. Sit down, have a cup of coffee. I'll be right with you. Enjoy the woolly mammoths. Yeah. Of course. It's polite. Right. Exactly. Okay. So we kind of get a feel for that, but it seems like these are, so I go and crack open just about any consumer level or even high end thing. It's going to have a Sigma Delta DAC in it. It sounds like, like it's the only game in town.

Nash Reilly: That seems like a pretty fair bet. I mean, they have a lot of things going for them, uh, on the chip side, as well as the performance side, just, uh, in terms of implementation, they're, they're not super complicated. Um, uh, and in some cases they can actually add a lot of functionality that you wouldn't want to do yourself. Like you can, you can load digital filter coefficients into them that, uh, you can, you can add PGA is you can add, um, like input PGA is or output PGA is, um, you could find some that are specifically DC coupled, like most audio, uh, most audio Sigma Delta ADCs don't have DC performance specifications just because they don't really bother with getting that part, right. Just because most audio applications look, uh, high pass at about 20 Hertz anyway. So you don't really care what the DC performance of it looks like. Just so long as you're not going to create some big pop when you plug something in.

Chris Gammell: Um, right, right, right. Yeah. Well, so talking about the specs of parts too, I mean, as a broad kind of thing, I mean, uh, it's an audio signal chain, right? There's lots of potential things you could put in that pathway, but like what differentiates a really high end or really good thing from another good thing? And what, what, great words there, Chris. You don't podcast. Uh, so like between different high end vendors, I'm sure that there's, you know, similarities and differences and then between high end and then low end, I'm sure that there's differences and similarities as well. Like what are, what, what starts to differentiate a really, really high end product like a Sonos thing? That's an excellent question. So I feel like I should also cut it. Sorry too. Oh yeah. Yeah. Because I get emails about Bluetooth speaker from, from AliExpress sellers all the time, you know, like, like the, there must be other magic in there or else it would just get ripped off. I mean, especially in a consumer level spot. So that's, that's kind of what I'm really getting at here is how do you differentiate and how do you, how do you maintain that differentiation?

Nash Reilly: So noise is something that we care about a lot. Like, like cheap, low quality speakers, um, get noise in the signal chain. I like you'll, you'll have it interfere. Like, like noise can couple in externally and, and interfere with the analog signal chain. Like, like putting cheapo Bluetooth speakers next to a microwave, like something will couple into that and, and like you'll start to hear a buzz or a hum or a crackle. Hear the, uh, hear the siren song of your burrito literally. Yes. Yeah. I like that. Uh, I wish my burrito could sing to me. Um, so another thing that is, is on my mind and I've had to deal with before our, our clicks and pops. So, um, most, most consumer products are super cost conscious. Um, as a result, you want to minimize the number of external components that you have. You, um, anything that you put onto the board is, is profit out of your pocket at the end of the day. Um, yeah. As a result, a lot of consumer audio components are designed to work off of a single rail just and to be AC coupled. So you don't have to pay with a voltage inverter. Like, like a lot of really high quality audio stuff will be DC coupled or have like an absolutely enormous AC blocking cap, like a huge electrolytic or, or film output blocking cap, um, to AC couple it with. Uh, but the thing is to DC couple and amplifier, you have to have a positive rail, you have to have a negative rail and it, it takes money to pay for that. And, and, and, and that is a problem.

Chris Gammell: And it seems like the prior, the parts in the ecosystem do not favor when I, at least looking at op amps these days and stuff like that. It seems like the, you know, the trend is towards single rail as well because of every, everything else that's moving in the industry.

Nash Reilly: Yeah, totally. Uh, it's, it's a pressure that's on everybody and not just the consumer section, uh, segment. It's, it's, it's anybody who really intensely cares about the cost of the electronics in their product. Um, so because of that AC coupling though, at some point you are going to have some part of that system have some kind of a transient that manifests in the speaker output as a click or a pop. That's not such a big deal in some products. Um, but it is a big deal for ours. Like we, we don't, we don't want you to be able to notice your amplifier turning on and clicking. Uh, and, and like we've made, we've, I've personally have spent a lot of time making very quiet amplifier pops, even quieter, just so that people don't notice them.

Chris Gammell: Are they, and they're, they're so that common that like you did something that you have to think about, or it's just a, is it just to be because of what's out in the field and what, what people, because people are interacting with this, like what is actually causing that ultimately?

Nash Reilly: Uh, great question. Lots of things can, can cause a click or a pop. Uh, frequently it's when you see some startup of some digital component in, in your audio signal chain, like a DAC startup, uh, is a pretty common thing that will cause a transient that will generate a pop. Um, uh, like a class D amplifier startup. Um, that's actually a really common source of a pop too, because, um, a class D amplifier output is basically just a motor driver, uh, like just a half Ridge motor driver. That's got LC filters in the output to control some of the higher harmonic content. Yeah. It's pretty crazy that that that's like, what's yeah, but that that's, yeah, that's right. Yeah. That's crazy. Yeah. That's all it is. It's, it's not, it's, it's not, uh, anything too terribly frightening. Um, but the thing is, if those two, uh, output impedance is of the, of each of those half bridges aren't perfectly balanced, there's going to be some discontinuity when they come up and that, that small voltage discontinuity can manifest as a pop. And it's something we've dealt with before. And it's something I'm sure we'll deal with again. Yeah.

Chris Gammell: Right.

