Mobility Electronics: Compatibility, Interference, Testing &Transmission w/ Gideon Kedem

Hello and welcome to the latest edition of the You Imagine Mobility Podcast series. I'm here with Gideon and very excited to talk about this field of mobility that he and his company are working in. So, Gideon, thank you very much for joining us today.

Maybe to start out with, introduce yourself to the listeners and to viewers, introduce your company, what you guys are doing, and then let's right after that, let's dive into the details here and how we together, from your perspective, see the mobility changing and reimagined going forward. Thank you, Stephane. It's a pleasure to be here today with you. My name is Gideon Kedem. I am the head of automotive business at Valens Semiconductor.

Valens is an Israeli company based in Hod Hasharon Israel and a little bit more than 250 employees focusing on high speed, high bandwidth, wired connectivity. We have two business units in Valens One is focusing on audio video, and the second is on auto automotive. I'm heading the automotive business unit. Our core competence is a fi, it's a digital fi and delivering to our customer, as I said, high speed, high speed, high bandwidth connectivity, which is very essential for automobiles today.

One of the key advantages of us being a digital our our core technology is based on the digital DSP, which is giving our customer very high noise, immunity and scalability for growth in the future. So that I would say in a nutshell, uh, is for less. We have, we have field offices around the world, in Japan, in Korea, in China, in Europe, in the US, around all the major automotive hubs. And we have very good momentum with our new chipset family, which is compatible with the MIP standard.

As data and communication networks and, and into data transfer between all the different modules independent if it's domain controllers or a distributed network, we have more and more technology, more and more controllers, more and more features in a vehicle. So obviously a very important thing that you guys are doing. Where do you see or where have you seen over the last 2 to 3 years that the biggest growth for us? Is it ADAS? Is it in the field of infotainment?

Is it electrification that maybe some don't realize requires a lot of communication between different components as well? Is it generally the vehicle? Explain a little bit where have you seen the growth and where do you see it's happening in the next 3 to 5 years? Well, I think all of what you've mentioned right, is, is the areas where there is tremendous changes and growth at the automotive industry.

If we take a step back and we just think about the passenger vehicle as a whole, I think for many years the automotive industry and the and the passenger vehicle was a masterpiece of engineering, of mechanical engineering. Right. For many decades, automotive automobile were masterpieces of mechanical engineering.

And what we recently see happening right is a huge transformation in that this masterpiece of mechanical engineering is actually transforming to a masterpiece of computer engineering, a masterpiece of computer engineering, an artificial, intelligent and a robotics. It's an amazing change. It's opening huge opportunities for growth in many areas.

The place where we are focusing on, as I said, is on connectivity, connecting all those new, uh, devices in the in this new, uh, in this new vehicle is in this new software defined vehicle of the future. Specifically, we are focusing on sensor to compute, uh, connectivity as part of the road to the automotive, to the automobile becoming an autonomous vehicle. Right.

Many, many sensors are deployed outside and inside the vehicle you have a combination of cameras, radar , LIDARs, monitoring the environment and processing the data, which is coming from the outside and to enable ADAS system supporting driver decision and in the future also full autonomous driving. But it's not only external, it's also internal, uh, following some regulation but also requirement of customers. You have both driver monitoring system, also passenger monitoring system.

A new man machine interfaces to activate the car, which is based on visual and sensing of of the of the movement of the passengers. So there's a lot a lot of sensing which is done inside the car and outside the car. And we are focusing on the connectivity, the connectivity from those sensors to the central compute of the vehicle.

Like very, very critical for sure. I mean, I've been involved myself in in many of those areas that you mentioned from the sensors to the compute platform for many years. And I've seen the you know, the amount of data that we're collecting is going up and up and up. And you mentioned A.I., right? So we're essentially creating our own, so to speak, data lake in a in a central compute platform that then can be used for a variety of different functions.

Obviously, Gideon explained maybe a little bit. What have you seen as it relates to the number of of of data? Are the amount of data that's being distributed along the vehicle or communicated along the vehicle and in the past we talked about kilobyte and we talked about megabytes and we talked about gigabytes. I think now we're probably talking about terabytes. Can you can you explain a little bit to that, how you have seen that progressing?

