#560 – High End Audio with Remco Stoutjesdijk

Remco Stardustite: This is The Amp Hour Podcast. Released October 3rd, 2021. Episode 560. High-end audio with Remco Stardustite.

Chris Gammell: Welcome to the Amp Hour. I'm Chris Gammell of Contextual Electronics. And I'm Remco Stardustite of It's Only Audio. It's Only Audio, Remco. Do people say that? Is that kind of like how you get your client engagements?

Dave Jones: No, it sort of started as a joke. I started a blog site because I call myself a recovering audiophile. I used to be one of the people that Dave talks about, you know, the audio fools with the golden plugs and the cables made by midnight by virgins in the forest. And I got out of that by the grace of science and at least the university insisting on me proving what I think is right.

Chris Gammell: And then... Yeah, yeah. Subjective becomes objective.

Dave Jones: Yeah. And I think the worst thing I did at some point is I trolled myself into thinking that something is better and I screwed up the test and I was listening to the same thing twice. So you can definitely hear differences between two of the same thing.

Chris Gammell: So that was... Like you did a double blind and you were indeed blind. Is that the idea? Exactly.

Dave Jones: That's sort of my message to the audiophiles, guys. It's only audio. And that harks back to what a customer of mine said, you know, audio is quasi slow moving DC voltage. You know, if people are playing around with 100 gigahertz and we're playing up to 20 kilohertz, it's essentially DC. Yeah. And at the same time, it's sort of a message to customers that this is really the only thing I'm good at. So don't hire me for video or RF. It's not going to work.

Chris Gammell: It's interesting to me that there, I mean, like it does feel like a solved problem, but yet it's not at all. Right. I mean, obviously things keep getting smaller. They get, you know, there's noise cancellation. There's detection of different speakers in terms of some of the ML stuff that's happening. And it's just like, it's just getting more and more complex on top of all this, you know, the base level stuff that you're talking about. It was like, yeah, zero to 20 kilohertz. Okay, cool. But within there, there's a whole lot of stuff going on, especially because the old, you know, the human ear holes are pretty well tuned to certain things. So, yeah.

Dave Jones: Yeah, that's absolutely true. I mean, if you look at, you know, stereos from the 1970s, they're actually pretty good. And you would be asking a very fair question when you say, you know, why didn't we just stop there? It was working perfectly fine. Then, of course, along came digital and we all decided that this is the perfect sound forever. And we found out that that's not really true because even though Nyquist says you need to sample so and so fast to be able to recover 20 kilohertz perfectly, Nyquist didn't actually tell us how to do that. So there were some practical downsides to that. And then along came, of course, the internet and streaming. And now every device needs to have an app on your phone. And what's happening currently is that AI gets involved and very heavy DSP for voice recognition and assistant stuff. So, yeah, I completely agree. Every time I think, why am I doing this? This is a dead-end career choice. Some new development comes along and it's super interesting every time.

Chris Gammell: Yeah. Well, and you have taken it all the way down to the chips as well. So you've worked on stuff on chip, like DSP, like you're saying, analog, things like that. So how did that transition happen? I mean, it sounded like you were kind of at the tube level before your college days and then you worked into the silicon side of things. How was that transition into the silicon side of things?

Dave Jones: Yeah. So I literally learned electronics starting in, let's say, 1905. So the first things I built for myself were vacuum tube amplifiers. And that wasn't necessarily because I thought that was much better than transistors. But if you look at the circuit diagram, it literally is five components. So I could think through it before even doing any unique classes. And then later on, I made those better and better and better and started to see more and more the advantages of putting transistors everywhere just to stabilize a current source or make sure that the bias conditions were always perfect. And more and more, I started to understand what makes sound good and understand what people think is good. And these are not the same thing. So some people think something sounds great, but you can demonstrably prove that it sounds horrible. But if you understand why that happens, and I have a lot of, let's say, scar tissue on that, trying to find that out. And that's sort of how I got interested in DSP technology. But I was studying analog electronics at university. So I kind of figured out that maybe I made the wrong specialization in uni. And then I got a job offer from a company in Scotland, Wolfson Microelectronics, that I actually sent a letter to because I found a chip in a DVD player. And I thought, this is an interesting chip. It sounds very good. And I've never heard of this company. And they actually hired me and gave me sort of a dream position. Based on that letter. Is that right? Yeah, based on that letter.

Chris Gammell: That's awesome.

Dave Jones: I love it.

Chris Gammell: They're like, someone's heard of us. Let's hire them.

Dave Jones: It was quite interesting because I literally did two job interviews there. The first one was a standard, you're a uni graduate. And the guy literally said, I am going to ask you questions until I break you. And don't worry, I will break you. The question is just when. Jesus. Sounds like a supervillain. No, he's absolutely.

Chris Gammell: You will break. Talked about Mr. Bond.

Dave Jones: He's actually great. But it's just, you really want to test someone. So that interview took about two hours. And I was thoroughly broken after that. And then I thought, okay, that's over. And he says, now I'll take you to your second interview. And that was literally the CTO of the company. He says, yeah, your CV was sent around. And I found it quite interesting that, yeah, you're studying, you know, in uni and you're doing the standard stuff. But outside of that, you're building vacuum tube amplifiers and electrostatic loudspeakers. Just tell us something about that. And we just had a pleasant chat for two hours. And based on that, he essentially wrote a job description for me. Wow. In the advanced technology department where you get to define how a DSP chain looks like. And when you're tired of that, you can do some analog design. And when you're tired of that, you go back to digital. You talk to some customers. So yeah, that's. Wow. That was absolutely.

Chris Gammell: Another vote in the column of put your hobbies on your resume, huh? Oh, yeah. Put your projects on your resume. Yeah. Yeah, that's great. Absolutely.

Dave Jones: So yeah, I played with that for about four years. And that was a very tumultuous time because we made a chip that was supposed to go into the first iPhone. And of course, in the end, you know, there was a little mismatch between marketing and design where marketing talks to the customer and the customer says, we need to have at least so many specs, so and so. Marketing adds, you know, a factor of two because they think, you know, the engineers are going to screw it up. More is better, right? Yeah. But the engineers add their own safety margins to that. And the thing worked perfectly. But of course, due to all the safety margins, it ended up being 30% more expensive than Apple was willing to pay. Yep. Yep. But luckily, Samsung wanted it. So it's in a few Samsung smartphones. Yeah, that was quite interesting. So you see the industry, the IC industry from front to back. But the one thing that I found out there is that if you like playing with hardware, you should probably work in software because if you work in hardware, you play only with simulators. So you're sitting behind the screen all day and that's all you do.

Chris Gammell: Interesting. So explain the software piece then, because I wouldn't have expected that if you want to work on hardware, you. If you want to play with hardware, you work in software. What does that mean?

Dave Jones: Yeah. So in a chip design company and you are designing the chip, you sit with cadence all day and you simulate that chip. Sure. Yeah, that makes sense. Yep. It was basically pretty hard to, if you want to simulate a chip with 100,000 gates and you want to get a proper FFT out of it, you need to have at least one millisecond of data with a timestamp of, I think, one nanosecond. So simulators were probably working a week before we could get any useful performance feedback. And only the people who do applications in the chip company, they get to play with evaluation boards and put the chip on there. And they were even the ones that were using all the measurement equipment that I was drooling over. But whenever a chip came back from the fab, for instance, we were locked out of the lab because they knew damn well that if you let the engineers into the lab, the whole test program goes to the dogs. So yeah, only later on in the second revision of the chip, you get to actually put your hands on it. And yeah, of course, with my history of building stuff and making PCBs or entire amplifiers, I just want to put a probe on a pin and measure something. So just simulating it after about four years and you see exactly what works and not, and then the cycle starts repeating again, you're looking for something different. So that's when I got the internal proposal to go to apps and I decided that's a good idea, but I'll do it back on the mainland of Europe because at the same time, of course, you walk around long enough, you find somebody who wants to walk around with you and have a child with you. So then you want to be closer to family.

