#476 – An Interview with Kendall Castor-Perry

Kendall Castor Perry: This is The Amp Hour Podcast. Released January 26th, 2020. Episode 476. An interview with Kendall Castor Perry.

Chris Gammell: Welcome to the Amp Hour. I'm Chris Gammell of Contextual Electronics. Hi, and I'm Kendall Castor Perry.

Dave Jones: They call me the filter wizard, though. I didn't invent that name.

Chris Gammell: Wow. I feel like if you did, though, we could have some kind of magical sound effect here or something like that. I don't know. We might put that in.

Dave Jones: Feel free to add that, yes. I mean, my daughter's a massive Harry Potter fan, so I'm sure when she was younger, she would have imagined me being in a kind of flowing robe with kind of arcane symbols over it. Of course, they wouldn't have been arcane at all. They would have just been transistors and op-amps and things. Right, right.

Chris Gammell: But to the layperson, then, you know, arcane, you know? Yeah. And so, actually, that's a great lead in here. I mean, filtering in general is kind of this, not, I mean, it is sort of Blackheart, but it's analog in general, and it's confusing for a lot of people. So, what is the genesis of being called the filter wizard, and what did you like working on that got you there?

Dave Jones: Oh, I mean, that's quite a difficult one. I mean, I've been doing filters my whole professional career. So, I mean, just as a kind of potted bio, when I was in the kind of that little gap period between the end of school, high school, senior school in the UK, and going to university, I'd already got my place at university. I was looking for a job. And because I've always been interested in audio and hi-fi, I looked around at the various hi-fi companies in my area and decided I was going to go and basically visit one of them, just basically wrap on the door and say, give me a job. I'm interested in hi-fi and I know a lot of this stuff already. So, I had to walk some distance to a particular bus stop to get a bus that went to the place where this hi-fi company was. And I walked past this storefront in a kind of parade of shops and, you know, nothing, you couldn't see much through the window, but there was a sign in the door saying, wire men wanted. You know, because I knew what a wire was, but I had no idea what a wire men wanted. And so, I was intrigued, you know, did the chin scratchy thing and went, well, I'm going to go in because if it involves wires. So, I went in and asked the secretary, I said, do you have any jobs for an aspiring young electronic engineer? And she kind of was a bit baffled, really. She said, well, I don't know, but I'll get Bill to ring you back. Give me your phone number. So, I did. And then I went on and I think I never actually even got to the hi-fi company. It's Rogers, the loudspeaker company, which was excellent in those days. And then Bill Meek, the guy, one of the founders of this company, Chemo, that I ended up working for, he rang me back and said, you know, tell me about yourself. And I said, you know, I've been interested in electronics since kind of I could barely walk and that sort of stuff. And he said, oh, that sounds interesting. Come in and have an interview. And so, I went in the next day to talk to them and it was the lunch hour. And there were all these guys that are kind of, you know, denim, long hair, and they were basically kind of checking out their home build stuff like amplifiers and electronic speedometers and stuff like that using their company's equipment. I thought, wow, fantastic. This is the place for me. You know, and the people that who run it said they explained that. They said, basically, yeah, we gave people to do anything they want and, you know, use any of the components they want, you know, as long as it's not crazy expensive or something that one of our customers actually gave to us basically because, you know, we want them to be interested in this kind of stuff. I thought, fantastic. So, basically, that's like catnip to a nerd, right? Yeah. Absolutely. It was sprinkled down. This catnip, electronic catnip was sprinkled down from above. And the long and the short of it is I stayed there for 21 years. Wow. So, basically, my first job was sweeping the floors and I ended up being the managing director.

Chris Gammell: So, did you ever find out what a wire man was?

Dave Jones: Yes. It's a person who does wiring. No, there we go. There we go. True to form. True to form. Yes. It's not a terrible insight. But no, and that was, and one of the main things that the company did was filters because it was the start of the kind of, I suppose, the kind of op-amp transistor era for filters and communications and things like that. And both of them had expertise, the guys who founded it, in filter design and kind of designing various kinds of industrial and scientific instrumentation. They decided to start out on their own. And so, a lot of the things the company made were filters. And some of them were active filters using op-amps, fairly early op-amps, kind of weird things you don't see that much. Some of them were passive filters, you know. So, we had old Ted, the guy who used to work for the British post office until he retired, with a fabulous winding machine. I'd call it Heath Robinson. I think you in America would call it Rube Goldberg. A kind of weird thing. It's like if you took a sewing machine and then rebuilt it inside out. That was kind of what it was. And he had, you know, shelves. And he did that. And he also selected the capacitors. He had an ancient capacitor measurement thing, the kind of thing that Michael Faraday probably would have used. And basically, because the company used to build things where you needed pretty precise inductors and pretty precisely known capacitors, he would get these boxes of capacitors and he'd measure their value and he'd write the value on the side in red Sharpie. So, basically, every capacitor in the place had gone through Ted and it's value written on it. So, I learned a lot about passive filtering and stuff like that. But basically, it was just a long, slow process. We had a fantastic library. We used to get regular, in the days well before the internet, we used to have a library service from the British Library. Basically, every week would come this thick sheaf of paper extracts. And basically, I could just get any of them I wanted. So, I just accumulated. It must be into the thousands of papers on filters. They had great books. And they said, if you ever see a book you like, buy it. Oh, man. That is the dream. That is the dream. It is. You know, and at the same time, I was using great components from, you know, people like National TI in the days when their data books and application handbooks were the gold standard for, you know, knowing anything. And I basically just, I lapped them up, really. I just covered the cover. They were my touchstones. And, you know, I mean, I'm in California now. But I remember in those days, you know, you'd see the names of where these companies were. And it would be places like Sunnyvale and Palo Alto. And in my mind, there was this kind of heaven of analog electronics. You know, when the sky is always blue, the sun always shines. The place names are fabulous. And, you know, there are super smart people who design op-hamps and then write books where they tell you how to use them. And I thought, that's just one day I'm going to go out there. I'm going to get out to California somehow.

Chris Gammell: Wow. And, yeah, and you did. I did, yeah, eventually. I'd love to hear about the interim, but I'd actually like to hear a little bit. Let's dive into the actual mechanism of filtering as well. I mean, so you mentioned passive filters. You mentioned active filters. And you mentioned, obviously, op-amps and active components are part of active filtering. But what is kind of your go-to? I mean, if someone comes up to you these days and says, how do I start learning about filters? Where do you usually point them?

Dave Jones: Wow. That's really difficult. I mean, in terms of books, I mean, I have some favorite books. They're all relatively dense in terms of simple books. I mean, I know a lot of people talk about Don Lancaster's active filter cookbook. And that is a great book, but that's really – it kind of sidesteps the understanding what's going on and just basically gives you a lot of clues about how you would make something if what you really want is just to make something.

Chris Gammell: Right. And if it goes wrong, then it's kind of you're on your own then, right?

Dave Jones: Yeah. In terms of understanding, and I talk to a lot of people, and it's very common these days that people kind of want to get that shortcut. They want to go – it's traveling without moving thing. They want to get to know how to do something without really putting the calories in. And I think you do need to have a certain amount of math and a certain amount of physics. You need to not be frightened of complex numbers for a start. A little bit of calculus doesn't hurt. Basic circuit theories, if you have the opportunity to learn that, that's quite a good idea. And I'm sure now with so many of the great educational establishments putting their material online, there are probably some great circuit theory courses that you can kind of sit through. And it does help. I mean, you know, I am a bit of a stickler for that. You know, there was a time when I was young when it was basically ready, fire, aim. And I was, you know, I would just kind of – and I'm still in a way a bit of an experimenter. I do believe in allowing a bit of creativity. But, you know, eventually you find out that you have to understand the reasons why something does the thing it does. Otherwise, you can't – you know, because if you don't understand where you are, you've got not much choice of understanding the shape of the surface of the space or whatever that you're in. So you don't know where – if you make us move, you don't know what's going to happen. And that always scared me a little bit. So anyway, back to the question. I mean, I think you could always start with a resistor and a capacitor, basically. Low-pass filtering and high-pass filtering. It's, I guess, a standard interview question. I guess you're going to get that when you go for an interview for a job where there is some analog thing. You need to have an intuition of, you know, if someone gives you a resistor and a capacitor and then, you know, you put a signal into it, a voltage signal or maybe a current signal, basically. What do the wave shapes look like? So in order to even really grok the idea of what a wave shape looks like, it really does help if you've got – learned how to use a scope. I think the scope just has to be the most important piece of equipment, you know, probably, you know, if you're talking about the 31 signs of a real electronics guy, basically one of them has to be you need to know your way around a scope and you need to understand what it is that it's showing you. So you need to be able to put in a waveform from a waveform generator and go, oh, I understand that. And then you need to put it through a circuit and the shape of it changes on the screen and you go, I kind of – I can understand what I'm seeing there and I can draw a graph or I can do a sum and go, yep, I know why that's happening. I think it's about building that kind of a core understanding because like architecture really and building foundations of things basically. If you haven't got good foundations, the thing you build on top of it will eventually fall over.