Nash Reilly: So, and like there, there's some tricks to, to get that down to a level where it's acceptable. Um, uh, common mode filtering, increasing common mode impedance. Uh, a lot of it is frankly in layout, uh, and a lot of it is situation dependent just because you'll be space constrained or you'll be thermally constrained just because, uh, you typically need a pretty big output inductor on some of these. That's going to start to get warm at max volume because it's got some finite resistance. It's going to start dissipating heat. Can you explain the, uh, the layout, the layout kind of stuff that you're talking about? One of the things that's really convenient about monolithic class D amps is that you don't have a ton of layout work to do. Um, you, you generally need a, uh, you need two second order LC filters, uh, to remove all of the higher harmonic content from the class D switching stage. Um, and, and that's just about, that's just about it. Um, but there are a lot of concerns that can jump out and bite you, uh, with those output filters. Like we've seen them on radiated emission scans. We've, we've seen them clicking and popping just because you have to figure out the right configuration of inductors and the correct configuration of capacitors to make sure that, uh, your fluxes are canceling out. So the impedances on each end are balanced. Um, like that big imbalance is what causes a click or a pop on a startup, on a class D amp, the output. And it can be absolutely maddening. If you catch it at a late stage and you don't have that much time to fix it before your mass production ramp.

Chris Gammell: Yeah. Right, right, right. Yeah. And that's, I mean, it sounds like too, like a lot of these requirements that you have are like, you know, a click or a pop does not blow up the product. However, it's a very noticeable, like you said, you have, you know, humans who can notice things and it's a measure of quality and everything else. And it's something I assume you have to chase down and just get out of there.

Nash Reilly: Click or a pop doesn't necessarily blow up a product. Um, I mean, like Sonos, we're, we're really well known for, uh, for like all in one speaker products where you have a speaker, an amplifier and, and a little computer inside to make it smart. Right. But there are plenty of products that click or pop that like we don't make that that sort of thing can actually be super, super harmful. Um, like headphones, for example, and hearing aids, like a click or a pop in somebody's ear canal, that's really small through that size of a transducer can be devastating. That can actually like damage your hearing just because you're putting this really high energy impulse right into somebody's ear canal. So like, it's not a, it's not a part of a field of audio that we play in, but it's something that like can be very serious. And another example of that is actually, um, uh, going back to the Sonos port, which was one of the products that I worked on, um, that we had to think about click and pop pretty intensely is that, uh, we expect that people will use this product to drive a downstream amp head. That's going to put 30 dB of voltage gain on our signal. Right. And then you might actually blow up a speaker. Yeah, yeah, exactly. And people, people connect like old vintage speakers to, to old vintage amp heads because they love them and they love how they sound and they're very emotionally attached to them. Like we don't want to blow up somebody's speaker. And also when it blows it up, it's going to make one hell of a noise. It's going to scare the crap out of you.

Chris Gammell: That's right. It's going to scare the crap out of someone. Yeah, of course, of course.

Nash Reilly: Yeah, no, I've, um, like, uh, let me tell you a, a half volt step at line level out of sight of 30 dB of gain sounds like somebody smashing the walls with a sledgehammer. It's terrifying. Wow. If you're not expecting it.

Chris Gammell: Well, I'm sure in the lab too, you've probably gotten one or two of those before and you're like, am I dead?

Nash Reilly: Yeah, it's not, it's not subtle. It is not subtle at all. I, unlike truthfully, um, I have thought, am I dead? Because, uh, some of the tower speakers that we have to test these on, I'm thinking of this one pair in particular, uh, belong to a, a guy who's been with the company for a decade plus at this point, since they were very, very small. And he has this pair of tower speakers that he has on indefinite loan to us that are like this pair of Boston acoustics towers that are taller than I am and are probably worth more than my annual salary. And if I pop them, he can't get another pair of them because they don't make them anymore. So, oh my God, I'm like, oh my God, if I, if I wreck my speakers, if I wreck my speakers, I am hosed. Um, uh, that particular instance, I, I did not break your speakers. Mike, if you're listening, I'm, I'm really sorry about that. I don't know if I've ever actually told you about this. This, this could be the end of me. Edit, edit. Yeah.

Chris Gammell: Well, you know, you had a good run. You had a good run. It was a, it was a productive career. Well, you know, now you can move back to Montana, you know, and do something else. Yeah, we can become a hermit. I don't know. I think he'll still find me there. Probably. Yeah. Um. Before we move on to more of the, I want to hear more about the products you've worked on because they, you know, they have a couple of different, uh, things, but, um, since we were talking about speakers and, you know, the craziness and costs of speakers and front ends and amps and everything else, um, how much does the audio file crowd, like, how does that impact you?

Nash Reilly: Uh. Is it, is it rough? You know, um, I have to say, I like, I've worked on two products where audio files pay very close attention to it. Uh, like I worked on the Sonos amp, uh, which is basically this really powerful class D amp that you can hook your own speakers up to. Uh, and I've worked on the Sonos port, uh, and I've worked on the Sonos port, which audio file and custom installer, uh, customers use to connect really high quality head end units and amplifiers to. Um, and I have to say all the feedback on the products, those particular products that I've received has been really generous and has been really kind. Um, uh, and I think, um, I, I can't really take full credit for that. I don't, I don't even think I deserve most of the credit for it just because, uh, I've had a lot of help from more experienced engineers at Sonos and both the analog and digital domains to help make those products really rad. Um, and I, I got a lot of mentorship on what things I should be paying very close attention to, uh, and what things don't necessarily merit so much attention. Cause like I, I would have never been able to tell you what a clicker pop was coming into this job. Yeah. Right. And now I like having heard what that half volt step signal sounds like out of a 30 dB amp head and it being terrifying and world ending. Um, that there are lots of things that I would never have thought of as, as a fresh graduate that I realized are kind of table stakes for high quality stuff like that.

