 Hello, my name is Michel Maurer and I'm working for Room Semiconductor. I'm acting as product marketeer and responsible for others application and with another head, I'm also product marketeer and responsible for power management ICs, DC-DC regulators which are going into the automotive industry. Today, we want to talk about our new power management IC, addressing the satellite cameras for others application. So, this is in the cars, it could be electric cars or self-driving cars or not or any car that has these cameras that do like where you park and go back and all this stuff or security and... It's for all the cars in the future which are driving autonomously or now in the transition phase where we have the mixture between autonomous cars and manual driven cars. You have everywhere cameras and the cameras are looking to the environment and trying to select and to detect objects which are there in the surrounding and to avoid any accidents. So, it's the eye of the brain of an autonomous driving car. So, a lot of these cameras will be mounted on a car. We're talking about six, eight and more cameras on each single vehicle. And so, Room is famous for doing billions and billions of chips, right? Yeah, correct, correct. We are exactly, especially in the automotive industry, we're selling or we have already sold in the past more than 3 billion LDOs, 400 million bug regulators, 200 million LDO drivers and on the camera, on the PMIC side, we sold more than 100 million devices. So, it's huge volumes which we are currently selling into the automotive industry and we are trying to expand this business significantly. Room is a very important significant company from Japan, right? And is it Japan headquarters, right? Correct, correct. All the manufacturing facilities, especially semiconductor facilities are located in Japan. So, we have a lot of manufacturing. We have our own in-god devices from the initial production part or from product part until the manufacturing, until the packaging, the testing, everything is done at Room. Semiconductor wise, everything is done in Japan. And if we are going towards the back end, means the packaging and the testing, that's going to the greater China. Nice, and Room is, I checked on the internet, it kind of stands for resistance and the Room is the famous, what do you call it? Yeah, that's really the past, yeah. So, it is almost 64 years ago where Room was founded and where the resistors, yeah, not invented but they were improved the manufacturing and the reliability of these devices were improved. And then within these, yeah, now history of 64 years, the company now is developing much more complex devices, but we still have huge manufacturing capacities on resistors. And this is your presentation, are you doing this during the Embedded World Digital? You talking about the stuff over there? Yeah, it's for the Embedded World or it is for the Embedded World, but it's also for all kind of opportunities we are gathering with our, or addressing with our products here. So, we are using that plus some more technical details also going to customers, discussing about the device and the pros and cons and everything what's around that. So, what goes into, is it PMIC? Is it Pro Management IC or what is the part you're talking about? I'm talking about Power Management IC. Power Management IC, you have to see it as a legal, yeah, legal. You have these switching regulators and these are let's say the green parts and you have there some, you have some LDOs which are orange parts and you have there some logic which is blue part and you're combining that and for this device, for this Power Management IC, we have there let's say one primary bug, it's a bug regulator which is connected directly to the car battery or can be connected to the car battery or here for this application it is connected to power over coax supply and then we have there secondary power management regulators which provide then the energy towards the serializer, towards the imager and so on and so on. So, there's more and more demand than the way these amazing cars are getting all these cameras and it's just fascinating to see what Elon Musk's company is doing with the self-driving and they're doing everything based on vision, right? So, it's really important to have all the system, everything working in that direction. Is that, are you just an important part of that or? Yeah, yeah, first of all, just one correction, it's not just vision or vision. Yeah, we also, all these autonomous cars, they are not just using camera stuff. They are also using radar and future leader. You probably know your sooner stuff when you're trying to park and you have all these beeps. Yeah, and at the end of the day, it is the combination of all these sensors. We call that sensor fusion, but the camera itself, so the vision will play an extreme role of it. You have all these cameras and they are trying to make a 360 degree picture of the surrounding and there you can see the impact of these camera stuff. And if you're talking about just considering the volumes, just assume you have one million cars which have this implemented and we are talking about eight cameras. Yeah, now we are all immediately talking about eight million power management ICs. So now let's assume that in Europe, we are selling, even in Corona, it was more than 10 million cars where it was really very low or very down the sales on the cars. We are talking about 80 billion devices and more. It's unbelievable mass. Yeah, it's unbelievable mass. This market is huge and it will become big and bigger. Now, if you think on a global basis, and I remember that since long years, people were trying to have their 100 million cars sold on a global base and just assume that you would have their 50% or 50% in, let's say, 30, 20, 35. With these cameras, 50 million multiplied with eight cameras. Can you imagine the volume? It's unbelievable. So you can see the big, big row these devices will have. There I see it's a very small chip. So it's not an ARM processor. What does it do? So I said that the input is a power over coax. So there is just one cable which carries the data, the exchange of the image data and the power. So we have one input, but these imager, the serializer and a few other things they need. They need energy and you have to generate out of the single input voltage, multiple output voltages in a very efficient way. So this is number one. Number two, you have to