Chris Gammell: Right. Yeah, it is that intuition. It's intuition effectively. And then like you're saying, I mean, you don't have to dive right into the math but at least understanding that like, oh, okay, well, the corners are going away on that square wave and now why is that? It's like, oh, maybe I should look at some Fourier stuff. You probably wouldn't think that. Eventually you'll get there and then you'll look at the math and you'll look at the this and the that.

Dave Jones: So if you come to it from a physics point of view and you kind of – you just look at what a capacitor does and, you know, then you can work out that, oh, there's some exponential thing going on there basically. And, you know, without even really thinking about sine waves and S domain or Z domain or whatever, it's just in the time domain. The physics takes you there. I'm a physicist by training really. So I guess that kind of thing, the ability to represent something with a bunch of equations to kind of actually be able to relate some independent variable to some dependent variable comes naturally to me. And it's quite helpful to be like that. I think if you come to it without a – you know, if you skimmed over physics at school because you kind of thought it was too difficult or you didn't like the teacher or something like that, it's kind of a bit more difficult to reverse back into that space because you haven't got a mirror to look to see what you're going to hit. If you see what I mean.

Chris Gammell: You had said you were – I mean, so you were working at this company right out of high school. I mean, did that – at that point, did you have kind of the math knowledge or the physics knowledge or was that like you were kind of working your way into that as you were going into college and then you also had this interest on the side?

Dave Jones: Incrementally, yes. Yes, I did. I mean, I started out with electronics fairly early on. I mean, my parents owned slightly against their will a very big house. And, you know, in order to make ends meet, they rented out rooms in it. So, you know, I grew up in a house where there were always a lot of tenants, you know, who would stay for a few months or a few years or whatever. And, you know, they were usually quite young people and quite young people in those days had record players because they would like to listen to their young people music. And their record players would often break. And I'd go, oh, let me have a go at fixing that, you know, with variable success, I have to admit. But I learned, you know, I was interested in music and what went on in those and in the old televisions that we had lying around the place as well. My dad was an engineer and he was quite interested in electronics. He kind of learned his way into it. So the house was full of electronics magazines. So, you know, right from, you know, when I was, you know, nine or 10 or whatever, you know, I would read these electronics magazines and there wasn't an awful lot of theory in there. And that was really mostly, you know, trans, nice hand, you know, transistors and resistors and stuff building things. So building things was the way I started. But, you know, because I was quite inquisitive, I wanted to find out how it worked. And I'd always been interested in science. It's the way that, I don't know if my parents steered me that way or, you know, my books when I was young were kind of like chemistry books and physics books. And the kind of monthly magazines you would get would be things like knowledge and that sort of stuff. And I would fall asleep, you know, I used to have to have these kind of reinforced glasses because I've been wearing spectacles since I was about three for various reasons. And so I would always fall asleep reading some kind of magazine or chemistry book or something, even when I was quite young, and always break my glasses. So that was made my parents upset. But anyway, I was interested in science from a very early age. And when I made my choices at school, the kind of the, you know, middle school and upper school, it was the same school for me, but we had kind of the various chunks of different chunks of school. I always leaned towards science as soon as I could. So and it's a shame, really, because I was also very interested in chemistry. I did at one point make the kind of transition from chemistry because my first lab wasn't an electronics lab. It was a chemistry lab. I had, you know, pretty amazing kind of chemistry set with all sorts of kind of things. Some of them really rather too poisonous for a 10 or 11 year old to have. They don't do that anymore. They really don't. They don't allow that anymore. I think, you know, I remember taking those. I selected the stuff that I thought was the kind of most, the most dangerous, you know, phenol, the potassium oxalate, potassium cyanide, that kind of stuff, and took it to the local pharmacists and said, can you dispose of this really? Because I don't want it anymore. And they just roll their eyes and go, oh, another chemistry geek these days. I think you're probably going to rest. But anyway, I made the transition from chemistry to electronics. And, you know, it's one of the sadnesses in my life because I had a lovely chemistry teacher, Mr. Mallinson, a kind of short guy, really, you know, just about to retire. And I was one of his kind of star pupils because I was quite keen and enthusiastic and kind of got the basics of chemistry. And I gave up for, I gave it up in order to concentrate on the maths and the physics. So I kind of went into the maths sixth form at school rather than the sort of general science sixth form. And, you know, he was quite upset, really. But anyway, I decided that, you know, physics was the way to go because it was at the time more of a way. The physics and the maths was more of a way into understanding how stuff worked and, you know, was already quite interested in, you know, transistors. I remember in our, you know, we did English right up to the end because it was, you know, you had to have a little bit of liberal arts stuff in there. And so English often, one of the tasks was that everyone had to prepare a little lecture to give to people. So we had the experience of basically explaining a subject we really understood or really had an appreciation for to your classmates. And I did it on a common emitter transistor amplifier. I guess this was when I was probably about 15 or something like that.

Chris Gammell: How did that go over the class? I imagine there was some defundlement.

Dave Jones: I had some really tolerant friends and they were all, you know, they were all doing maths and physics as well, basically. So they were, you know, it was a smart cohort, basically. You know, I was really lucky and I'm still friends with quite a few of them. And we met for the new year not that long ago. You know, we're all quite old now, but, you know, we still keep in touch. So, yeah, I think they were tolerant. But, you know, the English teacher kind of asked some kind of very kind of tangential questions about it because he wasn't quite so keen. But anyway, so, you know, right from another thing I did was a friend of mine in another class was quite keen on synthesizers. And I, you know, liked, I was into kind of Emerson, Lake and Palmer and all these kind of things where, you know, sort of aggressively played synthesizers with great kind of sounds were the thing. And so he said, you know, can you make me a synthesizer? You're interested in that. So I made him a synthesizer. So I learned quite a lot of stuff. And really the whole genesis, I suppose, of the filter was my enthusiasm for filters. And this was before I went to work in a company that made them was the Robert Moogs, Bob Moogs ladder filter, the voltage control filter that he designed for his synthesizers. Because the simple way of making those kind of sharp, very characteristic sounds, you know, you take a very harmonic, rich waveform and then you put them through a voltage control filter with a waveform that's kind of programming its cutoff frequency or its resonance or something like that during the duration of the note made a sound which was unique. And I felt very captivating. And so that was the thing. I wanted to really make a synthesizer because I wanted to play with a voltage control filter. And I think that my love of filtered sounds was what led me to sort of never quite be able to get filters out of my mind. And so going to work for a company that made filters, it was kind of like, you know, Kismet or Bichette or whatever some people call it, that it was kind of meant to be.

Chris Gammell: Yeah. I mean, it sounds like your feet kind of took you there. Yes, exactly. Yeah. And so at chemo, like what were people doing with the filters? And so those were going to like studios or they're going into commercial devices or just audio setups?

Dave Jones: Not a lot of audio work. I mean, all the audio work was just with the stuff that people were making in their spare time. The filters, I mean, over the period of time, the stuff that I designed went into a lot of areas. But when I joined, they made some general purpose lab instruments. I mean, we always used to say that the first, you know, the first two things you need in the lab, three things, okay, are a scope, a signal generator and a multimeter. But the fourth thing you need is a variable filter, basically. Because if you've got a signal and you need to basically chop out or process stuff to get rid of noise or get rid of a low frequency thing, you need to have a bench programmable thing with a lot of knobs on the front that you can just adjust the frequency. Yeah. And there were various companies making that.

Chris Gammell: So like a decade box, but kind of for filtering.

Dave Jones: Exactly that. And they look very similar to decade boxes with the big knobs on the front. And the company was already making stuff like that. Ones with continuously variable wire-bound pots, ones with switches. And when you looked inside, these switches were multibank switches with resistors kind of soldered onto them, very, very, you know, intensively manually made things. So there was that. And there were also kind of modular filters that would go into scientific instrumentation and some like commercial communication stuff where, you know, that the requirement for filters seemed to move ahead faster than engineers were being taught how to do it. So I guess the company was on the leading edge of finding that there was a market for that kind of stuff because, you know, someone would come along and they'd say they'd realize, oh, they needed a kind of phase equalized ninth order elliptic filter. How the hell do you do that? So basically we were the answer to that question. Basically, when you buy it from us, basically we make it for you. We put it in a little pot and we kind of cover it with goop so you can't tell them what we've done. And we sell it to you for a sort of super high margin price. Yeah, that's great. You know, so there was a lot of that. And over time, you know, in the era before the widespread use of Delta Sigma ADCs, for instance, which tend basically not to have an awful lot of analog filtering, you know, the data organization systems that people use for, you know, measuring strain and stuff and processing the data that they would acquire from some test, like sort of testing a rocket engine or something like that would all be done with a sampling ATD converter. So it needed to have a front end, you know, with some gain and, you know, transducer interfacing and then anti-aliasing filtering. So in the days when you had to put anti-aliasing filtering in front of an ATD converter to get good results, you know, there was a substantial market for that. And I suppose a significant number of the things that we did over the course of the time that I was there were for that purpose, you know, in some way or other, you know, and this was, you know, sometimes in the thousands or tens of thousands of channels. This wasn't, you know, it ceased to become a kind of, oh, we've got this, you know, there's a lab down the road and they want five. It's kind of, okay, Siemens are building a body scanner and they want 15,000.

