AAH #708 - The Data Super Highway in Your Car and How It Works

Speaker 1: I'll online After Hours is brought to you by bridge Stone Tires Solutions for your journey.

Speaker 2: Gary John, how are you?

Speaker 3: I'm doing good? Good?

Speaker 4: I mean, I know that's cliche. I say that every

single show, but I'm saying it because I want to hear what you're going to try to stump me with today. Well,

and not just ID but the other guys on the show here too.

Speaker 2: All right, So I gave you sort of a hard one last week. Yes, you did, right, So this one

is can be so easy. I mean it's it's almost

an embarrassment. It's so easy. Okay, So setting me up

the failing. So on July twenty fourth, seventeen oh one,

Detroit was founded. Yes, okay, right now, today is not

July twenty fourth, so I'm.

Speaker 4: Not it's not an anniversary.

Speaker 2: It's not an anniversary. But on August twenty second, nineteen

oh two, something happened that is related to the founding of Detroit in the auto industry. Nineteen oh two, August

twenty seconds. So this this is today's this is the anniversary.

Speaker 4: Oh interesting. So the only thing that was I know of,

automotive wise, that was going on in nineteen oh two Henry Ford was working on the Henry Ford Company, right, but I don't think that's what you're talking about. But

maybe Ransom E. Olds was building curve dash Olds mobiles.

Speaker 2: Maybe he was.

Speaker 4: Yeah, that's not it either, not either. Okay, Well hold on,

let's what we're going to see. We've got Dave Robbins

and Correa from in Trapid Controls here. Yes, we're going

to be talking about other stuff with you guys have any clue with.

Speaker 2: Basically John's calling for a lifeline.

Speaker 4: Yeah, yeah, that's right.

Speaker 3: Yeah.

Speaker 5: The only thing I was thinking of was one of like fordhead three companies. Yeah, think you too. I was

thinking it was the second or the first one perhaps, well.

Speaker 4: The first one was the Henry Ford Company. Yeah, we

learned this from Gary's question last week. And that was

already out of business I think by nineteen oh two.

That's why I think he had to hand for the company.

And then the Ford Motor Company was nineteen oh three.

If I remember, right, we've.

Speaker 2: Established it has nothing to do, it for nothing to.

Speaker 5: Do Billy Durant, No, that was really I think.

Speaker 2: All right, alright, alright, so on this day in nineteen oh two, which I guess wouldn't be this day. Cadillac

Automobile Company was founded and it was named after the guy who found out antion to Cadillac.

Speaker 3: Oh wow. Okay, So I began.

Speaker 2: To think about that for a while, and I mean, like, why didn't they name it Detroit the car Company?

Speaker 4: Yeah, good question. There were probably other Detroits something I

don't mean other cities, but other companies with Detroit in the name. So why not get something that's unique. And

I think Cadillac was not as real name either.

Speaker 3: But he was a fraud.

Speaker 4: Yeah, he was a fraud. He was such a fraud

that he was kicked out of the city and sent down to New Orleans, where they hated him so much he was murdered. Yeah yeah, yeah, no, no, I know

that people should look up. It was Antoine de la Mote,

the something or other, the Cadillac Cadillac.

Speaker 2: Yeah, yeah, he probably had several right.

Speaker 4: Right, he he he was a fraud, and but he did establish the city on orders of the King of France.

Speaker 2: And just and that he has this this brand that is venerable.

Speaker 6: Yeah, he's like Tim Martin's.

Speaker 4: So all right, that's not while we have you guys here, Yes, yeah, okay, so let me set the stage. Why I especially wanted

you guys in here. I know a few things about cars,

but my knowledge really harkens back to older cars. You know,

when I could cophood and understand everything that was under it.

Today I can't. So I know that there was a

wiring harness in a car and it connected everything that needed electricity. So when you turn the key, and that's

what you did, Remember you turned the key in the dash and it sent battery power to all the stuff, and you flick switches and they would the radio would come on, or the lights would come on, or the dome light would come on. And now we're into this

stuff called ethernet, and I don't have a clue what this is all about. And that's why I wanted you

guys on the show, because I figure, if I don't know, there's probably a lot of people out there in the audience that don't know the history of this. But that

stuff still exists, the wiring harness and the what it does, and.

Speaker 6: Now, at least for now, because wireless is entering the vehicle with battery management systems, so the actual batteries themselves are wireless or at least the controls the controls of the maget management. So like you know, it might evolve.

If it gets good, it could evolve further.

Speaker 5: I want to there's a story I used to get asked a question, why don't the cars just use wireless, you know? And I used to always answer, well, with

communication systems in the car. Now we've already we have

the ability to do communication and power on the same two wires. And my answer was always an ECU and

electronic control unit always needs power. So what there's never

going to be a case where you don't have at least two wires go into it for power and you can do data communication on that. But in what's very

interesting well wireless BMS is that the control unit is attached to a power source.

Speaker 2: All right, So THEREMS BMS.

Speaker 4: Is sorry all the yeah, yeah, yeah, no, but it's good for Gary to bring it up because we know but maybe people.

Speaker 6: Yeah, and like the voltages involved are dangerous, so it's like if you don't have wires, it's extra safety. And

there's a lot of very complicated wire harness to put together a battery that way. So there's a trend of

you know, like like the ultium for example.

Speaker 4: GM Multium, right, and you get of wires and weight and costs and all that kind of stuff. But let's

go back to the wiring artists.

Speaker 6: And then they came up they thinks go back, right, well, let's go back to like nineteen nineties. So like okay,

at the beginning of nineteen ninety or nineties. First of all,

the ECU started way before that, right, you had the message light that blinked and stuff like that. So the

engine controller and it was really driven by like regulations. Yeah,

it did a great I mean it made it would would cold last because he's like everyone's like, oh, regulations are bad for business, but if you look at that regulation, it really helped our business because like there was so much technology put into that and uh, it actually helped business.

Speaker 3: That regulation, I.

Speaker 4: Want to say mid seventies more or less is when the first engine And yeah, so like in the nineties, they had this wire.

Speaker 6: Harness and it's big, it's heavy, and one of the things they wanted to do is is kind of make it lighter. So the idea is that they'll have distributed

controllers or ECUs and then they connect them with you know, one or two wires and then they share share data.

So like like an example would be the the engine controller.

It's the engine and then you have your instrumentation, so it doesn't make sense to run the RPM from the engine compartment and when you can just run run wire just send a message that says the RPM is this, and then you just basically have a display. Even though

the displays are analog, they were still getting that digital message.

So like in the early early nineties, basically they invented CAN.

Bosch invented CAN, but like.

Speaker 4: And CAN is what car area network or controller.

Speaker 6: But like at the time that the silicon technology was was very expensive and CAN is a very complicated piece of silicon. So like it was kind of invented in

Germany and like you know, more of the higher end content vehicles and stuff like that. So so in Detroit

they came up with SA eighteen fifty, which was a standard actually turned out to be three separate standards.

Speaker 3: Because that's what they do.

Speaker 6: They shared like one piece of one page and that's kind of when we got started. So it would run

at like about ten kilobits, so if you remember your dial up modems, it'd be like half the speed of that.

Speaker 4: Yeah, slow in other words, but go back what is can do exactly?

Speaker 6: So what what the what what happens is is like you have you have these boxes and and and the actuators and sensors are all over the car, you know, so like the sun roofs over here, tail lights whatever.

So instead of running a wire from the tail light all the way to the front of the vehicle, you just you know, you just have a message it says turn on tail light, you know, and it gets rid of all that wire. And that's the whole point of

can And that's the whole point of janteen fifty. And

the evolution where we're at now is ethernet.

Speaker 2: So it's a it's a networking system for the.