Chris Gammell: Great. Great. And so on the application side then, so you're like an applications engineer or like an internal position? Like what is, how does that work in a chip company like Wolfson?

Dave Jones: Back in Wolfson, I was in the design department and never got to work on the apps level. Oh, you switched companies when you went to applications.

Chris Gammell: I see. Oh, I missed that. Sorry. Yeah.

Dave Jones: There were a couple of things going on at the same time. So I wanted to have more hands on the hardware. I wanted to be closer to the family, not across the sea if anything happens with, you know, aging grandparents and stuff like that. Yeah, of course. And of course we wanted to start a family and then you want to be closer to your relatives anyway. So I did a brief, I think one and a half year period in Rome Semiconductor where we designed LED drivers. And there I was already more on the app side also just working with the boards. But I very quickly found out this, this is why my company is called It's Only Audio. Even though you can do it and you can implement the specs, I felt no passion for it. We got the LED driver into the current Mercedes C-Class and E-Class. And I felt like 0% proud of that. It's like, yeah, you can't listen to a blinking LED, even though management, management loves

Chris Gammell: I mean, if you drive it poorly, you can. I mean, if it's a bad switcher. That's true.

Dave Jones: That is true. Yeah. No, so very quickly after that, I got an offer to work on a company for a company that was called ASK Industries. And that's, that's a name that nobody has heard of, but I think a lot of people have heard of Bang & Ollopsen or Boormester car audio systems. And yeah.

Chris Gammell: Well, if they have enough money, they've heard it. Yeah. They take really expensive Ubers.

Dave Jones: Exactly. That was a little anecdote that Dieter Boormester, the founder of the company, his audio systems are actually hundreds of thousands of dollars. And he joked that the cheapest Boormester audio system is a Porsche Panamera and you get a free car to go along with it.

Chris Gammell: Yeah. That's a, yeah. And, and I mean, that's, and that's really interesting too. Like just that culture of like super, super high-end audio. I mean, like it's almost big. I think it's, some of it is tied to brand and stuff like that, but some of it is just the amount of things that are out in the world. Like you're basically paying a significant chunk of NRE to have one of, you know, a few stereos that are out there, that sort of thing. It's just, it's just really, really intense.

Dave Jones: There's, there's definitely the culture of exclusivity. And then if you see, for instance, the hardware that Boormester puts in a car is, is obviously not designed by Boormester. It's just a run of the mill thing. And the moment they put their label on it, it suddenly becomes a 7,000 euro upgrade.

Chris Gammell: Yeah.

Dave Jones: Yeah. Of course, for the home equipment, it's, it's even worse. I mean, in, in the high-end scene, there are probably two directions. So one of them is, is stuff that is really just over-engineered and designed to never break. And a class A amplifier that can output three kilowatts and basically lights up all your cables is just something that is really well designed. But there's a second stream of people who spend more time on polishing the front panels and making sure that the cables are sitting on sticks because God help you if the cable would lie on the floor. And in that respect, you pay a lot for the brand image.

Chris Gammell: Yeah.

Dave Jones: Right. I used to have a few customers like that who brought me stuff that was modified by some companies. And then when you, when you measure it, they already start looking at you, like, are you destroying the thing by putting a scope on it? And that's, that's when you realize that that's not a business I want to be in.

Chris Gammell: I'll have to pay my shaman to be, to come and re-bless this, uh, this cable setup.

Dave Jones: Yeah. And, and I think there's actually a lot of people who really believe what they're selling. Um, I wouldn't, yeah, I wouldn't accuse them of being snake oil salesmen. They, they actually believe that stuff, but they get, they get very uncomfortable. Of course, when you start asking them questions and, um, you know, I have a science background. And so at some point the skeptic part of me got the better and said, I'm, I'm not against, you know, spending an extraordinary amount of money on good equipment, but it has to be provably better. And if I find a way to make something for a 10th of the price that does 95% of what the expensive thing does, that for me is a big win. Yep. Totally.

Speaker ?: Yeah.

Chris Gammell: So on the, I mean, on audio inside of a car, is there anything that's special about like, does it need to pass load dump testing? Does it need to have any kind of like specialized output? Or, I mean, like what, what is the difference between a setup inside a car? Like you mentioned, uh, you know, very expensive setup inside of a very expensive car versus something that would be sitting on a shelf at home.

Dave Jones: So there's, there's, there's two big differences. So one, the first one is of course the, the reliability. So these things need to handle all the hot and cold conditions and, and extreme humidity and all that. Um, I think we had our amplifiers in a thermal chamber at, at 75 degrees centigrade, 80%, uh, relative humidity. And they were shaking them for three months nonstop. This is accelerated lifetime testing. And half of them come out with the, with the coating of powder where one of the coils completely disintegrated. And, and, and you need to really design over every corner. I remember that Audi gave us hell, um, because they put a car under a light that simulates the sun. And after about 36 hours, we had a fail and the development leader said, can you name a country where the sun shines 36 hours nonstop? But that wasn't relevant for Audi. So that's, yeah, the stuff is really built to, to be indestructible. That's, that's the reliability side. The other side is that in a car, you know exactly what your acoustic environment looks like. So yeah, of course you can have one, two, three, or four people in the car, but we even, we even know how many people are in the car by reading out the seat occupancy sensors. So the acoustics can be very finely tuned. If you sell somebody a loudspeaker, you don't know what room they're going to put it in. And in a car, you know that exactly. So you have the opportunity of correcting for, let's say a resonance because a tweeter is firing straight into a piece of glass, but also to direct the sound to the people where they are sitting. But in the beginning, we just said, okay, this is a situation where the driver is sitting alone in his car. So we optimize for the driver. And then when you find out that the passenger is sitting there, you widen the stereo image, or you, you move it a little bit further to the back when you know that people are sitting there. And I think in recent years, they've actually employed a lot of DSP to, to even make it possible to send different sounds to different occupants. So the driver gets the, the navigation instructions, the passenger can do a hands-free phone call and the kids in the back can listen to some Looney tunes without dad going crazy up in the front. I mean, that's, that's pretty cool.

Chris Gammell: And that's without headphones. You're saying that's without headphones. Yeah. Yeah. And so like when they do that sort of stuff, is that like just constructive, destructive interference?

Dave Jones: Yeah, there's this, there's some, uh, wave, uh, form synthesis, of course, and they have added speakers in the headrest to make sure that, that you can do it more locally. Cause otherwise the car is just too small to, to have all these waves die out. And yeah. So the move has been to more and more channels of lesser power. So it used to be old cars have two or four speakers and you put 50 Watts in each of them. And if it's a high-end car, you add a subwoofer. I think cars nowadays have upwards of 20 speakers and probably more. So you can, you can fire them from, from all directions and without, you know, having too wide dispersion, you can make very localized sounds.

Chris Gammell: Yeah. That's really cool. I mean, and did you have an estimate on how many are in like average cars these days, or is it really just variable based on the car itself?

Dave Jones: It's sort of a question of price. So the, uh, the Mercedes S-Class that we did, I think had 21 speakers in the last generation. Oh wow. The current generation that just came out was, was developed after I left. So I don't exactly know, but it's probably not going to be less. Yeah.

Chris Gammell: Yeah.

Dave Jones: And there were, um, derivatives of that system in, in Porsche and in Aston Martin, but we also did a fairly simple systems for instance, for, for Fiat. Um, we did the Beats audio system and I think the Jeep Renegade has, has the same system in there. I think that's an eight channel system and that is these days pretty much the poverty line. So anything less than that is just not acceptable as an audio system.

Chris Gammell: Interesting. I mean, I could put up with some pretty bad audio, I guess. Yeah. Well, I drive low in cars.

