Brian Kaczynski with Second Sound

Welcome to the macro fab engineering podcast. I'm your guest Brian Kaczynski

and we are your hosts Parker Dolman

and Steven Craig.

This is episode 208. Brian Kaczynski

earned his Master's of Science in Electrical Engineering from Stanford University in 2001. And has worked since 1999 as a mixed signal RF ASIC designer, focusing mainly on land applications such as Wi Fi, Bluetooth, USB, etc. His passion for music and desire to bridge the gap between acoustic electric instruments and music synthesizers, both analog and digital inspire his current work.

Thank you, Brian, for coming on to our podcast.

Thank you for inviting me.

Yeah, we really appreciate it. Actually, two episodes ago, Episode 206. In our RFO section, we'd actually kind of spoke a little bit about one of the chips that you designed the ACO 160. And at that time, I was like, You know what, I've got this guy's email. Let's, let's reach out and see if he if he'd be willing to come on as a guest. So yeah, thanks so much for coming on.

Okay. Stephens totally not creeping on you?

Not at all. Not at

all. I have to check out the old podcast because I wasn't aware that it was mentioned on the podcast.

Well, so I had actually met Brian, a few months ago at nob con. Out in. Gosh, where was it? That was not there was Chicago. I was thinking Vegas, I went

to Vegas for out near out near O'Hare Airport. Right. Right.

And, and actually, months before that, I had to receive some some data sheets about this IC, and kind of some upcoming stuff about it. So yeah, that was it was really interesting. And so Brian, you're with or your company is second sound, right?

Yeah, well, I am second sound. It's just one person company. And it's basically like an IP provider. So it's based on my work is based on the pitch tracking algorithm that I envisioned, and I'm using to convert audio to CV for controlling analog sense and MIDI for controlling digital sense.

So a second sound, well, just being you is you've actually created hardware or heart ICS that, that incorporate this algorithm, which does pitch detection and converts the CV right.

Yeah, well, the ICs are, are big, like mixed analog and digital ICs. So, you know, part of it is I wouldn't know if it's correct to call that an algorithm. It's sort of based on based on my experience, designing clock generators, and, and RF synthesizers and PLLs. PLL is sort of, it's a basic building block in communications circuits. And I use a lot of my expertise from that world to to design kind of a audio music synthesizer, circuit, Chip.

So where does the name second sound come from?

It's actually second sound is actually like a real physical phenomenon. In low temperature physics. This probably is going to sound sound like it's coming out of left field. But I majored in physics in my undergrad. And I did a low temperature physics lab. And it's a phenomenon that happens with liquid helium, when it gets liquid helium can be a super fluid, which means it has no viscosity. And if you have like a little chamber with partially liquid, partially superfluid, helium and partially non superfluid helium, it's like two states of matter. And you can get waves where like the two different kinds of helium are kind of oscillating. And you can measure the speed of that that's called second sound. And I just liked the way it sounded too. And I figured it's not really trade markable because of the physical phenomenon has been around for so long. And, you know, I, I feel like I'm, I'm trying to offer a new way to make sound for people. So, you know, people like to play their traditional instruments. You know, people that aren't keyboard players, singers, guitarists. violinists all want to be able to control synthesizers for their instruments. And I have spoken to so many people who see what I'm doing and they I tried to insert with it and they say, Oh, I've been looking for something like this for for 10 years or even 20 years. And there are, of course, it's not anything totally new. There's been pitch tracking technologies before, but I think they're usually like really heavily targeted for guitar for a particular instrument. And a lot of other even some guitars say that they don't like the way that it sounds or, and other other instrumentalists just say that they feel like totally, totally passed over by that technology,

you will be talking about like a wah wah pedal. No, or is it TalkBox.

A wah wah pedal is, is a really basic thing. It's like it's like a filter, usually low pass, I guess, where you modulate the filter cut off with the with the pedal position. So that's even, you can even make a passive wah pedal, it's, it's like one of the simpler, simpler kind of effects out there. The this what I what I've done is really like generating a synthesizer voice based on the audio that comes in so it has an independent oscillator inside it. And, you know, some VCA the newer version that I made has a VCF as well that but that's that's getting into the firmware version with on the on a microcontroller for talking about the chips, then that ACR 160, that it basically has its own synthesizer voice that has a sine A sine wave, square wave and sawtooth wave, that are tracking the frequency of the audio coming in,

you know, I think it's a it's worth doing. Taking one quick step back and kind of mentioning where or how this is special. Because in most synthesizers, there's a control voltage that comes in and that that can be a DC voltage, or it could be anything, and that voltage is mapped to an output frequency. So your technology here takes some instrument that's already producing a pitch, and converts that back to a control voltage that you can then use to control another synthesizer. So you're going from a pitch to a control voltage and then a control voltage back to a pitch, right?

Yeah, well, that's if you choose to use that control voltage to control a synth using, you know, interpreting it as a frequency. So that converting that pitch to CV is one thing that it does.