Chris Gammell: Oh, wow. Okay. So those kind of volumes too.

Dave Jones: Yeah.

Chris Gammell: So. Is this company still around?

Dave Jones: It is, you know, it's, there is, it exists. The guy who's running it is the guy I actually appointed as a marketing director back in 1997, I think. Oh, okay. During the, you know, when I was at the tail end of my time there. And there are, there's still a market for it. In fact, you know, we, we talk occasionally sometimes, you know, when he's doing a redesign of a board, you know, he'll, he'll want me to help him out a little bit. And I'm quite encouraged by the fact that one of the projects that's, that is still quite successful for him. And he still sells quite a few systems a year. It's something I started designing in 1988 and went into first production in 1990. So that's quite, that's pretty good. I mean, he's done, he's done some changes to it with, you know, with, if not sometimes my help, then certainly with my help afterwards. If you see what I mean. It's still a great, it was a great product then and it's still a great product. So I'm quite, you know, pleased that the stuff that I was doing back then, when I suppose was in the kind of peak of that kind of work, it's still, you know, there are still people who want it.

Chris Gammell: That's an interesting point you brought up too about the switch to Sigma Delta ADCs and how that, that ended up impacting things. I mean, do you see that broadly or was that something that was very specific to this scenario?

Dave Jones: Obviously you can't avoid it in audio and I can't think of anyone using sampling ADCs in any realistic kind of audio application anymore. And sampling DACs, if you'd like to look at it that way, continuous time DACs are really relegated to the super high end, a particular strand of audiophile super high end. And I mean, I'm quite plugged into that kind of thing because, you know, back in the day, around the turn of the millennium, I was the audio kind of digital audio specialist and audio products manager for Burr Brown at the time. I mean, that's my, you know, after my kind of 20 year career, actually kind of designing stuff, proper equipment, I made the transition into the chip business where I've been for the last 20 years. Um, and, uh, the, some of the most celebrated, uh, D to a converters still use the, um, kind of R2R, uh, the 24 bit PCM 1704, uh, kind of a collaborative, uh, um, Burr Brown, Japan and Burr Brown design, very old school, very expensive, very old technology. But, you know, there are still people out there who love it, but that's, that's the exception that proves the rule. I mean, these things probably run 20 or 30 bucks a pop now. Um, which is kind of out of the range of most, uh, um, systems, but you know, broadly speaking.

Chris Gammell: High end systems, but maybe not a consumer level device like a Alexa or Amazon Echo might not have it in it, but, uh, something else.

Dave Jones: So, you know, some sampling ADCs still have their place, uh, in a lot of things, but, you know, more and more, uh, in, in audio, in interfacing to sensors, you know, the, the understanding of how to get good performance out of a Delta Sigma, Sigma converter. It's so refined these days that it's the preeminent choice. And of course that they do have filtering in them, but that filtering is kind of there for a different reason. So there's usually very little in the way of filtering at the front end. Um, usually the only residual reason you need to filter at the front end of a Delta Sigma converter is if you're interested in the very lowest noise and you've got a very, very wide band front end, which, which produces noise at such a high frequency that even the super fast sampling of the Delta Sigma could fold it back into the base band. Then you will kind of slice it off, uh, with a little bit of, uh, RC filtering, but it's, it's not, you know, ultra brick wall kind of super sharp filtering, uh, like it was in the old days.

Chris Gammell: So where, so where do you see then? I mean, obviously, you know, we can keep moving through your career and it's, you know, you've been at chip companies, like you said, for 20, 20 years now, but where, where then, you know, so someone thinks filtering in 2020, what, what, what, what are a lot of the applications that you see that are out there? Like what are, what are people using active high, um, high cutoff kind of filters for these days?

Dave Jones: Well, I mean, I'll, I'll kind of just break that up into several pieces. Um, first of all, um, you think where I'm going with this, you might, you might want to chop out as a little bit of sort of chin scratchy stuff. I was going to make three separate points. Well, we can, we can just filter it, you know, like this is like, uh, editing is like word filtering, filtering or noise gating or whatever. Um, for, for a start, you know, an awful lot of the electronics that you find has got filtering in it, but it's not filtering that the customer wants. It's filtering that is inherent to the functionality that the customer is buying. So for instance, when you buy an audio A to D or D to A converter, there's usually quite a lot of filtering in there, but that's, it's, it, that's not what the customer didn't buy it. Cause there was a filter there. The filter is all built in. And so if it does its job, you kind of don't even need to know it's there. Uh, and that's really, um, people, you know, in the transition from analog filtering to digital filtering, people often say, well, what's better analog filtering or digital filtering? And kind of basically, and then when you make a product like at one stage we did at Cypress, which has had a digital filtering engine in, they would say, well, how do we sell this digital filter? Can you sell it instead of an analog filter? And the broad answer there is no, because the digital filter in a, in a product like that serves a different role. It serves a role of enabling you to complete a function that the customer wants, but the function that the customer wants doesn't, doesn't objectively contain the filtering. So it's like, for instance, if you want to make an A to D converter with a certain bandwidth, you need the filtering, but the customer, the customer expects that to be done. They don't want to have to fiddle around with the filtering, the filtering themselves. So broadly speaking, um, the, the filtering has gone inside chips in such a way that customers don't need to know it's there anymore. They don't need to build them. So the requirement on people who aren't chip designers or, or kind of really rot sleeves, rolled up system designers is a lot less than it used to. It's if you, if you look at a circuit board, it's very rare that you'll actually find something on it, which is a digital filter. It just takes a digital signal that exists on the board and outputs another digital signal that exists on the board, because that's what you wanted to do. Typically speaking, some other guy already built that into the product or wrote some code or downloaded a library from ARM or something like that. Put an FPGA or something. It magically happens and you don't have to think about it. And the other thing that, the other point I was going to make is that, uh, and I kind of have said it in the various little kind of bylines and bios that I put there is that my job, I think in the end of the day has always been to chase signals around. Because when you build a thing like this, so your, your, you know, your first step is simply to transduce some activity in the real world into an electrical signal somehow. And that's basically what all sensors do. People say something, what, you know, what's the difference between a sensor and a transducer? And really a transducer is the little piece of, uh, of, of kind of instantiated physics that turns something like a strain or a temperature or a kind of flux of particles or something like that into a signal that you can actually plug into something. So that's the first stage. So sensors contain transducers, but they often contain other stuff as well, like linearization or computation or whatever, um, and interfacing circuits. Um, but really it's the signal that's important. So audio is a really good example of that basically. So if you think about all the different ways in which the, the, me talking to you is represented to you, I mean, you know, it's made by my mouth. It goes across the air to this mic, lovely microphone that we're talking goes down. It's a USB microphone, so it has to get crunched by the, the, the laptop. And so then internally there's going to be some compression because they don't send it uncompressed over the, the, the internet. And it gets, you know, it's being mashed around, but it hasn't lost the fact that I'm going, you know, Peter Piper peck to pick of whatever. Um, so the thing that's important is the signal and that's true really not just for voice, but for almost anything. And so I've always said, you know, my job is to chase signals and, you know, they can't, they can't hide just by going from analog to digital to one bit to compressed to wherever basically. And I think in one of my bylines, I said, you know, think of me as the Van Helsing of, uh, electronic design. Basically I, I'm, I'm chasing the vampire of, uh, you know, the, the Nosferatu of information and, you know, it can change into a pile of rats or bats or whatever, but it isn't going to escape me because it's, that's what's important. And I mean, I know that's a bit colorful as a, as a statement, but the, you know, you have to, then, then when you start thinking, well, what are the ways that you represent a signal and basically filtering in the way of kind of selecting the difference between what you do want to what you don't want appears in different guises in different systems. And I've had over the years to become familiar with the ways in which you apply that kind of differentiation and decision-making, not just in the analog domain, but in the sample domain, in the digital domain. And, and, you know, even in the physical domain as well, thinking about microphones and loudspeakers. So there's all sorts of different domains where the signal exists and you have to do something to it that you could call filtering. It's not just op amps and transistors or DSPs.