Dave Jones: People do spend a lot of time in cars and, and especially as people have smaller and smaller homes, it's, it's estimated that they do 80% of their music listening in the car. So it makes sense to spend some money on that. And also for the, uh, for the companies that do the audio tuning. So we just developed the hardware and the, the DSP modules. And then usually somebody from Bang & Olufsen or Burmester comes in and actually tunes the car. Um, and, and they put their, their little branding sound in there and their characteristics that they like.

Chris Gammell: But got it. They also, you're almost like setting up like a test stand for a Bang & Olufsen or a Burmester. Yeah. I'm saying his name's wrong. I'm sure. But, uh, yeah. So basically you're setting up like knobs that they could tweak effectively. Exactly. Yeah.

Dave Jones: We, we develop a tuning tool, which is things they can touch. We, we develop algorithms based on their requirements and our own ideas. And then we usually demonstrate like, well, we have a special algorithm that enhances the center channel where the voices come on and depending on what other channels are doing. So it just adds a little bit of gain when it decides it's a good idea. And when we demonstrate such an algorithm, of course, as a DSP engineer, you turn all the settings to the max and you make it super audible. And then when you give it to the acoustics people, you almost want to cry because they turn it all the way down to 0.01. That's right. Yeah. And they still think it makes a great difference. But yeah, these, these people are obviously very helpful in, in making sure that you don't go techno crazy. And then, you know, whenever in audio, you hear something that immediately impresses you, I can guarantee you that after two or three hours, you're going to get super bored of it. So a good audio system is just very unimpressive in the beginning. You go, yeah, nothing special.

Chris Gammell: And that's, that should be the way it is. Interesting. Yeah. I, like I said, I probably haven't been around high enough audio stuff to really care. Uh, what about like, so a car is obviously in a noisy environment as well. What about like adding in additional outside noise canceling? Is that something that's happening now? Yeah, that's definitely happening.

Dave Jones: That's, that's actually the reason I was recruited into that company because they wanted to develop noise cancellation. We obviously already do a very simple thing where we listen to the engine background noise and you either crank up the volume or you increase the compression a little bit to, to drown out the engine noise. The, the obvious next step is to do engine noise cancellation where you put sensors on the crankshaft and then you, you basically get a pulse back and you know that everything, um, that you hear on the microphone that is correlated to the engine is sound that you don't want to hear and you do know what you are playing back. So you can, you can make sure that that goes through and that's, that's already, let's say pretty accepted. For instance, if you buy an eight cylinder engine from Audi, it has this mode where it can turn off half of the cylinders. An engine with four disabled cylinders doesn't sound like a four cylinder engine. It sounds like a broken four cylinder engine. So they actually needed to remove a lot of harmonics and add in some others that are related to the engine crankshaft. And that technology, I think has been in cars already for a good 10 years, obviously starting at the high end, but it's now coming down to, to very simple, let's say compact cars and pretty much budget models.

Chris Gammell: Is that to like hit a fuel efficiency standards, things like that to, to go to four cylinders?

Dave Jones: Partly that the cylinder deactivation is definitely for fuel consumption. Then, and of course every brand does it differently. I mean, BMW has a very different way of, of doing the reduction of engine consumption. I think they, they stopped just adding the fuel to something and, and Audi is specifically turning them off. And one of them, one of these guys gave me a lecture of two hours and it went straight over my head.

Chris Gammell: And we all know how a Volkswagen does it too. Yeah.

Dave Jones: They turn on a debugger flag. That's right. That's right. Yeah. But the more recent noise cancellation is also a lot to do, especially since electric cars, of course, because engine noise cancellation is not an issue there. So now they're trying to do more broadband noise cancellation where you pick up the wind noise and the tire noise. And that has to do not only with just, you know, wanting to reduce the noise itself, but it's also a cost factor. So you can imagine that in, in a current car, there's about two to 300 kilos of sound dampening materials and that adds weight and it adds fuel consumption.

Chris Gammell: You're talking about like the rubberization around doors, things that are just sealing out wind and water and air and stuff like that.

Dave Jones: Yeah. And then under the engine bonnet, you have essentially this, this woolen blanket that just makes sure that the engine doesn't drone into the structure of the car. But if you can do that with, let's say four microphones and eight loudspeakers in the end, that is actually a lot cheaper than putting all this material in there. So there's definitely a cost saving factor there as well.

Chris Gammell: I'm just imagining like someone like, like driving my mom around and her being like, you know, Christopher, it's, it's so quiet in here, but there's a draft.

Dave Jones: You may, you may want to close the door, mom.

Chris Gammell: Right, right, right. Yeah. Yeah. No, it's interesting how that, I mean, yeah, that obviously a car is a huge system of, of inner, interwoven components, right? It's not just, just the stereo or just the ceiling or just the whatever. So it's, yeah, it's, you have to have a lot of interplay, I'm sure.

Dave Jones: Yeah. And of course, as soon as you have that technology, people start to think, you know, what, what else can I do with this? So I think pretty much every, every sports car today has a, has a, an eco drive button and then a normal and a sports setting. And that's when you play with the amount of noise cancellation where you, you turn it to zero for sport or even to negative where you pump extra engine sound into the cabinet, into the cabin.

Chris Gammell: Yeah.

Dave Jones: That's something that's, as soon as the technology becomes available, you know, the marketeers start going nuts and, and use it wherever they can.

Chris Gammell: Yeah. That's, that's a, that's a, I wouldn't have thought about the pumping it into the cabin, but that makes sense. You know, people want to, they want to be in that mode and they want to, they want to hear what's going on.

Dave Jones: Yeah. I mean, a lot of people, a lot of people say bad things about it, especially when you, when you buy a sports car, you, you want to hear the real engine and not some synthetic sound. But there is an example, for instance, where Lexus, they built these cars with the CVT drive chains or the, the one gearbox that changes ratio indefinitely, infinitely. And people find that to be really droning because the engine essentially runs at a constant RPM. So what's, what companies like that do is they, they literally simulate an engine with a gearbox. So the car goes. Yeah. And people know it's fake. You can tell them it's fake. You can, you can look up these reviews on YouTube. People say, yeah, I know it's fake, but I'm not turning it off because it's so much worse without it.

Chris Gammell: Yeah. Yeah. I mean, it, it does sound like in general, like audio has a very interwoven, like perception basis to it. Right. Like we talked about earlier about like what sounds better. One thing that might not empirically be better might sound better to someone and just, yeah, it's just based on taste and history and just so many human factors that are in there. It's really, really crazy.

Dave Jones: The question of course, in, in audio is do you want what is called high fi and high fidelity where if you put a loudspe, a microphone in front of the loudspeaker, you measure exactly the same waveform that went in, or do you want something that makes the listener happy? I'm fine with not reproducing perfectly as long as I know why I'm doing it. So what, what people are saying with, with playing vinyl and vacuum tube amplifiers, they always say you hear more of the music. And that's, that's one of these examples where I spent a lot of time figuring that out. And one of the things I learned is when you apply dynamic compression to audio, you would think that it starts to sound less dynamic because everything becomes equally loud in, in the limit case. But the perception of people is as soon as you start compressing audio, it sounds way more dynamic. That's what, what people tell you. You're like, no, I just made it far less dynamic. Right. Right. I can, I can hear them whispering better. Yeah, exactly. And if you think that through, it's actually pretty simple. So you're obviously not putting a concert amplifier into your living room where you can pump out 140 dB. But if you have a recording with a high dynamic range and your loudspeakers are turned up to a level where the maximum sound you can reproduce is, let's say 100 dB SPL. And you have things like a fridge in your kitchen or traffic going by, or just even, you know, a normal living room as a background noise. So it has a limited dynamic range that there are sounds that if you play them low enough, nobody will hear them. So then if you compress the audio to fit inside that window of playback, people will say, yeah, I start to hear a lot more. It starts to sound more dynamic. Yep.