Speaker 6: It's a networking system. And then basically the content of

the car is so much of a computer now, like the only way you know what's going on is through that network. So that's why our business is a big

business because like every engineer at like General Motors for example, has one of our tools that plugs in the car and like, okay, this is you know, this is doing this, and that's doing that. Even you know, even mechanical systems

or the h VRAX system. You know, seats that like

vibrate and massage you like that. It's it's really amazing stuff.

Speaker 2: But one of the things about about these these automotive bus systems is that I think people don't recognize is that some of the messages have to be very deterministic.

They have to get there in a certain amount of time.

So if say your radio doesn't tune as quickly as you want, that's not a big deal. But you want

your breaks to work all the time. So I mean,

is that that's right?

Speaker 3: Yeah?

Speaker 2: Is that in the hardware or is that in the networking system?

Speaker 3: Is that for safety?

Speaker 6: It's evolved, right, so, like you know, a lot of that stuff's hardwire. Like now you're kind of going into

the end.

Speaker 5: Of the Yeah, there's there are there are different networking technologies for different applications and for like the braking systems in the in the older days, going back ten years, there was some hard wiring there to because the signal couldn't be fast enough. But then it sort of evolved

so that the networking system became fast enough for that.

But again there's there's lots of different network uh used in automotive for different applications. And there's another one called

Local Interconnect Network that came along. It's just a single wire,

it's just twenty kilobits, but it works great for switches for lights.

Speaker 6: But it's a good point that you bring that because like latency is important and like so like you could have like you know, you hit the button on the door locks right, and if it if it's an unnatural delay, that's a problem. So like that, so they the people

at the OEMs designed their networks to fix those issues, and they buy our tools to debug those issues. And

one of the so one of the when we talk we're talking when we start talking about ethernet, the problem with ethernet is it's not designed. It was never designed

to be that kind of system.

Speaker 2: It's designed for offices.

Speaker 6: It was designed for offices. And uh I Triple Lee

is the organization that develops the standards for Ethernet. So

like around probably fifteen years ago, right is when it started.

That's right years ago, Like a an industry group worked with I Tripoli, not ISO and not jighteen fifty to come up with standards that added those real time of behaviors so you could you could push like for example, all the audio equipment here you can you can have a speaker and a speaker that connected to Ethernet and they can play and your ears can't tell that it's a digital you know, signal coming over. You know, it's

it's ones and zeros coming out to your ears, you know, in different spots. So but Colton expert that because you

have a chapter in the book about that one.

Speaker 5: Yeah, we're jumping around different time zones here. But that

that was sort of the start of Ethernet to how can we make Ethernet a real time system. There was

another industry that already did it before.

Speaker 3: The motive, and that was the audio.

Speaker 5: It's called audio video bridging, and it was really designed for large scale uh like concerts, Like if you want to set up a concert, you don't want to have to wire run a wire each speaker and there might be hundreds or hundreds of them around the concert.

Speaker 3: Uh.

Speaker 5: So they came up with an Ethernet system so that the the presentation, the time at which you know, one speaker output or the sound was synchronized with the other speakers and and and video as well. So audio video

bridging was actually designed outside of the automotive industry for this very similar application of bounded latencies and being able to to act on you know, physically act something synchronized across UH. You know, a wide range of different electronic

you know, electronic control units in the case of automotive, but in the case of A VB video and audio.

So that's still today, that's still being used in concert systems and is brought into the automotive industry.

Speaker 3: Now.

Speaker 5: The one the one caveat with that is the A VB system in a concert, if it goes wrong, nobody dies.

So the automotive industry actually took that as a base and created a bunch more technologies on top of that.

UH and now the whole family of protocols that exist on top of Ethernet to make it real time and safety criticals now called Time Sensitive Networking TSN. And that

was an automotive you know, it's fun to be in the industry because in specifically working at in tropic control systems, because we're we have to always be on the edge of what's happening because as soon as your technology development tools, development tools, so we've got to be always on the edge.

And the industry works together pretty well on standardization. Ever,

since Controller area network, and it's it's exciting because we're all over the place and the whole industry, whether you're in Asia or Europe, it kind of consolidates on. Okay,

let's do it this way so that we don't have proprietary technologies, which is always more expensive, and we do things a common way because it's just a way of reducing development costs.

Speaker 2: So you know, you're saying that not to have proprietary is CAN proprietary. I mean you mentioned earlier that Bosh

developed it. I mean did they did they do they

give it away or.

Speaker 6: Yeah, so it's not proprietary so they don't give it away.

They do charge a licensing fee and so yeah, they basically that pretty.

Speaker 3: Much after j eighteen fifty.

Speaker 6: You know, it had a certain amount of bandwidth, like ability to send things, like I said, dial up mode and like cars grow, features grow, so pretty much all that and the costs went down. The cost is a

huge driver, right, and that would be the cost is still an issue becing CAN and Ethernet. But like you know,

pretty much all everyone switched over to CAN in the early two thousands.

Speaker 4: And have they now switched to Ethernet or are they're in the process of.

Speaker 3: Doing that, or everyone's different.

Speaker 6: You know, it's like you know, like like we were talking earlier about central compute, like everyone's working towards things and they're figuring out and they're trying to figure out, you know, how it works in their system. But you know,

the cans. You know, if it's cost you know, some

of the vehicles, you know, the value vehicles are cost driven, and you know, ten cents is ten cents.

Speaker 3: So if you know, until.

Speaker 6: Ethernet CAN and ethernet's working towards you know, those kind of cost things, which we'll talk about. And you know,

so CAN is CAN still CAN is going to be around.

Speaker 4: For cars have both CAN and ethernet.

Speaker 2: Yes, yes, yeah, Okay, is it a fair statement that all cars have CAN, some cars have CAN in ethernet on the road today.

Speaker 3: I think that's a fair statement.

Speaker 2: Yeah, yeah, And so no car has all ethernet.

Speaker 3: No, and it may never happen. It probably won't happen.

Speaker 6: First of all, one of the reasons they have CAN is because you have to have CAN, uh, you know, for diagnostic tools. So it's legislated, uh that the CAN

is there, so so that the little connector entering at the dash that they plug in.

Speaker 3: It's got to have CAN.

Speaker 2: So the OBD two port where people.

Speaker 5: Plug Yeah, if the car has emissions, yeah, but electric cars don't. I think they changed it because Tesla didn't

figure out how to get get rid of it, and I think they brought it back. But you probably know

more about that.

Speaker 6: No I don't, no, no, no, no, I don't, because it affects the service people ability to you know, know what's going on.

Speaker 4: But right well they need to diagnose other things breaking.

Speaker 6: Because they have tools in the tools are can.

Speaker 4: You know, we're gonna have CAN for a while, but can we go pure ethernet some point?

Speaker 5: It's a it's a big discussion and a very exciting discussion in the industry, and it depends on which OEM you talk to. It's it's it's quite exciting because is.

Speaker 4: This legacy versus startup kind of a discussion?

Speaker 5: Not really, No, it's really not. It's it's more about there.

So the controller network, the way it functions, the way that the bits work on the wires versus Ethernet is very different. So it's you can't it's not possible to

have CAN talking on Ethernet the same network. So if

if you have CAN in the vehicle and any of that data needs to be shared on Ethernet or vice versa, there has to be some translator in the process, which adds costs and software. So there is a big push

to eliminate CAN for that reason. As you're you're removing translations,

you're removing software that does all this. You know, Gateway

was what we call it in the industry.

Speaker 3: So there's a big push for that.

Speaker 5: But others are saying, well, Ethernet's never going to be as good or low cost as CAN, And we have other technologies that that are so specialized that people don't eat.

Other networking technologies we haven't even talked about, like uh, Surdy Well these even higher speed ones. We call it

Surdy serializer d serializers and they're specifically designed, even outside of the automotive industry, for for uncompressed video transfer. And

so you see that using a lot of autonomous So so going back to an earlier question you asked, show, Yeah, exactly, I can.