So So you know, one of the one of the things that's that is interesting is the idea that you're using something that oscillates to control another thing that oscillates, because so take for instance, there's a lot of like you mentioned guitarists that want to, I guess you could say simplify their guitar sound. And in electronics, there's a lot of methods to do that in terms of modifying and shaping the waves. But with this technology, you're just picking up the frequency and then convert and then using that to produce another frequency. So you don't have to brutalize the wave, you just use its information.

Yeah, in fact, you don't even care what the wave looks like, the only thing the only information you're getting from it is the frequency, or period period is the same. And the amplitude, so it's just extracting the pitch in the amplitude. And the nice thing about the chip is that the the the frequency that are produced was perfectly in tune with with the audio input. So it's it's it's basically something called a frequency Locked Loop. It generates it measures the frequency coming in, and it sort of locks the frequency of the if its own oscillator to that so you have no no, like tuning error in the in the wave that comes out.

So your AC Oh 100 ic was your first kind of go round at this, right?

Yeah. So that chip I developed when I was working in Poland for companies Sonic Smith. And they have they have what's what we called Audio controlled synthesizers, which used that chip. They were like a kind of stomp box type form factors with synth functions like a VCF sub oscillators, PWM envelope generator and all kinds of other basic blocks. And I wanted to stop developing those end user products and just concentrate on on like what I knew best, which was the ship and the kind of underlying technology behind it. because I didn't want to, I got tired of making all these decisions like what, how do customers want to patch these things? That's why there's hundreds of analog sins out there. Because there's like, literally millions of ways you can patch them. And there's, if you design a product, you're automatically making some decisions on what the customer is going to want to do. And so many people are like, can you just bring all these patch points out and make all of these, you know, people just want to be able to do anything. And you're, you got, you guys know, because you work in this field a little bit here, you know, you're like, Okay, well, like it can cost $1,000. And they're like, no, they want it for $100. And they want to be able to do every patch, every single combination of things that you can do, and have knobs to control everything. And presets would be really fantastic to use. And I was like, finally, I was like, I'd have had it with that. I'm just gonna do, I'm gonna do what I do best and try to provide it as a service to people.

Yeah, but you're also going into, you're talking to to hardware engineers here where, like, I've definitely complained to companies where I'm like, Hey, your pan out sucks on your part. So it goes all the way down to the chip level to Yeah.

Well, I also do PCB layout. And I, I did all the layout on my chip, which is kind of like a little micro doing like a little micro universe of what's on the PCB. And it's much bigger, and many more, more wires and components and connections. So to me PCB layout, like, I find it really relaxing and nice. And I never really had a issue where I thought, Oh, this is like impossible layout or something is like, if your layout something with 10s of 1000s of components and 100,000 wires, and then you're like, and for layers of metal, it's it's like, then you really have to be careful of what goes on where and how everything's connected,

is that the number that you did,

it's around that it was something like 50,000, or on the order of 50,000 nets in that chip, I don't have an exact count. But I mean, in my industry work I've worked on on chips with millions of nets. But in analog circuits, we're usually on the order of 10,000. Because there you have, you know, you're doing things like amplifiers. And, and, you know, a PLL is probably the most complicated thing that you do over there.

Well, I'd love to get into the the layout stuff, but I think we'll do that a little bit later. Because I would first limited just talk about the technology behind the actual pitch detection. And if you wouldn't mind sharing, like what is actually happening under the hood with that,

well, in the earlier chip, the ACR 100 It was sort of a threshold based pitch deck detection. So, I filtered the input I low pass filter the input, and that was that just went through a threshold detector. So you had a high threshold and a low threshold to give you a hysteresis that gives you some rejection of noise and, and an errors in the pitch tracking. And then on top of that, we had a Pitch CV output, and we use that in a feedback loop to control to control that low pass filter that I was talking about. So it was kind of a special filter had had like it had a low it had sort of if it was locked, then you were kind of on the part of a low pass filter that was that was falling. And then it went to a shelf where where the attenuation reached a certain level and just stayed there. And on the low on the low side. There was there was game. So basically, if you if your frequency was too low, then if the let's say the low pass filter was too and too low, then at least you're guaranteed to have some signal coming out because the the filter flattened out in the past in the stop band. If the filter was tuned too high, then it would it would kind of just like track down close to the right area. And even if the second Herma if it couldn't track to the second harmonic because the if it did then the first harmonic would be at two times the lower frequency or the fundamental that is and we had gain there. So it would boost the gain up on that. And that worked pretty well. But any kind of approach like that with a feedback loop and a filter, it has issues with transient response. So if you especially if you want it to work with low frequency instruments, like if you play bass with it, or something is going to track down and then that filter is going to take some time to recover and move around to the new frequency if you play a new note. So, in the ACR 160, that I designed later, I tried to overcome that by not using any filtering at all. And instead, I use this double peak detector approach where I have peak detectors on the positive and negative peaks. And detecting a cycle requires detecting a positive peak, followed by a negative peak, and then another positive peak. And the the innovation there was that the time the delay, the decay time of the of the beat detector was proportional to the period. So as the period gets longer, the decay time gets longer, too. So basically, it's rejecting the same kind of content in the in the waveform, no matter what frequency, you're, you're playing into it. Hmm.