Chris Gammell: You know, that's, that's great that there's a good mentoring culture there too.

Nash Reilly: That, that really, I'm really fortunate in that respect.

Chris Gammell: I think that the other thing with like audiophile stuff too, I, I, you know, I generally, I'm curious because it's a, you know, obviously this is consumer level, but also high end. And, you know, so there's some interesting stuff in there, but then I also wonder with like the audiophile crowd if they're like, well, I couldn't possibly have anything digital in my speaker set up and they just count themselves out anyway. So I mean like it's, it's interesting. It's an interesting like human experiment.

Nash Reilly: I feel like, you know, like my reaction to that is a little bit like haters going to hate no matter what. And like, yeah, totally, totally. Yeah. Everybody's going to have a detractor no matter what you do. Like pleasing everybody is, is a recipe for insanity and a recipe for pleasing no one. Um, ultimately. And I, I've read people's reviews online that are like, I don't like Sonos for this reason. And like the, the one that you talk about, uh, like I don't let anything digital into my signal chain is, is like, that's fine. And if, if that's what you want, this, this product is clearly not for you because I, you probably are also dependent on vinyl and turntables. And if that floats your boat, that's right. Right. And gold, gold, uh, gold coated.

Chris Gammell: Yeah.

Nash Reilly: And, and you know what, if that, if that works for you and if that's, if that's, uh, if that's getting the job for you, like power to you, like I can't be upset at you finding a system that really truly works for you. Um, but like, there's also a fair amount of research that suggests that differences below, I don't know, something like 90 D 95 DB THDN or imperceptible. The human year, like there's maybe some truth to that, but at the same time, like 30 DB of amplifier gain, whatever I'm feeding into that has to be pretty, pretty damn good in order to sound pretty good under the 30 DB of amp gain, because the output is going to like, I know how noise figure works in amplifiers. Like it's going to amplify whatever noise is on there as well as whatever signal is there. Like it's gotta be good. Like the flaws, like the signal gets, gets bigger, but the flaws get bigger too.

Chris Gammell: That's right. That's right. Yep. Yeah. You know, you guys should really, uh, invent a thing that just takes out the noise. Have you thought about that?

Nash Reilly: You know, I, I've been working on that tirelessly for, for years now. Um, they, they, they've got us working in shifts. Good. Good. The, the, you know, you'll get there. You'll get the denoisificator. Actually, it's funny. You mentioned that I'm, uh, I, I, I don't want to shill too hard for art of electronics. Uh, actually that's, that's, that's a lie. I will show really, yeah, that's fine.

Chris Gammell: We do here too.

Nash Reilly: I'll show really hard for art of electronics. I've been going through a chapter eight on noise and low noise design, uh, starting before Christmas and like, it's, it's, it's something I should have read. It's re blowing your mind. It's something I should have read in much greater detail, much sooner than I finally did.

Chris Gammell: Yeah. You never know. Like, that's the thing about that book though. You never know what you're going to need it or which, you know, like, or really any book, I suppose. You don't know what you don't, you know, you, you need the depth in, right? Yeah. You need domain knowledge to go, go back and really be like, oh yeah, this is important.

Nash Reilly: The other thing I keep finding about art of electronics is just when I think it doesn't cover a certain topic in there, I've come to realize that that's actually not true. It's just in a different section or called something else. Like what's a good example? Oh man, the best one I can think of is actually like digital signal measurements and getting good digital signal measurements, uh, with cheapo probes. Um, uh, Horowitz and Hill refer to something that they call the El Cheapo special, which is their preferred method of digital probes, which is, uh, RJ, RG 174 coax with a kilohm resistor, um, soldered to a ground point and then to the test point. Uh, it's low inductance. It's relatively low capacitance. It's not going to degrade your edge and it's 20 to one digital probe. Wow. Yeah. That's interesting. Yeah. It's actually an overlap from, uh, Howard Johnson too. He's got a, he's got a chunk of that. And I think chapter three on tested measurement too. It's great stuff.

Chris Gammell: Is that because it's like the, the one K is because it's 20 X, uh, a 50 Ohm.

Nash Reilly: Yeah. 50 Ohm term. Transmission line. Exactly. You 50 Ohm terminate your scope. Uh, and then you have a kilohm resistance so that you have a high impedance that's broadband, uh, isolating the capacitance and the inductance of that cable going back to your scope probe. Yeah.

Chris Gammell: That's cool. Yeah.

Nash Reilly: It's, it's a, it's a nice trick. And I, I actually just did that, uh, as an experiment today, just, just cut a, cut a coax cable in half and compared it to an Agilent. Uh, I think it was an 1134 active probe and it's comparable. And actually, if you want to take the time to trim the leads really short, even a little bit better. If you can get the ground, Oh wow. Yeah. Getting the ground service loop smaller, it has a world of difference.