Speaker 3: You asked me a question.

Speaker 6: I can talk.

Speaker 5: But like when I wrote, you know, I wrote an Ethernet book because I was so excited about the technology.

And at that time we got the picture of the book cup. Yeah, this is a thousand pages.

Speaker 3: It was so happy to carry here. They don't carry

us too much. Yeah.

Speaker 5: Yeah, the truth is I forgot it unfortunately, but.

Speaker 3: He doesn't want to.

Speaker 5: He wants to forget it. Then writing a book is

a lot of you know, it's a lot of work, but it it the way you learn it is try to teach something. Try to teach your topic, and then

you're going to have to you know. So so that

was kind of it was great. It's a great experience

writing the book, and a lot of people helped. It

wasn't just me. But it's eleven hundred pages three hundred

pages shorter in its second edition than the first one because I wanted to. You know, there is a funny

thing about what.

Speaker 6: The first edition came out. So everybody thinks that Tesla

invented automotive ethernet, which is funny because we've been working out for fifteen years.

Speaker 3: Yeah, for a long time.

Speaker 5: Thousand and eight was the first Internet ten years two thousand and eight, kidding so like.

Speaker 4: It was the first one.

Speaker 5: BMW was the first and it was for it was for a special purpose.

Speaker 3: It was normally Ethernet.

Speaker 5: It wasn't specifically automotive Ethernet, was just standard Ethernet, but it was. BMW did it so that they could reflash

all the vehicles, update the software and all of the ECUs and a scene vehicle faster than what you could on can So, just by way of comparison, controller area network and practice is maximum five hundred kilobits per second.

It can go out, it can go to one megabit tops, but in practice nobody uses it that it's five hundred kilobits by way of comparison at the time, the slowest, the slowest version of Ethernet, it would use one hundred megabits, so we're talking two hundred percent increase in speed. So

that was the original one, but there was major problems with using it as a base networking technology. First of all,

the standard etherne is either four wire or eight wire, which is super expensive, right.

Speaker 6: And anyone that said, like I would explain my job to them and they're like, why don't they use Ethernet?

Like I heard that so many I heard yeah, and like there's reason why they couldn't, And that's that's.

Speaker 5: Number of wires was number one and number two was unfortunately, just by circumstance or chance, the frequencies that that the frequency at which it runs. Now I'm getting two technical maybe,

but the harmonics go right on top of the radio bands, so it would never meet a cause of statu I would ever meet the standards in automotive. In fact, automotive

electronic standards for electronics is a much higher standard than what you find outside of any any Commrently, like what do I want to say, I'm losing the word the you.

Speaker 3: Know, just electronics. You buying civilians, civilian.

Speaker 6: Anything that switches in electronics. It's basically it's like an antenna,

so it will spit out radio frequencies and noise.

Speaker 4: And like you know of no spec right military spec.

There's automotive spec. You know, because to what we've been

talking about. You know, your stereo doesn't work whatever, But

if your car crash is your real trouble?

Speaker 2: So okay.

Speaker 6: So but what automotive ethernet became is they had to come up with a way to send the ethernet frames over the wires that that was low power, that didn't cause the noise and reduce the wires. So so that's

that was kind of the foundational technology and started with I think I think BMW and Broadcom they kind of created an open called it's called open so and then you know BMW is more of a low volume relative, you know, manufacturer, so like in order to get that technology to pay off for broad Com, and they wanted to bring in like a big someone big like you know, you know, like like b W or GM or Toiletta or something so like, uh, you know that's that that was the foundation, that's what created it. Like they took

this two hour technology that they had called broad eye reach.

So someone at Broadcom must be in the sailing which I only learned that was a sailing term.

Speaker 3: I didn't know that.

Speaker 2: I didn't know that.

Speaker 6: Yeah, so someone there they had this technology, and I think it was you.

Speaker 3: I think it was you. You know, he knows. Let

him talk about this. He's an expert.

Speaker 2: To stop stop stocks hold that.

Speaker 6: I mean talking about this stuff.

Speaker 2: But but here's the point.

Speaker 3: Okay.

Speaker 2: John started out by talking about the wiring harnesses and all the wires that are on around the car. Right, yes, okay,

now you guys have jumped to ethernet. You to tell

the audience why they ought to care about ethernet versus having a whole bunch of wires in their car because they never see either of them.

Speaker 3: I don't know why they would care.

Speaker 6: That's the point, Like who cares what's under there?

Speaker 3: You just want to get in the car.

Speaker 2: But I mean, is it okay?

Speaker 3: So it's faster?

Speaker 6: Yeah, So like for example, like a radar, Like okay, so we have the technology now to never have a rear end collision almost, you know, and it's a radar.

They put a radar on the front. It's driving straight

and knows how far the car is. But that generates

a lot of data and it doesn't necessarily fit on CAN.

So it's like it pushes the limits of what can can do. So at some point you get these sensors

or like you have high content, you don't want to spend three days reflashing the whole vehicle with ethernet. You

can do it, you know, in four hours. I mean

that sounds terrible too, but like I'm not saying that it is that way. Every car will be different, but

I'm just saying like, at some point CAN runs out of steam and then you need to have something else, And what's it going to be like Ethernet. What's great

about ethernet is like every engineeringytriit now can fix their router because they had to learn it really well.

Speaker 3: You know, so.

Speaker 5: Just by way of example, on a typical CAN network, let's just it's hard to generalize, but a typical CAN network might have at least five thousand different parameters that are transferred from one EC to another. If you had

to do that with hard wiring, just with wires, you're talking about five thousand different wires. Okay, it's just not practical.

There had to be some technology so that different parts of the car that are electronic controlled, and we all know how much you know what the result of that is that everything is electronic controlled now, but could can communicate with each other. And the only way that you

can do that is what we call it serial technologies like network basically networking type different types of networking technologies, and controller area network was the dominant, almost exclusive, dominant technology for thirty years. So in thirty years time, the

car went from you know, two ECUs, maybe an engine and transmission controller. You know, if you read the original

like they released interesting enough in Detroit Trivia nineteen eighty nine, SAE was when Bosch released the technology in an SAE paper.

You can look it up. So that's some trivia maybe

for another show, but anyways, uh so, yeah, so the uh so, there had to be some serial technology first of all, and then there was this period where just content was added and added and added, and we got to the point where we had to send data like radar data or video data and a thirty year old technology that was five hundred kilobit just couldn't do it.

And so the whole industry is like, what are we going to do next? Cansab and Ethernet was the obvious

choice because it's so ubiquitous. All car companies want to

do things that are cool and you know, new features, and there were so many existing technologies from the Internet. Right,

this is ethernet if anybody doesn't know, is what is the is the backbone technology that drives all Internet traffic.

And so there's been so many technologies outside of automotive that were developed over the forty you know at the time, forty to fifty year period of time where Ethernet and you know, the Internet boom and all that, you can bring that right into the car. So that was another

big push and Ethernet. As I mentioned, just the first

form of ether, it was two hundred times faster than controller area network. And now we're talking about Ethernet isn't

just a single technology, it's a family technology. So now

we have a family technologies. Go from the low end

ten megabits per second, which is designed for cost competitiveness against can. This is where a lot of fights are

happening in this area. So it's really exciting to talk

about that. And then at the high end we already

we already have in silicon ten gigabits per second. Yeah,

and we're working on twenty five gigabits per second.

Speaker 3: So what is that for automotive?

Speaker 6: This actually converts the automotive version of ethernet to the regular version of ethernet. Why would you do that just

because you can buy all One of the benefits of ethernet is you can buy all kinds of off the shelf tools that you do ethernet. So like for example, a

groad shop, you know, just plug ethernet departs on your lap and you don't need a company like us to buy a tool from you know.