In that, I guess I'm a little bit a little bit curious about how that doesn't latch on to harmonics, how it how it kind of rejects that and sticks to the fundamental?

Well, if you look at a waveform, even with a lot of harmonics, if you look at kind of the maximum peaks, the maximum peaks of the signal always correspond to the fundamental. What happens if you have harmonics is that you'll have secondary peaks below, below that maximum level. And depending on how much you let the peak detectors decay, it will reject those harmonics up to a certain point. Of course, you can always give it like a signal with second harmonic 100 dB higher than the fundamental and then it will probably, it'll probably fail to work on that. But that's not very common in music signals. You can always break any system, there's always a way you know, you can always if you know how it works, you can construct a signal that will break it.

Yeah. But in the audio world, the second harmonic is never, you know, a higher peak. So, yeah, for its application, it's perfectly fine.

Yeah, so some people talk to me about what they call the missing fundamental problem where you have a second and third harmonic, but you have no fundamental. So that can actually happen. Fundamental can be can be non existent. But still, even in that case, you see the peaks, if you if you track the maximum peaks, then they they follow the fundamental, they allow you to recreate that fundamental

is that it kind of reminds me of signals, classes, and all the all that like fundamental math that goes behind it that allows you to reconstruct things like that. Those were those were not fun glasses. I remember,

yeah. Well, I'm not going to talk about Fourier transforms.

I've certainly seen that before. You know, take take sample a chunk of signal, take the FFT, and then do some some algorithm to pick the fundamental, and then you kind of get that information. But that's a really slow method of doing it.

I don't like that method. Because basically, depending on the lowest frequency, you wanted to tag that sets the window of the FFT. So if you wanted to take down to 30 hertz, then the window has to be 30 milliseconds. And certainly, you're going to hear 30 milliseconds latency. Or even if it's like half of that 15 milliseconds. I'm not sure what exactly the latency is, but you will hear it

right right. What is the latency of your system,

especially one cycle, so, it depends on the frequency you play, if you play 30 hertz, it will be it will be 30 milliseconds. If you play a kilohertz that will be one millisecond. But sort of psycho acoustics teach that the latency that we can tolerate is actually proportional to the period of the audio. That's something that not many people have studied. Because most people when they do latency experiments, they don't they don't pay that much attention to how it how it depends on frequency. But I was I was at a yes convention this year in March in Dublin last year, I guess now, and I spoke for a long time with with with like a leading figure in in the frequency in the pitch detection field and And, and she knew a lot about this and explain that, that there was like one study out of many, many that's that studied the the impact of pitch of frequency on latency that we can tolerate.

Hmm. Well, and like you said, the it's it's always one cycle right or best case if it's around,

it's around. Yeah, the latency for my protection is one cycle. And the latency that we can tolerate is around one cycle, I don't want to say like exactly one cycle, but it's, it's that order of magnitude, because our ears can't really detect a pitch anyway, until it's, they've heard one cycle of it. It's sort of like our ears are kind of like a Fourier transform themselves. Physiologically,

sounds like it could have a lot of applications outside the audio world with like, tracking filters, and, you know, frequency detection and, and will even back in, like the audio world, like faster tuners, and a lot of tuners that I have have a very noticeable latency before they pick up your actual tone.

Yeah, wasn't that one of your it was one of your projects, right? So even in college?

Yeah, I had, I had a math class called waves and wavelets. And it was it was all the math behind the FFT in that class was actually really, really awesome. And for one of the projects, I made a, I made a guitar tuner, but I basically just did the FFT, and then put it in an array and sorted everything and found the highest one and say, That's my fundamental and, and I got, you know, it was close enough for guitar.

Close enough for the person grading it right.

I close enough to get me an A.

But I actually, on that project, I actually did build like an entire GUI, and I put my computer up in front of the class and like, had like a graphical thing that would show like, if you're sharp or flat or, or, you know, spot on and stuff. So that was fun. But it seems like it seems like if it was all done in hardware with this IC like that would have been like you just automatically get an A in the entire class.

Now, yeah. Okay, if you that would work if you want to spend several years designing an IC. Here's a master's. that would that would be a big labor of love for a class.

Oh, for sure. So, your so your first I see is was the ACO 100. But you've moved on from that to the ACO 160, which you've mentioned a handful of times, you want to kind of give like a description of what's different between those.

So So before we get into the super technical actual stuff, yeah. Is was there 60 versions of or 59 versions of this chip before? No,

there were not that many. There were two versions of the ACR 160. I just wanted to make a big increment. I don't even remember where I pulled that number from 160. I think I just thought it sounded good.

Product Design 101. It just sounds good.

So 100 was the first one once for the 160s pretty simple to describe, I just tried to integrate a lot of the features that were off chip in the earlier version, like that tracking filter, I wanted to get rid of the tracking filter and make it or make it unnecessary. I added also envelope detector. So it has an envelope follower built into the chip instead of having to do that with diodes and op amps like we usually do. And it works totally differently that envelope follower is what is like is more like a sampling. It's uses sampling and some digital filtering, not digital filtering. It uses switch capacitor filtering, which is like discrete time filtering instead of continuous time. So what else it it? I think the sine Yeah, the sine wave generator was new. I didn't have the sine wave in the AC 100. So I wanted to see how good of a sine wave I could generate with with like in a in the integrated circuit.