Chris Gammell: Okay. That's great. That's great. Well, going back to the, uh, product stuff. So what, uh, what have you worked on there? I guess we've kind of danced around these things. Oh yeah. What were some of the products that you've, you've worked on in the past?

Nash Reilly: So the most recent one that I launched was, uh, was the Sonos port. Uh, and that, uh, that, that just came out in the last couple months here. Um, actually, no, that might've been even longer ago. That was probably earlier this year. When was the Sonos port release?

Chris Gammell: And so what is, what is the Sonos port for people haven't seen it?

Nash Reilly: Yeah. The Sonos port, uh, is, is a very high quality DAC and a very high quality ADC that allows you to connect any RCA enabled audio input or output into the Sonos system. So if you have a record player, uh, the port services, your line into a Sonos system. So you can rebroadcast your, uh, your vinyl over a Sonos system. Uh, it also serves as a really high quality output that could go into an amp head. Um, so that if you have a beloved rack mount amplifier or beloved receiver that you want to, uh, get Sonos output through, uh, that's an option for getting it out.

Chris Gammell: Nice. Okay. So basically digitally enabling a lot of old gear and high, high end gear that you might have sitting around.

Nash Reilly: Yeah, exactly. Uh, it's also really popular with, um, with custom installers who like to go into houses and do big custom AV installs where they're like rack mounting lots of equipment and doing things where they have like amplifiers that route to in wall speakers or in ceiling speakers. Cause you don't need wires then? Is that the idea? I mean, like, or, or what are they? Well, it's, it's an interesting thing. Uh, uh, you would need wires in that case cause you would need to do a home run, uh, twisted pair or a home run speaker wire connection from like your in ceiling speakers or your in wall speakers back to some simple control amplifier and a cabinet. But there's actually this whole range of consumers, um, that are basically buying audio and video in their houses, prepackaged solutions from small businesses that go in and install all of these in their houses.

Chris Gammell: Oh, wow. Yeah.

Nash Reilly: It's, it's a really interesting.

Chris Gammell: That's a higher end than I ever need, but yes.

Nash Reilly: Like truth be told, it's probably higher end than I will ever need to, but it's, it's a fascinating little market and they really like Sonos stuff because it's very easy to install and integrate with these racks because all they have to do is connect it to the internet and connect it to the piece of equipment that they want it to run with and it's done. Yeah. It's, it's, it's a fascinating little market and I've, I've, I've actually spent most of my career at Sonos working with those types of customers and not the types that are, uh, not like consumer customers like you and me who are buying like a play one or a soundbar or something like that. Right, right, right.

Chris Gammell: The lower end of the Sonos field, but still high end in general. Oh yeah. Yeah. Like, like, I mean, if I'm going to be honest here, I mean, I, I, I buy Boothoo speaker. I mean, like I have a cheapo cheapo, you know, AliExpress kind of thing. I don't really, I don't care. I care about audio sound for amp for the amp hour, but you know, not really otherwise.

Nash Reilly: And like, and that's fine. Like there's, there's nothing wrong with that. Like there's, uh, there are people who care about just having sound and just wanting something that they can like carry around with them. Like we, we carry a Bluetooth speaker, but like it's fairly large. Like it couldn't fit in your pocket and there are Bluetooth speakers out there that fit in your pocket. Um, and, and we were, we are not a company that can do that. Um, and an AliExpress speaker is almost certainly going to be much cheaper than our speakers.

Chris Gammell: And that's right. Yeah. And, and, and with the cons, the, the similar, uh, low quality audio, blah, blah, blah.

Nash Reilly: It's a, it's a spectrum. And like, there's, uh, there's nothing wrong with going to the lower end, but if you ever want to move up, we'll be here. That sounds good. That sounds good.

Chris Gammell: So, okay. So you have the port. Um, well, I mean, what are, so you're a digital designer, but you've been saying a lot of analog words here too. So where does, you know, I guess I don't really have a feel for how big your teams are either. Like, you know, are you part of a 10 person team, 20 person team, one person team? You know, like what is, what is the, uh, what's the relative size of these things?

Nash Reilly: Yeah. I say the analog words and I actually even know what some of them mean. Um, so I feel the same way. You heard my thing about the DAC earlier. So yeah.

Chris Gammell: Yeah.

Nash Reilly: We're all, I even know how to hook up grounds with broad wires. Hey man, we're all, we're all just faking this together. It's all good. Uh, so Sonos is, is a, is matrixed, uh, in the way that most large companies are these days where we have, uh, an electrical engineering group. Like I report to a manager who manages electrical engineers, uh, and we have mechanical engineers, radio engineers, software engineers, legions of software engineers. It seems like. Yeah. The stack, I usually need a lot more software people than, you know, hardware is never really

Chris Gammell: done, but it, you know, it's, you get to, you get to move on to the next thing a little faster than the software folks do, I think.

Nash Reilly: Yeah. I I've noticed this, uh, in a lot of folks that I know in Boston who work in, in smaller startups. I'm in Boston, by the way, uh, if that's not clear yet. So there's, there's a lot of little hardware startups and a lot of smaller businesses around here. And it always seems like the, the electrical engineers are outnumbered three to one, if not five to one or 10 to one.