Speaker 3: The boxes like that?

Speaker 5: Is that that like all of our other boxes with can and everything else.

Speaker 3: It's simple. Why is that? Why is that it?

Speaker 5: It converts automotive Ethernet to regular e thurnt. Why would

you want to do that? So you can plug in

your laptop and communicate with the car. We sell to

engineers at car companies. They want to plug in something

to be able to see that things are going right with their new software versions. And we're talking about in

development here, so that's that's the mainstay of our business.

Speaker 6: Yeah, and software evolves over the development process. So it's

it's like you have suppliers that work together and they have to integrate all the pieces to make sure they all communicate.

Speaker 5: And we know the electronics is already a multi billion I don't know what the number.

Speaker 3: But in the back to.

Speaker 6: Your question about what value that provides the customer the networking, there's some features you get for free just from having all the data there. So like I just like wanted

to be like, you know, how your radio changes the volume based on your road speed, Like that's something that they were able to add in software with no costs, and that that would be one customer benefit that maybe customers don't know about. But there's a lot of instances

of that type of stuff that go on, and maybe even more with the software defined vehicle. But if we'll

get the same technology.

Speaker 4: You know, maybe not directly to the customer. The customer

doesn't experience it, but for the automaker, if you're getting rid of all these wires, if you're able to increase the amount of data and the you know, how much data you can put through the pipeline and all.

Speaker 6: That safety system, everything is all put part of it. Yeah,

so there is a lot of value and you know your competitors are doing it. So if you weren't, like

you know, you know, if you're you have a higher technology system and you can deliver more features versus you know, another customer.

Speaker 2: So how do they make the determination of what gets ethernet and what stays on? Can well.

Speaker 6: Somebody not where an architect will do that, but it would, it would, it would It would be really based on well cult spotter answering this question. Actually with the zone

the architecture changes from candid z on all to STV.

Speaker 5: I mean, the simple answer is that in some ways it's simple. If you if you're transferring some parameter based

data or signal based data, like RPM engine temperature right, imagine temperature is not going to change that fast?

Speaker 3: Right, how much bandwidth do you need for that? Not

a lot so controllery network for that.

Speaker 5: Oh, I want to transfer video from my you know, my my center console system here to display in the back.

Guess what you're not doing that on can So so a lot of the calculations on what technology needs to be used is based on just simply on what what physical bandwidth do I.

Speaker 3: Need to transfer?

Speaker 5: The amount of data I need to transfer from point A to B. And there's lots of course, levels of

decisions there, but that's the main one is how much bandwidth and latent bandwidthin latency is really the two things.

Latency is architecture like as evolve. Like I said, there's

lots of layers architecture decisions.

Speaker 6: And because like a radar, right it could it's like borderline can like it will lead up a whole can network, but like it could be Ethernet lighters are all ethernet because you can't put that kind of data any type of screens. You know, all the displays now are soft displays,

but they actually use a different technology that's related to really the iPhone. So but but that's some I can

talk about that for a second. If we have to,

we're going to take all the cameras with all the cameras that you know have been made real cheap by the iPhone. You know, the phones, they have a specific

type of interface called c S I two or three, I don't remember, but basically it just spits out data and then that data is designed to go right into a what they call a GPU, which is a type of like displays graphic process and so so like if you images, so it's kind of like a dedicated network, and so you have low costs. So I'm like, like

an autonomous system for example, you have you have your GPU or your AI computer, which is you know in video everyone that follows stock market and knows why that's important.

And then they have the camera interface, but they don't have what's in the middle. So a couple of companies

Texas Instruments and uh analyg Devices, now they've created like a way where you can take a camera and locate that very very high speed camera interface and put it anywhere in the car. So like that would be an

example where Ethernet. Although Ethernet is trying to match that

same thing is like, once you get to a certain speed, who cares what's on the wire, It's just the technology.

So but like literally the data goes right in and then it goes right into the the the memory of the GPU and then it you know, it's paralyzed process and everything. So that would be like Ethernet. Maybe there's

a competition there.

Speaker 4: Okay, we're going to take a quick commercial break here, but I I want to hear from you guys on software defined vehicles, zonal and centralized compute and I want to learn from you guys. But first a shout out

to Bridgestone.

Speaker 7: When the peace and quiet of your morning commute is as comforting as your morning machiato, that's what really matters.

Bridge down toronzo Evy tires, less noise for more quiet comfort.

Speaker 4: All right, we're back talking with Dave Robbins, the founder and CEO of Intrepid Controls, and Cold Correa, the COO of the company.

Speaker 3: And yeah, okay, so I want to give a message out.

Speaker 6: Oh goodness, I can see Gary has an anxiety about all this.

Speaker 3: My dad rests in peace.

Speaker 6: He was a car guy and he had corvettes and he had sixty five Corvette and all he would talk about is like, how come I can't work on this anymore?

And like I just feel bad, you know, I just feel bad. I understand. So I managed a lot of

your viewers had that same sentiment.

Speaker 4: Yeah, you can't work on like you could on old ones.

Speaker 3: Yeah he had.

Speaker 6: He had like a ninety seven of that and it had it had eighteen fifty on it. And I helped

him a little bit with it because they had some issues.

But he's like, what is this? You know, like all

you can fix is yeah, stuff on a computer.

Speaker 4: But okay, SDVS software defined vehicle minunderstanding is it's not possible to do this without what you guys do.

Speaker 5: The central idea to well, software defined vehicles, I mean that that defines it pretty well. But when you when

you scratch under the hood, most car car makers they want to be they want to produce a software defined vehicle, but what what is that exactly? When you look look

underneath in the electronics, First of all, it's more centralized computing because if you if you want to update software in the vehicle after production, like Tesla does and many companies do now, it's much easier to do that with one ECU rather than a hundred around the vehicle. And

then in addition, if you can have one central compute module that has superpower, you're you're dealing with with one major e c U and not one hundred. So so

so the the weight of the vehicle goes down, which is a huge benefit, and the weight of the wiring harness and complexity of the wiring harness and and believe it or not, wiring harnesses are still made by hand today.

There's not a machine that can still make wiring harnesses.

So we're talking about human interaction and a lot that's that's where a lot of the technical issues with vehicles after production come in is with harnesses.

Speaker 4: And take a lot of room.

Speaker 5: So there's lots of big benefits of going towards what we call central compute and uh uh. And for US engineers,

most car companies that are STV or their SDV platform is what we call a zonal controller and it's a whole different paradigm where in the past we call domain control, meaning you had an ECU that's performed a specific function, and engine controller a brake controller that that worked in the break domain and the engine domain separately, and they might be networked together.

Speaker 3: But in.

Speaker 5: This new paradigm, what we call zono controller, you link everything up not based on function of the of the electronics, but rather based on location in the vehicle. And so

all sensor they're not all sensors, they're like safety critical ones aren't aren't part of that at the moment, But in general, as much as possible, you link in all the electronics within a certain region to a single zone controller.

Speaker 4: A zone might be left front, left, front of the car, right fire car.

Speaker 5: Some cars have four zones, some have six, depends on the but but you're basically reducing the number of number of major issues in the car from from let's say fifty down to five or six where there's just one in each zone. And then they all link back to

a central compute module and as much as possible, the central compute module has all of the processing and decision making.

Speaker 3: What does that mean? That means that.

Speaker 5: The central compute module has to take inputs from sensors that are far away.

Speaker 3: How do you do that? Networking?

Speaker 5: So so the network becomes so much more important and complex that the protocols how data transfers is much more bandwidth.

First of all, much more, much more data on the network, because everything has to be transferred across you know, from one from one wherever it's at in the in the actuator and sensor sensor systems where you're talking about headlights or internal lights in the car or door knobs to the central compute module, and the output of the central compute module. It's got to be transferred over the network too,

so you get inputs and outputs that didn't exist before all over the network. So now the amount of bandwidth

that go that that that on the network of a of a SDV vehicle or or technology speak more zonal controller vehicle is much more than what we have with the older with the older platforms.

