#443 – An Interview with JP Norair

Chris Gammell: This is The Amp Hour Podcast. Released May 19th, 2019. Episode 443. An interview with J.P. Norair. Welcome to The Amp Hour. I'm Chris Gammell of Contextual Electronics. And I'm J.P. Norair of Haystack Technologies. Hey, J.P., how are you doing? Not bad. Thanks for joining us here today. I know you actually from the consulting forum. You've joined that and you've been a very valuable contributor there. But the thing that I was really interested in is that we had a conversation going on there roughly that was about, it was something about LoRaWAN effectively. And you had started to kind of give me some education about stuff that I didn't quite understand of like different networks around LoRa and LoRaWAN and other things that are out there. And maybe that's a good way to start and then we can kind of go back to where all that came from.

Dave Jones: Sure. This is a topic that can be extremely expansive or, you know, extremely summarized. I'll try to go somewhere in the middle ground and, of course, ask questions in between if you want to go deeper. But so LoRa is a product from Semtech. And LoRa, the origins are from the mid-2000s from a couple of French engineers who had a little company. And Semtech ended up buying their technology and productizing it into LoRa. It's a unique modulation among other things of its class. It's chirp spread spectrum. It's good. It has good attributes of it. And it's been an improvement over low power radios in the past for a lot of very technical reasons. But, you know, it's been an improvement over what had been before in terms of propagation, quality of service. So that means the sort of likelihood of receiving a correct message, performance, and multipath. But, you know, LoRa is, it's a phi layer is what we call it in the industry. Right.

Chris Gammell: I kind of think about it like the actual low level, like what's actually happening to that 900 megahertz in order to make it into a data packet. Is that a good way to say it?

Dave Jones: Well, it's kind of the other way around. It's kind of, well, yeah, if you're transmitting, it's how you take some bits and turn them into waves. And from receiving, it's how you take some waves and turn them into bits. But, yeah, so it's quite novel in that respect. And it's not the first chirp spread spectrum technology ever done. But it's probably one of the first ones that's been put into a cheap little low power, low cost device. So that's what makes it unique from a technology perspective. LoRaWAN is a protocol that utilizes LoRa. And LoRaWAN, I mean, the origins of it, it comes from IBM originally. It was really just, I think, sort of an afterthought research project from a Zurich branch of IBM. And it's grown into more. SimTech started to adopt it. That's the company that sort of brought LoRaWAN to market to begin with. And they formed an alliance around it. And they have partners like the Things Network and maybe a few others. Yeah, it's pretty expansive.

Chris Gammell: I think the number, like Intel's in there. There's a bunch of people in there.

Dave Jones: Yeah, Comcast is trying to get in there too. Well, they just had a press release, I think maybe yesterday, about how they're not going to go after becoming effectively a carrier. That they're...

Chris Gammell: Oh. Yeah, because they were talking about putting it in every cable box at one point.

Dave Jones: Yeah, it seems as though they're going more after campus installations now. So their strategy is shifting. Which, I mean, is totally sensible. I think it's absolutely fine. I don't think that's going to slow down LoRa or LoRaWAN. Yeah. But yeah, so... But it does sort of tell you about the situation, sort of the market demand of these technologies is that, you know, a lot of these users are willing to install their own infrastructure, especially when it's low cost. So especially when it's easy to install. So, you know, the value of paying $3 a month or maybe even $10 a year per device or $1 a month, I don't know what... I don't know. Everybody has different prices. But, you know, Sigfox was trying to... Sigfox was a competitor of LoRaWAN, really, not so much LoRa. They are also in... They're a French company. I don't know if they're quite as strong as they used to be. They have a very different technology. And they've had some uptake in France, mostly due to the French government really going in on... Going big on Sigfox. But they've installed in a few other metros and so forth. And they've had some vertical success in security. Like home security, basically. But they raise too much money. And it just doesn't seem... Yeah, they raise hundreds of millions of dollars.

Chris Gammell: Yeah, yeah.

Dave Jones: You know, so it doesn't seem like selling $1 a month subscriptions per device is going to cover what was put into it. I don't know. You make it up in volume.

Chris Gammell: Well, I mean, it takes one heck of a lot of volume. Yeah. Well, and that one, too, is... Sigfox, it always reminded me more of, like, they wanted to install what was effectively like cell towers, but for low-power devices. And then you sell the service to the users.

Dave Jones: Yeah, I mean, but that's sort of... That's sort of LoRaWAN as well. So LoRaWAN is not... They're not as vertically integrated. I mean, it's a technology. It's an alliance. It's an industry alliance backed by Semtech and some others, but some more heterogeneous mix of customers and companies. And they have service providers. They have kind of networking companies out there. I think... What's that company? Senate, I think, is one. And they set up a lot of LoRaWAN networks. They're just sort of a LoRaWAN network set up company.

Chris Gammell: A specialist, yeah.

Dave Jones: Yeah, that's sort of what they do. And then Things Network is... Things Network would be actually kind of a direct competitor of Sigfox. Right, yeah. So that would be a direct competitor. But anyway, so getting back to that at the beginning, you know, if you're building a device and you want to... You have certain things you need to do and you need some kind of low-power, long-range wireless networking. And LoRaWAN sort of meets maybe 50% of your requirements or maybe it meets almost all of them or maybe it meets almost none. But nonetheless, you need low-power, sort of long-range machine-to-machine networking. You can still use LoRa. I mean, you can still use the LoRa hardware. And there are a number of other networking stacks that you can install into your firmware and into your software that are not LoRaWAN. So that's kind of the main difference between some of these, like, vertical solution providers, like the Things Network or Sigfox, where you're locked into a technology stack and, you know, your business arrangement is with them. It's with the carrier. And you get certified hardware or maybe you can make your own and then certify it. And you certify it against the Things Network or you certify it against Sigfox versus, you know, you have a lot of freedom if you're just using LoRa. You have an enormous amount of freedom. You just need to make sure it works. And if you're selling your product into, like, campus deployments. So campus deployment just means that someone's going to buy your hardware. Someone's going to buy your infrastructure. And they're going to set it up and they're going to manage it. And, you know, you don't, roaming isn't an issue because it's a, it's just, you care about it. Right.

Chris Gammell: You leave the campus, you're no longer covered. Right.

Dave Jones: Right. Right. And so there's also, and you can extend it from there. I mean, the reality is that both of these things can coexist. Like if I, if I have some technology and I really need it for my campus deployment, like for example, LoRaWAN, you cannot move that much data. The frequency of moving data is quite limited. The maximum you can do is about once per two minutes. So I can move one, one piece of data upstream every two minutes. And, you know, downstream is not that practical either. Moving data downstream. It's, it's kind of, you know, this is a technology that was really designed, LoRaWAN, with kind of meter reading in mind. I think meter reading was, was maybe the major use case when, when they went in and designed it. And that's understandable. It's a big kind of, a big use case. It's got not a huge number of customers. So, you know, really good. If you're, if you're an enterprise and you want to go deploy some technology, it's, it's nice not having to have a lot of different customers. Like, like retail would be the ultimate opposite of that. You know, when you've got millions of independent customers, like the Bluetooth tracker market would be like the absolute opposite of this. And so, so it's nice if you don't, the fewer customers you have, I think it's, it's a lot easier to, you know, have a market strategy. What they want in version two, you ask them, right? Yeah, right. Exactly. And then they'll tell you. It's not intuiting and trying. No, yeah. There's no, there's no focus groups. There's no, you know, there's no rooms of 50 people with a half silvered mirror. That's right. Silicon Valley. You've seen it in Silicon Valley. Yeah. You don't have any of that. So that's, that's real nice. And especially for industries that are pretty young. You know, the ones that don't, don't have a lot of history and capital and behind them.

Chris Gammell: Yeah. Yeah. Well, let's talk about how this relates then to what you designed for Haystack and more on contracts. Sure. So what did you do as a result of this?