Chris Gammell: Yep.

Dave Jones: The question is, is that hi-fi? Because you've distorted the sound, but you've made more audible of the original recording.

Chris Gammell: Right. People might just be saying, I want to hear all of it. Yeah. Right. I want to hear, like, I think about there's, there's this one Bowie record that I listened to that, like, they, it just wasn't, you know, it's before mass, before like remastering and it's, you know, you got really high dynamic range and there's some parts where you just have to turn it up. And it's like, I don't want to do that. I want to hear it like without having to touch anything. You know, that's what it comes down to. And it's just like, okay, well, that's not how they recorded this. Not how it was, you know?

Dave Jones: That's, that's how they get you at audio shows, right? They, they play super loud and you're like, wow, I can hear so many details. And you buy the same stuff. You put it in your living room. You turn it down to where your neighbor stops coming around to, to tell you, turn it down. And then you can only hear half of it. That's right.

Chris Gammell: Yeah. To go back to the, the car thing real quick. So what is the actual mechanism? So when you say there's cancellation happening, how do you actually remove like engine noise, for example, from an audio signal? Is it a transform? Is it just a true subtraction? Like if I'm looking at like a, a waveform and audacity or something like that, is it, is it like a copy paste or cut a cut and paste kind of thing? How does that actually work?

Dave Jones: It's based on an estimator. And what you do is you, you put either microphones or other sensors on the places where, where the noise is the loudest. So in cars, we are using accelerometers that are sitting on the car suspension that, that really pick up vibrations that we know will go into the chassis. And when they come out of the chassis inside the cabin, it will become audible. Or you just have microphones that pick up the wind noise. And what you're then doing is you are, you know, that there is a transfer function, an acoustic transfer function in real life from that spot where, where the noise is to the spot where your ears are. And you're trying to estimate that transfer function. And if you have that transfer function, you, you put your picked up noise through it and then you add a minus one. And, and if it then comes out at the same spot as where the listener is, those two will cancel each other out. The, the more accurate your estimation of the transfer function from the noise source to the listener is, the better your cancellation is.

Chris Gammell: Is it done in the frequency domain? Is it in the time domain? I guess I'm trying to visualize my DSP stuff from many, many years ago and it's very weak. So just thinking about like, how does it actually do the math in there? Is it like going through an FPGA or is it in a DSP chip, like a specialized DSP chip or how, how did, where, where does that actually happen?

Dave Jones: Well, you can, you can do it either in a DSP chip or an FPGA. There's essentially two parts to it. So there's the real time part where you say I'm picking up the noise. I'm running it through a certain filter with, with coefficients that we'll figure out later and then we'll play it back. That needs to happen very quickly because essentially you're picking up the noise somewhere and the real noise will travel at 340 meters a second to the listener. And you're essentially going at light speed, but you have a few latencies through an A to D, D to A and the DSP. In that time, you need to apply the filtering. Now, how you determine the coefficients of that filter, that is of course, for many different algorithms, that's, that's the secret of the whole cake. But essentially what you're doing is you're recording, you're looking at the sound at the disturbance point. You're looking at the sound that comes out at the, at the other end. So the, only the disturbing sound, not the wanted sound. And based on that difference, you, you run an iteration loop. So it's, it's usually a quite complicated matrix inversion. Um, what, one of the well-known algorithms is the so-called Levinson-Durbin recursion, but basically you're, it's all control theory in the end. So you're, you're trying to work that out. And from the differences, you're, you're doing a, uh, a gradient crawl back to optimize the, uh, minimize the, the error that comes out of the estimating filter. So essentially what you need is a microphone at the one side that picks up the noise and a microphone at the listener side that picks up your error. And then an algorithm that, that has a, a stable minimum that you can iterate towards.

Chris Gammell: Got it. Okay. So you're saying there's like a, there's a bang that happens in the engine compartment. And then you basically, so you actually then go and measure, do you try and isolate that bang inside how, how it like propagates through all the stuff in between the engine and the human's ears? And then you kind of back calculate what that is.

Dave Jones: So you're not actually reacting to specific noises. Okay. When, when you are, when you're setting up one of these ANC things and when you're tuning them, you're just playing noise. ANC is a active noise cancellation. Sorry. Active noise control. I think cancellation is a word that's, that creates a little bit too high expectations because even if you remove, if you remove 90% of the noise, it's only minus 20 dB. And that's not that impressive, but yeah, you're, you're, you're stimulating your noise microphone at that point to either broadband or a limited band. You put in a microphone on the other side that is measuring the, the, the error function and you, you try and find a stable setting for your, for your filters and you store that in there. And the next time people actually go driving, the filter starts from these coefficients and then has a limited range in, in which it can move to adapt itself a little bit. But the adaptation is, is prone to instability. So you have to do that very slowly. That's well, let's say that the audio transfer function from noise to your ears, it's usually fairly stable. So if, if you keep your windows closed and you're just driving, it might be speed dependent or something, but essentially it stays roughly the same. So that's why you, you, you pre, I don't want to say train because that sounds like AI, but you, you pre-tune the car for the rough transfer function. And then you, you give it an adaptive range where it can move to account for, for different, different conditions. So if somebody would be driving and opens the window, it might be that the ANC says, well, listen, I've been, I've been pre-tuned for a closed window and I can do only so much. I can't do much of that. Yeah, exactly. Right. Exactly.

Chris Gammell: Right. Yeah. You know, I think, I think if, you know, the inner marketer in you was really hyper active, you would slap a label on each microphone and you'd call it a machine. And then you'd say, we're doing machine learning. And then you charge three times as much, you know?

Dave Jones: I have a very bad track record of coming up with ideas and patents and then somebody else coming up with the name and actually getting twice the bonus for the name that I got for the patent itself.

Chris Gammell: Yeah. Yeah. The, the scourge of, of the engineering versus the marketing mindset.

Dave Jones: Yeah. You spend six months developing a DA converter thing. And then the guy who decided to call it the smart deck decides that he gets a 5k bonus.

Chris Gammell: Wow. Wow. Yeah. That's a great name too. I know. So that's really, no, it's really interesting though. I mean, it is, it, I like the characterization of it's all a control system, right? It is. I feel like that's a, maybe not easy to visualize, but it, it, it's like a feed forward at that point, you're measuring error, you're feeding it back into the system. What about instabilities that I always think of happening inside control systems? Does that, does that ever happen that you're, you know, you're basically feeding back too much and then it, it, it, it just goes to the rails?

Dave Jones: Yeah. I mean, that's, that's the entire trick of coming up with a stable system. I think everybody can look up, you know, this is roughly how an ANC algorithm works. And we implement the matrix inverter and the, the, the gradient algorithm. And then you find in practice that, yeah, you want to adapt as quick as possible, but if you adapt too quickly, the whole thing might go into a certain oscillation. So, um, yeah, I think in the beginning when we did that, it was, it was fairly pragmatic where you just say, okay, well, you know, your, your lambda coefficient that that's essentially a multiplier with the, uh, the error function. We set it to 0.001 and it's, it's suddenly stable. Yeah, that's great. But of course, you know, now it barely adapts. So that's where you get the real control system experts in and they come up with root loci and tau loci. And I, and I remember from uni that control systems was the course I hated the most until I got into spectral estimation and, uh, statistical, uh, signal processing. And then I'm like, oh, this actually has a purpose. And then it started to get interesting.

Chris Gammell: Yeah. I mean, I feel like this is, you know, like I talked to engineering students and they're like, you know, all this math and like, why do I need to do matrix math and stuff like that? And it's like, well, you know, you're going to need it at some point. And I feel like, you know, pointing at this kind of application, like this is a very, not that matrices aren't everywhere anyways, but I'm just saying that, you know, this is a very practical, very cool application of that sort of thing. You know, it's a building filters based on a lot of math. Uh, it's really neat.