Speaker 2: Okay, so so you said, you've got you've got sensors that are sending data. The data eventually cape so why

not just go straight from the sensor to the central computer.

Speaker 3: The number of wires skip the zonal thing.

Speaker 2: A number of wires, I mean, but if if I have a sensor that is in my heaviness, that goes to a zonal computer, right, so wire there which is a short wire, so it's just it's a distance. So

it's just a matter of distance.

Speaker 3: It's absolutely yes, yes.

Speaker 2: But if if okay, So, so you're sending the sensor data and then going back and forth because presumably an actuator is involved to make it do something, just taking the information, so it goes to the zonal thing. The

zonal does nothing to it except for package it and then send it on a wire.

Speaker 5: The zonal predominantly is just an aggregator of information to send it central computer.

Speaker 3: That's where is the key.

Speaker 6: That's the key because you can there's well, if you worked on your house ethernet, you have a router and a switch and those are pieces of old they're outer to software. But in these aggregate things that aggregate the sensors,

they're just a switch. Well, they want the dream. The

dream is there's no software anywhere but the central compute, and software is the bane of you know, like of of.

Speaker 5: Our companies hate software. I mean they want to provide.

Speaker 6: Everybody hates I mean software. Yeah, Like software is hard,

it's really hard, and especially as a growth, you know, like it's just get rid of the software. It simplifies,

you know, proves everything.

Speaker 3: So it's like the.

Speaker 6: Dream of STV ethernet is you know software in the central one spot, you know, and they can sit with Ethernet.

If you do it right, you can simulate everything at design times, so you know every how everything is going to work down to the microsecond.

Speaker 2: And and the digital twins, the simulation of what's going on.

Speaker 6: Yeah, I don't really understand that, so I Don't'm not expert.

Speaker 3: I'm sure he does. I don't know a digital twin.

Speaker 6: I'm uh, I mean, I mean, I guess I understand.

The digital twin is that you have a cloud version and then and then real version, and then you can kind of predict things, which I don't think that applies to Ethernet. This is just like a digital simulation of

of like if if if you push a button here, it's going to be in it, you know, in the software is going to handle it within this amount of time, and you can absolutely guarantee that when you design the systems. Yeah,

which is very That's one very big, one big weakness of CAN is that if you as you load it up with data, it becomes less deterministic, and you know, it's it's it's hard not to like add stuff, you know.

Speaker 5: Yeah, So so summarize what they've said basically the zonal control Ethernet tech. Let's just the zonal controllers are based

on transferring data based on Ethernet to the technology. And

the one thing that Ethernet does two things. First of all,

just the bandwidth a loan. The amount of bandwidth is

much higher than what can or any other networking technology for automotive can support. So it's it's automatic Ethernet just

because of the amount of bandwidth. And like I said,

we got from ten megabit all the way up to ten gigabit in Silicon now that you can choose from based on your bandwidth requirements.

Speaker 3: That's number one.

Speaker 5: Number two is Ethernet has the most astonishing ability to address and route data than any other technology that has ever existed. I mean, you can get on your phone

right now and you can send a message to your body in India and it's going to be there like this, Okay, So how does that work?

Speaker 3: It's it's magic.

Speaker 5: I'd have to go into so many technical details that we don't have time portage, but.

Speaker 3: I'd love to do. Let's just say it's magic.

Speaker 5: The Ethernet magic is that it can you can send a message from here to there with like you didn't have to change software to you know, change a message from from you to another body in Japan. Right, you

have to change nothing about the software. And there's magic

and ethernet that enables addressing and routing to happen with no software changes, just a single set of hardware. And

that also ties into the STV because car companies want to be able to change the software. They want to

be able to change the messages after production as they improve features of the vehicle after production, just like Tesla.

That's the revenue you know, the holy grail of revenue stream afterwards. And Ethernet provides that framework where you can

do that. Because if you need a message going from

point A to point B now and then you change the software around and that message needs to go from point A to point C, zero software changes because of this magic of ethernet.

Speaker 2: So why did they call it software to find vehicle If everybody hates software, they don't call it central compute vehicle.

Speaker 5: Because car companies want to be software companies because of stock market for software companies.

Speaker 3: So everyone, that's the real answer. It's marketing.

Speaker 6: The problem is now is every company is a tech company.

There's no other kind of company. If you don't think

you're a tech company, you're not going You're not staying sticking around. Every technology has changed.

Speaker 3: Everything is a tech food and everything else. But as

far as stock price and that goes.

Speaker 4: You know. But really what we're talking about is if

you can do these over the year updates after the car leaves the plant, in fact, ideally once the second or third owners got them, now you can monetize your used car.

Speaker 3: Net Uh. It's that's the table stakes. You got it.

Speaker 6: If you if you want to do autonomy, there's no way you cannot update software because like because like, what are you gonna have.

Speaker 5: The environmental drive after the car is sold?

Speaker 3: Right?

Speaker 6: Drive the dealership every time they find something, you know, and then you know, cybersecurity is part of that.

Speaker 4: So how do you guys see you think? So, we

we had domain computing where every single component on a car had a microprocessor. Then we went to zonal or

was it right, zonal centralized.

Speaker 5: There's another word for centralized. Okay, it's just the same

thing centralized com We're.

Speaker 4: Not just going to go pures centralized. There's still going

to be zones.

Speaker 5: Not there's going to be zones because zones aggregate data.

All there's like hundreds of wires going into the zone controller and you want the zono controller as close as possible to those sensors in that region. And there's only

one wire coming out to the control to the central one.

Got it, And that's an ethernet Okay, so my understanding.

So so just the amount of wiring you have is so much less, so the Ethernet wire.

Speaker 6: The CEOs all know about how many ECUs they have now, Yeah, because of the chip crisis, right, that's something that like as a CEO, you're always like, well, what's our problems?

Like that was never a problem, right right, right, So it's like why don't we not have it?

Speaker 3: Why do we have so many? Like I would ask

that question. You know.

Speaker 4: Well, what they learned is in the domain computing, everything had a chip in it. They all got it from suppliers.

They had no idea what chips the suppliers were building, and they had no idea what.

Speaker 6: The It's very simple that they have. They have a

switch from there. There there's a force upponyres, you know, Marvel,

Broadcom Etheranovia, you know, and they have a switch and they have some parts and and you know, maybe NXP and that's just really it really kind of produces reduces the bomb.

Speaker 5: What you're doing with the zonal controller, which is also used predominantly for sdv SO software defined vehicles and is another term is just central computing. What you're doing is

is you're you're trading off complexity of the hardware, number of vcus, number of wires, with complexity in the network itself, the protocols, the bandwidth on the network.

Speaker 3: It's a trade off.

Speaker 5: But bandwidth doesn't cost you know, peace wise costs on every single car, you know, so it's a it's a big the big cost benefit.

Speaker 2: So I mean, does does the and weight does the zonal computer require Ethernet to send the information to the central computing?

Speaker 5: Ethernet is by far the best technology that can be used for that, the lowest cost, best technology with the best bandwidth and the addressing of.

Speaker 2: Ethernet because I'm thinking, you know, again, if.

Speaker 5: It's not practical to choose anything else.

Speaker 2: If we have the zone, and there are some things that have to happen, like in real time in some things that you know, if the wiper.

Speaker 4: Doesn't idea so fast, it's all real time.

Speaker 5: Yeah, you got to remember, you're going from five hundred kilobits now when you when you do the calculations, and we've done it, and it's in the book and the Ethernet book that we did, and everyone's done done these calculations.

We're talking about uh ONET, just gigabit Ethernet, which is sort of like the standard baseline for for centralized computing.