Just out of curiosity, what was your method of creating a sine wave?

Okay, that's interesting, I think so it basically they're the, the internal oscillator on that chip runs, runs eight 8192 times faster than the audio. And that number is there's a reason for that number. It's two to the power of 13. So in, in, in engineering, we like to work in powers of two it's no coincidence. It's easy to make, like clock dividers that are divided by powers of two. So that chip also has a harmony feature where you can tune like to different intervals offset from the unison, it can go from minus two to plus two octaves, and it has like eight notes that can hit on kind of adjust intonation scale between the octaves. And to do that I, I generated, basically a divided downclock, which is divided down by just some integer related to that really fast 808,000 something times clock. And the net result is that you get a lower frequency clock that's roughly 128 times faster than the audio, but it varies because it's a harmony. So it's varies from, from basically, basically 32 times to 512 times faster than the audio. And then that's used to clock a DAC, which goes through 128 cycles, which you're you're probably used to thinking though, 128 is seven bits. So for the the sawtooth wave, it just goes through a staircase, basically, of 128 levels. And for the sine wave, it goes through a cycle of 128 values. And I designed sort of a sign as a special sine wave DAC, which, which generates exactly like points close to the sine wave all the way around one cycle. So it's kind of like an analog wave table. Actually, if you want to think of

it like that, was it just kind of like a built in ROM that had values? No, it's

It's totally analog. So it uses a big resistor divider, and the points on the resistor divider just correspond to those voltages on that sine wave, oh, that must have been fun to calculate. And I used, I use basically resistors all of the same size. So for example, I want to if you want a really small resistor, you put 20 of them in parallel. And then you know, for any other value you want, you can use parallel of the series combinations. And it's something like it's it's a lot of math, and it looks really hairy when you try to put it all together. And then especially if you want to generate 128 levels. But the nice thing is that in an IC, O you can do things like that, and it ends up being really small. Because a resistor is like two microns by by eight microns or something, let's say which is I mean, a micron is 1,000th of a millimeter. So it's like nothing really it's like a smaller than a human hair.

And so are the I guess the resistors then are are they just different? Slightly different trace for trace with inside the IC then? Or is it? Are they? Are they different doped materials or

resistors? That's there, they use poly silicon. So, you know, in terms of what physically are the resistors? They're not metal? They're, they're using the same material that's used for the gates of the transistors. Okay. And they are they are, it depends like what resistance you want, you can get something that gives you typically like 1000 ohms per square, which means if you have, you know, two microns by two microns, if it's a square, the resistance of it will be around 1000 ohms. That's typical. I mean, it varies from 200 to 2000, maybe, depending on the process.

You know, okay, let's, let's, let's dig into that a little bit. How do you control that as the designer? If you have an IC and you say I want it to be 1k instead of 1k? Ish? What what knobs do you have to turn to be able to control that?

Well, usually we just don't worry about it, they say that it can vary. It can vary by plus or minus 15%. It's around 50%. And usually you just do the design so that you don't care about something that small because for a deck, for example, a resistor deck, it's like a, it's like a big voltage divider, the voltage is at the top, you know, because it's some it's you can make a reference and we know how to make pretty accurate references, or you use the supply voltage or something like that. The bottom can be ground which is zero volts, as most people know. And because the only thing determining the voltages on the on the resistor divider are the ratios between the resistors. And the good thing about ICS is that even if the, the resistivity of that material isn't well controlled, the ratio is generally very well controlled. So you can that's that's how we design accurate things by sickly with ICS.

Just don't don't don't, your design shouldn't care about the absolute value.

That's right, yeah, you can make you can double an analog voltage by making a invert a non inverting op amp circuit with a voltage divider where the resistors are integrated, and it will be pretty close to two times the voltage. Alright, you know, just expand that to any ratio and you can generate any voltage you want really

don't want to guess that's because when they put down the resistors, I personally put down when they do the masking for the resistant part, that all gets applied at once so that your materials all the same, but the absolute I want to say doping, because I don't know what the process is of like making that poly. What was it? Poly? Silicone, poly silicone? Like how, like when they lay that down?

Yeah, usually the what makes it higher or lower is is etching. So sometimes it can be edge too much. And then the resistor gets narrower than you expect. Or edge too little. Okay? I don't it is. Some of them are dope, some of them aren't. So you can have different doping. The doping can be slightly higher or lower. There's, there's n type and p type poly silicon. So it's the numbers can be different depending on what which one you use.

Just as long as it's not straight to low, then yeah, fuses.

Oh, yeah. It's a, you know, they, they guarantee that it will be within a certain limits. And generally, if you track, you know, if you measure them over many, many lots of an IC, usually they come out much, much more accurate than the fab wants to guarantee it is because like, it's like anything, when you're meant in manufacturing, you guarantee something, you're gonna pad that you're gonna pad that like a couple of percent out. If not 10%? You know, I don't know. So we, we don't really see silicone that varies as much as the fab say that it can. But you don't want to depend on that either. Because someday, maybe it will come out high or low,

and someone will sneeze on it. Yeah.