Dave Jones: Yeah. So Haystack is a company. It's been around for a little while, you know, on paper since about 2010, but in reality since about 2013. Um, so Haystack was a company that was born out of work I had done for a different company. I was an employee of a company called Savvy SIVI, uh, technology and Savvy was eventually bought by Lockheed Martin, the big defense conglomerate. Um, Savvy made wireless tags, uh, active tags, not passive RFID tags, but they made, um, you know, basically wireless sensor tags that go on shipping containers and go on shipping pallets. And the military used these, uh, quite extensively between about 1998 maybe, and probably even up still today. Um, but these are, these were used for their in-transit visibility network and, um, eventually.

Chris Gammell: So like, uh, asset tracking, but on a lower scale.

Dave Jones: Yeah, asset tracking. Asset tracking on a global scale. So, um, like in a typical deployment, um, you know, they'd have containers, shipping containers full of stuff. They'd dump them somewhere in the desert or whatever, especially in during the, the sort of Iraq and Afghanistan war era, um, where they just go and dump a bunch of containers full of stuff, uh, somewhere. And the supply people in the military would have to find the stuff. Uh, and so this helped them a lot, you know, it's.

Chris Gammell: It's like finding a cache in a video game, you know?

Dave Jones: Well, yeah, uh, pretty much, um, the, the, you may have gone past or gone inside military surplus stores in your lifetime and you've seen them and they're still around. Yeah. Um, but the old, the old way that the military used to deal with supply chain, uh, challenges is to have 10 times as much stuff as you need.

Chris Gammell: Right.

Dave Jones: No, that's actually the number 10 times. Um, so through in transit visibility, I think they got it down to about three times as much stuff. So that's, yeah, three times, only, only three times. Only three. That's a big improvement. That's a big improvement. So, so, you know, through, through technology, they were able to improve their supply chain and make it more efficient. So that's, that's good. That's good for everyone. Um, but, uh, so. So I guess the reason why I'm bringing this up is that, uh, that was a cash cow for savvy because they were a sole supplier of this technology that they built. And eventually if you're a sole supplier and you have a DOD customer, what they're going to do is require you to standardize it and, uh, license it to your competitors, uh, in order to. So that they can have a standardize it among other people.

Chris Gammell: Yeah, exactly. Yeah.

Dave Jones: So I was involved in that. I, so I got, I got hired to that company to support, uh, licensees of this technology. And it was a strange relationship. Like the, the, the rest of the company did not like our group. Uh, you know, the engineers did not like us. The, the, so we were, we were not very popular. Yeah. Um, but, um, why, why, why weren't you popular though? Just because you had to support? Well, I mean, we were, we were basically, you know, we, we had to exist because the military said you, you guys need to license your tech.

Chris Gammell: Um, but, but you were also butchering the cash cow. Is that kind of the problem?

Dave Jones: Yeah, exactly. We were, we were kind of, and it wasn't a lot because we, I mean, you hire me to do a job. I'm going to do my job. But, um, sure, but it was cool because it, it exposed me a lot to standardization, technology standardization, um, and, uh, wireless technology. It, it, a relatively young age. So this was my late twenties when I was doing this. Um, and we did go through, we did get the technology they had standardized. So it is an ISO standard or actually ISO. It's not ISO. So, but, um, so there's, there's a standard called ISO 18,000 dash seven, which was the first version of, uh, of that technology. So, uh, and that's the one the military uses. And after we finished that. Yeah. I was going to say, just figure, just figure out the product name. Okay. Yeah. We'll get there. Yeah. So, so after we finished that, um, we finished doing that. I think that was around 2010. We got all wrapped up and all of the other suppliers and vendors were happy. And we said, okay, well, you know, what's next? Uh, so let's, let's, let's take all the things we've learned over the last 12 years, uh, all the new technologies, um, and make a second version two of this, of this, uh, of this technology. Uh, and, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh, uh version of that. And so, um, then the benefit is that we had all of this, this marketing, we had this huge, uh, I mean, calling it marketing is, is maybe not exactly what you'd think, but brand behind it. Yeah. I mean, we had this, we had this, no, we had this huge group of people who had years and years of experience with, um, this sort of product. And we were all in a room and we were all trying to, you know, build, build products to sell the military. And so there was just an enormous amount of, of what works, what doesn't work, just knowledge about how to deploy. Yeah. How to deploy and how to design these, uh, these sort of wire, mobile wireless data, uh, systems. And, um, so because we had all of that, it enabled us to build a spec very easily. We didn't have to go out and figure out, um, didn't take like 10 years, like Bluetooth did or whatever. Well, I mean, well, maybe it does, but, um, Bluetooth, you know, Bluetooth also has a lot of people in a room and they, they, they, uh, they're become very sophisticated. They become very efficient. It didn't used to be that way, but, um, the dash seven is a way smaller and it's way more grassroots and it's, uh, you know, but the dash seven Alliance, uh, remains, uh, it's still out there. Um, it's mostly a European presence at this point. Uh, the technology itself is used. I mean, I think mainly in indoor kind of mapping and indoor location, uh, also in parking lots, uh, other really unsexy things like this, but, um, you know, stuff, stuff that needs to get done. And it's at a campus level. It's not, I mean, you could call it a WAN since the range can get quite long. Um, but it's not, it's not like a cellular system. It's not generally deployed in a roaming way. It's not tempting to be that it's attempting to be a campus, uh, almost like a LP LAN, if you want to call it that. So, okay. So, yeah, so, so it, it has LAN like features as bi-directional communication, um, and kind of bursty asynchronous, um, access. So it's, it's very different in how it behaves, uh, as a networking technology than LoRaWAN. And, um, um, so it, it makes a good partner with LoRaWAN is the bottom line. If you have a LoRaWAN network and you want to add some, some LAN features, um, you know, Dash7 is a good option. It's not something that is so mature that you can just go and buy parts for it, uh, at this point. But, uh, usually the kind of people who are going and, um, contacting me or contacting various people in the Dash7, uh, community, uh, about Dash7 are people who have tried LoRaWAN and it just hasn't worked for their application, uh, from one reason or another. And, uh, you know, for the last seven years or so, we've been adding features as needed, um, adding technologies as, as people demand them. Uh, so we, we need to have, you know, firmware over the air. Okay, well, we'll, we'll, we'll put that in there. And lately over the last three years, really, it's been all about range. So it's just been, um, how, how do I get more range? How do I get more quality of service? Uh, Semtech and the LoRaWAN Alliance have done a fantastic marketing job. And, uh, so we're using a lot, the LoRaFi right now. So we're using the LoRa hardware, not necessarily the LoRaWAN stack. And, um,

Chris Gammell: can you give us a, an idea of what the, uh, relative distances are like for just maybe a, a LoRa, uh, LoRa chip or transmitter by itself versus when there's, you know, maybe quality of service on top of it or when it's using something else. So quality of service just means, um, it's,

Dave Jones: it's kind of a more of a high level term. It's not really a technical term. Uh, and it's used heavily in the cellular industry. And it's, it just sort of means like, okay, I have packets and I'm transmitting them and you're receiving them correctly and you're receiving them correctly most of the time and in most conditions. So that's good QOS. That's good quality of service. And if you're not receiving packets, if you were dropping packets, um, then that's bad quality of service. So you want high quality of service and there are a lot of ways to, to get there. Um, the dominant ways of improving quality of service are having, uh, methods for retransmission or efficient retransmission and also, uh, error correction, forward error correction. So you can use, these are, this is not, um, I mean, there's nothing mysterious or secret about this. Every, you know, every cellular system is, is, I mean, probably hundreds of thousands of man hours are spent by the big players in the cellular industry, uh, to improve quality of service through better error correction technology or through, you know, countless other ways and proving the antennas and the, um, you know, the, the cellular industry is so mature, you know, they're chasing tenths of decibels. You know, they're, they're really, they're really kind of at the 99%. Um, and they're trying to get to a hundred percent, uh, LP when is, is not there. I mean, it is not that mature. Um, you know, you can implement like something that Haystack did recently. Um, uh, we, we have, uh, sort of a history of taking technology from deep space telemetry. So that's, uh, technology used for error correction to ground to space probes. And a lot of the stuff that's just available. I mean, you can just go and, and download it. Uh, not usually code. It's, it's usually some form of,

Chris Gammell: uh, math documentation. Yeah. Algorithms and similar kind of stuff. Yeah. Uh, so I, I'm,

Dave Jones: I'm not a, uh, I'm good at math, but I'm not one of these people who's building these error correction, um, schemes. That is, it blows my mind. Uh, how, how they come up with this stuff, but I, I do know. Error correction?