Everyone has at least gigabit Ethernet going between the zones.

You're talking about sub millisecond latencies. Just because the bandwidth

is so high, the speed is so high, you can do it so much faster than you could even on CAN.

Speaker 3: You know we can, we can.

Speaker 5: We can transfer a message on Ethernet, and a modern Ethernet network and automotive Ethernet transfer it from point A to point B across the zones faster than the and the message even exists on CAN. So there's a certain

length of time that a message is on controller area network, and it's this amount of time on Ethernet. It's you know,

a thousand times less than that two two thousand yeah four full yeah, so full technical term, but you know, it's like it's so small that you can actually transfer that same amount of data faster across the network. Then

that that amount of same amount of data existed on a on a previous controller area network. So there's just

no other practical app, practical decision. There's just no no

other technology that's best suited for this type of architecture than ethernet.

Speaker 4: Teesla was the first to come out with this zonal centralized right, and then Rivian's developed a system. In fact,

Volkswagon liked it so much they're paying them five billion dollars.

And then Gary and I just learned this week Volvo is there, and which surprised me because they did it all in Sweden. Here's a little, I mean pretty little company.

Speaker 3: They're high tech.

Speaker 5: I mean in the automotive ethernet world, there are certain companies that are heavily involved in heavily pushed technology, and is one of them.

Speaker 6: So I triple. They have the show every year for

Automotive Ethernet and this year it's a Detroit so maybe you can stop by. Yeah, but but Fovo regularly presents.

Speaker 5: And we can introduce you to the Volvo. You know

architects that designed that system.

Speaker 6: They were the keynote speed, they were the keynote speaker of the show and Munich.

Speaker 4: They're very here, we're talking all about and you guys are into the weeds on this stuff. And uh, but

only three automakers have got it at least out on the roade or Gary you were singing. But what let

me finish. I mean, everybody's buying your tools to work

on this right this way, everyone's working on but they're struggling with it, right, a lot of them.

Speaker 3: Yeah, because of our tools.

Speaker 4: No, no, no, I don't because it's very even.

Speaker 2: I could probably figure that out.

Speaker 3: Big architecture changes are difficult. In fact, it's easier to

do it at.

Speaker 5: A smaller company like Vovol than a larger company because you got so many shared platforms and so many you know, let me let me there's a centeral.

Speaker 4: Thousands of engineers right the wait, wait, I want to hear.

Speaker 5: Yeah, yeah, in a domain architecture that is the old style, right, you could look at the domain architecture on a piece of paper and you could almost one hundred percent guests the organization of the company.

Speaker 3: So you had to break group here.

Speaker 5: You had an engine group here, you had you know, lighting group here, and all of those different groups. The

way the company was organized, had their own domain.

Speaker 4: Industrial anthropology.

Speaker 5: Okay, yeah, so so that really drove that, and so bigger companies it's harder. It's harder to get those larger

amounts and you know ingrained silos integrated together and guess what in central compute, it means those guys have to talk together actually put everything in on ECU. So that's

a that's one of the major challenges. And the larger,

the larger you are as a car company, the more challenge you have.

Speaker 3: In doing that. Okay.

Speaker 6: So Rivian the transition between like J eighteen fifty to can was a huge transition for all our customers, and I think this one's bigger, you know, and it's like it's gonna take time, and it's just human inertia, you know.

Speaker 4: So it's more than human in inertia because.

Speaker 3: You know the benefit of ethernet is everyone kind of knows that.

Speaker 4: But we've got some horrific examples out there of automakers who have tried to do this and fallen flat. This

is why Volkswagen's paying five billion dollars to Rivian. It

couldn't figure it out.

Speaker 6: Yeah, yeah, I mean, I don't know if that's why.

But I don't want to guess.

Speaker 2: Yeah, yeah, yeah, Okay, so we mentioned three car companies, all of which sell premium cars, right, right, So it's not cheap to buy a Rivian or right Tesla comparably speaking. Yeah,

and Volvo is that has something to do with the.

Speaker 5: Technology always always comes out almost always in the high in vehicles, and then over time, you know, quote goes any technologies that way. That's not even unique to the

automotive industry, right, I mean, the brightest and shiniest new thing right is on the you know, on the highest end vehicle typically, So at.

Speaker 2: This point it's still a cost thing. So the mass

manufacturers are not able to necessarily do it.

Speaker 3: Now they are, they're pushing towards it.

Speaker 5: But I think in fact the big argument for ZONE and STV or centralized computing kind of all the same.

Speaker 3: Is a big push is cost. That's a big push.

Speaker 2: So it costs a lot to do it or saves it costs a lot to.

Speaker 3: Do it, don't be be queer about that.

Speaker 5: But the peace costs per vehicle, you're reducing the weight, you're reducing the number of ECUs.

Speaker 3: It's going to have major.

Speaker 5: Impact on the on the supply chain of like and.

Speaker 6: Like they're all they're all worried because.

Speaker 3: They need to be.

Speaker 5: They're going from fifty ECUs per car down to like.

Speaker 3: They don't have a rolling software.

Speaker 4: Not only that the size compute that's all going to the software is going to come in the house and they've invested all the software case.

Speaker 6: Well, I mean, I mean there could be a software model where they license the software that sits in that you know, the central compute, because it's like that. The

problem is is like you know, I mean, I don't know how many like just say like a breaking system, like I don't think that will be in centerview, but just for example, like like someone like Bosh might have a hundred man years of software. So it's right, like

if you if you don't understand software enough, you know, you're not going to be able to start from scratch and write it. So so I would guess there would

be a licensing model business.

Speaker 4: What I've heard is and I don't know.

Speaker 3: I don't want to get I don't like guessing.

Speaker 4: I just want to talk about anything safety related.

Speaker 3: But I know about software and how hard it is.

Speaker 6: It's it's something that you can pour millions of dollars in and then nothing comes out at the other end.

Speaker 4: What I've heard from even for the end, it's just what they said, be a bad design. If it's breaking,

if it's steering, if it's air bags, Oh we're gonna we're going to stick with the tried.

Speaker 5: And true sigma. We're not going to risk it's a

new architecture. You're not going to do a new architecture

for the whole car and add the safety systems to that. Right,

it's not you know, and those aren't the systems that customers.

You know, the money stream after after UH market is not going to be oh my breaks, you know, perform one millisecond faster or something. It's gonna be some interaction

feature with the Internet or something like that that they're going to try to.

Speaker 6: I think fature software on a vehicle is it's not going to be clear where the where the car stops and the cloud begins. It's just like a feature could

be sitting here or there, and it just depends on the bandwidth between the links. So the bandwidth is fast enough.

Like v TV, for example, we were talking about V to V. If it's the bandwidth is fast enough that

you can just do it over the internet. You're gonna

be like I'm here, you're there, and the latencies this, I know where you're at, you know.

Speaker 3: So it's s.

Speaker 6: A huge part of it is that link to the cloud and and like I said, like and that and that and that technology is evolving very fast, you know, so like you know, you know, Silicon Valley, you know, whatever involves faster and like that's just going to.

Speaker 3: Be stuffed in the vehicle with STV.

Speaker 2: So are you guys involved in V t V or V two X or no.

Speaker 5: We'd like to be, but we're so busy with the change from from domain to STV for our for our products, for our tools is a big change and we're pretty consumed by that with our own engineers developing our own products just for this.

Speaker 6: A lot of our knowledge is in work is just based on OEMs and tier ones. You know, we need this,

we need that. So so if they're not focused on it,

neither are us. Not saying that they aren't focused on

B tow d D. But it might be like research

or you know whatever.

Speaker 4: Yeah, but it'll come, but it's not there right now.

Speaker 3: Yeah.

Speaker 4: Yeah, yeah, So let's go back to these centralized compute things.