Cool. Well, so you also have some development boards that you offer for these, the ACO 101 60. Right.

Right now just for the 160, the, the AC 100. I have, like we went to production with that chip. And so I have chips available that I could sell, but I didn't make evaluation board for it. Okay. Basically, some people have asked, I've received emails asking if I'm selling that, and I and I deal with those on a case by case basis, I tried to figure out what that person wants to or thinks they want to do with it. And, you know, then explain how it works and the datasheet. And it's, it's not like it's not something that I would expect to be like a cash cow business for me because that chip is older. And although you can do some cool things with it. It has its own characteristics. And some people might be interested in the way it malfunctions versus something that works more, more reliably, quote unquote. So you know, experimental, like experimentalists want all kinds of things, and they don't necessarily want something that's going to be perfect.

Sure, sure. So you have to know how to design imperfection into it.

Yeah. So the I have an evaluation board. Now for the ACO 160. I have a few. So I made for the chip, I made an evaluation board. And I sampled a bunch of them out, I sampled about, I had about 25. And a lot of them went to musicians, some of them I sold. And a couple months later, I don't know if you want to get into this topic now yet, but I decided to try to make a fully digital version of it. So I thought, you know, instead of designing this ship, which takes a huge amount of resources and time, and it can have bugs, if it has bugs, then you have to fix it. Or just tell everybody, I'm sorry, I tried to make this chip, but it doesn't work. Like I wanted

to want to run a datasheet before. Yeah.

Instead of that, I thought, what if I could make a digital emulation of the chip and something like a microcontroller? And what if it worked just as well, or even better than that chip? I thought, well, it would be kind of sad because I spent so many years of my life making these chips. But on the other hand, if it's Paul assembled do, I better be the one that does it? Because otherwise, somebody else will do it and say, Hey, we've got these things that work better than the ACO 160. You don't have to, you don't have to risk buying custom silicone that somebody, you know, a one person team has to test. Because the, honestly, the hardest part about designing chips is is Production and testing. Because if like, even if one out of 10 chips is bad, if you make 1000 of them, then you got to test 1000 chips to throw out the 100 chips that are bad. And that's something that I never got into. And I was always kind of the Becca, in the back of my mind worried if I go into chip production? How am I going to do this production test? And I thought, you know, if I can use off the shelf, microcontroller or something equivalent, then those are already tested. They're already like two bucks in volume. So what am I what am I doing with my life? You know, what have I been doing?

Since you're implementing this in firmware now?

Yeah, so now we have firmware that is basically an emulation of the ACR 160. And most of the evaluation boards I have now I call the de ACO 160 Because it's digital ACO. And those are the ones I'm promoting most now. That's what I'm providing manufacturers and selling some to to musicians hobbyists, demoing them to people that that produce video content that I can use to help promote it.

So I'm looking at your website. And I'm noticing some of the boards have a, a a quad pack socket on it that has looks like your chip on it.

Yes, that's

got a clear top on it.

Yeah, it's gold. Or let me see. Let me try to find that. So if you go to ACR 160 Evie K page that has a Yeah. So here, the picture that was taken had the the cover removed. So you can see right into the chip on some of these pictures. That's a test doesn't test chips there. They were made just for as a prototype. And I made about 50 of them. In production, you never see something like that, because Chips are always encapsulated in black plastic.

And that's for because photons hitting it will cause all sorts of craziness.

Well, I found that they didn't I, I found that I could put this under a bright light, and it had no impact whatsoever on the performance. I always heard you know, yeah, photons, it'll make like the PN junctions leak. Not not true in this. So maybe in this process, the I don't know why it's like that. But it had very little difference in performance, whether the lead was on or not,

or just sounded cooler. So if anyone wants to see it, the website is second sound.com. And you can go over to the ACO 160 EBk. Product page.

Yeah, and all of the all of the products are under under products you can find right now all four of them. So I have product pages for the ACR 101 60 And the two EV Ks, which are using the AC 160. Chip and are using this digital version we're talking about with a microcontroller.

I love your website, by the way, because your product pages have like flow diagrams on them and system diagrams, which is awesome. Yeah,

I mean, because I'm from an engineering background, I thought what are what do people want to know when they

look at this? Right? Like who's gonna visit this site? That's what they want to see. I try to provide that

and like some audio demos, because I'm trying to cover what I what I think all the people that are going to look at it what they will want to see.

Yeah, speaking of audio demos, you have something that you can show us, right?

Yeah, so I can you can at least hear an audio demo. What I have is I have my voice controlling the Behringer Model D via the pitch the Pitch CV envelope CV and Gate City. So I have an envelope. This is my favorite way to patch it. You can't really see it, but I patch it in a way that the the Model D responds most organically to my voice so it fairly follows the envelope and I don't use any of the envelope generator that's built into it. I just

let it follow what I'm doing. So this is your voice going into a microphone and then that controlling multiple aspects of the synthesizer.

That's right. So I have the microphone here. I'll turn up the volume on that model D. So I can do some feedback here if you want

that sounds like that sounds really aliased.