Chris Gammell: Can you explain what that is real quick for people that don't know what that is? Yeah. Yeah. I, um, I remember I saw Mike Osman talking about it, uh, talking about it at like a, SDR presentation and I didn't, I didn't understand it prior to that, but, uh, it's,

Dave Jones: it's a pretty crazy technology. I mean, it's, it's not a new technology. Uh, it's stuff that's been around for a long time. It all, it all goes back, I think to the thirties or forties, 1930s or forties, I guess I got to say now. Right. Um, there was a guy called Shannon. Yeah. Information series. So he started it all and, um, he came up with these, this thing called the Shannon limit, which is, um, basically how, what is the best an error correction scheme can be, you know, theoretical maximum that, that you can get of gain of improvement, uh, you can get from error correction. And it's a, it's a curve because, um, you know, I can send a lot of redundant information and, uh, and so that way it's padding the length of my, of my packet. It's padding the length of it and I can get like really impressive gain. However, it's not really genuine gain because if I just slow down the baud rate, I can also get gain. And that's just because like, if I just slow down the baud rate of a transmission, like, so if I take, if I take something that's at a hundred K bits and I bring it down to 10 K bits, every bit has 10 times the length of it that you previously had, 10 times the duration. So it's 10 times the energy, if you think about it that way. And I have 10 times the energy per bit, then I'm going to be 10 times as resilience against noise. So, you know, just by increasing the, um, just by increasing the, the, or decreasing the data rate, increasing the length of the bits, I can improve, uh, you know, the chance of sending a correct message. I can, I can decrease my chance of packet loss, but that takes more energy, uh, to transmit because it's, it's longer. So, you know, I, I mean, I am trying to find this trade-offs. I'm trying to build new technologies and I'm using I loosely and I assume that I'm just some guy, some communications engineer, right? I'm some communication that you're trying to improve the power efficiency and trying to improve the quality of service of my, of my connection. And, uh, I can, you know, I can, I can drop the data rate. So that's, that's kind of what Sigfox and LoRaWAN have done. Uh, and LoRaWAN is very good at just bringing the data rate down, uh, because of the way the modulation works. So the modulation of LoRa, uh, it's, it's very easy for the receiver to track the, uh, the transmission. And so that means not to lose synchronization with the transmission. It's very, very good, uh, for that. And so it enables, um, low data rates to be used, lower data rates than were really feasible, um, with more basic modulation schemes, uh, and cheap radios. So if I have like a basic modulation scheme and, uh, I want to run it at a really low data rate, I need a pretty sophisticated receiver, um, to track it over such a long period of time. And, um, but LoRa, LoRa works that way. I mean, with, with LoRa, I can, I can bring it down, I think to like 18 BPS if I want or something. Yeah. I remember seeing some really low ones. Yeah. Yeah. So that's, that's kind of the way that LoRaWAN, um, uh, deals with, uh, trying to improve, um, to decrease packet loss. And it's just to bring down the data rate and, and Sigfox to a large degree as well. Um, another way to do it is to add error correction and error correction. I mean, there is a little bit of built-in error correction in the LoRa hardware, although it doesn't really work. Uh, that's, that's, I don't, yeah, I think, I don't know. Uh, actually, so there's a guy, his name is Matt Knight, and he was tasked at some point, um, a few years ago to, uh, try to reverse engineer the LoRa 5 specification and did. And, um, I mean, we all had our suspicions about it and then, you know, he kind of erased all those suspicions. So credit to Matt Knight. Uh, if you were interested in this stuff, go, go hit, uh, Google search for, um, I think it's called Matt Knight reversing LoRa. I think that's, he has a bunch of

Chris Gammell: presentations and, uh, or, uh, actually hear him on the Amp Hour. Uh, I interviewed him at TorCamp, uh, uh, short, 20, 30 minute interview. Super. Yeah. Check that out. Check that out too. So, uh,

Dave Jones: so the, the, the reality is, is that the, the integrated error correction in LoRa doesn't really work. Um, it's, it's kind of parasitic actually. So, uh, don't use it. Um, okay. Yeah. Don't, or don't, don't use it. And, uh, so if you want to put error correction on LoRa, you got to do it yourself. Uh, if you want to put good error correction on LoRa, you have to do it yourself. And, um, so that's something that, that I've spent a lot of time in, uh, is doing very good, uh, firmware error correction. And, uh, that is to add data to the message. So adding bits to the message so that I can reconstruct erroneous bits. And, um, it's, it's very complicated, but like the idea is kind of, I mean, it's, it's like human intelligence is pretty good at error correcting. I mean, we do it all the time and we don't think about it. Pattern matching and then like, and really in this case, so in, in data error correction, I think the analogy is about, you know, having a vector in a multidimensional space. And it sounds like it's going to go off into some horrible story, but, uh, it's, it's pretty, you can think about it in a 3d space and that's, that's very visualizable. And, um, and you know, if some, if this vector gets sort of corrupted a little bit, uh, you know, if I can add points on the vector, if it's not just two points, if I can add some points, then, you know, I can still kind of figure out which direction it's heading in. Uh, if it's not two points, if it's more than two points. So, um, you know, if one of those points gets skewed off a little bit, I can still kind of extrapolate if I'm, you know, heading north or south, uh, or one or zero. And yeah, that's the fundamental concept of error correction. And there's so many different ways to do it. There's so many different mathematical, you know, kind of back ends to it. And I don't want to, I don't want to get into that, um, too much.

Chris Gammell: The one that I had heard is like, uh, and I think this is the, from the Mike's presentation I saw was like, okay, you're trying to transmit a one. You can transmit one, or you can transmit one, one, or you can transmit one, one, zero, one, or you can transmit one, one, zero, one, zero, one. Right. And like the longer that number is, and it's a unique number and you have that. And then maybe the inverse of it is the other thing you could start to basically, as you approximate that chunk of a number, it looks more, it's, it's easier and easier to suss out whether that long, long number you've created and this pattern you've created is supposed to be a one.