Dave Jones: And I know that this is your soapbox, but I too, in uni, I struggled a lot with mathematics until about, I think, end of the third, beginning of the fourth year. And suddenly I saw applications of it. I'm like, oh, but this is actually cool. Yeah.

Chris Gammell: Yeah. I mean, if they could start with this, that would be a great that you're right. That is my soapbox. Yeah. But why don't they start with this? Why don't they show you the Mercedes not being whisper quiet? You know, that's just like, that's, that's a really cool application.

Dave Jones: I think not all students are the same. I spent a few years as a teacher's assistant teaching the basic courses in, in transistor design and op amp design. And in the beginning, I would explain it to people the way I understand it. You know, you take an op amp and the difference on the two input pins is always zero volts. And from there on, you work out the input current, the output current. And I found out that there were students who just didn't grasp it that way. They wanted to have the equations written out for them. And as long as the equations made sense, I said, yeah, I'm fine with it. So it takes all sorts, I guess. But yeah, the mathematics types, they have a very easy first two years in uni. And the people who were, you know, soldering the back of TVs when they were 12, they have it quite hard and they quite often drop out. And I think it's a loss.

Chris Gammell: Yeah. It is too bad. Well, how has this progressed then from here? So you, you left the company, the ASK, but you're still doing audio stuff. So now you do it. Were there other companies you're working for, or it was just like out on your own and building your own systems then?

Dave Jones: Well, yes and yes. So later on in ASK, I think our son was growing up and my wife is originally from Poland. So the logical thing would be to, to put him to school in Poland. And because that's, face it, the family of your wife is far more important than, than your own family at that point. And especially if you, if you want to have babysitters and stuff like that.

Chris Gammell: And I was going to say, you're not going to get a lot of Polish speaking around the child if they're, if you're living in Scotland, for instance, actually barely going to get English. Sorry about that. It's Scottish listeners.

Dave Jones: No, you, you find out, you'll, you'll figure out how far you get with Polish in Scotland. There's quite a lot of them.

Chris Gammell: Oh, that's interesting that there, I mean, there's a large Polish population in Scotland you're saying.

Dave Jones: Yeah. Yeah. That's great. And then it's, it's a very hard language to learn. I, I, I, yes, I was under the impression that I was a language miracle because I can speak Dutch and German and English. And I was studying Danish for a while because Denmark is the country of audio. I thought never hurts to learn that language. So I was under the impression I can pick up any language in, in two months. And when I tried Polish and found out that Slavic languages have nothing to do with regular grammar and they have different times and different words. It's a, it's still a nightmare, but.

Chris Gammell: It's actually really cool to, if people have never seen it, there's a language tree that shows like all the branching of different, uh, of different languages. It's so cool. Like how they all, and what was it? I think it's Lithuanian. There's one where it's just like, oh, Hungarian. It's like out on its own.

Dave Jones: Um, no, together with Finnish, Hungarian and Finnish are very similar. That's right.

Chris Gammell: Yeah. Or a wrong language as it's saying, I'll, I'll link a image, a really cool image here, but uh, yeah. Wow. Yeah.

Dave Jones: And the funny thing now is if, if we go to the Czech Republic or even if I listened to a Russian speaker, my wife will say, I understand nothing. And I'm like, how do you not understand this? It's clear what the guy is talking about. And she goes, well, I, I only hear two or three words in a sentence that I understand. I'm like, well, that's sort of the way I run around all day long. Yeah.

Chris Gammell: I've learned to deal with that. Yeah. Yeah. Yeah. Be comfortable in not knowing. Right.

Dave Jones: Yeah. And, and just smiling and nodding and then gauging the reaction and going, no, that I was not supposed to nod. Yes. I was not supposed to smile at that. Okay. Well, yeah.

Chris Gammell: It's a machine learning, I believe is the term. Yeah. Yeah, exactly. Yeah. Artificial, artificial intelligence. Exactly. Yeah.

Dave Jones: So the reason I decided to, to come here was partly family and also partly finding out that, that working in big companies would never really be my thing. And, um, the job I had in, in the audio company was the speed team leader. So you're, you're spending a lot of time on, uh, sickness forms and, and strategy meetings and all that. And at some point you just want to get your hands dirty again. And, um, I was reminded of what Steve Wozniak said, if you want to do great things, you know, work alone. I think Tesla said the same thing. So I decided if I work alone, I will apparently do great things. It turns out it doesn't work the other way automatically, but.

Chris Gammell: There, there may be a couple more ingredients in, in that cake mix you're saying. Yeah, exactly.

Dave Jones: It's a many to one mapping that it doesn't map the other way. But, um, yeah, I found a company in the Netherlands that was actually a, a spinoff out of my university and they said, well, we want to hire you, but we still have no office or anything. It was a startup. So it doesn't matter much where you work. And I said, well, then, then I'll, I would prefer to work from Poland so I can be close to the family. And I think that was about six years ago. And, um, I worked for them for a few years until they got on their own feet. And in the meantime, I, I got to know a few other companies. And again, there was one of them that essentially had very small electronics departments. And, you know, you, you come in and you go, why don't you just do this? Why don't you just do that? And they go, why don't you just work for us?

Chris Gammell: No, I did that. I am always curious in this, uh, you know, like the ecosystem side of things. So Poland, I'm not sure on the electronics side of things. I know one very fantastic firmware engineer, past guest, Jan Richter is also out in Poland, I believe, you know, software side of things, but. Yeah. We met through you actually through Twitter. Oh, great. Great. Yeah. But yeah. So first off, what is Poland like? And then what is the, what is the audio industry like more broadly as well?

Dave Jones: Yeah. So I definitely didn't come here because of the great network and infrastructure and all that. I think, uh, you'd be far better off in Germany.

Chris Gammell: Okay.

Dave Jones: And, and when I came here, I think six years ago, there wasn't that much going on, but Poland is growing like, like fire.

Chris Gammell: Oh, I'm sorry. I should also say past guests, Ant Micro is out in Poland, not in Warsaw, but there, I forget what city they're in. Ant Micro is, yeah, they, cause they're growing like crazy too. Uh, and, uh, yeah, they've been on the show a couple of times.

Dave Jones: Yeah. I think you find it a lot with, with former Soviet bloc company, uh, countries where back in the days, they used to put a lot of emphasis on physics and math in school. So these, these people, I think Polish programmers, for instance, are, are world known to be really good programmers. Yeah. And it's, it's really growing quickly now. So from my window, I can see a big tower where Samsung has a large office where they're doing the, the assistant voice recognition development. Oh, cool. And we have a, a Google campus now that, that a lot of startups are going to. So there's, there's a lot going on. There's, there's also a company that, uh, they're called seven sensing. They, they work on essentially only audio stuff for smart applications. They have a development center in Belgium and one in here and one in France. So it's a good feeling right now that these things are growing, are coming up. So, you know, if I would be without customers for a while, like I could always go there. Go knock at some doors. Yeah. It's a lot less scary than, than when I first came here. I mean, the first time you set up your own office as a, as a consultant, you sit there essentially having nonstop panic attacks. I've got a buffer of about two months. I have no customer that's interested. I, there's nobody that I know. And I'm, I'm going to be sleeping under a bridge in a month. And then you figure out like, well, you know, there's always somebody. And then of course, for a while, that's, that's a very good income source. Then they go away. And then the company comes around and over time you build up a little bit more of a buffer. And at this point I'm in it for about five years. And not only did I build up a buffer to, to, you know, get across the, the times with no customers, but I also built up a good library of IP stuff. So DSP algorithms, frameworks, the number of platforms that you know, and can work with. That's, that's just super useful. And it makes you far less anxious to talk to a customer and say, well, I can try this. Or in the beginning, I simply told the customer, yeah, I'm an expert at this and I'd never seen it. Then you know that for every hour that you can invoice to a customer, you need to work for 10 hours to actually teach yourself this stuff. So in the beginning, it was, it was really crazy.