Chris Gammell: I'd put a 10 to one. Yeah. 10 to one. Yeah. And there are at least a hundred people that know JavaScript. Then to, you know, to the one person that knows VHDL or Verilog, I think. Oh yeah. I believe that. And that's still software E, I think, you know, I just think that there's just a lot, there's just a lot more need in the software world, to be honest. Oh yeah. It's not like a measure of value. It's just a need.

Nash Reilly: Yeah. And I think I have a lot of theories about that and that's, that's, that's probably its own conversation, frankly. Um, uh, but anyway, um, so yeah, I work in an electrical engineering group and there's probably about 20 of us in total. Uh, and that constitutes PCB designers, digital designers, analog designers, and tons of different people have tons of different specialties within that. Like some people know firmware. Some people are really into high speed digital. Some people are power supply experts. We, we have a lot of really sharp, really great power supply designers, uh, at Sonos who are very lucky to have that. Um, a lot of audio electronics people, obviously. Um, yeah, of course. Uh, a lot of really talented radio engineers, a lot of really talented antenna engineers. Um, yeah, no, I, if there's, there's one thing I feel fortunate about, it's that I work with a lot of people who are really smart and really driven and are always willing to answer my questions, no matter how dumb they are. Uh, so I'm, I'm really grateful for that.

Chris Gammell: I mean, like that is, I, that is the number one thing I tell people to like, and this is, I mean, you've been here, you've been at Sonos for what, like eight years now? It sounds like. Uh, I'll be, I'll be at seven years in July. Oh, seven. Okay. So seven years, but even, but it's your first, effectively your first job, your first full fledged engineering job.

Nash Reilly: Yeah. That's the data thing. Yeah. I think about it. That's how I think of it.

Chris Gammell: That is like the most important thing. Like whenever I tell people who are looking for a new gig or for their first gig, rather, it's like, go anywhere you can learn from that many people. Like that is just, that is like being in a crucible of, of like learning and, and like, you know, there's stress obviously, but like it is the right kind of stress. It's going to be like, that's going to be where you just learn so much stuff.

Nash Reilly: Oh yeah. I, I, I could not agree more strongly with that a hundred percent.

Chris Gammell: And you had an article on your blog about something about, uh, what was it? Uh, something about, uh, talking to impressing an engineer at a career fair. I suppose that's, that's a little different.

Nash Reilly: Yeah. I mean, like it's, it's funny cause like, I definitely realized in hindsight that coming out of engineering school, I had no idea what I didn't know. And like, I, I got out of school and I was like, yeah, I know stuff. I'm an engineer now. And like, I went to work at a company and I was like, oh God, I've, I've, I've been seriously misled. I don't, I don't know.

Chris Gammell: I gotta, I gotta go study some more.

Nash Reilly: Yeah, exactly.

Chris Gammell: And I mean, that's when I started reading the art of electronics too. Like, that's the thing. Like, you know, like it wasn't until I realized how much I had to learn and like, not look like a dummy and you know, yeah. I wanted to learn it too.

Nash Reilly: There's, there's, there's so much that I, that I know now that I had never even heard of when, when I started, when I started my engineering career. Like, like if I told you, if I, if I told my, my graduating from college self that I was going to go on an engineering podcast and try to explain Sigma Delta ADCs to somebody, I'd be like, that's cool. What's a podcast. What's a Sigma Delta ADC. That's right. Yeah. Um, so it's, it's not impossible to learn all this stuff, but like the good mentorship that I've had and like the strong teammates that I've had have made it bounds easier, like bounds easier. Um, yeah. And, and like our, our project teams kind of get organized along those lines where we like pick a person from each discipline and then work together. I think the project that I'm working on right now, I work with about 10 people in total on the day to day, uh, mechanical engineers, radio engineers, project managers, product managers, some software engineers, mechanical engineers. Um, and it's, it's very nice in that. Like I, I, I spend more time than I should on Reddit and like, and I see a lot of people talking about how like they go to certain larger companies, um, and they don't feel like they're doing technical stuff. Like they feel like they're doing a lot of paperwork and they're doing a lot of things like change control. Um, like I don't want to rag unnecessarily, but it seems like this is kind of a theme in the defense industry. Oh, totally. Yeah.

Chris Gammell: But yeah, I think just big company, like big co in general, you know, where it is, you know, lots of hierarchy and stuff like that.

Nash Reilly: Yeah. And I feel really lucky that like Sonos is an order of magnitude bigger now than it was when I was an intern. Like it was a hundred, 150 people when I was interning there. It's right about 1600 people right now, I think. Oh wow. That's big. Yeah. And yet still like I'm 31 years old and I can be the lead electrical engineer on a consumer product. Um, I'm very fortunate to be able to do that. And I think a lot of that is just as much luck as luck and patience and, and as it is work. Um, frankly, I, I, I really got lucky in a lot of these things.

Chris Gammell: No, it's, it's good to, to recognize that too. And like, and just, you know, try your best and, and see what you can make.

Nash Reilly: Yeah, definitely.

Chris Gammell: So, okay. So you have the, uh, the port we talked about and then what are the other ones?

Nash Reilly: Uh, the one I worked on before the port was the Sonos amp. And, uh, I was, I was the digital lead for that. That was, that was kind of a bigger beast than one engineer could rightly tackle. Uh, I, I was working on that with a very senior, uh, analog electrical engineer who knows an awful lot about power supply design and an awful lot about audio electronics. I liked to think of my job on that project is basically doing as much electrical engineering as I could to let him focus on the power supply and the amplifier. Cause those were the stars of this show.