Dave Jones: Is that, is that a good explanation? Well, that's, that's, yeah, that's sort of one. Yeah. That's basically, that's the Claude Shannon, um, explanation more or less. And, uh, one, one implementation, one of the earlier implementations of error correction code is something called a convolutional code. And that's just, I've got, I encode my data in a certain way based on a generator, you know, which is like a bit pattern and I can take a moving window across my signal and look at the, you know, look at the ones and the zeros and in my message that I'm receiving and I can map them onto a state, a system of states. And, um, you know, if, if I don't have too many errors, uh, I can fix a certain amount of errors because they will like my, my signal will sort of coalesce around states, you know, one state, one state or zero state, uh, effectively. So that's, that's an early form of error correction. A lot of, so there's a, another standard, you're going to get alphabet soup. Um, if you go into the, the wireless standards, um, cause you know, like there's 802.11, everyone knows that there's so many more, uh, a lot of 802. 802.15.4, right. So 802.15.4. And then you got 802.15.4, like a through X or whatever. Uh, so in one of those, I think it's F G F or G maybe, um, they specify a convolutional code, uh, for error correction and the convolutional code that actually, this was something that was implemented on a TI chip many years ago. And I think they just brought it into the standard, but since then it's been put on a bunch of other chips. I know it's on some ST transceivers and probably others as well. Um, so that, that convolutional code is pretty simple. Uh, it works reasonably well. I mean, you will get some gain out of it. Um, you can get maybe like three DB of gain with a fairly naive convolutional code. Um, and so that means like, like I'm improving my sensitivity by two times cause I've got three DB. Yeah. That's pretty good. Uh, I can make it even better if I do it in a non-binary way. And so if I, instead of, if I, if instead of my inputs being zero and one, I can make my inputs like, you know, one or zero to 255 or zero to 64. I mean, often it's just zero to 16. That seems like it's enough for a lot of these, uh, they call it soft decision, uh, instead of hard decision, meaning bits one and zero. Um, so in these soft dishes and I did more gain, you know, I get another two or three DB just, just from having it soft decision, I get better, uh, coalescing around states, uh, in my decoder. So, um, that's always good. It's something that it's hard to do in software. It's hard to have a soft decision, um, um, soft decision decoder written in software using a cheap transceiver, because they don't usually give you this information. They don't usually tell you, all right, you know, I've demodulated this part of the signal and I'm going to tell you 15 out of 16, usually it's just one or zero. They, they, they clip it for you. So, um, if you're doing, if you're trying to build firmware that does, uh, um, error correction of a low cost transceiver, and you want to get that extra two to three DB of soft decision decoding, you have to be kind of clever. Usually you have to do things that, um, usually you're not going to find in any example code or no, nobody, nobody, nobody's LoRaWAN library is going to be like flogging the chip is real hard to, to get that soft decision information. Um, but you know, if you want to, if you want to really push it, then you can, it's there. Um, so this is, this is the kind of stuff that, that you, you've developed

Chris Gammell: then you've developed like a alternative to the, the basic stacks that are out there, you're saying, and that's like something that you would license out as a, as a design.

Dave Jones: Yeah. I mean, that's something that we, we do license out and often, you know, often it's just something that sweetens the deal, you know, in the consulting it's, it's tricky to do to license technology directly. It's tricky to be arm, you know, it's tricky to be, it is tricky to be arm, unless you have like soft bank money, that sweet, sweet soft bank money. And then you can do a lot of things. Yeah. Or just a huge backlog of success, right? People really have to trust you to just do pure licensing. And usually your customers are going to have to be pretty big and they're going to have to develop their own stuff. So a lot of times the, the technology, the proprietary stuff that we've done is, is just, it sweetens the deal, right? It's, it's almost for marketing purposes. It's, it's, uh, like if someone's going around trying to find a solution to their LoRa problem, you know, why don't we look at the guys who implemented, um, like LDPC error correction over LoRa and get 14 DB gain over, you know, LoRaWAN? Like that sounds incredible. Right. Let's talk to

Chris Gammell: these guys. Yeah. We should probably talk to them instead of the people who just follow the regular

Dave Jones: manual, right? Yeah. Right. You know, if, if you're trying to get the best range, you know, go after the technology that at least claims, you know, give it a shot, right? You don't, nobody's ever going to just believe you, but if you send them a dev kit or, you know, in our case, like the first, the first thing we usually do is, is, um, uh, there's an ST dev kit and it's like, you know, just get the ST dev kit, install this, install this binary on the ST dev kit. And, you know, we have a very simple range test and, um, you know, go, go off and test it yourself. Yeah. And if you don't get, if you don't get better range than LoRaWAN, like, I don't know what you're

Chris Gammell: doing wrong. But what is, what are the relative ranges though too? So like if you send someone, they have this dev kit, it gets what, like 500 meters, a thousand meters. It depends a lot on environment. Um, sure. Sure. So just give me any, any numbers.

Dave Jones: Yeah, sure. So in the 900 and 900 megahertz range, if you do, if you do like a test, that's like a stupid test. So this, this, our device is running at one K or our reference one K bit per second, which is, um, it sounds really low. It is, it is low, but in the LoRa, in the LoRaWAN world, that's sort of in the middle. Um, so, uh, we do one K, we get better performance than any observable LoRaWAN, um, uh, data rate. So, you know, LoRaWAN at 125 bits per second is worse, uh, than us at one K. So if we put it on a high location and you go driving and you go driving, um, basically it's until you hit the horizon. And, um, if you put it on a mountain, that's going to be 36 miles or whatever. Yeah. If you, if you put it on, um, the roof of a house, it's, it's going to be less, uh, you know, maybe like two miles, three miles, but, but, uh, so yeah, I should also say that it depends a lot on the power transmit. Um, but we just, we're, we're using like the, just the LoRa chip. We're not putting any amplifier on it. So, um, you can amplify it. You could get more. I think we would still beat LoRa even with an amp, if they had an amp or LoRaWAN, if they had an amp, but, um,

Chris Gammell: but, uh, okay. So this is like software making, making the hardware perform better.

Dave Jones: Yeah. No, it's just, it's just, yeah. Just pulling bits out of, uh, you know, where bits didn't used to be, I guess. Yeah. Just like information, man. It's crazy. Yeah. Just, just finding it. It's, it's like, it was shocking. Uh, the first, the first time I tested it, I actually couldn't, couldn't believe how, how well it worked. Um, uh, a lot of times with these low cost radios, one of the issues is that the, the tracking is, is just a problem. So if I just don't have that strong of a signal, it's just going to lose timing. It's going to lose synchronization. It's going to lose bit synchronization, but, uh, LoRa seems to not have that problem. And, um, this really enables some strong error correction, much stronger than on like a standard, um, like PSK channel, a phase shift keying or an FSK channel. Um, so that's really cool. I mean, like I would love, uh, I think, I think there's a huge opportunity for, for, you know, LoRa to, to have future versions with much better error correction. I mean, they, they could improve their throughput enormously, um, because the FI is, can take it. The FI can, can, um, uh, you know, it won't lose tracking. It's, it's really the, the, the encoding itself becomes the, um, like the, I don't know the weakest

Chris Gammell: link, I guess. Yeah. Yeah. That's great. Well, you mentioned, so you mentioned you started, you work with a lot of startups on this kind of stuff as well. So like, how does that end up manifesting with, you know, is it, is it, that's who's pushing for the range in this case, or like, what is it like working with startups? Um, yeah, I mean, it's oftentimes startups are