Chris Gammell: I mean, it sounded like you were kind of on the border of like information theory type of stuff too there. Like where you're talking about these, you know, this information, well, you were saying signal, but I was thinking information because it's at the end of the day, you were saying Peter Piper picks and I, it had to get to my ear or our listeners ears. And that ultimately was the, that was the thing that needed to be transferred.

Dave Jones: That's the thing. Basically it's, it's not, uh, and a guy whose stuff I read on the, uh, um, on LinkedIn made a good point about, uh, his, his systems. He said, you know, the hardware is the bottle, but they kind of information or the software is the wine. Um, and you know, so the hard, it's, it's interesting studying the hardware, but hardware is always just a particular instantiation of something that does something to a signal, which is meaningful to you. And so there is an information thing. I mean, I don't spend much of my time thinking about kind of Shannon and his equations and what's the maximum information I can get through, because that's not typically relevant to really almost any of the things we do. I mean, I know that there are people doing, who will kind of squirm listening to that, you know, trying to get the maximum information throughput through a given channel. And one of the things that's, that's so fascinating about the modern, um, high speed interfaces between things is just how analog they become without people thinking about it. You know, you think about, uh, you know, DRAM, for instance, RAM memory. It's as digital as it gets right. Except once it gets out of the package, you have to send it somewhere else and you really need to know a lot about the kind of dynamics of, uh, of, of copper traces on FR4 and that kind of thing to understand that. And what's fascinating to me is that it's, it's a return almost to filter theory because that, you know, what's important is the frequency response, the time response, how that works. And if you, if you want to correct for it to equalize for it, all of a sudden you're working with almost with filter theory to try and fix up this connection between two entirely digital chips. So the information contained on there is essentially unknown to you or should be, if it's appropriately encrypted, basically it's just a bunch of ones and zeros or in the slightly more sophisticated case, a small number of, uh, of symbols for 16, maybe 64. But that's going across an analog channel and you need to understand the analog situation. And, you know, that's where a lot of the analog interest is at the moment. I mean, if you're, you know, looking for a job in the analog, uh, world, the thing that most people are looking for at the moment is what they call SERDES or, you know, the serializer deserializer, because, you know, people realize that the most coherent way of getting, you know, lots of data across was using a smaller number of lines as possible. So you go very fast, but you have to understand how this data goes across. And, you know, now that people are talking about things up to 112 gigabits per second, basically, this is like, uh, radio frequencies that are outside the experience of, of most radio designers going across a piece of copper trace on a circuit board as an analog signal. And you get, you know, it's amazing, the kind of rat holes that you can get into. I mean, you, you discover things like different board manufacturers use different roughnesses of the copper underneath that, you know, between the copper and the FR4 to get peel strength. And basically the way that the electromagnetic wave of those frequencies travels, it's exquisitely dependent on the smoothness of the underside of the track. So you can double, you can double the, you know, the loss effectively by having a surface that's too rough. And so you have the people now, they use material with a, you know, you sacrifice peel strength. So these tracks, basically, they almost come off if you blow too hard, but basically they're better at getting the fast signals across. It's just endlessly fascinating and, and all analog, really.

Chris Gammell: I'd like to go back a little bit, if that's okay. Because you were at Burr Brown, you were doing, I mean, like, so, so we were talking about these big, broad topics here, but what were some of the, the implications of these, like on an everyday basis, right? So you said, well, I, you know, I'm trying to help people figure out the solutions and stuff like that. It sounds like that means that there's a variety of, of methods to help people filter things. But what kind of chips did you end up designing or working on as a result of all this that, you know, I guess, because as technology was progressing as well?

Dave Jones: Well, I guess that the, the biggest job all the time that I was there, the time between joining Burr Brown and then Burr Brown being actually bought up by TI, it boiled down to an applications job, really, because people wanted to know how to get the best out of their system. And there were kind of several, sorry, several elements to, to that. One is that, you know, a lot of the applications are quite low end applications. And so they were serviced by quite low end products. And one of the places where a mixed signal designer of a, of a converter will often skimp is on the filtering. So, you know, you need, you need filtering both when you upsample to put stuff into a, a Delsig, D2A converter and down sample when you've finished converting analog with a modulator in an A to D. And that filter, you know, can often take up a fair amount of space and power on the chip. And so people are always looking for ways to make that kind of smaller. And so they make a value judgment about how good the end performance of the product needs to be in order to make a decision to, you know, priority call on how much space to allow that to take up. And sometimes the, the, the, the quality wasn't quite good enough for certain applications and you would end up getting some high frequency stuff out that you didn't want. And so all of these things typically needed a bit of analog filtering afterwards. And so one of the things that I would do, because the, you know, ad designers, you know, they understood a lot of things, but they didn't really know how to optimize the, the analog filtering back end. They did a, you know, a good job designing the, the converters with their internal filters, you know, cost optimized and therefore limited, but very predictable in performance. But the filters that often got designed onto the reference boards, which were the ones often that customers used, maybe didn't do quite the trick. And so I brought, you know, as soon as I joined, I brought some of that expert expertise, basically going, okay, if you use a slightly different topology, maybe add another, you know, another R and C here connected up like this. You know, you can make something that has a notch there, which is going to notch out that thing that you don't want. And overall make something which gave a better performance if people, when people were interested in the kind of the high frequency rubbish that was coming out. So some of it was that some of it was just basically guiding people to do a better job of the analog circuitry they had to add. And some of it was related to the architecture of D2A converters that people were using. And to this day, it's still kind of a major issue, I suppose, for most of my career, but certainly in terms of applications for you were talking about Burr-Brown, the D2A converters, an audio D2A, you know, produces an output which is equal to some digital fraction times a reference voltage. And overwhelmingly, that reference voltage is still the power supply. So you need to have a really, really good power supply. I mean, a fantastically good power supply if you want good performance. And I think that kind of escaped a lot of people's attentions. So most of the problems that I've had, and I have them still. I mean, only a few years ago, I had to work with someone making a made-for-ipod accessory. It was a guitar interface. They had a weird distortion problem, which turned out to be a power supply issue.

Chris Gammell: And they're using the power supply instead of like a reference, like a voltage reference?

Dave Jones: That's exactly it. That's basically, you have to get pretty much into kind of the audiophile range of things before the intrinsic power supply rejection of a D2A converter is actually independent, is actually good and independent of the local supplies. I mean, people are, you know, broadly wise to that kind of thing now. But still at the low end, you still find D2A architectures, which actually just use the supply as a reference. Because they figure, well, gain stability isn't really important at that application. So if the supply voltage goes from 3.3 to 3.4, who cares? But it's what happens if the supply is going from 3.3 to 3.31 at a particular frequency. This kind of thing is clear as anything when you go looking for it, provided you know what you're looking for. So, you know, that's the kind of thing, you know, in making an audio product much better in a way that the issue is discovered through measurement. But, you know, the people might say, well, you know, it doesn't actually measure bad, but I want to know why it's doing that slightly weird thing. And having a suspicion that that slightly weird thing, which doesn't necessarily look bad on the spec table, but you go, shouldn't be doing that. It turns out that it is something that can significantly affect the way that the product sounds. And I've had similar experiences with the different kinds of clocks that are used, for instance, in D2A converters and different architectures are differently sensitive to the quality of the clock, the level of jitter, whether the jitter is white noise or whether it's structural jitter coming from some clock synthesis process. You know, it's a real minefield. And, you know, I was solving those problems for people at Burr-Brown and TI. When I went to TI, I focused more on the digital amplifier stuff, which was the kind of early days of digital class D, if you want to call it that. And even now with, you know, audio, you know, most recently with the work that I was doing at Cypress, that the quality of the clocks that you had available was often the dominant driver of measurement performance. And also what the kind of, you might call them the golden ears people at various customers would think about the sound quality.

Chris Gammell: Yeah, it's interesting how much that does feed through over and over. And yet I still find myself like not checking the voltages. You know, it's just like one of those things where it's like you just assume, you assume that the power is going to be fine. Oh, the light's on. It's great. You know, but it's still, it's like, no, no, no. It's got to, it's got to be a nice clean power supply. It's got to be a nice, well, well apportioned clock and all these things like you're saying. And, you know, for some reason it should be the number one thing on a troubleshooting list. And yet it's, it, it still seems to fall down a little bit. I don't know why.