Yeah. And I think I think you have to look at that from several dimension. One is the total amount of data, which is really becoming enormous because the vehicle today is not just sensing the environment and doing the relative processing and also sensing the near environment and the environment and also the inside of the car. And there is also a lot of data which is being collected and monitored by the vehicle. Right.

If you talk to if you talk to OEMs or developing the software defined vehicle of the future, and if you see and monitor what they do, they actually talk about also about the concept of a digital twin. So every vehicle, which is actually moving on the ground, we have in the cloud an image, right, monitoring their the behavioral, the vehicle monitoring the environment and then implementing via over-the-air updates improvement to the behavior of those of the of this vehicle.

We don't see that on the road today, but this is coming. But, you know, this is just another example of a huge amount of data that need to be transferred if talking about where our company is focusing. Again, I said we are focusing on the sensing and the the amount of data from our perspective is less relevant. What is important is the bandwidth which is required, which of course is related to the amount of data.

But as you increase the number of sensor and as you increase the bandwidth rights of of those sensors, the connectivity becoming more and more challenging. So you see the transition from a 2 to 2 three megapixel cameras, two, two, five and eight and 12 and 16. Right. So you see the growth in this, you see the growth in the number of sensors.

So we are talking to OEMs today about use cases, about architectures which are involving 16, sometimes 24 cameras per vehicle and on top of that, radars and LIDARS. So the overall bandwidth that you need to to transmit on the on the wires is growing exponentially today cause we are on the road, maybe needs 1 to 2 a gigabyte per second. We are talking now about about a four, eight, 16 and even 32. So our car technology is supporting eight gigabyte.

We are talking about 16 gigabyte in the next generation. And this bandwidth is is coming with, I would say, a lot of challenges and also limitation to how you can operate in the advancement of the vehicle. So that that I think there is a challenge with the amount of the data but there is also a challenge with the bandwidth which is required to to transfer this data over the cables.

Sure. That’s a very, very good distinction. I appreciate that. Absolutely. So you're saying you're looking already at 16 gigabyte, right. Bandwidth.

So where are we going to be in five years Is there, Is there ever a point where we say we don't need more or is it are you is going to say, are you foreseeing right now that as we continue to re-imagine mobility and bring new technologies and features into the vehicle and make the the truly software defined vehicle a reality, is, is it ever going to end and how much bandwidth increases we need or are we going to change in architecture that we may be do much more again, go

instead of to central computing, maybe change it again and go to more compute at edge computing again to to be able to limit that constant increase in bandwidth needs. What do you guys see there? When we talk to customers, they are all talking about central compute, right? I think edge compute is something which is more expensive, not efficient, and also is causing a problem for, uh, I would say creating a reliable autonomous vehicle.

You need to bring all the data into a single point in aggregate, what you see around the car, right, to make decision. If you are relying on edge compute, uh, eventually you are creating more latency and you are also limiting the ability of the central compute to make sure to make the right decision. But there are also people who are looking into intermediate solution because, you know, central computing, it's all a limitation.

And I think we will end up with a hybrid model where the critical ADAS related data will all come to the central compute, the less critical data, the less critical data will be processed probably in a zonal controller. Right. And and then fed into the central, but definitely a critical data moving to the central compute is something that we hear from all customers.

Some of them would say, uh, you know, they said that they believe in total cycle compute, everything is going to sound compute and of course they will do it with redundancy. So you cannot rely just on a single element. They will do it with redundancy. But some of them are saying that they will do a hybrid, right, of a central and a zonal, uh, controller compute.

But I think from from you were asking about, about bandwidth, I think that that increase in bandwidth for a single sense or uh, is less dramatic and fast than the, the increase in bandwidth, which is the impacted by the number of sensors. Right. Because uh, when you are, I would say for, for the sensors today and, and tomorrow, a full gigabit per second at gigabit per second, even, you know, 12 might be sufficient for a single sensor.

But then when you have multiple sensors and you want to aggregate them and you want to do it in an efficient way, you want to combine a few sets or into a single line, this is where the challenge is going. But I think if you look into that in the next decade, probably 16 gigabit per second or maybe 30 to get gigabit per second, I don't think there will be a requirements for higher bandwidth than this. When you are talking about write sensor to compute connectivity to.