Dave Jones: willing to take a risk. So, uh, if you, if the big companies, um, they're not usually worried about cost, but they're worried about timing, market timing. They're worried about, um, they're worried about failing and they're worried about doing something that, you know, maybe if it's different, you know, there's the old saying, nobody ever got fired with IBM. That's right. Yeah. So, yeah, I mean, it's, it's doesn't change. I mean, the IBM part changes, but, but, um, that's sort of the, nobody wants to get fired. And if you're the kind of person who is corporate, um, you know, you, you're there because you don't want to have risk in that part of your life. You know, you don't, you don't really, and that's like an absolutely fine life choice. Um, but, um, if you're, if you're some person who's like a serial startup or, I mean, you're, you're like kind of a risk junkie. Yeah. Right. Absolutely fine too. Too boring. If I just used the off the shelf thing. Yeah. It's like, cause I'm, I'm doing this cause I want to do, you know, my idea the best. I want to, I want to be better than everybody. Um, so, so you're willing to take a risk. I mean, that's the whole point of startups. And so if you have like some technology that's, that's different, uh, but it claims to be better than, you know, startups are often where you're going to, um, you know, get the most interest and, you know, working with startups, it's, it's, uh, startups have the issue with, with working with startups is that they usually don't want to be time and materials. Like if you're working with a larger business, then, then time and materials is, is kind of standard and they expect it. And it's, it's, you know, it's low risk, right? You get paid for the time you put into it. It's nice. It's real nice. Uh, but, um, yeah, sometimes you got to go and do a fixed price or a, um, you know, a contract that's sort of quasi fixed price where estimated hours with the ceiling or something like that. The way that I've usually gone about it is that, um, you know, I'll present a list of deliverables, milestones, if you want to call them milestones, and there will be a payment on milestone of some, some amount, some prearranged amount. And, um, you know, if, if you guys decide to part ways, then, you know, we don't continue, right? We stop, but at least you get something. And, um, the other thing you can do in these milestone things, that's, it's sort of a trick. It's not, it's not a trick, but it's hedging, uh, is that I would recommend as a strategy to anyone who wants to be a startup consultant or a consultant who works with startups. There's a bunch of things I would recommend, but one of them that's pretty simple is just to make sure don't pad your schedule, but, um, there are going to be certain milestones that are gonna be really hard and they're going to take a lot of time. And there are going to be certain milestones that are going to be not hard and they're not going to take a lot of time. My advice is to kind of normalize the hours on, on these things. So, you know, that way you can, you know, you're hedging your bets. So if, you know, there might be one milestone where you're, you're effectively getting paid, you know, 50 an hour, and there might be another one where you're effectively getting paid, you know, 200 an hour or, or whatever. Right. Uh, but you know, you just want to make sure that, that at the end of the day or at the end of the contract, um, your average billable hours is what it needs to be. So I, I think that that's, that's the other thing to support. The other, the other little tip is, um, a lot of times if you can sell hardware, you know, sell dev kits, I don't care, sell anything. Uh, you can put cost into that. You can put, um, you can offset your hours, uh, by adding cost to the dev kits. A lot of times that's okay. Sometimes it's not, um, a lot of times people are happier, particularly in sort of like what I would call small businesses, but not startups. Um, you know, they're happier to get something. Yeah. They're happier to get something, uh, than to just, you know, pay for hours or pay for NRE. Um, these are just tactics. And, and, uh, as long as everyone's happy at the end of it, you know, you, you don't have to feel dirty. Right. Um, if someone's not happy, then yeah, you, you feel bad, but, and you're not going to always win. You're never going to always win. It's not every, every contract is going to be a total victory. You're going to have the occasional disappointed, you know, customer, and it's just going to happen. Right. Um, but, uh, you know, if you can, if you can, if you can have happy customers most of the time, then, then yeah, you're good. Yeah. What are the,

Chris Gammell: uh, what are the different areas that you see people? So specifically like using, like using Laura type stuff or the dash seven stuff and trying to get long range. What, what are like some of the

Dave Jones: areas that you see people using this technology in? Uh, well, uh, it's for, for dash seven over Laura. Yeah. It's a lot of mobile stuff, not fixed assets. So if you have, if I got fixed meters, then or fixed anything, if something is fixed in space, I can build my network around it. And I can make sure that the quality of service is good enough. I don't have to worry about, you know, what's going to happen tomorrow because the thing is going to, if it's not connecting, it's because of one of a number of issues that, you know, are usually not, or it's either temporal like, okay, a FedEx truck parked in front of the meter. It's in between me and the, and the meter, or it's going to go away soon. Or, you know, something else was like, oh, you got hit by lightning and you laugh. Just that little thing. Yeah. But like, uh, in, in lightning prone areas, this can happen, you know, in Florida, you know, like I lived in Florida briefly, um, like a while ago, but, uh, uh, you would see, uh, traffic lights occasionally that were just on the ground and you're like, what, what the heck is going on? I mean, it's cause they got hit by lightning. Okay. So, so yeah, I mean, stop. I'm sure that cellular base stations get hit by lightning all the time down there. Yeah. Um, so, so, you know, it's possible or, or it could be something like somewhere in between, right? Somewhere in between getting hit by lightning and getting blocked by a FedEx truck. But you know, nonetheless in the base case, so in the, in the, for most of the time you can figure out what kind of link budget you need, what kind of sensitivity I need, what kind of power I need to transmit at, you know, what kind of antenna I need to build, how many gateways I need. You can figure all that out beforehand and you can build your network to meet that requirement. Um, when you start to have stuff that moves around, it gets a lot harder, uh, because you can't predict where something's going to be. It's like, if I have like animal tracking for like of a cow, where is my cow going to be tomorrow? Um, you know, if I have, uh, like if, if I have a, um, a bunch of equipment, like, uh, um, if I'm in a mine and I have all this mining equipment and it's all over the place, um, you know, I don't know necessarily where something is one day versus the next day. Right. Uh, or, or, you know, especially up in Canada, it could snow and then I've got snow on top of it and I've got attenuation. So these are the sort of cases where people tend to come to Haystack. They tend to come to us because it's, it's like, we have a really challenging environment and we need all the pro we need all the QoS we can get. So that's, that's in a nutshell, that's it. I don't want to, we have a couple, um, we have a couple customers at the moment, you know, at the, at this exact moment, I don't want to reveal their exact, uh, verticals, but you know, I can, it's fights to say they're all moving things.

Chris Gammell: Okay. Yeah. That's, that's a, that's a good, good overarching type thing. I was just thinking like, so when I think about a lot of the lore stuff that's out there too, or the lower land, I guess, um, you know, it is like, yeah, it's like sensors in a field, right? It's maybe like agriculture type stuff or it's a, it's a meter, like you said, or just a sensor that's sitting in the middle of, in the middle of nowhere and trying to transmit back. But at a certain point, it's, it's not, it's not, it's pretty fixed. So that, that helps.

Dave Jones: Yeah. It's just, it makes your network deployment much simpler. I can build a good, solid network. I can test it. I can make sure it works. Um, but you know, as soon as stuff starts

Chris Gammell: to move, then everything gets harder. Yeah. And so you mentioned link budget too. Could you explain

Dave Jones: what that is? Yeah. So link budget is just, it's a term used by us radio people, uh, to describe the amount of energy that is between let's take the difference of the transmission power and the received sensitivity, the best case or minimum received sensitivity of the receiver. So if I, if I take the difference of those two things, I get a link budget. Usually it's expressed in terms of decibels because, uh, communication engineers put everything in decibels. That's just what we do. Um, and, uh, uh, so like a Laura link budget, you know, if in America I can, I can use, or I should say in ITU region two, uh, which is, which is North America. And I think certain parts of South America as well. Um, you know, I, I have a regulatory environment that allows me to use, uh, let's say in excess of a hundred milliwatts if I, if I really want of output power. So a hundred milliwatts is, uh, it's 20 DBMs. So that's like a DB milliwatts. Um, and you know, a Laura receiver, uh, you know, the book value might say, oh yeah, we can, we can, we can receive it down to about minus one, two, nine DBM, uh, 129 minus 129 decibel, which is pretty, pretty low, pretty low. It's, it's not a lot of, it's like, you know, 10 to the minus a lot, right. Uh, uh, uh, milliwatts. It's all, it's not a lot of power. It's really impressive that radio works at all. Right. Yep. Uh, uh, but you know, in reality, what, what you can observe is about, I think with, with a typical Laura WAN at one K, I mean, you'll, you'll be lucky to get about minus one, two, one minus 121 decibel. So, so if we have, if I have 20 DBM, um, at the, at the transmit and I have minus 121 at the, uh, receiver, so then I've got, uh, 141 decibel link budget. That's, that's pretty good, right? That's, that's not bad. Um, something like, like wifi. I mean, you can, you can pull up a lot of war driving apps. Um, you know, there, there are quite a few of these. I have one on the Mac called iStumbler, but I'm sure there are dozens of others and, uh, you can see all the different wifi networks around you. And they usually tell you certain things like, uh, since like, um, they'll give you something called an RSSI, which is just receive signal strength. And that's in a, in a DB, it's going to be a negative number. Uh, like a good, a good, strong wifi signal is going to be better than usually minus 60 DBM. And they'll usually give you some SNR, which is like signal to noise. So, um, you know, that's just, if you have a lot of wifi network interference around you, that might go up, you know, you're, or might go down, your signals of noise will be worse because you have, um, other wifi devices in the environment that are interfering with yours and they're creating noise. Um, so in the 900 band, there is noise, but it's not, it's not at all like the case with wifi is much, much less. So, uh, you don't have to worry that much about, about noise in the 900 band. It's, it's not nearly the problem. It is at 2.4 gigahertz. And, uh, and that's good. And especially outdoors. Yeah,