Chris Gammell: Oh yeah. Right. Right. Yeah. And I mean that you need to have a good debounced power switch or cap touch or whatever's got like it just needs to have everything else that makes it a real product. But like you said, the sound path has to be flawless.

Nash Reilly: Yeah. A hundred percent. And like, and that, uh, and that was something that I got to learn a whole bunch about class D amplifiers for through just because, um, at that point we were designing, we're designing a class D amplifier in, in that product that is so powerful and is so large that it's hard to find a, an integrated solution for it. So to get the audio quality that we wanted in the space that we wanted without it getting so hot that it caught on fire, we basically had to go with a discrete class D amp design. And that offered a whole, whole lot of challenges.

Chris Gammell: Um, just clarify that you mean that you have a non monolithic chip. That's non integrated fats, non, but not even what would like, what is, what is the non?

Nash Reilly: Oh yeah. You, you, you got it. You're selecting, you're selecting individual fats for your output drive stage. You're designing custom magnetics for your output filters. Uh, you're getting, uh, you're getting, uh, you're getting, um, uh, like charge pumped drivers to switch your switching stage quickly.

Chris Gammell: So basically all the things that people usually want to avoid doing when they're buying class D amps off the shelf was like, I just want to chip that hook a speaker to it with maybe a filter in between. Yeah. Yeah.

Nash Reilly: This, this was like that blown up and bigger and with more power. Um, and that was actually a fun project because the way I fit into that was, so one of the nice things that, uh, monolithic integrated class D amp chips have is they have protections built into them. Um, they will typically switch themselves off. If they get too hot, they will typically shut off the outputs. If they start sinking too much current, um, in this discrete design, we didn't have that opportunity. Uh, so my role in that was basically taking this little microcontroller to do a bunch of real time control and basically making that whole discrete system look to software, like a packaged IC. And that was, that was like a wild journey down the mixed signal design rabbit hole. Um, like that was a chance to do a lot of firmware, uh, a lot of low level C coding, uh, a lot of like software architecture work, which I never, ever thought I would be doing kind of like, designing, like writing design documents and design specs for like this bare metal firmware program. And then I actually bit off a lot more than I could chew with that project. I had to end up asking for help later on in the project and getting some assistance from another really talented firmware engineer and a couple of other very talented, uh, signal processing engineers to kind of get us over the finish line for that. Um, but it was, it was, it was, it was, I was like, again, I keep saying I'm very fortunate in this case. Um, the, the guy, uh, the guy who actually did the bulk of, of that, like talented firmware stuff is actually on his honeymoon right now. So Jake, if you listen to the amp hour, thank you. It was, it was all you, man. Yeah. Thanks, Jake.

Chris Gammell: Um, so, so when you say, you know, when you say digital design too, is that what you usually mean? You mean you're putting in micros and control elements or like, what is the, I guess, what is the smarts inside of this thing as well? Like, so, you know, you, you mentioned the class D amp that, that needed some, some low level control. You mentioned, uh, you know, the DAC and stuff like that as well, but is there like a main, like a higher, higher level, like Linux kind of system that's driving this all? Or is it, you know, our toss in there on a larger chip as well? Like what else, what else is in there that's on the digital realm?

Nash Reilly: Yeah, that's a great question. Uh, at the heart, all Sims products are, are tiny little Linux computers. They're, they're basically just, uh, relatively beefy Linux machines with class D amplifiers and high quality power supplies built into them. I mean, like that's a radical oversimplification, but. Sure, sure.

Chris Gammell: But architecturally that makes sense, right? There's, there's memory management. There's, you, you have people that are writing software that can access these things. And I guess if they're running, they're on networks anyway. So that makes a lot of sense from a security perspective and a overhead perspective and everything else.

Nash Reilly: And in this particular case, talking about this like particular mixed signal project, um, we had to go with that solution just because, um, Linux is terrible at doing real time stuff. Yeah, that's right. Yep. It, like it, it can't respond quickly enough. It can't do, it can't make any sort of time guarantees about when it does stuff. So you need something, if something in your system could fail by accidentally sinking 10 amps through an output connection into a, into like a sneak ground path, you need something that's going to shut that off post haste.

Chris Gammell: Uh, and wait, wait, wait, I've got, I've got a wifi packet here. I've got a process. So I'll get to you in a second thing that's on fire. Yeah, exactly.

Nash Reilly: So you need something that's a little faster and maybe not, maybe not as, uh, uh, shall we say higher order of functioning as a Linux computer, but we'll get the job done when they say they're going to get the job done. If you will.

Chris Gammell: Does that mean that you then, uh, so you said it's kind of a standardized Linux computer internally. Do you have like standardized interfaces then to get down to a lower level thing, like a, over like a serial serial bus or something like that to talk to? Like, I assume like a proprietary, like Sonos based communication method.