ensure that this is working. Means any abnormality needs to be informed or an higher instance needs to be informed that something is happening because it's part of ADAS. You have to ensure that it's operational. So functional safety plays an extreme role and that in a very, very small form factor. If we talk about these satellite cameras, these camera systems which are connected via power over coax, we are talking about an application size of roughly two euro coin. Means estimate to 20 by 20 millimeters. Is that the area of the chip or the area of the camera and everything? It's the area of the sensor plus the serializer plus the PMIC plus all the externals. It's so, so small and our chip comes in a package of a size of three and a half by three and a half millimeter. You can see that here in the picture where we zoomed it out. And it's just a few external components which are necessary. So we are able to bring it down at the size of less than 100 square millimeters. This is the size of less than 10 by 10 millimeter. We are already working on a solution which provides an or which which can be on an area of around 80 square millimeters. So this is really a must being very, very small. And another point which comes into the game for this very small area we have existing is the power dissipation. Means all the power rails we are generating needs to be generated in an extreme efficient way. So the architecture plays a role. The way how you're doing that, the efficiency means the size of the transistors and all the things that plays an extreme role. And we have the solution for that. We think the device what we have at the moment is the smallest power management I see for this application in the world. When I look at the zoom in that you have there, you have the little black part which is a chip, right? How about this other stuff that's around it? Is that also part of your architecture or your design or? Yeah, exactly. For switching bug regulators and a lot of these regulators which are here show or which are necessary, they are switching regulators due to efficiency reason and therefore you need inductors and some capacitors. What's shown over there is more than the application would need. We have there some spare parts in it, some spare placements for additional capacitors as an example just to play with it. So that the designers can evaluate by themselves what is the optimum in their solution. But we already tested everything in such a way that we will come to today's solution of around 85 square millimeters. This is really outstanding. So there's a need for power management in everything because we're doing everything that's using power everywhere all the time. And I guess to try to understand the power management is it also has to do to get it's not easy. It's like you need to get exactly the right amount of power to every electrical component. If they get too much power, it doesn't really work or too little power or is that part of the deal? You're absolutely right. You really have to pay attention there. What we have done, we have implemented a kind of flexibility in our chips. So we can adjust the output voltages. We have a lot of or we have some kind of headroom in terms of output current. So we are able to define the power sequence means how the PMIC starts up, what rail comes after the other and so on. There are a lot of parameters which are configurable via so-called OTP, one-time programmable logic. And this gives us the freedom to address a great variety of implementations, means different sensors, different serializers and so on. And one important thing is these functional safety features what we have there on board. In contradiction to discrete solutions or to some competitors, we have there already some features included or the necessary features included. And that allows them to really shrink the application to the given form factor. So here on your slide, is that what you're talking about? The functional safety, UV, LO, all this? Okay, yeah. So we have these typical things implemented like the undervoltage lockout. This is the initial voltage or the minimum voltage you need to start up. This is the undervoltage lockout. We have a short circuit protection which protects the chip itself that a shorting doesn't damage it. Or we have an undervoltage detection or overvoltage detection. This is we are detecting the output voltage if the output voltage is in the given range. So if it is slightly above or outside this given range, we are notifying an higher instance that something is going wrong. And that's important for these functional safety features. It's important that then the sense of fusion processor can take then the action to do there something that this is no more happening, that your application or this application at the end of the day is stable and can contribute to the autonomous driving. So this doesn't ship with any software. There's no risk that it will be hacked. One of the weird things that I saw a presentation a few years ago, I don't know if it got better, where people are saying that all these cars are so easy to be hacked, all the ones that were made until like two or three years ago. And I don't know if they're improving in that they're there. They're absolutely improving. And I think they are also learning from the former telecommunication stuff where people used to work and were trying to avoid some hackings. But here on this device, on this PMIC, there is no software included. Everything internally is done by registers. We have state machines, which is just logic. There is no software included. So in principle, there is no possibility in hacking the behavior. For sure, you could try to modify the behavior or the functionalities via the I2C access, but even this I2C access is protected so that with no deep know how the PMIC operates, there is in principle no possibility to hack this PMIC device. Because when you do something, and I'm looking forward to it in some way, but I mean, it's fun to drive a car, right? But it's going to be weird when we don't drive the car anymore. And then these cameras, they just need to be 100%, 99.999999% reliable. And you are providing an important part of that safety. Yeah, exactly. And this is a real very, very important point. And we have the solution for that. Everything is protected. Even the generation of the voltages internally is done in such a way that we have redundant references, redundant or dual clock systems