Chris Gammell: definitely. Especially outdoors. Well, when you're trying to get like lots of range too,

Dave Jones: you want to make sure you cover that area. So it, and, and, um, so anyway, the, if I can improve that minus one 21 to let's say like minus one 33, uh, that's, that's 12 DB more of link budget and, uh, in open space, uh, that means like do the way that fall off works. So, uh, like fall off is, is kind of quadratic. It's a relationship. And, um, if I have six DB more of link budget and open space, I can double the range. So if I have 12 DB more of link budget, I have four times the range and, uh, you know, that's just to clarify that the numbers too, it's because the,

Chris Gammell: it's the minus it's going down, right. It's going from minus 21 to minus 30, one 33, whatever you said, that's actually just the, how sensitive it is. Right. That's like saying, yeah, that's like,

Dave Jones: the signal's that low, I can still get it. Right. If I add the error correction, um, you know, I can get some extra sensitivity. That's what that's saying. So if I, if I don't have the error correction, the best I can get the, the, the weakest signal I can receive without errors, uh, is maybe minus 121 DBM. And if I add the error correction, the weakest signal I can receive reliably is minus 133. Well, in this particular case, uh, and different error corrections and different modulations are, it's all going to be different. But, uh, in this particular case, I can get a 12, uh, DB gain, um, from, from adding error correction. And that's going to result in an enormous, uh, improvement of range and quality of service and, um, and, and particularly just reduced packet loss, uh, in sort of fringe areas. So, um, yeah, it, it, it's not going to be four times in most scenarios, like most scenarios where we're going to do deploying any kind of LP WAN or LP LAN and it's not space, right. It's not, we're not, we're not shooting this to the moon. So, so we have other stuff in the way, we have obstructions, you know, I have ground bounce, like just the ground, the ground is the worst. Uh, the ground is, is your enemy. Um, if you can put your antennas high, I mean, you can make a huge difference, even just one antenna high. Uh, if you can put one antenna on a roof or if you can put, you know, on a pole is even better. But, um, if you can, if you can raise an antenna in your system and that's why all the cellular systems you'll ever see are on, are on poles. That's why they're not on the ground. Um, and just because the ground is, is, is such an enemy, uh, it, uh, the, the reason for this is something called the Fresnel zone, or I think if you, if you want to be actually correct, uh, it's the Fresnel zone, zone, but I, I'm not French. So, so, uh, but that's just that, uh, we, when we see light with our eyes, we, we tend to think of, of it being just like a ray, like a point, you know, a line, like a laser or whatever. And, uh, we tend to think of line of sight as really like a direct geometric line, but, um, that's just because light is such a small wavelength, visible light. I mean, you, a lot of you guys probably know is, is between, um, I think like 400 and 660 nanometers. I think that's, that's the wavelength of visible light. Right. That's, that's, that's small. It's, it's, we're talking terahertz. Yep. Yep. Um, so, so if we, if we take this down, if we take this, this thing down to from, from terahertz into sub one gigahertz, um, you know, it's, it's definitely not a ray, uh, it's, it's a field and the field can get impacted by, uh, by things that you do not think are in the line of sight, you know, the line of visible sight, but are in the way of, of a lower frequency. Yeah. So the ground, like if I have, if I have, uh, um, yeah, you know, if I have two antennas and they're both raised by two meters, um, you know, I will, that's not necessarily line of sight. Uh, that's probably going to be non-line of sight. Um, you know, if I go beyond maybe, you know, 10 or 20 meters distance, it starts to become non-line of sight at,

Chris Gammell: at 900 megahertz. And that's because you're saying the ground kind of cuts off the, the, it rounds off the bottom of that, the transmission path kind of thing. Yeah. There, there's sort of this,

Dave Jones: like, I think people will, might, might analogize this as like a football shaped, that's a American football shaped, uh, uh, you know, sort of, um, field that goes between the two antennas. Yeah. And anything that gets in the way of that is, is going to make it a non-line of sight, uh, transmission. So if that, if that's not, if that's interrupted in some way in the bottom or the top or in the middle, um, yeah, it's going to block, it's going to reduce the, it's going to cause attenuation is the, the technical term. So we're going to attenuate the, the signal and attenuating the signal is means less range. Yeah. Huh? Well, um,

Chris Gammell: the, the long distance stuff is definitely interesting to me. I mean, like the fact that you're able to extend it like that. Um, I want to also talk about the other thing you've been working on the, uh, soul pad. Could you, so this is also in the RF domain, but it's, it's, you're, you're doing the RF stuff on it, but it's, it's kind of a broader, a broader project,

Dave Jones: right? Yeah, sort of. It's, that's a good way to put it. I'll explain, I guess. Um, so one of the things that, uh, one of the sort of sub threads of this, if this conversation is about, uh, being a contractor or being a consultant and, you know, one of the things that I think you kind of have to be ready for is bouncing in between, you know, kind of full-time and consulting and, and, you know, sort of the hybrid of that, like full-time consulting. Yeah, that's true. Yeah. Being

Chris Gammell: like on contract for like your, your entire chunk of time. Right. Yeah, exactly. So sometimes,

Dave Jones: you know, you, you, sometimes you're going to, you're going to go dry and you're going to, you're going to need to take up work or sometimes something might just come to you that, that you really want to do and it's full-time and maybe it's a little bit different than what you've been doing. And that's the case with, with Solpad. So Solpad is, Solpad is not a solar company. Uh, it's sort of seems like a solar company, but it's, uh, it's got soul in the name. So yeah, it's got soul. Um, it's a, it's a company that builds battery backed storage and conversion, uh, solutions that are, you could use them, I guess, with any DC input, but solar is sort of the predominant, uh, input. And, uh, this is a startup company, although it's, um, you could call it like it's a fairly mature startup company and, uh, it's a very, very close to launch, uh, of the product and has been, been on for a while. And it is like some people lament kind of the death of Silicon Valley, you know, Oh, it's just a bunch of media and, you know, a bunch of software and a bunch of selling ads to extracting ad revenue. There's still a privacy problem. Don't forget those. Yeah. There are, there are, there are still some, some, you know, hardware companies here. Uh, there are still some, um, you know, I, and I consider Solpad to be kind of one, one of those companies that's, that's trying to do something, uh, kind of really bold and challenging and different. And I mean, the core, the core idea here is that there's a really sophisticated, uh, inverter technology that is being pioneered, um, by Solpad and, um, they, it uses a, uh, uh, a gallium nitride chemistry for the switching electronics. And it's just, it's just a new type of semiconductor. It's a new, well, it's probably been around for, for 10 years. I don't know.

Chris Gammell: Yeah, I think so. We had, uh, I'll throw people back to John. Uh, oh man, I forgot his last name. No, he was from Cree and he was here and he was, he was on the show probably about six or seven years ago now. And he was talking about when, when they started making gallium nitride, uh, switches, um, at Cree.