Maybe shifting a little bit for a second. And over the years I've been involved in in electrical architecture development programs in in, in communication development programs, some of the software staff to that to the to the integration to two gateways, both to the cloud and internally to the vehicle cybersecurity. And the discussion always came back and forth again, what are we going to use? What's the what's the next version Are we going to remain with CAN right? Are we going to go?

CAN F.D., are we going to go to Flexray is it going to be automotive Ethernet? And over the last ten years he was over and over again quite a bit of out of a discussion about. And yet generally speaking, I personally have seen only a few relative to the entire industry move away from CAN and go to either Flexray or automotive Ethernet unit works specifically in that space and work with all different types of OEMs. What do you see as relates to that?

What is what is that that the future choice of the CAN community either yeah, the communication protocol or physical layer protocol that's going to be used. So so I think that we will see in the vehicle will be forward in an environment with multiple protocols. Right. Because you need to design the car in an optimal way. Right.

So, so you will see CAN for certain, uh, for certain things, uh, you will see automotive Ethernet for certain things and you will see and technologies like MIPI A-PHY like Valens MIPI A-PHY, verify it for the high speed, high bandwidth, what we call asymmetric link between sensor and, and uh, central compute, the growing need for this says so so-connectivity was actually the main motivation behind the definition of the MIPI A-PHY standard because the protocol that you have mentioned, right,

they do not answer the requirement of those type of links in an optimal way. You know, one one of the one of the aspect that, uh, I think people are not still fully internalizing and understanding is the, I would say, the tremendous growth or the exponential growth in, in noise immunity challenges and EMC and the I a challenges, right? Because today, again today most of the vehicles are still using relatively low bandwith a 1 to 2 gigabit per second.

If you are looking into the cause, that should leave the production line in 25/26 with multiple SAS or with four gigabit per second, eight gigabit per second, a bandwith length, then the sensitivity to noise, the sensitivity to to EMI in EMC interference is becoming very, very critical. We have designed our devices right and then the MIPI A-PHY specification are addressing the next the the next generation challenges of EMC EMI One of the things that we have

implemented very successfully is a connectivity based on a digital file. A digital file with a DSP core inside is enabling you to implement mechanism like noise cancellation, like retransmit at the FI level. Right?

And those are actually enabling overcoming very severe EMC environment because, you know, today in the vehicle with the growth, with the tremendous growth of electronic components and the bandwidth and and despite the noise which has been created inside the car is much more significant than in the past. The car is also the vehicle is also exposed to strong noise coming from the outside cellular antennas radar. The outside right of the cars and near.

So it's uh, the environment is becoming much more challenging and I think one of the one of the things that, that we are also asking customer to consider is the way they test for EMC immunity. Many of the test and the standards today are very appropriate for the tests, but not for the future. Right? And the tests that are done today are done for cables which are relatively short to make two meters of a cable for cables which are not taking into account the impact of aging.

So if you test today, if you want to put, uh, a element into a vehicle today and you test it with a cable in the length of two meter and your use using a new cable. Right. And then you put it in the car and actually the cable, which is connected to the system, is ten or 15 meter at length. And with time the cable is aging. You end up with a problem. You may have a problem in the field.

A two or three years after the vehicle have left the production line, I think that we are building our technology and we have testing our technology for much longer a cable situation than the standard we are also building it and testing it to be able to overcome the challenge of the cable aging issue. Right. And when we compare it to competitors technology, we see we see huge differences. And I think we are in the process of educating the market.

Hey, it's great that everybody are moving up the scale. You put many sensors, you take higher bandwidth, everything is going going faster. But what about your noise immunity? What about the EMC sensitivity? Are you sure you have checked it? Then don't check it for the situation which is now in the lab. Check it for the environment that the car we need to face two or three years after it left the production light. It's it's it's a process

we are going through. And I think I think that we see the future in a very clear way and we are trying to share that with our customers.

And that's very good. Now, I completely agree with you. I think EMC and I in the in the 25 years I've been in the industry has as consistently increased in importance and increasingly also increased in in the challenge that it poses to to to the OEM and and the end product. Right.

To really we have a product that is protected against that at least or limited its impact of it to to the to the product because of all the different components, the different features and functions we've brought into the vehicles that are absolutely right. So I go with that question. We've talk now, I think maybe mainly passenger vehicles and you talked about cable lengths of 15 meters or longer.