Chris Gammell: Less so these days, hopefully.

Dave Jones: Yeah. Yeah. No, now you can really say, yeah, I have worked on this and you don't have to spend all night making sure that what you say is true.

Chris Gammell: Yeah. Yeah. That's great. And then, I mean, on the IP side of things, I mean, like, obviously I'm very interested in the consulting side of things. You're part of the consulting forum and a contributor there. How does that go then? So, I mean, you show up to someone's door and you say, Hey, look, I've got this IP library. Have at it. I mean, like, what does that, what does that go like?

Dave Jones: At the moment, of course, you have far less of these conferences and audio shows, but usually you meet people there and you go, well, what do you do? And who do you know? And it also goes through ex colleagues that worked on other companies and they go, oh, you know, if we want to build this little widget, Remco is the guy who can actually build it for you. And you go there and usually these people ask, you know, do you have something, you know, we need a DSP platform because our loudspeakers are analog and now we want them to be digital and have algorithms in it and talk to the internet, but we're not going to, you know, give you five years to develop that. So do you have something that's ready to go?

Chris Gammell: Do they ever, do they ever say anything like, can you digitally transform us? I feel like depending how far up in the business chain you go, that someone would say that eventually we need to do digital transformation.

Dave Jones: Yeah. Usually you have your little bullshit bingo book on you and if you hear synergy and transformation. Oh yeah. Yeah. Transformation and synergy. And as soon as you hear the word blockchain, you know, it's time to go.

Chris Gammell: Oh, well, time to go.

Dave Jones: Yeah. Or double, triple your prices, you know, why not? Yeah. But that's obviously having developed these things, most of the stuff you develop is, becomes the intellectual property of your customer. So you can't resell it, but if, if you have to develop it again, you know, it's going to take a fraction of the time. So whenever I'm, I'm not fully engaged with the customer, I now will develop my own stuff. Yeah. And it's just great to have a lot of that stuff ready to go. And in the past, I have missed, let's say certain customers because they said, well, we would love to hire you, but we need to have something that's ready to go immediately. So if you didn't have that stuff, a prototype, a general purpose DSP board with a library you wrote yourself, then you can easily say, well, I'll make a version for you and I'll adapt the board and you pay the NRE on the adaptation, but on the stuff that was pre-developed, you just buy a license and that's it.

Chris Gammell: Yep.

Dave Jones: Yep. Yep. And they're pretty, pretty okay with that usually. Well, I've had customers that were not okay with not having that. It doesn't automatically, if you have it, that they're ready to go, but it's, it's going to be interesting because I think the last audio show I went to was two years ago. So, um, we'll, we'll see how it goes this time when I say, well, if you want to do this, that algorithm is already developed and that thing is just ready to go.

Chris Gammell: Right. If it's not a, uh, the catalog is not just your time. It's now literally a, it's like a product catalog at that point. That is a, yeah, that's a very interesting. And I imagine at an audio show too, you could bring a sample, bring up, you know, demo, that sort of thing, and like actually show them at that time. And then you really start to get interest and it just changes the whole dynamic. I'm sure.

Dave Jones: Yeah. So I'm, I'm definitely hoping that we get back to that kind of stuff. The other way, of course, to demonstrate is to, to, you know, record product demonstrations on, on YouTube and things like that. So I have been working on that slowly just releasing some, some videos of, of hobby projects I did used to get used to, uh, producing videos and all that. And that's actually a lot of work in that.

Chris Gammell: Yes, it is. Yeah. Especially, yeah, to make it like kind of get where you expect it, you know, you see other people's videos and you expect it to look like, and it's like, oh yeah, that there's, there's a lot of other stuff going on. Yeah.

Dave Jones: I'm starting to appreciate how much work is in there. Yeah. Yeah. Yeah. There are people in the space who say they just go off the cuff and record something, but it's, it's not that easy.

Chris Gammell: Yeah. Yeah. You'd mentioned a DSP platform as well. So like, what does, what does that look like? Is that like a software framework or, I mean, what would you actually present? So like if, if someone listening is, is like, oh yeah, I need that. What does that actually end up being for them? Like a dashboard or a, I don't know.

Dave Jones: Well, usually it's, it's, it's a few circuit boards. So I think depending on how big your algorithm is, you're going to make a choice for a certain processor. So if you, if you need a lot of processing power and you're going to make a fairly low number of products, there is the well-known analog devices, shark DSP, which is essentially the, the, the world number one DSP. It's, it does 32 bit floating point operations in one clock cycle. But these things cost upwards of $20 in, in small quantities. So if, if you're going to be making a little widget or a loudspeaker that that's supposed to retail for a few hundred bucks, that, that just doesn't make sense. So I, I also developed platform that runs on a very simple Cortex M4. Oh, cool. I, I now have that at the point where we're doing 32 bit floating point audio as well. Of course, not, you know, in, in one clock cycle per instruction, but we still get a fairly impressive number of filtering operations done. So it's easily workable for an active loudspeaker or even, even let's say a noise cancellation system with up to four channels.

Chris Gammell: Yeah.

Dave Jones: That was also one of the considerations that I saw, you know, starting to be self-employed like all of these things are getting cheaper and cheaper and algorithms are getting more and more intensive. So there's, there's a natural crossover point where you can afford to develop fairly complex algorithms by yourself, either on an FPGA or a shark or a Cortex M4 M7 and really come up with something that, that has a very reasonable per product price.

Chris Gammell: And then it kind of just goes into everything you're saying, right? Yeah. Yeah. That's like a Shannon parks was on the show a couple of years ago now, I think. And Shannon was talking about, I think the same thing he was doing the phono preamp stuff and STM 32 Cortex M4. And I don't remember that. I think he was talking about developing some algorithmic stuff as well, but like, again, I, I, I kind of struggled with that. Like, so the, is it like just like a block of code? Like what does an algorithm actually look like on a Cortex M7 chip? Is it just, are there like standard libraries that you're developing? Is that kind of the idea? Yeah.

Dave Jones: Yeah. Well, I'm, I'm making very heavy use of the Cortex DMC's DSP library. So. Okay. There are, there are other companies, you might be surprised. There are other companies that, that sell, you know, consultancy in DSP services. And some of them have contributed and together with ARM, they developed, let's say the base library for the, for the Cortex DSPs. Whenever you're building a platform, the first thing you want to develop, let's say, is, is the bi-quad. So bi-quad is a universal filtering component with programmable coefficients. And for every platform, you want to have a very optimized version of that because your code is going to be 90% running that bi-quad and everything around it is just, you know, doing the housekeeping. So it's, it's grabbing DMA buffers from an A to D converter and sticking it in a ring buffer. Then the bi-quad picks up the ring buffer and puts it to another ring buffer. And then on the output side, you, you stuff the process data back into a deck. So the, the, the algorithms quite interestingly are just a set of C and H files that call on that library. And you can, you can put the CMSIS library either pre-compiled in, or you can get the source code and optimize it a little bit. But having, let's say a gain algorithm, a bi-quad algorithm, and very simple things like getting the square root of a signal or the power of a signal, you can use those as basic building blocks and say, well, my algorithm will amplify the sound. As long as a light sensor is on. And when the light sensor goes off, it decides that it's evening and it will compress the sound. And you can build all of that with those simple blocks. So, um, you read out the light sensor.

Chris Gammell: Yeah. That's a great demo too, right? That's like very demonstrable, like at a show, like you're saying.