Dave Jones: Yes. I mean, gardeners have often said, you know, a penny for the plant, a shilling for the whole. I mean, a shilling is an old unit of currency in England. It's a bit like saying, you know, a penny, a penny for the plant, a nickel for the whole. And, and, you know, I've certainly met hi-fi designers. There are some hi-fi designers who just want to buy the, use the best stack available and use their own kind of creative smarts in pursuit of some imagined issue that they're going to fix and therefore achieve the holy grail of fantastic sound quality. But there are others who realize that, that really you can get fantastic sound quality out of the most humble components if you actually do your design right. And so that's another one of the kind of schisms or splits in, in, in audio design. You can either go very complex or you can go very simple, but carefully designed to actually be good at doing things that other people ignore. Power supplies is definitely one of those. And that's definitely a dividing factor, I think, in the, the quality stakes, you know, generally speaking. I mean, I can't, can't claim to have listened to every amplifier on the market, of course, but the ones that seem to do better aren't the ones where some guy has some fantastic idea for a new topology, but where someone paid really, really close attention to the power supplies, because if the power supplies are, you know, a bad, you know, a bad, a bad diet gives you an unhealthy person, really. And it's like, it's like that with electronics as well. So, you know, power is important.

Chris Gammell: Yeah. Well, and I think the other thing that I've, you know, had hammered into me over my much shorter career is like that the circuit board is a component, right? In your design and that the component and its drift is a component in your design and the, you know, the case and the heat and everything is a factor at the very least, if not a entirely beast, beast of its own. And I think what I'm hearing is that like, you know, when you are paying that close attention, for example, with the power supply, you're integrating all these other things that seem like they just should work, you know, these mental models of what these things are, you know, you have to pay much closer attention because of like you're saying, like you said with the roughness of a, obviously that was much higher speed, but like the roughness on a surges connector or the, you know, the, the weave of a FR4 or whatever, whatever the small thing that might end up impacting your design, but you got to really pay close attention to it.

Dave Jones: That's exactly it. I mean, that, that, that takes me, you know, more into the direction of stuff that I've been writing about more recently, but it's also the case that power integrity in general has become, you know, a significant issue as people are doing more and more at high speeds and at high powers with switching power supplies and whatever. And I've kind of been trying to bang the drum for that for, for, for many years really, because you're right in saying that they, the, the circuit, that the circuit board is a component, but you know, it's not only physically a component because of what it's made of, but the layout itself is a component and that, you know, certainly in the case of audio, you could take the same circuit and give it to two different designers. They'll build two different amplifiers and they will, they sometimes will measure differently. And often people will say that they sound different. It's the same circuit, even if you use the same components, the build, the build has an amazing effect and, you know, got sort of great city kind of audio design feel like Doug self. I don't know if you've had done an interview with Doug himself.

Chris Gammell: We've not interviewed him. Dave actually knows him. I'm like a host, Dave knows him. So, yeah.

Dave Jones: So, you know, and, you know, he, he, you know, very early on, you know, was able to demonstrate that, you know, you would look at a typical push pull amplifier, old school amplifier would need low value emitter resistors. And so the easiest way of making a small resistor that doesn't burn out readily is you get some resistance wire and you wind it into a kind of a coil. It's got some inductance. That's not intentional, but you wind it into a coil and you stick it in. And he was able to show that the, that kind of thing is coupled enough to the distorted currents that flow in the two halves of a push pull circuit, that it's a source of distortion. It's a distortion that's very difficult to get rid of as well. You're right. Right. So just little things like that. The physical instantiation is not what you see on the schematic diagram. And so I, you know, I've written a lot about the kind of myths of ground, you know, people assume that ground or naught bolts or something, they have the one, the one symbol for it. They assume it's a magical place where you can connect something to and all your problems go away.

Chris Gammell: The current disappears into this ether and then it just, you know, well, you don't have to worry about it anymore.

Dave Jones: Kirchhoff, yes, Kirchhoff be damned. And especially, you know, when, you know, when, when op-amp macro models that you get from, from manufacturers don't obey Kirchhoff's laws. That's the thing about Spice. I love Spice and I've LTE Spice always have done, but it's not bound by physics. You know, it's models don't, they're not thermodynamically accurate. Usually they say in the sense that they can do things that the things in the real world can't do for things like noise, noise and energy. And also the models that you get from manufacturers, even sometimes from linear tech, although they were usually good at that.

Chris Gammell: I always think about the, the, the, some of the, the sources, like the current source that's in there and stuff like that. It's like, oh, look, I, I have one mega amp of current flowing through this thing now. I don't think that, I don't think that's actually right.

Dave Jones: Yes. I mean, I, I don't, I, I make macro models where you just use them as kind of instantiations of math and that's fine because it is a floating point system. You can have the most wacky versions of numbers if you're deliberately building that and you don't mind. But when you're actually doing a macro model. And so I showed, I mean, I, there's a six part piece of mine on EDN, which is, it was a kind of a reprint of that on basically I was determined to demonstrate that it made a difference how you, what, what bypass capacitors you used on an op-amp circuit, you know, because there was always this kind of idea that bypass capacitors don't matter. You just need a capacitor and then the job's done. Or it's, it's just there to stop the amplifier from going unstable for various reasons. But apart from that, it doesn't really matter. So, you know, I worked through with simulations to show with actual simulations that it does actually make a difference. You can change the capacitor and it would change the, the output voltage. You know, you don't talk about small effects, but when you're in audio, the small effects are the only things that you've got left.

Chris Gammell: I've been tracking and learning, learning more about RF lately. And if there's one, one lesson that I've been learning is nothing is as it seems with, with, with components, every, every, every parasitic, every, everything really matters. And it seems like that could impact your.

Dave Jones: Yes. I mean, I'll come on to that when you said parasitic and I'll come on to that in a moment, but just to kind of wrap up on the, on the amplifier side of things, you know, an amplifier in a typical circuit, it's output pin is delivering current into, into your load. Right. But where did that current come from? And you get, there's plenty of macro models there on the market where the current doesn't actually come from the power supply pins. So you're, you're completely, you know, you can't do any simulation of what's happening at the power supply pins because the op amp is a magical source of current, you know, works as an op amp, but the current that it produces comes out of a magical nowhere. And that's no good. You can't. And so, you know, I would, you know, work through very carefully to filter out these models who, pardon the pun, that, that were not that, you know, if it doesn't, if it doesn't obey Kirchhoff's laws, then basically you can't take it seriously because it's not really a model of what's going on. If you're interested in a whole level system simulation. And so you, you, you say parasitic and that, and Paul Rayko is a great guy. I don't know if you've interviewed him. You probably have.

Chris Gammell: I know Paul, I actually haven't had him on the show yet, but yeah.

Dave Jones: Yeah, he's, he's good. And then, you know, and he wrote a piece, you know, I usually basically just agree with everything that he wrote, but I had a slight kind of amber light on the one thing he wrote recently about, you know, adding capacitors to things. I think it was about snubber capacitors. He was quite right about what he was describing. But the thing is that when you start thinking about adding components, you lose the, the, the systems thinking side, because really what's important is the interaction between every component that you've got. An adding component changes your complete system. It basically, it's not incremental. So I mean, it's, it's like chefs. Well, basically they'll, they'll kind of add something to the pot until they think it's right. One of my hobbies is mixology. I make cocktails and I've got a friend up in Seattle who does the same thing. And his approach basically is he just has a, you know, you have the mixing thing in front of him, loads of bottles on the list, and he'll just be tipping things in and dripping things in and stirring around until he gets to something he likes. And he's designed some fantastic cocktails that way. But the, the downside is that it's very difficult to keep track of what you did, which means it's almost impossible to replicate it. So it is the one perfect cocktail. It's like, you know, the, the, the joke that kills everyone, basically it's kind of, there's only one of them. You, so, you know, and then, so you have to put a lot of effort in. So I'm much more of a kind of, okay, I get the measuring cups out and the droppers and things like that. I will put everything in to a known recipe. And if it's good, it's good. And if it's not, well, there's a little bit of like, okay, I can. Right. You tweak instead of ripping it up. I can tweak, but I tweak with numbers. And so the thing about adding a capacitor, and I, and that's one of the things I posted on LinkedIn when I started my burst of, of, of stuff earlier last year was something, a presentation that I made at Cyprus, which is, you know, just the rhetorical question a capacitor can fix that. Right. But the thing is that the system already should have comprehended the fact that some component was needed there. Just adding something is not a, once you've done the system design is kind of, okay, what you're saying is the system design you had isn't right. So now you're going into empirical mode and you've lost the, the credibility of any of the analysis that you did, because just adding a capacitor to something changes everything. So really what you need to have done is in your system design go, okay, well, there's going to be a capacitor there and its value might be zero, but it might not be. So let's have it in the design and think what are the consequences of doing that? So I'm not a kind of anymore, the kind of person, maybe I suppose I used to be as a, as a youngster in, in the sort of thrall of the ability to fork find. And you've probably done this. You've got a circuit on the bench and the guy, the test engineer is basically going, it's oscillating. And you come up to it and you put your finger on something and it stops. You put your finger on it, it stops. You go, okay, well, you need a 2.2 path capacitor there. Job done. You wrap your hands together. You think I'm so smart. You know, that's why one day they'll pay me the big bucks. But you know, that's, you know, that might get you out of a hole that day. But if you don't realize why, then that's just, you know, you can't build, you know, credible electronics on that. You need to go away and go, why did that happen? What did I actually do? Was it, was it, I mean, did I put a pie of capacitors? I put some capacitors across that resistor, but actually I put some capacitors to ground on both of those nodes. And simulation is, is useful for that kind of thing. I must admit, you know, as well as experimentation, I know the great Bob Pease, much lamented Bob Pease was skeptical about, about spice and about how it could take you down the wrong direction. But, you know, I do think it's that, you know, it's like a calculator or, or X or a spreadsheet for engineers, basically, because if you're, you know, up late at night at home without a breadboard in front of you, and you want to know what would happen if I put a capacitor there is, you know, no substitute for, for building the circuit and going, okay, without capacitor with capacitor. That's interesting. Do I understand that? And if you do, that's good. If you don't, you then have to, you do a hand analysis. I'm a great fan of hand analysis as well, even though now you can get a project, um, programs like SAPWIN, S-A-P-W-I-N, which is a symbolic, uh, analyzer, which basically, if you basically draw a circuit with a component indices on it, and it'll give you the equation for the transfer function, which is really quite cool and saves you a lot of bother. But, you know, I, I grew up without that kind of thing. And so if in doubt, you draw the circuit. That's great. And Johnny will do the analysis. Yeah.