which are supervising each other. So we are trying to provide the highest safety level at the end of the day for this kind of application. So you're providing the power to the camera and the data that goes through back to the main chip that does all the processing later somewhere else? Now, as you can see it here in the slide, what you have shown there, we are covering just the top part there, which is meant to be power supply. And that's just generating the energy which is needed for the image and for the serializer and the other devices which are here in the surrounding. The data itself is generated, first of all, by the image. This is then transferred to the serializer and then from the serializer towards the coax cable. Here we have the AC, which is the data. And here we are talking about a huge number of data. We are talking about assuming an 8 megapixel camera and 16-bit color depth and 30 frame rates per second. We are talking about 4 gigabits per second. It's a huge amount of data which is transferred via this cable and we have also the energy which is going for that. For sure, there is some minor data going also in the direction from the sensor fusion towards the PMIC just to control the things. But this is really very minor. It's neglectable versus the image data which is transferred over there. When you talk about 8 cameras, it's 8 times 4 gigabit per second that has to be powered and sent through and to the big little iron processor that does... So probably these 8 megapixel cameras, they are a little bit the future but we can see that even today people are considering the 8 megapixel cameras on the front. You will have two cameras on the front and then you have six other cameras which are somewhere in the surrounding of your car. So to get the 360 degree surround view and these are mainly 2 megapixel cameras, 3 megapixel cameras probably in future 4. If we are talking about 20, 30, 25, 20, 40, 8 megapixel, you never know. Because when I see and it's very interesting how confident Elon Musk is that his vision system is just what people need. Of course you talk about sensor fusion and LiDAR and everything but I always wonder that the human resolution when you drive the car, it's probably more than 8 megapixel. So I'm thinking that maybe... I don't know if he's future proof of all these Tesla cars. Maybe the next gen cars will have so many more megapixel to have more safety. They need to see more, not only pixels but also other stuff. You will have a lot of these parallel processing at the end of the day. As I said, it's not just a vision. At a certain moment these 8 megapixel probably they will be replaced by 16 megapixel. You never know. You never can say that this is the end of the development but we have more and more of these sensors which are working together. So the disadvantage or one of the disadvantages for these cameras is in light load mode, means in the darkness. So you can't see there too much. So therefore we have there other solutions like LiDAR, like radar or one of the biggest challenges what we can think of is driving into a tunnel. So cameras will have there really some problems because there is an average for the brightness and the tunnel itself. If you go there, if you would zoom there in, it's really dark. So this is big trade-off and it's really very difficult to find here good solutions. And at the end of the day, I would say sense of fusion. That will be the solution as we can see it today. So for sure the camera will play an extreme role for parking, for the blind spot, for everything for lane departure warnings and so on and so on. But at the end for the application for other real autonomous driving cars, it is the fusion of it. And therefore we need huge processing power. We are talking about billions of operations per second and yeah, also here, Rome is working on solutions for that for supplies of such kind of extreme high processing SOCs which can then operate and in a first step include all the data which is generated by the sensors and then the software can operate. And this is a huge challenge. And you can see this is the reason why it takes so long that the real autonomous driving will be fully implemented. And one of the biggest risks, but this is my view, is the mixture of autonomous driving cars and the cars which are driven by humans. Hybrids. Yeah, hybrids. So I've been doing lots of videos about ARM CPUs, ARM SOCs. And in my understanding, one of the things that they often include directly on the ARM chips is the power management. How does that differ compared to what you're doing? Is Rome doing so many different solutions? Are there also some ARM SOCs in there that are having this stuff integrated, but it's better to have it separate sometimes? Yeah, so first of all, it's not really concentrating on solutions which have processes inside. So the majority, and this is what I said before, is in power management generation. We have for sure some other devices like serializers and so on where you have some kind of processes, but it's more on controller base on controller level. So the integration of power management into processors. Yeah, we have to go there, the step into the semiconductor world, into the semiconductor technology. So these modern processes, they are using technology, semiconductor technologies which are 28 nanometer in the automotive industry, probably in future going down to the 10 or 7 in long term. And these are very expensive processes. And if we would do analog designs in these kind of processes, we can't really shrink the transistors because these transistors, they are generating the output voltages. And these transistors would remain roughly the same size and this would be extremely expensive. So the sweet spot in the power management as of today is in the range of 180 nanometer down to let's say 90 nanometer as we can see today is the cost of the initial cost of the semiconductor technology. Sorry, you're muted. Sorry, I forgot to unmute. So there's no point in talking about 7 nanometer power management I see or something like that. Not at the moment. Not in 2020, not in 2030. And I even don't think that it will be realized somewhere in 2040. So as I mentioned, it's so expensive we are talking about masks of a million of dollars wafer cost of thousands of dollars. And if you compare then it in an economic way so it won't worth it. We know that some of the SOC vendors are implementing such kind of solutions on their