Dave Jones: Exactly. Yeah. That was one of the things that's, uh, I think my wife was always complaining about when we're listening to movies in the evening, that either the explosions wake up your sleeping child or you can't hear the dialogue. Yes. Right. Came up with an algorithm that's either looks at the clock or at the light, or it's just manually programmable that in those conditions, it compresses the sound a lot further. And it's at some point it becomes so natural that nobody notices that this thing is running. That's perfect.

Chris Gammell: It tests whether or not the rock or Bruce Willis is in the movie. And then if it is, then it knows there's going to be explosions and then it cranks those down too. Yeah.

Dave Jones: That would be a good idea. Look up the IMDB database. Yeah. There you go. Yeah. This is going to be loud. Yeah.

Chris Gammell: Yeah. Yeah. You'd think there'd almost be like a rating system too for, for movies. Like how much, how likely is something to explode or where even like where in, like you could probably timestamp when explosions are going to happen.

Dave Jones: And then, I don't know, maybe what you see quite often, and it happens in the music industry quite a lot is that the producers start to think about who's going to listen to this stuff and, and recompress it. So a lot of pop music these days is so compressed because they, they understand that people are going to be listening to it on their phones in a, in a crowded bus. And I've, I've always maintained that's, you know, that kind of adaptation should happen on the client side, right? So you should distribute, you know, as highly dynamic and undistorted music as possible and let the playback device decide what the environment requires. So if you play it in, in a perfect room at a very high level, you wouldn't have to compress anything because the full dynamic range is available. And if it's, if it's been pre-compressed, it's really hard to uncompress something like that.

Chris Gammell: Yeah, I get it. And I know Kesha is a very, uh, actually talented artist on her own. Uh, but you know, her songs are basically like bleep blorps and, uh, you know, similar, you know, I'm, there's just so many songs where it's like, why even bother?

Dave Jones: Uh, you know? Yeah. I mean, it's just a crying shame. Um, and then obviously the, the, the intelligent DSP algorithms came out, you know, 10, 15 years after we started using digital audio in the 1980s, you wouldn't dream of putting a full DSP in an audio system. I mean, are you crazy? You can fly to the moon on a DSP, but now they're essentially free. So yeah, it's, it's a, it's a shame. I mean, there are definitely people who care about that. Um, there, there's, I think I triple E and an AES committee are working on loudness levels and loudness meters. And there's an entire website about the loudness war. But I think the, the realization that, you know, loudness adaptation should happen in the playback device. I'm, you know, skeptical whether that's ever going to happen, but that would be perfect.

Chris Gammell: Yeah. That'd be nice. Uh, as I get older, you know, my, my tender, my tender eardrums. Yes. Uh, so what about, uh, what about like custom chips then? So now as, as someone out on your own, I mean, are you being approached to like, then take these algorithms and then make them even faster by putting them onto Silicon? Or is it more not worth it? You can still get it. If you're pipeline enough, you can get it fast enough. Like what, what is the push towards custom Silicon?

Dave Jones: There is a push to custom Silicon for those things that are really hard real time. And then I, I mean on more than, you know, one sample basis, uh, even on, on a fast enough DSP between two samples of audio, you can do a thousand, uh, processing operations. But I worked for a customer that wants to digitize the output of a loudspeaker back to the input of the amplifier. And, um, at that point you have to do such low latency stuff that you need to divide, uh, need to develop a custom chip. And then another application.

Chris Gammell: Just for the, the numbers again, sorry. The, the numbers you're saying one sample. So there'd be like one over 48,000 or one over 44, 100. Is that, is that like the single sample you're talking about? Yeah. Okay.

Dave Jones: So that's, that's usually when you do straight feed forward stuff. Even, even if you have a loudspeaker model, those models are parameterized offline. And then, you know, okay, according to this sample and this sample and the history of that, we need to now need to change the gain by so-and-so, uh, that that's easy to do. I mean, the DSP that runs at 200 megahertz, even if it doesn't do one operation per clock cycle can do quite a lot, you know, in, in eight microseconds. I mean, that's, that's half a week in DSP terms.

Chris Gammell: Yeah. Right. Right.

Dave Jones: Yeah. But when you're doing really face sensitive stuff, then you need far, far quicker feedback. The other application field, for instance, if you would do an FPGA or a custom chip is if you want to do massively parallel stuff, that's all time aligned or even stuff like, uh, PTP time protocols over the ethernet, where you have two loudspeakers that are connected over an ethernet link and you want to play them exactly in sync, but you need to bounce time steps of each other. And that quite easily can end up in, in an FPGA like chip.

Chris Gammell: And that would be like, you're running through your house and you wanted it to sound like you had the exact same song playing in every single room in the house. Is that kind of the idea?

Dave Jones: Yeah. There are of course many network algorithms that just stream to every loudspeaker. Um, but they have no idea if there is any delay in that. So if you then want to play a stereo signal and your left loudspeaker is, let's say a few microseconds or even a millisecond behind the right loudspeaker, it's going to sound absolutely terrible for anything over one kilohertz essentially. So, um, to have any kind of stereo image, they, they really need to be time aligned.

Chris Gammell: I've heard that I think Disney world, they have like really ubiquitous audio. I don't know if that's like a specialized system there. I'm sure that they've spent a lot of money on that sort of thing, but I remember hearing about that where like, if you go to, you know, basically the sound is kind of the same everywhere and maybe it's doing something like that. I'm not sure.

Dave Jones: Yeah. There's the, there are a number of network protocols. I think the most well-known one is the Dante protocol. I think it runs over normal switches, so you don't need any hardware in your switches, but the, the loudspeakers have the, uh, the PTP protocols or the precision timestamp protocol and they, they synchronize themselves. So one of the customers that I work for, um, are big proponents of the Dante standard. And it's, it's quite amazing when you plug a single ethernet wire into a loudspeaker that's 64 digital audio channels can come out of that. Yeah. Yeah. That's nuts.

Chris Gammell: Yeah. I mean, and, and also like if it's a regular switch too, like I, is it actually like IP traffic? Cause like, I just think about collisions and. You know, like the non-deterministic nature of IP. Yeah.

Dave Jones: As far as I know, Dante is based on, on UDP. So, you know, you might miss a packet here or there. Okay. There are other protocols like AVB that actually require the, the PTP to be implemented at a switch level. So then you need an ethernet chip that can actually do that. But that's, that is mostly used in broadcasting industry. It's one of these things where I'm quite disappointed that there isn't an open standard for it. So everybody makes their own devices. In the beginning, you, you even had to buy the chips from Dante. Now they're slowly releasing a FPGA code and you can then integrate it with your other code, but you get a fragmented ecosystem. So that's a bit of a shame.

Chris Gammell: Yeah. So actually that's a great segue back to ecosystem. I asked about Poland, but not about the audio ecosystem. So like, so now you're, you know, you're out there as an independent consultant. I mean, how, how many audio companies are there? So now it was like, you're trying to shop around and say that sort of thing. I guess there's probably different tiers of, and you've even mentioned, you know, integration levels of like BNO versus the actual hardware makers, that sort of thing. But like, what does the audio ecosystem look like to someone who wanted to work in that industry?

Dave Jones: So I think the, the audio world is fairly small and it's really, and everybody knows everybody world. So sometimes you meet a new company and you go, oh, you know, that guy from that company. Yeah. I used to work with him at this company. Oh yeah. So you always have to play the, the, the name dropping game a bit here in Poland. There aren't, you know, the well-known audio companies, but there are a lot of small startups doing, doing small stuff because of course production is a bit more available, a bit more affordable here than let's say in Western Europe or even in the States. So that's, that's quite interesting. And there's, there's a vibrant hobby scene, but you know, that's, that's not going to pay the bills. Sure. So I think most of my customers are going to be in Germany. In my case, Netherlands of all, because I have the language advantage and the States is where I get most of my inquiries from. Yeah. Most of this stuff you do remotely. So it doesn't even matter that, that you can't be on site more than once a month or something. But yeah, you do have to really watch out in this industry that there, there's no screwing up because everybody knows everybody. That's right.