So yeah, because what's happening there is that the, you know, the audio from the Model D is going back in as the audio that it's going to convert. So if you have a frequency error, then it will kind of, like drift around. And it'll kind of try to follow that around. And, you know, I have, I have like some sub oscillators here. And it's Sir, I'm using all three oscillators, I was using all three oscillators from the Model D for that, that voice. So I know that's fun. That's what I had the most fun doing when I went to trade shows. It's just like sitting there with a microphone. And like, just don't do it. Dude. I can You can tune in do all kinds of crazy stuff like that?

Oh, that's fantastic. And I had nob Khan, I got to play. I think you brought a bass guitar and and got to play. Was it a Model D that you had there? Also? Yeah, I

had the Model D and the microkorg. Right. Here's the microkorg for MIDI to convert audio to MIDI and and, you know, because it's it's just a small synthesizer. It's very popular. And it's like, I was thinking what is the most compact rep most widely recognized thing I can use to convert MIDI to some cool synth sounds?

Well, and I think one of the one of the big things is this conversion to synth is isn't anything new, which you mentioned a lot earlier on. But but many times it involves, you know, installing new hardware on your guitar or your instrument like, like a MIDI pickup or something like that. Yeah. And 90% of the time, the latency is awful, especially for bass guitar. And you're certainly wasn't.

Yeah, so I mean, it doesn't work for polyphony, yet. If you want to do polyphony, then you have to install some kind of polyphonic pickups. So you want to have hex pickups, and then some six wire interface. And you want to, you know, have six instances of this pitch tracking thing going. And I believe people there are people that are working on that now without naming any names. But it's not me I'm not like that's that's kind of exactly what I was saying earlier, I, I didn't want to be in that business of trying to decide what users wanted.

I've talked to a handful of people about doing something like that with with guitar. The biggest problem with it is, if you if you don't go digital with it, then you're exactly what you're saying. You have to have a six coil pickup and individual per each string on the guitar. And then you have to have the ability to get six individual signals out of the guitar. And most guitarists would want that to be analog and it ends up you're having to uproot some basic functions of the guitar that have been around for decades. And you would be alone in this sea of quarter inch cables with your special guitar that has a six or seven conductor cable. Yeah, it gets impossible real fast. Yeah,

there's a lot of custom hardware that goes into it and but there's a huge community of people that want to do this and they are happy to modify their guitar. And I think it's kind of an expensive field. But you know, there is an audience for it. There's if you look on Facebook, there's a group called SIG fi s, Cy CFI. And all of its like all people that are interested in doing polyphonic guitar synth stuff. That's not the only one there's there's other there's other communities that people into it as well I think. So,

I want to do a touch real quick on just the IC creation. We've talked a good bit about it as we've jumped around, but so I wanted to just ask, like, how did you go about designing these chips? What what software did you use for that?

I looked around a long time for like a like a bargain, a bargaining chip design EDA software. And a few years ago what I found was Tanner tools. Tanner itself Like, they were kind of a small company, based in California, that they were providing an alternative to like the industrial options like cadence, synopsis, mentor, those are the big names that that everybody in this kind of this kind of EDA field has heard. And that was, you're able to basically lease the tools for a couple $1,000 for a few months, or like, maybe it was around something like around five to $10,000 for a year. So that's something around that. So you could design a chip in that time, like, as one person, I could design a chip in a year. And it was, you know, it was a big chunk of it was it was not cheap, I wouldn't say but it's something that if you've saved some money, you has something in the savings account, you could do it yourself, you could find it yourself. Since then, unfortunately, Mentor Graphics has purchased tender tools, and they've completely changed the pricing structure. So that was another thing that was a huge bummer for me that using those tools became quite expensive. And also why I was kind of happy to go down the path of of a firmware based design. Now, you're basically talking about $50,000 to lease those tools. And it's like, not very different from the pricing that you could get with cadence or some of those big industrial providers. I've heard people have talked to me about like freeware chip design tools. So you can you know, spice is supposed to be free. And you can find, certainly find free spice simulators. But if you want to put everything together and have something where you're able to verify that schematic matches layout, that's the big thing, by the way, is that how do you how do you know that the layout you did is the same functionally as your schematic, if, if you think you're gonna do that by hand, you're out of your mind, it's like a, it's a huge thing that you need, very specialized. And that's the software that really ends up costing money that you can't get for free. And the other thing is that you need support, you need technical support to set up all set these things up, because it requires a lot of IT expertise and setting up like network servers and stuff like that, which I didn't have the expertise to do. So I ended up shelling out a lot of money at some times to lease these tools and to get access to the support. And if I didn't do that there was there was no other way I could have done it. So it's like, I don't recommend actually design your own ship unless you have a big budget for it. And you really need to do something that is impossible to do otherwise.