SOCs but they are doing it just because they need sometimes some extreme accuracy which they can only generate have on the silicon itself. And there's a big story right now going on with the chipset what's called there's not enough supply all these car makers are not getting enough chips are you part of the bottleneck or are you supplying with no problem what you need to supply? I wouldn't say that we are not supplying by no problem but we are supplying we are trying to support our customers as good as we can and I don't think that we are one of the root causes why some of these manufacturers are stalled at the moment. This is one of the advantages of Rome. As I started at the very beginning we have everything enhanced from the input until the packaging. So there is no foundry service, there is nothing we do not have to go to somewhere outside we have everything in the house so we have huge manufacturing facilities regarding the semiconductor for sure we are going to call it offshore going to China, Philippines, Malaysia for the baghunt but this is easy to be done we also are owning there the FABS for that and the testing facilities and we are not using as of today any external services so that's a good thing here. And I guess sometimes to invest in these FABS it costs maybe billions or at least billions of Japanese yen it costs a lot of money sometimes right? Yes exactly, this costs really money but it will worth it at the end of the day so sooner or later while going to foundry services you have to pay that money in your way for cost and this is the advantage that we do not have that so we have to invest once and then we have the things available but we are expanding significantly our technology, our bandwidths here and we are investing in new technologies and we have that, we have now a 130 nanometer facility FAP opened which has huge wafer throughput and we are well positioned for the future there. And you are coming here to the embedded world 21 digital so talking about automotive stuff but I guess this is relevant for everybody who does stuff in the embedded world you have solutions for all the PCBs So the main focus of the company itself is automotive and industrial so we moved a little bit away from the consumer parts but you can use all these devices in any kind of application so one of these examples what we were talking about on the camera was the application sorting as an example the letters so there is somewhere a camera which looks to the address on your letter on the post office and trying to figure out the addresses and this can be done with such kind of camera systems because they are also connected via cable somewhere to a processing unit and I guess the R&D departments and let's say where you are connected maybe with all these European car manufacturers who probably want to the next gen, they want more innovation they are just asking for pushing the boundary all the time so it's probably an interesting roadmap also maybe what could come out in the future to optimize everything everything more self-driving, more secure and the governments need to start to legalize this because as far as I understand it's already in a level where it's possibly safer than humans so at that point it just needs to become law and you'll get busy So this is one of the biggest challenges I would say who is reliable at the end of the day for the decision an engine has done so this is one of the biggest challenges and they are the insurance companies I think they have to really consider how that's done and you are absolutely right it's also a task of the government to find solutions for that because human has programmed the software which is deciding whether they are crashing into A or into B so this is a huge decision and really complex decision not easy and what we can see at the moment is that even the OEMs are trying to get involved in the development of such kind of software into this ECU part of the ADAS application it's no more the big tier once it will become also the OEMs what we can see today So thanks a lot for this overview and I guess all these SMT lines and stuff that are part of the automation of getting these cars done more and more automatically they have all your chips lined up and put them on the boards and things are getting more and more accelerating and making more of them every year So we see a big big rose here and I think we have now nice products which are currently existing or which we have in the pipe and probably we can talk about a new generation somewhere in autumn of power management ICs which are which have an approach out of the box this is totally different what we had before and I think this is also the way how we can contribute to this extreme growing market and this market which needs a lot of innovation because it is clear we are at the beginning of this autonomous driving it's absolutely the beginning so this autonomous driving is divided into five different segments we are at the moment we are in the segment two where we can park autonomously sometimes where you have your lane departure warning next steps is that you can drive but the complex things they will come and they need the innovation from us they need the innovation from the SOC vendors there is a lot of innovation necessary to find the solution going the next steps Sorry to bring back Elon Musk all the time but I think he says on Twitter that his level 5 is definitely kind of ready but kind of for sure this year so things are coming fast he is really a pioneer but if you are looking to the news and what has gone wrong sometimes on the 101 I was very often in the US so there will be a lot of hurdles and it will need some time until it will be really established and I think one of these I think already two times it is the mixture of the autonomous driving and the human driven cars and this will become the complex things if everything would be autonomous I think in 2030 we would have the majority driving like that but as soon as we have the mixture it is much more complex Cool alright so thanks a lot thanks for this embedded world 2021 digital presentation right here Many thanks and also giving me the opportunity talking to you Thanks a lot and thanks for doing this awesome technology that is changing the world that everybody can integrate and looking forward to these self driving cars it is going to be so comfortable to sit down and watch YouTube videos while we are driving around Cool thanks a lot