Dave Jones: Yeah. I think, I think it came from blue LEDs originally, but maybe, maybe I'm off on that. Uh, but you can move a lot of current fast. I mean, you can switch fast. Uh, you can move a lot of current without a lot of drive current. Um, and you can switch pretty fast. And so if I can switch fast, it means I can reduce the size of the inductors in my, uh, you know, power converter, um, whether I'm going DC to DC or AC to DC or DC to AC. Um, you know, if I, if I have a faster switching, I can have smaller inductors. And that's really important because inductors are heavy and expensive and, you know, create a lot of magnetic fields and, you know, all of that, all that bad stuff. So, um, if I, if I want to have small inductors, I need faster switching. And in addition, if I want to have less loss, less, uh, less heat loss in my, uh, my, my transistors, you know, I need somehow more efficient, you know, power transistor. And so again, again, does these things and, uh, the power team here is, you know, done a great job, uh, building this, uh, really, really efficient, um, it's called a micro inverter, which just means it's, um, small, I guess. That's all it means. It's, it's, it's not micro in terms of power because it's, it can move, I think, two, two kilowatts. And, uh, and, uh, it's, it's like, you can put two kilowatts through it and it's cold. Well, I should say it's not cold, but it's not hot. It doesn't give off. It doesn't give off a lot of, a lot of heat, like almost none. I always think about micro

Chris Gammell: inverters too, because you can basically, you can put them on the back of solar panels and then you're not combining DC power and making sure everything's kind of matched up. Then you're kind of combining AC power, which allows you to, it works a little bit better usually.

Dave Jones: Yeah, that's, that's how, that's how this works. So that's the idea behind, behind SolPad is that it's, it's roof mounted inverters and batteries, which is not something that, uh, usually people think of as a good idea. Oh yeah. We're going to put batteries on our roof. That sounds like a great idea, right? You know, it's, they're just going to burn my roof off. Those things that catch on fire, burn my roof off. So a lot of, a lot of effort had to be put into, you know, validating that the temperatures never get hot enough. And that that's been done, uh, to, to validate that, that you're not going to have a fire risk, uh, with this. And, and, but part of that is that they have to be light. They have to be small. They have to, you know, not crash through the roof. Um, and, and so forth. So that was the engineering challenge of the inverter. And, and, uh, it's super impressive piece of technology that I had nothing to do with is, I guess the, you know, is, is the, is the, the short of it, but they all need to be linked up. They all need to be connected, uh, all, all together. This is sounding familiar now. Here we go. Yeah. So this is, this is, this is, this is where, where I came in and this is where they brought me, uh, they brought me in. And, uh, there was, there was actually originally supposed to be a LoRa element to it. Um, but that's been postponed I think to, to maybe V2. Um, but there is still a, uh, let's just say it's a non LoRa IOT, uh, um, data backhaul to get, to get these, to get the control, uh, signals and get the telemetry of all these, these inverters, which are usually deployed on a roof. And maybe there are five of them. I think they're never more than 25 on a roof, but, uh, like, let's say a typical is like, like five. So that's 10, 10 kilowatts. Um, standard, standard size of a residential solar deployment. And, and, uh, so they have to be brought down to a, uh, a gateway, which pretty standard, you know, pretty standard concept of operations is, is to have a gateway and then the gateway is connected to the internet. So then I don't have all of my embedded electronics, um, you know, don't have to burden them with, you know, IP stacks and, you know, uh, all, all that, all that jazz. So if I can bring that down into a gateway and I can use some kind of industrial networking, uh, in between the gateway and the, the, uh, the inverters, then, you know, that's what I want to do. That's, that's what, right. That's the solution that that's going to work. And, um, you know, one of the, the major issue here, like the major challenge is just having everything be, um, decentralized. And I use the term decentralized. It's not blockchain, right? There's no blockchain happening here. Okay, good. Uh, but, but, you know, like Dropbox is also decentralized. Git is decentralized and they're not, they're not, they're not blockchain. It's just exerting

Chris Gammell: control without a, without a central master kind of thing. Yeah. Because you'd like, the last thing you

Dave Jones: want is to make a 30 year investment in your house and then have the company go out of business. Uh, the cloud site goes down and your investment is bricked, right? Right. So you don't, you don't want that. So that was, this has been, um, this is, this, this has been an example of something where I've gone and done and I've learned a lot out of it. Uh, this is like that, you know, I, and I recommend to any, any person who gets like, like, I think the reason why I ended up taking this job is because I'd been doing wireless for so long that I felt like, I mean, there are people out there, uh, who are like, you know, put me to shame in terms of wireless and communications technology. And, um, um, I don't know who's that guy, that guy, uh, Shiri are on the signal path. So he's, he's, he is like absolutely legit. I mean, he's sort of at the top of the pyramid. Yeah. You know, when

Chris Gammell: you're at Bell Labs in charge of the chip design group, that'll, that'll, that'll do it. He is,

Dave Jones: he is absolutely at the top of the pyramid of like signal processing and communication. And, you know, so like that's one direction you can go, uh, and the other direction you can go is to try to broaden your, your expertise. And so I've, I've tried to broaden my expertise by, um, you know, learning a lot about, uh, I guess what, what falls under the term of like SCADA, they call it, which is just like industrial stuff linked together and community industrial communications

Chris Gammell: for, I mean something, I don't even remember what it means. Control. And I remember the C is control.

Dave Jones: I know that for SCADA systems. Yeah. There's such a huge, like anybody who's listening and thinking like, I want to do embedded electronics and I want to find a way to make cool stuff and sell cool stuff. Like look into SCADA because there's, there's all this old technology and a lot of it's getting revamped. Um, just take a look and see if there's any low hanging fruit there because I mean, certainly, you know, I found it to be sort of a fun challenge to try to modernize a SCADA network for, for, for this technology. And I'm sure there are many other opportunities to it. Um, you know, so, uh, and, and so I'm trying to integrate that into, you know, what is, what is generally called the internet of things, but what really is just, you know, a bunch of devices talking to each other on the internet. Yeah. Yeah. Getting data on the internet. Yeah. So forwarding my, my field, uh, data to a place where I can look at it on a dashboard where marketing can look at it on a

Chris Gammell: dashboard. So you're, you're basically taking each individual panel and then concentrating the data from each panel. And then also, you know, you're pushing data back and forth, I guess, as well. And, uh, and then, and then there's an internet element that basically says, you know, collecting all the data and stuff like that. Right.

Dave Jones: Well, yeah. I mean, the, the real short answer of it is that it's, um, it's dash seven over a wire. So, uh, it's, we, we're using a wire instead of wireless for this. And part of that is just because these things put out a lot of noise. Oh, interesting. Okay. And yeah. So, but it's, uh, it's RS 45 and it's half duplex and it's, um, like RS 45 is kind of like a party line. So in reality, it's a lot like a wireless link. It's just wired. Yeah. You can't have two people talking at once, that kind of thing. Yeah. Yeah, exactly. Uh, there's termination in both as well.

Chris Gammell: Right. Uh, antenna is nothing but a weird terminator. Yes. You know, it's almost the same.

Dave Jones: So, so, you know, you can take, you can find yourself in a place where you can take all the experience that you had and, and, you know, find a different, I guess, vernacular for it. And, uh, that's always really fun. Um, when, you know, you just realize you shortcutted, uh, you know, a big, um, learning curve is very satisfying. Yeah.

Chris Gammell: It's like a superpower. Yeah. You know, honestly for industrial applications too, it's interesting with, you know, talking about SCADA systems and stuff. A lot of them are still allergic to wireless just because any amount of packet loss is not really allowed. You know, it's like super controlled, like retries. What's a retry? If I retry, the thing frigging blows up, you know, you know, there, you need to, you need to have a lot of, a lot of certainty and wires usually at least have the, the, uh, the view of that in my experience.