Chris Gammell: You know, in defense of Bob Peace, I remember reading, you know, I've read a lot of his, you know, what's all this spice anyways, and all his, his similar things. And I have his, his, his, some of his, a couple of his books. I think he always said, he laments the fact that, that youngsters like me would go straight to spice. And then instead it was more like what you're talking about anyways of, you know, you're, you're, you're, you have a model that in your mind, you want to try something and it's more convenient to switch it like that instead of going and soldering, desoldering, soldering, desoldering. You know, I think that was the essence of what he was writing about a lot of the times.

Dave Jones: I, in fairness, I think that's right. I mean, he has, I, I know people who don't so much care for him. I, you know, I think. Yeah, totally. Totally. Contributed to our, to our art. If nothing more than as a character, you know, like he. To read.

Chris Gammell: Yes. Him and, uh, him and Jim going with him the same week was, was pretty tough.

Dave Jones: Oh, that was, yeah. Yeah. That was terrible. Yeah. I, I, you know, spice is an important tool. You need to know how to use it. Spreadsheets. Excel is an important tool. You need how to use that. I, I, I use Excel and a whole hell of a lot in engineering. Really just as a kind of score. Well, uh, yeah. A score, a scorecard thing, basically. It's just a great way of, you know, seeing what's happening. I, you know, I think it's, uh, um, an essential tool you need to understand. You know, it really helps to be able to solve a problem in several different ways. So if I'm designing a filter, I like to have several different ways in which I can look at issues that, uh, it might present. And sometimes, you know, if you're, you've got your coefficients in Excel, you can do things like for instance, um, if you want to work out what's the maximum possible signal level out of something, you know, basically you, if you're an FIR filter, for instance, the, uh, the trick is, is you, you take the absolute value of the coefficients and add them all up. So in a, in a, in an FIR filter, if you, if you add all the coefficients together, you, you, you know, that is the, if you like the DC case, you get the DC gain. And if it's a filter that's designed to have unity gain at DC, you add all the coefficients together, the answer will be one. Um, but if you add all the inverse, if the, uh, absolute values of the coefficients together, you get an answer that's greater than one. And that is the peak value that what I call the pessimal signal is the opposite of optimal. The pessimal signal will give you. And that's just, if you, the signal is if you take the impulse response of the, of a, of a filter, uh, and just take the sign, the S I G N, the sign of that, of that, to make it all kind of chirp, a square wavy burst, if you like. And you put that into a filter that will peak at the maximum value, which is the, the, the sum of the absolute values. And it's greater than unity. So if you happen to encounter a signal that, that, and you, you're basically what you did with your filter is you said, well, I don't need to worry about overload because it can never go greater than, look at the frequency response. It doesn't go with one, but actually you'll find that you could, it could get up to two, for instance. So you need to be very careful about, uh, uh, understanding what the absolute worst case signal is you can put into a filter and design defensively around that. You know, if you're not using floating point, I mean, that's the people do floating point these days because it's so cheap. And basically the dynamic range is unbounded because, you know, you can have voltages up to, you know, three times, 10 times 10 to the three Oh eight volts, which is, but you know, that's cheating really. If you want to make, do a real, if you want to do a real thing and yeah, another thing. And I remember actually I went for a job interview, believe it or not, not that long ago. And one of the guys, uh, questions they asked me was basically, it was just simple stuff. It was the R and the R and C I mentioned, you know, going right back to what you need to know about filters. You take an R and a C. And so he spun it around and said, what happens if you put a square wave into this, the, um, capacitor, the CR circuit into a capacitor resistor down to ground. And one of the things that people don't really grok until you've seen it happening on a scope is that if you put a one volt peak to peak square wave into a, uh, an RC high pass, you know, it's quite easy to get two volts peak to peak out of it. If you hadn't taken that into account in your, uh, in your design, basically what it means is that your, um, circuit, your, your system can overload on a signal that you thought was safe. So you need to take that into account. And then the corollary of that is, uh, uh, you know, for anyone out there who's ever used the state variable filter, that's the standard, what's called the KHN, Kermin, Hulzman, Newcombe, uh, uh, state variable filter, three op amps. Uh, it's, it has three outputs, a high pass, band pass, and a low pass. They used to use it a lot. It's, it has some tremendous advantages if you don't mind spending a lot of op amps. Um, but one of the problems is that if you don't, if you're not looking at the high pass filter, if you're only taking the low pass output, what can happen is you put a square wave in, inappropriate square wave, the high pass output has this capability of doubling the, uh, or in fact, more than doubling, there's an equation, there's an expression for it, but you can, you know, it can get well more than two times if the circuit has actually, uh, uh, if it's a complex pole pair, um, you're never looking at that. And so what happens is that the high pass output clips. And if I, if the filter is clipping internally, it's not actually doing its job properly, but you can't see that because the output just still kind of looks all right. Uh, and so, and that's something that, uh, I had to fix in, in someone's design that involved, you know, some poor chap in Germany having to change 128, uh, when it would have been 394 resistors in a system. Basically we had to send him a lot of beer to do that basically because they were, they, they, the customer was experiencing an unexpected and weird error. And it turned out that the signal they were putting in was, um, clipping the high pass filter, which wasn't actually part of the output. And so that's one of the other things I've written about. I think possibly in one of my, um, electronic design pieces is basically, you know, when you're doing a simulation and it has got 10 op amps, plot the outputs of all 10 op amps, not just the one you want, because if you don't look, you're probably not going to notice. You may assume that something's okay, but you know, find, think about what the pessimal solution will be, uh, uh, and, uh, and see what's going on. It's really important. Otherwise this kind of unexamined, you know, it's like, uh, the tree falling in the forest with no one seeing, does it make a sound? And I can tell you that the, the, the op amp overloading in a circuit, you know, no one can be looking, but it really does make the equivalent. Right. Right.

Chris Gammell: Well, and I think, I think at the, at the end of the day, it's like, you're standing outside the forest and you're looking to see if all the trees are falling down. If, and if a crap ton of the ones in the middle are, you might actually care about that because, you know, they might just be falling the wrong way or something like that. You just don't know. Yeah. Yeah. It's yeah. And that is the, I mean, that is one of the big, big benefits of spice in the, in the anyways, you know, if you're simulating some larger system, it's, you could see all the intermediates, you could see all the weird thing that might be happening, you know, to, to make it akin to code. It's like, you're, you're seeing all of the, the problem states that you might not be exciting, but you, that are definitely there and you just don't happen to see on the outside.

Dave Jones: And that's a good way of putting it because, you know, spice just like, you know, you can treat Excel as a programming tool. I mean, I don't program in VBA, but I mean, I could do, and it's a great, it's a great tool, even if you can just express everything in, in, in, in formulae and just have a big mess of formulae there, but spice, you can do the same thing. You can have behavioral sources, which have quite complicated behavior. And I've used spice quite regularly as a design tool. It's a kind of de novo, you know, if I, if I have no idea how something's actually even going to be built, I can do, I can do the work.

Kendall Castor Perry: Yeah. It's very low risk.