Chris Gammell: Yeah. Yeah. You can't bad mouth people either. You gotta, yeah, just be very civil and high, a high performer.

Dave Jones: Well, civil isn't, isn't even necessary. I mean, a lot of these audio people, you know, are, are music industry executives. So they, they know that there are some very, let's say special personalities in the world. If you go to a conference, half of the people are wearing suits and the other half are wearing torn jeans and have long hair.

Chris Gammell: So that's the rocker. Yeah. That's absolutely more. Aging rocker. Yeah.

Dave Jones: There are some, some really, really colorful characters in the space and I suppose I'm one of them. Interesting.

Chris Gammell: Yeah. Well, that's great. That's great. So then, so you'd mentioned as well, adding internet to things. I would imagine with the rise of smart speakers and just everybody wanting to have a Google backend or Amazon Alexa backend and that sort of thing. Like what is, how does that end up impacting the stuff that you're doing?

Dave Jones: Yeah. So in the beginning when, when all this stuff came out, I think the audio industry was very shaken. There's a, there's a loudspeaker company in the U S that, that literally went into a tailspin because they thought, you know, our bread and butter is internet streaming loudspeakers. And then the voice assistants came out and they were just caught off guard. So I think that's ecosystem has sort of panned out now. There's Google, there's Amazon and there's Apple essentially. And you have to put one of these three into your system. But you see that a lot of, a lot of people who offer services now start to make a platform where they can plug that in. And then, you know, you just use the services of Amazon. So they have APIs ready and all you need to do is implement those APIs and then you can play your music. You get your stream from somewhere and you can still build your device.

Chris Gammell: So then like the, so then the audio, the audio guts are basically still similar. They're just kind of taking new inputs, like you're saying from the, like an internet based digital analog converter.

Dave Jones: Yeah. I mean, there's, there's that interactivity where, um, the API just said, well, the user asked for the volume to be increased. So you increase the volume and you, you hand off all the speech that you record back to the, uh, to the, to the server somewhere in the cloud. And if no command comes back to you, then you can just keep playing audio as is. Got it. Hmm. And there's another development, I think in the pro audio business where people are trying to unify the command infrastructure. So there's a protocol called OCA open control Alliance, where essentially now with you get the remote control from company a, you plug it into speaker from company B and it recognizes that the thing has a gain control and it has a name and you can, you can tell it which stream to listen to. That's, that's actually quite a cool development and it's, it's all open standard. So I'm a huge fan of that.

Chris Gammell: Yeah. I mean, that just cuts down on like having to integrate with three different types of standards and what the preferences of each engineer at each is, you know, like, yeah, that's yeah. Yeah.

Dave Jones: You might even have all three standards fail because they, they refuse to interoperate. So it's very good that they're proposing to make this open.

Chris Gammell: Yeah. That's good. Well, I mean, what do you see? What is the future of audio? I mean, obviously you've mentioned from your own perspective, you're going to be developing algorithms and, you know, being able to offer that to clients and things, but what is 10 years from now? What, what are, what are people going to be tweaking with in the audio realm?

Dave Jones: Well, I'm notoriously bad at predicting the impact of trends. I remember when, when cell phones came around, I thought I'm not a veterinarian. And I won't ever need a cell phone, but it happened anyway, but, but I personally really want to do and would like to see the industry develop. And I think we've made, let's say signal processing and signal fidelity about as good as it can get. And honestly, even in back in the eighties, it was already pretty good. What is really limiting the experiences of people to, to enjoy music is having a good playback environment. So now that noise cancellation is, for instance, a thing in cars, I would really like to work on putting that inside normal environments. So if you're, if you're sitting in your nice little loft in the middle of the city and you're trying to enjoy your music, but you can't open the window because if a car drives by, you can't hear anything. That's still not a good thing. So if we want to increase signal to noise ratio, I think we've, we've done a lot about the signal. We need to work on noise now. And I think there's a huge market in, in just improving, let's say the, the amount of noise comfort that you have in your own home. So people start to live in smaller homes, closer to each other, and you're going to have more and more noise. And I think you can really see that it's, it's getting to people. So if I have a time where everything is quiet and I can hear myself think, I feel that I'm getting calmer and more balanced. And, and I would like to offer that to a lot of other people.

Chris Gammell: Yeah. It's almost like, yeah. Sometimes when I'm like really need to concentrate, I'll put my noise canceling quote unquote headphones on, but no music. And just like, yeah, it just kind of blanks out everything. And like, yeah, that really helps to center me sometimes. We're all like, I'll, I'll realize that the music stopped playing if I had been playing and I'll just been working for another hour.

Dave Jones: I'm like, yeah, if that's not good enough, you use these, these brainwave patterns or, or. Oh yeah. Yeah. Binaural beats. Exactly. Yeah. And that's just people trying to, to drown out the noise. And, you know, at some point there's only so much you can do.

Chris Gammell: Frigging open office plans. That's the problem here. That was, that's when I started doing binaural beats is when I was in an open office and I was like, this isn't going to work for me. I needed to distract the hell out of myself.

Dave Jones: Yeah. But still, even with binaural beats at some point, you're just going to get so tired of listening to noise all the time.

Chris Gammell: Yeah. I mean, yeah. Brains are used to nature, right? I mean, like that's, we've evolved to be in nature and quiet. And then we, we knew that there was something dangerous when there was a snapping tree branch and it's not just someone at the coffee machine, you know, half the office over it's, uh, you know, so like having quiet can actually help a lot. Yeah.

Dave Jones: Even if you're sitting in your nice quiet office and everything is okay. If you have, for instance, an ambulance driving by every 15 minutes, you're going to be in some, some kind of dress state all the time.

Chris Gammell: Yeah. Yeah. Like heightened, heightened awareness, right? Yeah. Yeah. Yeah. Sign me up, man. I'll, uh, I'll, I'll, I'll beta test. Yeah. Even moving to a smaller city. It's, it's, it's not quieter here. That's for sure.

Dave Jones: Exactly. So the plan is to, to implement a system like that, get filthy rich and then buy a house in the middle of nowhere that's quiet anyway.

Chris Gammell: Yeah. Right. All right. You know, some of the houses in the middle of nowhere are pretty cheap. I'm just saying.

Dave Jones: Yeah. And, and well, interestingly, I found that out this, this summer on holiday that in the middle of nowhere, it can be pretty loud as well because sound travels much further. And your neighbor a mile down the road might be hammering his roof and you can hear every bit of it.

Chris Gammell: Got it. Yeah. Yeah. And I guess your brain does kind of pick up on it then too. Just the difference. Yeah. I guess we're back to perception talking there, huh? It's still a lot better than being in the city, obviously. Yeah. Yeah. Yeah. Well, where can people find you online if they want to hire you or see your work or your YouTube videos, like you've mentioned, you've been making, where's the best place to check that out?

Dave Jones: Yeah. So I'm, I'm trying to put everything under the it's only audio brand. So there's it's only audio.com. It's only audio on Twitter. And I think my YouTube channel is also called it's only audio. There's only like three small movies on there. I'm only testing it out. Yeah. On the website, there's a, there's a contact page. There's every receivable way to get to me. So you can't miss me there. Awesome.

Chris Gammell: Well, Remke, thanks so much for being on the show and telling us about audio. I think this is a, you know, like you said, it's a audio is not going away anytime soon. And I think there's a lot of interesting things coming up. I'm really excited to see about the, the future of noise, noise, uh, modulation or whatever it's called. Noise control. Yeah. Noise control. That's right. All right.

Dave Jones: Well, thanks for being on the show. Thanks for having me, Chris. It's been a pleasure.

Speaker ?: x x