It's actually what I wanted to get into as your first couple designs were in silicone, and then you move to a firmware design, what changed that allowed you to do that,

probably nothing, probably I could have done the former design from the beginning, what changed was kind of my philosophy about it, where I thought finally, like, you know, I know how to make, for example, a digital version of a analog filter, you know, I took I took digital filters and university and I know how to do by linear transform and stuff like that. And I you know, something that I found always found fun and never really did that much in my in my professional career. So I was happy to play with it. And I knew some I knew some programmers, some embedded programmers in Poland and Krakow from, you know, I, I, we didn't talk about in the podcast, but I lived in Krakow for 11 years before moving to Miami and about a year ago. And there's, it's like a really great place for it. And for, you know, finding talent in AI in it web design programming. So I had, I had, you know, friends there that were good embedded programmers, and I told them that idea and they said, Yeah, we can do it. So we spent a few months on it and, and ended up working pretty well. That's what you heard in that audio demo before it's that that digital firmware version controlling the the Model D and you know it for people say oh, you know, analog, digital, you know, you have these these steps, you know, you have a DAC with that, that that particular microcontroller has a 12 bit DAC, and if you read the fine print in the datasheet it turns out that the performance isn't even really 12 bit performance. It's really more like 10 bit performance and, and it looks like from the if you read it carefully, it looks like the steps can actually, you know, not even be monotonic in the 12 bit deck. So, I was thinking okay, we really have a 10 bit deck here. How Alright, make analog signals with that, because the CVS that come out come from those decks. And I use another technique from my sick from my, you know, signal processing, which is sigma delta modulation. This is a way, it's a really, you know, decades old way that people use and in fact, pretty much all of these 24 bit audio decks that you that you can buy. Now, they're not really they're not really 24 bit DAX, there's some lower resolution DAC that's being modulated, where the the code is being modulated very fast, and it gets filtered. And what it looks like 24 bit, it looks like it's giving you 24 bit levels. But inside, you really don't have a 24 bit DAC. And so I did the same thing. I used the 10 bit DAC and I modulated the code, and I low pass filtered it. And it's good enough to sound like an analog signal controlling these, these, these sent these sins. So people should look up sigma delta modulation, it's really, really cool technique. And it's very simple. If you, you know, it doesn't take that much reading or math understand how it works, it's basically, you know, to get, I can explain it simply, if you if you want to generate like, a code that's halfway between two, two codes of your deck, basically, what you do is you switch really fast between the two, those two, you know, two codes, one LSB, apart from each other, and you filter it, so you get a really high frequency squarewave. I don't remember, I think we were we were outputting it at 48 kilohertz, because we were using the same, the audio sample rate for everything. And we just, if you, if you're generating a control voltage, if you low pass filter, a signal that's been that's changing, you know, at 48 kilohertz sampling rate, you're never going to hear that, that square wave, you know, you're going to filter it and it's going to sound like a constant voltage.

See, you can you can buy yourself some more bits with code.

Yeah. And basically, you can, because the microcontrollers usually work with 32 bits. So it's technically a 32, bit DAC, if you want to be precise. Now, it doesn't know doesn't get performance if you if you try to change the code by one LSB of 32 bits. Is it going to be like that? Are you going to measure that one? microvolt, or whatever it is? No, of course not. Now, there, now you're talking about like getting into the noise, the thermal noise of resistors. So it but it's good enough for controlling your your Model D or your x VCOs or anything else, because if you think about it, you have one volt per octave. one semitone is about 83 millivolts. So you have to have really, really high noise, even if your noise is like one millivolt, which is very high. So the noise is less than that, it's still pretty low, when you convert it to frequency,

right? I can't remember what I think it was. The average human ear is capable of perceiving four sense of pitch variation in terms of tuning, and that that results in a still fairly high in control voltage, higher than the noise for

sure. four cents would be about 83 divided by 25. It's about three, three and a half millivolts or

so. Right? So as long as you're beneath that you're beneath the average humans pitch, ability to good or ability to determine a pitch variation. Yeah.

So Brian, what is next for a second sound.

So I I'm just gonna be the

161 or

something, I am working on a newer version of the of the firmware based based pitch tracking stuff. So I'm also kind of fielding fielding video content from demo artists, I'm still doing that I'm still expecting to release some, some cool demo videos that show you know, really good musicians playing with it because I'm, I'm an amateur musician. I'm, I'm an engineer that happens to like playing keyboard and guitar, you know, just for fun. But I am excited about a new version of the of the, the technology I'm working on. The main thing that I'm going to add is I will do a teaser for it. I want to add automatic gain control, because that was one of the things that was probably my most frequent requests from users. People don't like to have to tweak the game setting for their instrument. To get it, you know, sometimes it's hard to hit that sweet spot. Some instruments like guitar or bass that have pretty low level signal, they require high gain around 40 DB or so. And the way that the preamp is designed, if you tune it to a very high gain, it becomes very sensitive to the knob position. So I want to basically I want to get rid of that preamp control altogether. And I want the microcontroller to set the gain itself. So it's kind of a challenge. And I have basically just in the past few days, I've I've kind of been polishing off a automatic gain control kind of algorithm that I hope will do it pretty well. The other things are, I'm thinking about making it more enclosure friendly. So a lot of people have asked for an enclosure for the the evaluation kit, because people can can use it at home. But if you talk about taking it to gigs, playing live with it, it's not like the most practical thing for that because it's a bare PCB on metal standoffs. And at the very least, it needs to have a solid base so that it sits like it doesn't slide around. I recognize that and I apologize to my users who have struggled with this, some people have been creative and, and screwed it because you know, you can get if you get screws of the right, the right diameter, you can just screw them in from below the standoffs. And you can mount it easily to something, but I didn't provide that kind of hardware with it.