As you said, very expensive to develop. Takes years to

develop a full STD right, But once you get over the hump, and you want to put it on your expensive cars because you can amortize the costs and you know, bake that into their price custom rules of paper. But

once you get over the hump of developing it, how easy is it to scale then into other platforms?

Speaker 5: Relatively easy, relatively easy, So you just got to get over the hump. And oh yeah, I mean there's there's

big plans for for you know, of course we have we're under India. I can't give any specific examples, but

but of course most most of the customers that we work with from a car company standpoint, multinational car companies have you know, big plans for for STV and they may not have all the software features in the lower end vehicles, but it's still lower cost architecture because it's less wiring, less number of ECU. So so regardless, there's

a big push for that across the board. And let's

say Western markets, including you know, fully developed markets. It

might be a little different in India where or you know, other countries that aren't fully developed because you're not you don't have the cost uh and it.

Speaker 3: Might take fifteen years.

Speaker 6: It's not going to happen over the point because.

Speaker 3: Can was that way.

Speaker 6: I can like you know, like one of our one of the reasons we started our business in China was because like the can't they were vehicles. They were still

sourcing Jting fifty ECUs for a very long time, and we were kind of like the J ten fifty tool provider.

Speaker 3: So we got we are.

Speaker 6: Our business grew in China very big in early two thousands because they were still sourcing those and like they were just it was just focused on costs, so you know, depending on you know, but eventually it just be you know, all all the supply bases like well we have a can ecu done, you can use this one at the Jting fifty one, and then it gets cheaper just to do that, right, But it takes time, and it'll probably also depend on the feature. Right, if there's a feature

like you know, like a radar requires Ethernet, then it's going to be you know, so.

Speaker 2: If then OEM develops an SDV so they have the rest, does that go to all different types of vehicles.

Speaker 5: As much as possible. Of course, they don't want to

have to develop something new again. I mean that's that's so.

Speaker 2: It's just like leaving things out right.

Speaker 5: Exactly, and oms have done that for on domain architectures as well. You have a base software with base components,

and the high end vehicles get more content, the lower end vehicles get less content.

Speaker 6: Always been millions of lines of code. Yeah, first of all,

I would no one should ever brag about how many lines of code.

Speaker 3: That's y.

Speaker 6: But like a lot of those lines of code are like the interface to the network, So that interface driver is like stamped out in every ECU. So like if

there is no software in the ac except for that, it would still be probably millions of lines.

Speaker 5: But once you do but like they're saying, once you develop software to consume and route consume it, the transmit and receive Ethernet data, you've got some software in there that interacts with the chip, right, Are you gonna redo that for no reason for your low in vehicles. No way,

of course, You're gonna just transfer that software over and it's gonna be you, you know, biquitous across the entire company.

Speaker 6: That's advantage of learning that's sitting on the high end.

Speaker 3: You know.

Speaker 4: So I've had some software experts tell me it can take anywhere from eight to ten years for a legacy auto maker to develop an STV from scratch.

Speaker 3: I don't I don't know.

Speaker 6: Software projects are tough. Like I said, you can put

millions of dollars in software and nothing comes out.

Speaker 5: I can tell you this, that's not the plan. Yeah,

it's a lot shorter than that, a lot shorter.

Speaker 4: So when I talk to the Volvo guys, they were suggesting that it took them maybe four to five years to do it.

Speaker 3: That sounds I've already been working on it for two years. Yeah,

I mean, I don't know.

Speaker 2: But yeah, well, but you know, going to what you just said about software projects not necessarily paying off of the end, I mean, is there the possibility that some of these projects that are going on now that they think they're going to be done in four.

Speaker 3: Years, that.

Speaker 2: Four years after that they're still saying it'll be four years from now.

Speaker 5: Once you have a sunk costs in the plan, soware dump it right.

Speaker 3: Right off the ship or no? Right? That's the problem

with software.

Speaker 6: It's not logical, Like there's this book a mythical man month and then and then one of them famous COVID is it takes nine months for women to make a baby, and like if you take one hundred women, it still takes nine months. So it's like there's something about software

if you don't have someone architecting it right, and you can like spend a lot of money, and like the end result is like doesn't work well or it's buggy, or doesn't work at all. No one knows why. You know,

then you have like people that leave, and then like like who who wrote this stuff? And then all engineers

want to rewrite it. You know, I just spent in

some of those products. Not I've been in those projects.

Speaker 5: But if that we know about we're talking in a couple of years time, we're not I'm not saying I'm not coming, is going to be very at least the architectures we're talking about around going to be on there.

Speaker 3: Like I don't think that just a.

Speaker 6: Couple of years that kind of because because like the idea is to I mean, I have my own doubts about some of the stuff like the supercomputer and stuff like that, because I don't know how that's going to work as an engineer. But but I think, I mean,

I don't I don't think the software. I don't think

it's a big software problem. It's more of a hardware problem.

But I feel like they've been working on it, and there's so much sharing in the industry. Like you know

this conference coming up in October, uh, you know, every every OEM is there, and they're like, well, we're doing it this way, We're doing it that way because they're not seeing it as something they're competing on.

Speaker 5: It is astonishing how much information is transferred between the different OEMs that compete with each other and the supply base is really at these conferences, they're openly talking about trying.

That makes it so exciting for for for me actually going.

Speaker 3: There, you learn a lot.

Speaker 6: You learn a lot because like Bulbo, like they did their presentation, they're like, these are the reasons why ether not it's going to work, and like you know they're struggling with and you know, like can is there's like the cost differential between a canoe and ethernet out that's you know, that's they share it.

Speaker 3: They share it.

Speaker 2: You mentioned hardware, and earlier you mentioned everybody knows in video because you know they're the stock market and that so.

Speaker 6: Many people have told me how much money they made on in video and I'm just like, I don't know.

Speaker 2: So is there sufficient processors out there for the auto industry that's affordable and robust enough that can do yes?

Speaker 3: Yes, Well I don't know. I don't know.

Speaker 5: Yes, yes easily because because in the automotive industry, I you know, I love the electronics and so forth, but in terms of processing power and like the engine or the motor on the car only spends so fast, right, and so you add up like how fast do we really need to react to the mechanical reality the physical reality of the vehicle, and that physical reality of the vehicle for control systems itself, which sounds complicated, really is less computing.

Speaker 3: Than what you see in other industries. And so it's

it's it's not as.

Speaker 5: It's not as high speed in my opinion, it's not as well.

Speaker 3: We're not doing what.

Speaker 5: About you know, or yeah, for uncompressed video data real time analyzing that for AI that's going to be like the Navidio system that actually work together with the Bulbo.

That's the Bulbo system at least in some of their cars.

Speaker 2: Isn't right, But I mean, I basically crossed an arm and a leg and.

Speaker 5: It's it's on high right, it's autonomous.

Speaker 3: That's built their own chips.

Speaker 6: If you look at what they've done that you know they had the autonomy day and you can look at them as someone who's because what happens is when when you start this stuff, you start at a and then you learn stuff along the way, and then it changes and eventually you get it and they're just far far ahead with that and then they you know, making their own chip. Like the problem with Nvideo that like it's

it's more of a general purpose compute, so like it's very it's very inefficient. Well, it starts out with f PJ,

which is like a circuit where you can design in software.

So it's like a XYE links and Altera which we've been bought up by am be Intel. But that's very

inefficient because since you can divine it, software has all these things that you have to do. It has a

lot of transistors to do something generic. And then the

next step would be like in Nvidia where they have like these these pieces in the chip that that all they're like little processors, but they're still also generic.

Speaker 3: So Tesla took it to the next step.

Speaker 6: Where they they did an application specific circuit, which is there's a lot of matrix matrix eldeerra and like, like the most important slide they had was they talked about pico jewels per multipli.