Dave Jones: There are a lot of people who'd say, Oh, use map, use MATLAB or Simulink, you know, SILAB or, you know, something like that. But my personal tool has always been, well, I'll do it in spice because look at, I can build a little behavioral thing, which has my version of a, of an FIR filter, but I can also put op-amps on there. Basically I can start building some stuff. I can prototype using that. And it's not quite the same thing as soldering components onto the ball, but the advantage is I can put things in there that I don't have in my parts bin. I can make stuff up.

Chris Gammell: You're not, you're not resoldering 348 resistors either. Right. I mean, that's the downside to a hardware solution is it's pretty fixed in, you know, meat space or atom space, you know, it's like, it's tough to redo. Exactly.

Dave Jones: And I can also find out what the performance of a million of them is going to be with, you know, random variations. You know, I remember once that, you know, I had to, for various reasons, I had to interview twice at Cyprus. Oh, really? In order to, yes, it's a little too, that's too long a story for, for this. But I had, it wasn't a standup argument because I was sitting, but basically one of the guys who was there at the time, very, very smart guy, basically argued intensely against the need to have spice simulation for something like the kind of embedded micros we have. Basically he said, because it's so easy just to build what you want on the desk and look at it in front of you. And I kind of, I am not at one with that approach because for a start, you may not, if, you know, you may not know what you're looking for if you're a novice, so you can build it. But if all you do in your engineering life is build something and then basically do a combination of hacking the code and crossing your fingers until it works, that's not engineering. And also, you don't really know what's going to happen if you build a million of them, because, you know, you know, a thousand of them might not work and you can't get an indication of that by, you know, sitting at your bench. So, you know, I don't, I don't mind.

Chris Gammell: By the time you figure that out, it's, it's much, much too late. You've already built a million. Exactly.

Dave Jones: So, you know, I don't know how you could really design something with, without the ability to validate it very thoroughly. And it doesn't, you know, it doesn't have to take a lot of time. You know, I think if you, if you don't simulate because you don't think you know how, then you probably don't understand your circuit well enough. You should be able to express it in some form, which is amenable to having it crunched on a computer in some way. Otherwise, kind of, you're more just a kind of gifted amateur, really, that, that kind of dilettante approach, you know, and I speak from experience because I'm sure that I was more like that when I was young in the days before computers and stuff like that. As I said, I was quite ready, firing, you know, and was lucky to do something. But over time, I realized that that's not the way to do it.

Chris Gammell: Right. I think Adam Savage or someone says like a difference between science and messing around is writing it down. And, you know, if you add rigor and all the other experimental method in there too, I think that kind of gets towards it, right? You know, it's having some process and having some, some rigor really, really is what makes engineering a thing.

Dave Jones: I know. And it's kind of, you know, basically if you do something, that something has to align with your expectations because there's always two, two things could be wrong. Either the thing you're doing is wrong or your expectations are wrong. Yeah. Right. You know, the only time it really works is when you have an expectation, you do something and it actually, it aligns with it, you know, because if it doesn't align with it, it's like, you know, I'm a great, from a philosophical point of view, great believer in Karl Popper with basically, you know, when that, you know, effectively falsification or disproof is the only thing because you can do many, many things that, that, you know, agree with what you're doing. But the first thing you do that disagrees with what you're doing, you know, that difference is the important thing. Basically, you can disprove something in that way, in a way that you can't prove it. So, and I think, you know, that kind of thing's, you know, lost to a lot of quasi-scientific thinking that is kind of going around the world on a lot of subjects at the moment. You know, experimentation is important and, you know, you need, if you do the experiment and you're completely happy with the, you know, that you did it right and that's the result and it doesn't correspond to your theory, then your theory is wrong.

Kendall Castor Perry: That's right.

Dave Jones: Yeah. End of story. Right, right. And I think that, you know, you have to, that's why you have to be prepared to be wrong. You absolutely need to be prepared to be wrong. You know, you can't go through your engineering life hoping that everything you do matches, you know, because that's, you know, you're not doing anything new. You know, one of the things that my dad used to say to me is, you know, the man who's never made a mistake has never made anything.

Speaker ?: Yeah.

Chris Gammell: That's a great phrase. Yeah.

Dave Jones: And I saw, I think, was it, I was reading Eric Bogartin's 20, um, sort of rules for engineers or something was linked to, in LinkedIn, you know, one of them is basically, you know, an expert is someone who's made all the mistakes or something like that.

Chris Gammell: Yeah. Right. Seen everything.

Dave Jones: And it's true. You know, I'm, so, you know, I remember one of my, I was a industrial supervisor for someone's PhD program while they were working for me back at the, in the chemo days. Uh, and, uh, you know, we, uh, got into a disagreement about something and I was pretty sure that something went a certain way and he was sure it went a different way. And so he thought about it again, you know, uh, and eventually, you know, he, he made some mistake, you know, in his hand analysis and what actually came out was what my intuition said. And he said, well, thank goodness for that. I was like, I'm the boy, you know, I couldn't imagine that you were wrong, you know, and I thought that was, and I was kind of a little bit concerned about that really, because you don't want people to, to, you know, assume that, you know, you swing by their desk and you say something and it's always right. Cause that's kind of, that stops people from being, uh, you know, creative and, you know, thinking about things for themselves. It kind of, if you, you know, if we can't solve this problem, we'll get Kendall in and he'll fix it and then we'll move on. That's kind of, I've over, over the years kind of tried to get out of that situation because I, you know, I'm, I love mentoring people, but the whole point is to give them the tools they need to solve their own problem, not just to give them a channel to ask me how to solve it. That's what often happens on the, on the web. Basically, you know, as soon as I link with someone and he says, I've got this problem, can you, can you solve it? You know, as if I was kind of homework machine. So I really, I, you know, I spent a lot of my time at the moment kind of guiding people to kind of thinking of things a little bit more clearly, maybe getting the right kind of books or looking at the right things, but it's kind of just solving someone's circuit problems for them. It's not, it's not going to help anyone.

Chris Gammell: When I think in the age of Google too, it's kind of the expectation, Google stack exchange, all those things. It's like, well, I just need this, this thing, then I'll be done. So, well, yeah, but then what if it doesn't work? You're kind of screwed. So, yeah. I wanted to briefly touch on like the, I mean, there's a lot of types of filters out there and I'm sure that that's one thing you guide people towards is just like topologies and, and similar things. How do you, how do you go about guiding people towards the, you know, what feels like a wide, a wide field of ways to filter, but, you know, like differentiating between the, the Chevy chevs and everything that's out there.

Dave Jones: I guess it's, it's sort of like a, um, what I imagine because I haven't experienced is the personal shopper experience. If you're, you know, you want to get some help in a, in a fancy department store or something like that. And so you'll have a little interview with this person who's going to help you find the, um, the, the, the right shirt or tie or perfume or whatever like that. And they'll ask you questions, you know, in order to narrow in on your kind of personality type and all that kind of thing. And it's the same, I think when people ask that general question is you have to start digging into what it is, where are their kind of, uh, priorities, where are the boundaries? I mean, there's always kind of rudimentary spider diagram that you can, uh, um, pull out on, you know, power, power consumption, solution size, cost, uh, what's, what's the signal? What's your expectation? You know, uh, you know, I have genuinely had people who say, well, yeah, I want to, I want to filter. I want to filter the signal, but I don't want it to change the signal shape. And I'm going, well, that's impossible because basically the signal that you've got contains some stuff that in your mind you think is the right stuff and some stuff that in your mind you think is the wrong stuff. And now your ideal filter is the thing that's going to take the wrong stuff away, but inherently it's going to change the shape because if you take the wrong stuff away from the signal, it isn't going to be the same shape anymore. So whatever you do, if the filter is having any effect, it's changing the shape of the signal. So let's refine that. What do you actually mean you don't want to happen to the signal? And then you get into lots of things about, well, I don't want there to be any overshoot, even though they're maybe not putting any kind of a signal in that's going to provoke an overshoot. Or they'll say, well, I absolutely have to have a linear phase because then I'm not going to get any waveform distortion, but then they choose something which has got a roll off in frequency response. So you take away some of the harmonics and it changes the shape of the signal. So really the process of solving the right filter for people, first of all, is basically digging in what do they think is going to happen when they filter the signal? What for them, what does success look like? They're going to filter the signal, then they're going to do something else to it. And they've got another black box in there further down the chain that does something else with the signal. And maybe it tells them something useful, like power or the kind of second moment of some impulse. Or there's something ultimately they're going to need to know. But typically when people ask about filters, I often find that they haven't thought that bit through enough yet. So they don't really know what they want. It's not usually a case that all they really need is, say, an anti-aliasing filter. And sometimes when it is an anti-aliasing filter, I mean, for instance, I've had people who would say to me, oh, they don't want to use the kind of a sharp brick wall, to use the old-fashioned term, brick wall filter, because it's going to make the impulse response have more overshoots. And I'll point out that their system already has a brick wall filter in it. So adding another brick wall filter is completely invisible. It's not going to do anything at all to the impulse response. Because two brick wall filters together, provided they're kind of linear phase and fairly sharp, it basically just looks like another brick wall filter. Or they'll say, well, I've got the system bandwidth limited to 20 kilohertz by some sharp filter. And they're worried about the impact of a 250 kilohertz RC pole. They say, well, is that going to have some effect? And I go, well, no, it's not going to have any effect. Now, it might do something useful in your system. But what is it that you actually wanted to do? And actually persuading people to think about what they want to do often almost it defines what they need. And that often gets them there. In terms of the detail, I mean, people say, oh, you know a lot about active filters. Which topology should I use? Should I use multiple feedback? Or should I use selling the key? Or Norbert Flieger's lovely GIC-based thing, which is a great filter. Or the three op-amp per state variable. Or maybe I need to do something that's based on the very good properties of a doubly terminated ladder. And so I could use FDNRs. Or I could use LeapFrog. You know, but I often, those are people that just kind of, they're reeling off the stuff they read. You know? Right. It's kind of not.