Nash Reilly: Yeah. So, um, so your original question was like, uh, what is, what does digital design mean in the context of my job? So basically I'm, I am the owner of the digital subsystem in one of our products. So I'm responsible typically for designing the electrical layout of that computer. Like I'm, I'm responsible for writing test plans on the memory interfaces and the PCIe interfaces and the ethernet interfaces. And then all of the little interfaces like I2C, SPI, UART, what, what not that go to all the other little, uh, subsystems that we have in the product and making sure those are good. Um, it also comes into like power integrity and power stability. So I'm doing things like managing all of the low level DC DC conversions, making sure that we're getting the proper reset sequences and voltage conversion rights and, um, making sure that our noise tolerances are good and our, our bucks are stable. It's, it's, it's a pretty multifaceted role. Um, uh, and a lot of that, uh, a lot of that in the front end is actually working with PCB designers and like architecting like the fan out from a, from a chip and like the power routing and, uh, uh, placement of chips so that you're not going to be like, like stitching vias. Like, like you're not doing like ridiculous jumps between layers. Like you've got your DRAM interface fanning out in a sensible way and your PCIe interface fanning out in a sensible way and that sort of thing.

Chris Gammell: Yeah. So definitely like architectural, like low level decisions. Um, that sounds like it's, it sounds like it's very front heavy as well. Like the research and like you had mentioned talking to vendors as well and working with PCB and stuff like that. What is the usual like time to a first, to a first rev kind of thing?

Nash Reilly: Like, that's, that's a really good question. Um, so I've spent a lot of the first half of 2019 bringing up a new chip, a new ARM chip from, uh, from a vendor of ours. Um, I can't really go into the specifics of the, of the actual chip cause it's, uh, it's, it's under a non-disclosure, but I started talking to that vendor about Christmas 2018 and getting details on their chip and getting details on like the reference design and the interconnects available. And like the IO fabric and the IO controls and like the power sequence and the, the associated power management I see, uh, in January. And it was, it was a ton of back and forth and that, that ended up turning into a schematic in about April, which got in in May, which I brought up in June. So, and that I think was actually one of the faster ones, believe it or not. Yeah.

Chris Gammell: Well, and I think the other thing too, is like thinking that this is, you know, you guys are, you know, consumer electronics company too. And so like, that is faster, you need faster turnaround in general, I would think versus like an industrial where a two year might be okay. If not normal.

Nash Reilly: Yeah. And like, and, and this is, um, it's, it's definitely a different mindset for how you invest your employees time. Like, like you mentioned the industrial case, like, um, I don't think any industrial engineering firm, I don't know a ton about industrial applications, but I would say that it doesn't sound to me like any industrial electronics designer is going to go out and spin his own arm cortex, a nine processor design and his own board when he could just go buy a really tried and true one in a module form.

Chris Gammell: That's right. Yeah. I think that's right. And I think it's depends on volume, but usually because it is lower volume, because, you know, you're wanting to hand off some of that, um, you know, I guarantee there's people out that are doing that as well, but probably the most, having worked on a lot of industrial electronics, the, the more important thing is like, Hey, can I get that chip in 10 years? Because that's all I really care about, you know, like, and if I have to put it on my own board because of that, so be it. But, you know, like thinking about like the constraints being a little different there.

Nash Reilly: Yeah. And, and, and you have a different profile and, and that sort of thing for risk and reliability. Like, like we, like we design things to be Sonos designs electronics products to be very reliable and like really high quality electronics. Uh, but like if, if, if your wireless speaker fails, it doesn't have quite the same impact as like, say, uh, a nuclear reactor failing. Like, like I have friends who work in the pharmaceutical industry and if, if their industrial controller for their manufacturing goes down, they lose $25 million of precursor for medicine. Yeah. Like that'd be bad. Yeah.

Chris Gammell: The scale of, you know, someone's party might be ruined if a Sonos speaker, um, blanks out. So I'm just saying. Which is, which is terrible. That's tragic.

Nash Reilly: I, I, I don't, I hate to be the buzzkill at a party, but like, uh, I mean, anybody who's actually.

Chris Gammell: I don't think you're allowed to use that as an excuse though either. You can't be like, well, sorry, boss. I know that the design I did ended up, you know, not working out for all of the units we have in the field, but it's not a nuclear reactor. Yeah.

Nash Reilly: So are we cool? It's just a five millivolt RMS voltage transient just because you can hear a tiny little pop like that. No, that doesn't fly. Uh, quality is important wherever you go and in, in whatever your application, all the way from wireless speaker to nuclear reactor.

Chris Gammell: So how much then is the firmware aspect? And you mentioned you're writing some firmware and doing that, um, you know, stuff as well. But like, are you, are you writing stuff for the Linux computer?

Nash Reilly: Are you doing it more for the low level stuff like you mentioned? Or I, I typically, um, do not do much firmware on the day to day. Like I, it, it comes in cycles and it comes in waves. Like every now and again, I'll need to do some sort of evaluation on like a, like a amp reference design or some sort of chip reference design, like a sensor or something where generally like I, like I keep a couple of like microchip eval boards around. I keep a couple of like STM micro, like STM 32 eval boards around just when I need something that can drive an I2C interface or a SPI interface. Like, uh, that's the sort of coding I do generally shipping production code. Uh, Sonos samples is, is, is probably the one major example I have of that. Um, but day to day writing Linux code, not really my thing. We have a, a bunch of very talented embedded system, embedded software engineers at Sonos. And I'm very grateful to be able to leverage their support when I need it.

Chris Gammell: Yeah, that's cool. It sounds like you have a really well built out team. And, uh, you know, you'd mentioned there's one other product that you had worked on as well. What was the last one? Was it the, the move?