Speaker 2: So they really were just talking about earlier I.

Speaker 5: Know, po juechnology exist.

Speaker 3: I was not referring to the A for me.

Speaker 6: That means that they One of the one of the interesting problems that's happening with autonomy is like and ethernet is like like when we had gas vehicles, like the stuff that omis would do to save like a percentage of economy, and now there's a direct link between computation and fuel economy.

Speaker 3: So like if.

Speaker 5: You central are water cooled, for example, there's just so much heat, so much.

Speaker 6: I mean, but like, but like one of the solutions of that is to design a chip that just has the transitionors.

Speaker 3: You need to do the algorithm specialized chips for that.

So but like if you if.

Speaker 6: You spend five hundred watts on the autonomy system, that's going to reduce your fuel economy.

Speaker 4: So like I don't know, or you're a range as an electric vehicle.

Speaker 3: Yeah, yeah, that's what I'm saying.

Speaker 2: But I'm guessing. I'm guessing Florid or General Motors is

not going to get in the chip business.

Speaker 6: So they're basically to be stuck with buy do they have to I don't.

Speaker 2: Know, buying commodity?

Speaker 6: No, no, No, they have partners. They have partners, so

like someone that's partners arm, you know. They there's chips

coming out which probably are I don't know what's used in the central compute because we don't we don't work in the box. We only work from the outside of

the box. But like there are chips that have like

thirteen CPUs and then they have they have basically GPU and they have like dedicated AI units, which is similar to what Tesla did. You know.

Speaker 5: So we've seen we've seen the industry for a long time partnership between the silicon world and the automotive world, where the automotive role says, you know who can do this? Okay,

this guy can do this, and they see money over here, and these guys get the features that they want, you know, so and that that continues.

Speaker 6: But my thought it's got to be custom chips because like so the traditional way, like a candy suit the CAN message would come in, and it's like the CPU itself, the clock rate on the CPU would be like one hundred times faster than the CAN bus. But now with Ethernet,

like a gigabit Ethernet, the clock rate of the CPU is is is the same or like very close to the actual bus rate. So it's kind of like this,

the age of the CPU is over, So you have to have some sort of hardware fanciness which ship fundors can make. But it's but that's what I'm saying, Like

when they talk about central compute, it's got to be something special which they are working on, and but I don't know what it is. So but like but I

did it. We did like data log we do data

loggers for these high speeds works. And it was very

hard to get it working because the scale of the problem.

It's like going back to nineteen eighty like when they wrote when they wrote code in nineteen eighty, they only had like a tiny computer and they did so much if you look at like a one megahurtz computer, and like the coding style that they had to do is was just like you had to like really know everything and and it's kind of a similar problem now because like the data rates are so fast, like the CPU is, you know, it's really at its limit.

Speaker 5: With Dave's with Dave's not talking about gets more and more and more and more like the future world with autonomous driving, right, the central compute. But but the central

compute for like a car that doesn't have the most advanced autonomous features that you know, general control problems about motor control or battery control or even temperature control, those types of things are solved for a long time with just sent with just basic processors that are relatively low cost, so you can you can centralize that and in good micro controllers today easily. But Dave's right, the next sort

of the next sort of uh step would be how do we do this autonomy? You know, in more autonomy

where the car can actually see things happening and in that world You're right, the processing well there's really two different Is.

Speaker 6: It fast enough where you can eliminate the micro controller, the little computer in the node that the remote node, So like when like you measure, you measure whatever you measure, and then that Ethernet message goes all the way back to central compute, and then in the central compute you execute some software in time to like create a response that goes all the way back, and that that would be like the holy grail of of of like STV, because you wouldn't need any software anywhere else. But but

like I don't know if that's gonna happen, but it's gonna be an evolution, right, like the commoter control, Like that seems crazy to me that they could do it over But I don't know. Like if it's fast enough,

it doesn't matter. Like you know, when you talk on

the phone, right, you don't know that your voice is being chopped up into a million pieces digitized, but like your ears, you know, so like there's no reason why that same concept can't work with any remote noe.

Speaker 3: Like it's fast enough, there'll be an evolution, for sure.

Speaker 5: I mean it's not never is the case where all we got the first STV and we're successful and then let's we're done forget about it. You know, of course,

there's always new frontiers and that's what you know, that's what that's what the business is, any business in technology.

Speaker 4: I want to go back to the micros and then we're going to have to wrap this up. But you're

saying Tesla's designing designing its own chips.

Speaker 6: So I only know about the central central compute.

Speaker 4: Yes, okay for the central compute.

Speaker 6: And the huge advantage of that is the power efficiency.

Speaker 3: Correct, I think.

Speaker 4: But can other OEMs who are not going to get into chip design? Can they go to chip makers?

Speaker 3: Yeah, they absolutely can, but they.

Speaker 4: Have the knowledge of how to of what FI that they need or are the chip company is going to have to tell them this if you want to do this.

Speaker 5: This, it's a conversation. Okay, come to I triple and

you'll see the conversation happening.

Speaker 3: Honestly.

Speaker 5: Well, that's what makes it so exciting, these conversations is what happens.

Speaker 6: How to implement in hardware. I think people understand that.

So I feel like, like, you know, an OEM and Troy has those people like that, Like, I mean definitely, I know our customers have transformed in terms of the people they're hiring and and and like I would not count them out.

Speaker 2: So they could design it and send it to a found read the foundering makes it pretty to go.

Speaker 6: Well, they could specify the.

Speaker 3: Features that we need in this.

Speaker 6: Yeah, we have like a you know camera input. You know,

here's like here's like the AI matrixs that we're doing.

Speaker 5: I mean, and then you'll get silicon manufacturers that are specialized in that area, which you know.

Speaker 6: People thought maybe someone could love to they could have it, they could license their their chips, you know, I mean, but like they're making too much money other places, I guess.

Speaker 5: In the Ethernet book, the last reference, yeah, one of the big Yeah, that's three hundred pages shorter, second edition, three hundred pages shorter, so it's a lighter read than the first edition, but it is a reference.

Speaker 3: But nonetheless, just just.

Speaker 5: By way of comparison, Oh, I wish I had this number of memorized. I forgot it now I used to

have a memorized But let's say the number of Ethernet chips, or let's say Ethernet ports that are in the automotive industry now outnumber what is the rest of the world.

And so that reason alone is why silicon manufactures. When

when a when a big OEM says we want to do X, Y and Z, all o ems are going to want to do that same thing with networking, and so the you know, there's a lot of attention, a lot of communication, a lot of this conversation happening between the silicon world and the automotive world. And that's what

makes uh this, you know, the conferences and you know, working at a trapics exciting.

Speaker 3: Yeah, and we're right there.

Speaker 6: We don't make any of the decisions. We just we're

on along the ride. Our goal is to whatever our

customers dream up, we make them successful.

Speaker 3: And that's our goal.

Speaker 6: It's like, how do we make it easy to develop? So, like,

you know, if we can, if we can make General Motors get it to market. You know, you're that's what

we want to do. And like we you know, we

want to do a good job and help them and help Detroit, you know.

Speaker 2: So that's awesome.

Speaker 4: Look, this has been a fantastic show. I so appreciate

you guys coming on. I've learned a bunch of this.

I'm sure the audience has to, uh, but thank you, thank you for coming on.

Speaker 2: Thank you.

Speaker 4: And Gary. We'll do another show next week. We will,

like we always do. So I want to thank all

of you who have tuned in.

Speaker 1: OLL Online After Hours is brought to you by Bridge Don't Tires Solutions for your journey

AAH #708 - The Data Super Highway in Your Car and How It Works

About this episode

Oldsmobile Curve Dash

BMW 5 Bmw

Chevrolet Corvette

Request an Explanation

Want to learn more?

Help improve this episode

More from Autoline After Hours