Dave Jones: Yeah. I mean, that's, that's definitely a marketing, I would say that's marketing first. Wireless stuff now is so sophisticated. Um, uh, even like off the show, like if I, if I, if I'm happy to use wifi, like wifi is very sophisticated and, um, no, honestly, I'll tell you that I think that in many cases they're wireless would have been more reliable than RS45. Really? Interesting. And it's just, yeah. And it's just, it's just because there's a lot of areas where RS45 can go wrong, not because of anything wrong with the wires, but because, um, it's sort of assumed that there's so many solutions for it. You know, uh, there's so many ways to do it wrong. And it's not something that is, is often really, uh, monolithic. It's not like, oh, I just get a RS45 transceiver and this thing.

Chris Gammell: Sure. Yeah. It depends who it's coming from and how you're implementing it. Right. And every division thing is the one over 256 or whatever it is. Right.

Dave Jones: Well, it's kind of just that every, every microcontroller that's going to be hooked up to this thing is a different UART driver. And ultimately it hooks up to the UART driver. And, um, like if I have a wireless stack, if I have like a really mature little wireless module, whether it's wifi or LoRaWAN or Dash 7 or whatever, um, you know, that, that little piece of technology is responsible for, you know, sending my data and getting it right. Yeah. And yeah, I guess there's standards bodies there too, right? I guess that also. Yeah. There's standards. And usually the company has been as nice to make some kind of API for you, you know, even if it's just as simple as AT commands and that's all been done. And, um, so like a lot of times you'll have problems in an RS45 or any, any kind of wire network, just because of, you know, there's, there's work to do that you maybe didn't realize you had to do, uh, to get it to be reliable. And there's more testing because you, you can't off, you can't put the testing into somebody else's responsibility. You can't buy necessarily a module that is pre-certified, pre-tested, you know, been used in a million deployments or whatever. Uh, you can't necessarily do that, uh, with, with RS45. And, and so because of that, sometimes there are cases where wireless can be, I won't say more reliable, but easier to

Chris Gammell: integrate in a reliable way. Yeah. Well, uh, where can people find out more about you and

Dave Jones: the work that you've been doing? Well, we blog a lot about it at Haystack. Uh, I have a, you know, I have a business partner who does a lot of the marketing element of it. Um, and he, he, uh, writes extensively about, you know, what we're doing at Haystack. Um, uh, the SoulPad stuff you can check out at the SoulPad website, which is soulpad, S-O-L-P-A-D.com. Um, and I, there's a lot of information there. And as far as me personally, uh, you know, I don't really have, I don't really have a rich, uh, personal life. I, I, my hobby is, is, um, antennas really, you know, like my hobby is, is designing antennas and that's, that's kind of part of my work as well. So, so, uh, you know, I, sometimes, uh, sometimes I like to cook, I have a blog. Uh, it doesn't get updated very much, uh, personal blog. It doesn't get updated very much. Um, that's, uh, that's indigresso.com. And, uh, I don't know, I don't know if there's anything that interesting. There's a big data repository for Dash 7 there, but, um, uh, yeah, I mean, you can, you can check on you that out if you're, if you're curious, but, uh, if you're, if you want to know about, uh, Haystack, Haystack website, Haystack technologies.com soul pad, soul pad website, uh, soul pad.com. And, uh, I,

Chris Gammell: I wish I would have asked about the antennas earlier now. Uh, do you, do you write about that

Dave Jones: too? Or is that just for fun, fun in the evenings? Uh, you know, I should write about it more. Um, but I, I, I should, I haven't, I I'm going to start, I haven't had a really good way of, um, kind of making it interesting until recently. So somewhat recently I started to use, uh, MATLAB to simulate antennas and, um, it might even be possible on Octave, which is open source. I don't know. I know it's possible in MATLAB and I know that MATLAB has a, like a hobbyist version as well. That's, I mean, I wouldn't say it's cheap, but it's accessible, right? It's not, it's not thousands of dollars. I think it's hundreds of dollars. Um, and so that for me is, is something that I could maybe take and, and blogging about it. Otherwise you're just like taking photos of messy boards. This is the shape that I cut out. Like I did this because, you know, crazy reasons and, you know, a little bit of math, I scrolled on the back of an envelope. Uh, but, but, you know, with MATLAB, you get charts and you get, you know, current flow and everything and, and you can, you can simulate all that stuff. And it's, it's not, it's not like there's higher cost stuff. There's like, and soft and things that are really expensive. Right. Um, like 30, I think it's like 30,000 bucks for a license for that. But you know, MATLAB ain't bad. Like if I'm not trying to do something ridiculous, if I'm trying to do like, let's say I can build a, like a really simple antenna, like just wires, right? I can, I can do that quite, quite easily. Uh, that's like a one out of 10 in terms of complexity. And then a 10 out of 10 is like something NASA puts on a space probe. Right. It's like baked into an iPhone or something crazy. Yeah. Yeah. Like where they have some genetic algorithm running on a supercomputer to like eke out the last, you know, one hundredth of a decibel of gain, you know? So, right. So like, yeah, MATLAB is pretty, I think it can get you to like seven out of 10. So like that NASA antenna is a 10 out of 10 and you know, like a wire dipole is like a one out of 10 and you can kind of get to seven out of 10 with MATLAB, which is still really awesome. I mean, it's like anything short of like, what would you go into a cell phone? So like, certainly what would go into any of the industrial wireless stuff that I'd build. That's great. Yeah. I'd love to see that

Chris Gammell: kind of stuff that I'm yeah, I'm doing much research around antennas stuff right now as well. So I'm looking for help. I probably will keep you on the line after we stop recording here.

Dave Jones: Sure. Well, my two cents of advice, if you, if you want to get an antenna done cheaply and quickly, buy it. Well, you can do better than that. Go, go to, there are other, there are many vendors, but I'm just going to recommend one and that's Johansson. They make a lot of ceramic resonators. And the reason why I tell you to get them is because they make some really simple ones. They make some ones that's basically just like a ceramic core with a coil around it. You know, it's like a helical coil with a ceramic core. And that's going to get you a lot of the way. So I can take that and maybe it's, maybe it's for like, maybe it's a three gigahertz resonator. Maybe it's, you know, an antenna designed for 2.4 gigahertz and, you know, and you have to do something additional to it to get it from three gigahertz to 2.4. But let's say I want to build a 900 megahertz antenna. Like often all you really have to do is run a wire off of it. Right. So I just extend, they'll have a little diagram in their data sheet that's, that says for 2.4 gigahertz add, you know, five millimeters of wire. Yeah. Well, you know, what if I add 20 millimeters of wire, you know, what happens then? You know, like it'll bring the, it'll bring the frequency down and it'll usually, they're easier to tune because that, like that little chip part that you buy is kind of doing a lot of the tuning for you. So you have a lot more flexibility on, you don't have to get it perfect and you don't have to simulate it as much. So that's, that's my recommendation for like, you know, if you're just trying to hack an antenna out and you don't want it to be huge, you know, start with a high frequency, like a higher frequency ceramic chip antenna, like for example, from, from Johansson, but maybe there are some others, extend a wire off of it. And, you know, see if that works for you. There are some ceramic chip antennas that are more complicated in terms of how they're designed and there do not work with this, you know, the strategy. Uh, there's some that are like, you know, embedded like PIFAs, that's a type of antenna. There's some that are like something called an, I think what's called like an integrated magnetic dipole, you know, that's different type of antenna. Uh, those ones can be really good if you follow the instructions, like to the letter on the data sheet, um, that they're not really good for hacking with. Uh, so for hacking like the, just the, the really simple ones from Johansson or Johansson, I think they're called, uh, are really, they're easy to get, they're cheap and they're easy to hack. So I recommend those.

Chris Gammell: That's great. That's a great tip. Awesome. Well, JP, thank you so much for being on the show and telling us about Dash 7 and your SoulPad stuff. And now antennas too. That's been, it's been a wealth of

Dave Jones: knowledge and I appreciate hearing about all this stuff. Yeah. I hope everybody's awake after the end.

Chris Gammell: I'm sure they are. All right. Thanks. We'll talk to you soon. Okay. Thank you.

Speaker ?: Bye.