Now you get into, you know, where, where are the connectors supposed to go? And the knobs supposed to go?

Well, yeah, you have to be careful about designing everything that you know, everything has to be the right height off the clearances all matter. And I I dealt with some of that with my work with Sonic Smith from before when we were designing those those basically similar to this, but within closers. Yeah, you have to maybe panel mount some things because the the evaluation kit, now it has a combo it has a quarter inch XLR combo jack for the audio input, because I wanted to be able to plug microphones directly into it. And that thing is a is a beast, it's like twice as tall as all the knobs. And it's you know, it basically makes it impossible to put it in closure. How it is. So I'd have to think about panel mounting some things and doing some wiring inside it. I'd have to move around the MIDI Jack, maybe because the mini jack is on the side and the power and audio out jacks are on the back. So it's like it's some mechanical design. And I not positive how it will end up but it's just I just want to say something I'm considering. Very cool. Let me see, I might get rid of the the high pass and low pass filters or just make them fixed because it has switches for high pass filter and low pass filter frequencies. And I found that it really doesn't. I can set them to be wide open. Like the I can tune the filters to give the widest possible bandwidth and it doesn't really have any impact on the performance. I haven't really seen a case where I had to close the filters to reject some noise.

What else was I thinking to do? I don't know if anybody has some ideas. I'm, I'm open.

We'll where can people find more about you and and send you that information.

I know there was another one some people requested to use expression pedal. Some people want to use an expression pedal for the VCF frequency. That's really cool, you know, good violin players or guitarist they don't have a hand free to turn the knob. So you know that we were all familiar with what? Why are expression pedals. Nice. There's a sustain feature where if you if you push a momentary switch to one side, it will freeze the pitch and the envelope at wherever it wherever it is currently, that would be nice to have to be accessible from a pedal as well. Just like I was trying to minimize the clutter in my lab so I didn't want to have cables I already have enough cables on my on my desks on my desktop. So I was thinking oh another cable going to the floor and and more pedals. Why do I want to do that if I can just have a switch so that the evaluation kit I designed is more of a proof of concept that shows Yes, you can do it. And in a product you would of course make some other designs sessions to make it more ergonomic for users.

Right, right. Yeah. And Jack for just a standard 50k pot style expression pedal would be nice for sure.

Yeah, that's easy to do. And, you know, the only thing is, with that, that standard expression pedal interface, you have to be careful because somebody might plug in a mono cable to it. And I always worried about that. If you plug in a model cable, it basically shorts out the reference voltage. So you have to have a reference voltages that can be shorted to ground without damaging it. And that can be done, but it's just like, you have to think about stuff like that.

It that that will happen. Yeah,

I have a solution for that. It just requires another op amp and a resistor. So people have different solutions. Some people use, like a resettable fuse. I, you know, I like using the op amp if you if you don't mind paying for another op amp?

Yeah, for sure. One that can sustain a continuous short output.

Yeah. And you use a resistor to limit the current. So it's like, it's, you know, you can limit it to whatever current you want a milliamp if you want, right, right.

Well, and the output, the output accuracy doesn't matter as much, if it's already going to be warbled with someone's foot, you know,

yeah. And you can even make it accurate. If you want by using feedback in the Op Ed, there's, there's ways to make a DC accurate circuit that's sort circuit protected, I use that I use that on actually on this evaluation kit for the pitch TV out the pitch TV out because pitch is really important because you want it to be you don't want it to the DC accuracy is very important, no matter what load you put on it. So if you put 100k or a 10k, even load on, you don't want that to affect the the pitch voltage output. So I use a feedback circuit and an op amp to make the DC value independent of the of the load. And it requires some compensation. It needs to be you have to do some tricks. But it's it's those are well known tricks in op amp circuits.

Well, great. Where can people find out more about you? And what what's a way that they might be able to give you some some of those hints, like you were asking for?

So definitely visit my website, second sound.com. There's contact info there, or you can use the contact form. You can look at my YouTube. I mean, the website has links to every all the other social medias to Facebook, YouTube, Instagram. Those are the those are the big social medias that I'm on. I'm not I haven't been on any others. It's enough for me to keep up with those three. And definitely like my Facebook, you know, I don't have that many. I don't have that much of a following on Facebook. Because I think because I'm not really making end user products. I'm more of like a technology provider. But I definitely love to talk to users and people have written me and asked questions. So you know, message me on Facebook, write to me from the website, whatever you you can even is using some message direct messaging on Instagram. I've seen you know all kinds of contacts. Very cool.

Thank you, Brian, for coming on to our podcast and discussing chip design and firmware control and oscillation.

Yes, thank you for hosting me. It's been a pleasure for me.

Well, great. Well that would you like to sign us out?

That was the macro fab engineering podcast. I was your guest Brian Kaczynski

and we're your hosts Parker Dolman and

Steven Craig. Later everyone take it easy.