Nash Reilly: Oh, so, um, so I didn't work on the, so it's an interesting thing. So the Sonos move is, is our first, uh, is our first battery powered product. It's our, it's our first Bluetooth audio. I didn't realize that. Okay. Yeah. Yeah. It's the first thing you can actually stream Bluetooth audio to. Um, and it's our first battery powered product. And it's actually something that we've been working on for a very long time. Uh, and I, I, I've actually worked on an earlier incarnation of this, um, way before, way before it, way before its current form. Um, and where we were trying out a lot of the concepts that we were working with now and trying out a lot of the technologies in that. Um, unfortunately at that point in time, we decided as a company that it wasn't quite right for us then. And, and we couldn't get the experience of, of, of a wireless Sonos product and like a mobile Sonos product quite right. So that ended up being a canceled project, which was not fun and not, not a, not a fun experience for me to deal with. I was still, I was still pretty new to, uh, electronics design then. I'd never really had something that I'd worked on like so constantly. And so like thrown myself into just be canceled like that. That was, that was actually something I kind of had to learn to deal with. It was like, like, like practice a little separation and the kind of like, like let go of these things. It's not fully within your control. Cause it's, it's a little jarring to spend so much time trying to get something right on, on a technical angle and then, um, kind of just feel like the decisions handed down that like, Oh, no, this, this isn't really a, this isn't a thing anymore. You got to find something else to be useful on.

Chris Gammell: Yeah. I've, I've heard, uh, I think it was like a, you know, I, I'm a kind of a self-help junkie, but there was one I was, I was, uh, listening to like a, a talk about it. And someone was talking about like, you know, your product failed. You are not a failure, you know, that's kind of like a big, but like, because we internalize a lot of the things that we do, like personally, I, you know, I identify as an engineer. Like I tell people that I work on electronics and like, so then when something doesn't work, I take it pretty hard or if it gets canceled as well, you know, it's just like, yeah, that, that, that hurts, that hurts a lot. I've had, you know, gigs canceled and similar things and it, it sucks.

Nash Reilly: Yeah, it does suck. And like, I, and there are times where engineering is really hard. Like deadlines are real and, and the stress associated with them is, is real too. Like I've, I've learned so much about being a better engineer and like technical details and like learning about technology and learning the best practice and the right way to do things. But like, I feel like I've become a much better engineer just by learning how to, how to like take care of myself and like eating better and, and like going running and going to the gym. Like those things have benefits too, and they pay off. And, and like, I, like, there are some people that, that love engineering and do it all the time. And like, like you hear about the, the kind of meme of software developers going home and then like writing open source code. Like, I mean, I, I, I'm not quite there yet. Like I, I, I am, have spent the last month, like working my way through a chapter of the art of electronics. So it's not, that's not totally me, but like, I, I'm also a musician in my spare time. I play bluegrass music and like, I kind of need that to decouple from being an engineer every once in a while.

Chris Gammell: Yeah, that's good. And it still hits your, like the, the important parts of your brain. I mean, there's a huge crossover of musicians, musicians and engineers and stuff like that. I think it helps soothe the mind, but, um, that's awesome. Einstein was a fiddle player. Was he really? Oh, I didn't know that.

Nash Reilly: Yeah. Well, you can't call it a fiddle. You, you, uh, he was a violin player. You don't spill beer on a violin. That's a bluegrass joke. Sorry. Oh, is that, that's what a fiddle is when you've. Yeah, that. I get it. I get it. Yeah. I might be, I might be swinging for the wrong fence here.

Chris Gammell: No, that's okay. I mean, I listened to a lot of bluegrass music. I just don't know that. Yeah. I don't know the insides and the inside jokes and stuff. So the fiddle, the fiddle players get like the short end of the stick, like drummers always do. They're always like, uh, jokes about drummers.

Nash Reilly: Oh no, no. It's the banjo players that really get it. It's a banjo. Okay. Yeah. It's like, how do you, how do you achieve perfect pitch on a banjo? How's that? You throw it in the dumpster without hitting the rim. Sorry. That's good. No, that's good. Yeah. Cool. So what else should people know about you? Uh, I read a blog. Uh, I blog for fun.

Chris Gammell: Yeah. I like these.

Nash Reilly: I, I'm glad you like it. Uh, it's, if you, uh, if you want to read it, it's at cushychicken.github.io. Um, I'm on, I'm on, uh, I'm on Reddit and you can, uh, you can find my Reddit handle on my blog and send me strong opinions or, uh, comments or flame bait, uh, about my blog post there. Um, I should probably get an email handle for that blog post at some point so people can

Chris Gammell: send me an email there too. Twitter account too. That's usually a good one, you know.

Nash Reilly: Oh yeah. I am on Twitter, even though I, I haven't checked it on a year, but it's also at cushychicken because like you, like you said, people were really fighting over that Twitter handle. Yeah.

Chris Gammell: It's good. You got it early. You know, that's, that's important.

Nash Reilly: It was a struggle.

Chris Gammell: Well, Nash, thank you for, uh, thanks for being on here. I mean, this is a great look at, uh, audio and obviously the digital side of things and, uh, you know, just the system level design is, is, is fascinating. And especially at a, you know, a growing, having grown, I don't know where, Sonos is in the life cycle, but it's, uh, it's, uh, it seems like a household brand at this point and consumer level electronics that are, are pretty cool. So thanks for telling us about it.

Nash Reilly: Yeah. Thank you so much. It was great to talk to you. Same. We'll talk to you soon.