So I'm thinking now autonomous trucks that are at least in in the U.S. as well as to some degree in Europe as well, really looked at may be ready to lead right at the trailblazer, so to speak, as it relates to level four or level five, autonomy for for good deliveries. So specifically to you, I assume you guys are working with heavy duty truck manufacturers as well. Look at look at again, the the bandwidth challenges we have there.

Well, we maybe even need more sensors to have a level four, level five heavy duty truck to drive deliver goods. So how long do you think with with your technology can these wires be? Because we may need sensors that are in the back of the trailer that is way more than 50 meters. So what what do you guys see there? Or how are some of those challenges overcome in in that space? Yeah, So so I think what you're raising is a very good and valid point.

And we do have some interesting activities, uh, in this area. A I must say, you know, you started the discussion asking me about the, the growth potential, etc.. You know, the, the passenger vehicle is a much larger market. The growth potential there is huge.

And I think also the focus and the interest of of the I would say the industry is much more on the passenger vehicle than on on the heavy trucks, etc.. But there's definitely a very growing and demanding market for what we call the commercial vehicle. And there are many challenges in this commercial vehicle. You mentioned a few of them. Um, you know, it's not just for the autonomous car, also for direct control the trucks.

Also if you look today into a regular trucks, right, Uh, it's usually a combination of a truck and trailer and in the different fleets, they want to, uh, disconnect and mix and match, uh, a trailer and truck. They take a trailer to a certain location, then the disconnect and and connect to another one. And, uh, all of those trucks today on most of them don't have very few cameras because of the combination of the, of the challenge of the length.

But also they need to be able to disconnect and connect in a in a fluid way. And we are working with, uh, with a partner in the US company by name of Stonebridge, and we are actually rolling out to the market a solution for this challenge. This is something that is unique to us and Stonebridge not only with our technology, you can actually overcome those two challenges, but also looking into the future.

Uh, what you have mentioned, we are already qualified and our new products are family that we have first rolled out into the passenger vehicle uh, market, and we're already qualifying it for lengths of 40 meters and even a bit above. Right, to address exactly what you've said. Right. Because in those commercial vehicle, uh, market you must have a solution that can sustain the signals and overcome the interference over a much longer distance. Right. So we're already qualifying this.

We are working with, uh, with a few leading truck vendors, but I think in this market, things are moving a little bit slower than the market in general is is a bit so-so. Definitely we have focused on that. We have some things happening there, but our main focus is on the passenger vehicle market now. Very good, very good. So maybe the last question then, not going to ask you when you believe the true mass deployment of level five vehicles for autonomous driving will happen.

But maybe I'm maybe I'm going to ask you a slightly different question. You talked a lot about central compute platform, and I know a lot of all your arms are most OEMs are looking into that, maybe even working on it. When do we start seeing mass deployment of vehicles that are truly having a central compute platform? Uh, 27/28 right. As So be 27/28 is where you will see this thing happening in a larger scale and you'll see some of it already happening in 25/26.

But I think 27/28 is more the most part. And you know, when you go around the world, right in different territories, things happen in a different base. If you go to China and I've been to China a couple of months ago, uh, it's amazing, like the the pass and how quickly they do things.

What not always successful is right, but they, they move really fast because they are trying to to get ahead of the pack right now if you talk if you talk to the traditional OEMs like the Japanese or the European traditional OEMs, they move a bit slower. Doesn't mean that the Chinese will actually beat them, right? They may end up being first because of their capability. And I and I would say and heritage and also experience.

But but this is why I say it's not that all OEMs will be there at the same time. Some you know, some waves are already doing that today. Right. And rolling up car was included. But I think you will see this coming in a very significant way from an industry perspective. 2728 okay. Very interesting. Thank you very much, Gideon. Very, very interesting topic and very relevant to where we're going with with advanced levels of 8 hours or again, level four, level five autonomy going forward.

And again, communication and data are more and more important than it ever was in a in a vehicle today. So thank you very much for your insight and helping us reimagine mobility together here. Thank you. Thank you for watching this clip of Reimagined Mobility podcast to see the entire episode and more, check out the links in the description.