Chris Gammell: There's like a laundry list. It's a laundry list. They have like a list, a checklist. Yeah.

Dave Jones: You know, and you know, if you're commercially in the business of making stuff, you have to make money. You don't want to spend more money than you have to. You know, someone said that an engineer is someone who can do for 10 cents what any damn fool could do for a dollar. And you can, you know, typically people don't want to spend much money. Uh, so it's okay. Well, you know, here's a single op amp thing. And I don't, I don't want to have any, I don't want to have any accurate components. I don't, I don't, my, I don't want to have any close tolerance components. So I say, well, okay, you can have, I don't know pay. I don't know pay. Yeah. You can have components that aren't close tolerance, but then if you make a million of these things, they're not all going to be the same. Is that all right? Um, and then you can go down and you say, okay, well, I mean, they're going to be all over the place up here at this frequency range, but that's not important. They're going to be very closely matched down here. Guess what? Simulation is a great way of doing that with Monte Carlo analysis. So sometimes you have to get past people's kind of, they've already decided something they don't want, but they haven't really thought it through. So, you know, it, it is just an organic process really. Um, uh, and, and at the end you come up with, you know, something hopefully which, which meets their goals. I mean, I did have one, one customer who was adamant that they wanted, he said, I've only got room and I've only got a space and a power for one op amp. So can you solve this problem? Which wasn't a, it wasn't a filter so much, although it was trans impedance amplifier with a varying input capacitance load because it was a monitoring a photo diode that was rattling backwards and forwards inside a disc drive. Um, but you know, eventually because it happened, this was at Burr Brown, we happened to have a nice fast dual decompensated dual amplifier in a small package. Each of whose amplifiers were very low current that I could meet the budget that he imagined he could only meet with a single op amp with two op amps and therefore design a much better circuit. And what I was then able to do was do an analysis of his system and build something, which is can we do a filter analysis such that I could position, uh, the, the capacity input capacitance as a variable and adjust the transfer function. So that as that capacitance varied, it had a, I feel like a parabolic error curve rather than a linear error curve. So the transfer function of the whole trans, the trans impedance amplifier varied a lot less because the thing was designed as a filter where you needed to have one parameter, which is one of the input capacities was variable. Um, but it couldn't, I couldn't have done it with a single amplifier needed to have this two amplifier design, but I had to break the customer's specification that he only wanted to have a single amplifier because what he really meant was he only wanted to have a single package. And also because the, was the amplifiers were low power amplifiers. I still met his power consumption requirements. So sometimes you have to, you hear what the customer says and you have to go, well, okay, I will infer from that what you need and I'll design something that fits what you need.

Chris Gammell: Um, well, speaking of hearing things, uh, I think we have to definitely call out the, probably the only time this will ever happen. Uh, the microphone that you were speaking into right now, you helped design one of the chips that at least was formerly designed into the microphone that we shipped to people.

Dave Jones: It's quite possible. Yes. The particular manufacturer of this that I'm talking to, I can't say for certain whether now several years on, they're still using it or whether they've moved to something cheaper or, or, or whatever, but this particular manufacturer, um, there was a time when, uh, uh, Cypress was very active in the USB audio space, particularly for Apple accessories. Um, and, uh, um, as was the case, we, you know, always had to try and make our part, the PSOC great, great part, um, in all its variants, um, do something that people wanted to do. And so, uh, I designed the, effectively the clock recovery system for early, early generations of what's called the synchronous mode of USB audio. And that's a mode where the peripheral needs to be able to create its own local audio clock effectively from timing information it receives from the, from the host, the USB host PC or, or, or, or whatever. Um, and the only information you ever get from the host in a, in a full speed, the USB is the one kilohertz, uh, the one millisecond packets. So basically what you have to do is you have to create something like a, um, 22.5792 megahertz clock. And the only thing you've got is one megahertz, one kilohertz. Uh, so people might say, well, why don't you multiply it up on a PLL? But the answer is you get rubbish because that's too much to multiply it. So I, you know, starting too slow. Right. So starting with the bits and pieces that we had available, uh, I designed a system which enabled the, the, uh, thing to exactly create the, you know, identically create the, uh, required clocks, um, on the assumption that at the other end, the, uh, uh, the 22.5792 megahertz, uh, clock would have been created the same way. I either PC only have one timing system. So that was what, what the downside of that is, of course, you're generating a local clock and we were doing it with a chip, which wasn't really designed to do that. So it had its shortcomings if you were doing ultra audiophile stuff, a little bit too much jitter for some people's cases. Right.

Chris Gammell: But we actually, I was going to say, we don't, we don't really ship, uh, ultra audiophile mics to our guests, unfortunately. Maybe someday.

Dave Jones: It would be fine for this. I mean, what they do, what, what is now used is the asynchronous mode. It's made, it suits people much better because now you use the smarts that are available in, in, in better USB hosts and better drivers and things like that, where the peripheral can say to the host, oh, hang on, um, slow down or speed up a bit. And so you occasionally get a packet with one, with one, two fewer, one, you know, one less or one extra, um, word in it basically. And so you throttle the link so you can have your own local clock that's as high quality as you like. So the audiophiles guys like that. So they can spend hundreds of dollars on a super, super low phase noise clock for their D to A converter. And basically the, the, the USB interface basically monitors the, uh, some, some pointer, you know, sort of lagging leading thing. And we'll occasionally tell the host to speed up or slow down, uh, incrementally. And that, you know, so that's a much higher quality. And it turned out that, uh, you know, we could, we could do that solution with our, with our chip. But by the time we implemented that on a, on a subsequent chip, um, you know, the market was already moving away from wired connections to wireless. So, you know, in a way I'm surprised that this microphone isn't already a wifi microphone. I'm sure that there are options for that where you don't even have to have any wires at all.

Chris Gammell: I don't trust wireless. No, me neither. Yeah. Nothing like a, nothing like a bit of copper.

Dave Jones: But there we are. Yeah. So that's the, you know, maybe there is still a little bit of me in this, in the, in the microphone in some way. So, you know, this thing about, you know, transfusing, tracing the signal, you know, at some point the signal gets chopped up into little pieces using a piece of circuitry that I designed. Yeah.

Chris Gammell: That's super cool. Well, Kendall, where can people find more of your articles and more about you and what you've been working on as the filter wizard?

Dave Jones: Well, if you want the fire hose approach, you can just Google my name. Uh, if you want to see some of the stuff that I've been sort of just basically fire hosing onto LinkedIn, you can go to my profile on LinkedIn, uh, and LinkedIn supports hashtags, something that a lot of people don't, uh, know. So if you just search for the hashtag filter wizard, so hash filter wizard, or one word, you, uh, because I try and remember to put that in all the pieces I put on LinkedIn, you'll find that there. Um, there's the stuff on planet analog. It's a growing list. You just go there, search for my name. You'll find my author page. Likewise, uh, EDN, um, electronic design have also gone through a bit of a kind of thing. You can find my pieces on electronic design, but there isn't an author page with an index right at the moment. But if you search maybe for the chronicles of ground as in GND, chronicles of GND, you'll find the kind of stuff, my musings on ground. Um, yeah. And so, yeah, there's, there's all, all sorts of stuff out there spread all over the place, not, not in one place. So, you know, you have to go searching for it.

Chris Gammell: Well, Kendall, thank you for being on the show, sharing your broad knowledge about audio and filtering and just, yeah, it's been great.

Dave Jones: I look forward to hearing it with all the kind of boring stuff chopped out. No, no, this has been great.

Chris Gammell: Well, thank you for being on the show. We'll talk to you soon.

Dave Jones: Yeah. Take care.