 So this is the Embedded World 2019 and hi, so who are you? Hi there, my name is Roman and it's my great pleasure to welcome you on our booth at Embedded World. You have really big booths this year, right? Yeah, of course, we are trying to bring it richer and nicer every year to bring the latest technologies of ST and of our partners. Let's check it out what we have there. You have lots of cool stuff going on. Let's go in through here. Sorry. So hello. Hello. Hi, so who are you? Okay, so I'm an ST enthusiast and maker and I have created this project here that is an old one replica software emulator that runs inside a discovery board by ST and I created this to show the potential of this development board because I wanted a real hardware system emulated via software inside a normal embedded system board. So as you can see here there are several hardware components that are interfaced to the STM32 discovery board and this is a common Macintosh classic keyboard from 1983. What is the screen? The screen is a demo that Steve Wozniak wrote to show Steve Jobs the potential of this computer because in those ages it was impossible to use a computer with a monitor and he thought that this would have been a real home computer for real people. And what is this? This is a new system that I am developing and I will launch on Kickstarter in this year and it's a new idea of embedded system. This is a STM32469 chip by ST and it's a big brother of the chip inside. You put the real games there? No, the real games can be plugged in the expansion bus with a normal home made product board like this. The home made is on Kickstarter? It will be everything? Yes, yes, yes. And then old games can just work? Yes, yes. In this case there is a special emulator of the Sega SG-1000 early console by Sega in 1983 and this game is Space Invaders that is fully emulated and if you see it's possible to play and there's also the audio generated in real time. So ST is perfect for emulation? Yes. So the benefit of STM32 products, they are really GP products that can host and manage many different applications including the emulations of these legacy beautiful games and platforms. So here at the show you are launching some new solutions, right? Absolutely, so of course we give a space to our makers and fans but always we use embedded word as the main show where we try to bring new products and new solutions for our customers. One of the biggest launch of this year for us is STM32 MP1 which is extending our STM32 family to microprocessor domain. So we are coming to Linux operating systems with STM32 MP1 and you see we have a great demos already demonstrating the product in the live applications. So this is, what is MP1? MP1 is a product, it's basically like STM32 and reached with A7 cores being able to run Linux in typical microprocessor application but the beauty of this product is that it keeps and retains all the peripheries, ecosystem and support from regular STM32s. So it's ARM Cortex M4 together with the A7? So we have dual A7 plus Cortex M4 inside so the M4 can be used for real-time applications while the A7 will be used to run Linux on Android. We can demonstrate this here. Hi Roman. So who are you? I'm Gerard Bayza, I'm the MP1 software architect. So this is MP1 right here? Yes, exactly. This is a demonstration with our evaluation board. This is an example, we can see that on the right we have the motor. This motor is fully controlled by the Cortex M4. So the Cortex M4 is dedicated to real-time activities whereas on the left, on the screen, we have a user interface that is controlled by the Cortex A7 and we break it in black because of the processing unit. So you have two different OS running at the same time? Exactly. We have a QBAS firmware on the right on Cortex M4 for the motor control and on the left we have Linux running on the Cortex A7. Which kind of Linux? This is a fully standard Linux. ST started several years ago to develop Linux drivers, having in mind to port them after on the MP1 and now we are fully ready because today we announced the product last week and we already have 80% of our drivers that are upstream to Linux.org. So what's the considerations? Is it kind of like a big new move for ST in the Linux world? Yes, of course. Our intention was to extend the existing MCU family, the very well-known STM32 MCU family with a step above. Yes, on graphical side and also with all connectivity that can be breached by Linux. And there's a lot of other MP1 demos, right? Absolutely, so you see we won't have only this demo. We have some additional ones where we demonstrate again the strength points of this architecture. Here we demonstrate the power of the 3D GPU unit inside or another demo is the AI engine running on the Cortex M. So there's a 3D going on here. Is there a GPU? Yes, of course. Everything here is done by the GPU. Which one? This is a Divante GPU. Divante GPU? Yes. And these little boards right here. So you have a camera. And it's putting it into the 3D. The GPU is putting the image on the cube faces. Yes, we have another demonstration. What is this? It is used to show what we can do with the Cortex M4. This time the Cortex M4 is not really used for real-time tasks. It is used this time for Resolve, Bopower, and TVT. In that case, you can see that we have here our discovery package. This is the smaller version of our evaluation board. It is running Linux, like on the other side. And on the Cortex M4 side, this time we put two connected booties, one in Lora 1. And if you want to make a Lina Rover, the standard, you can use the information and send them packaged to the Cortex A7 to display them on the screen. And it's very power efficient architecture. Some things only go when it needs to go on a bigger core. Yes, exactly, because you can imagine to completely power off the display when there is nothing to show. And during this period, the Cortex M4 can continue to gather the information. And is it a smooth process to develop for this platform? Yes, maybe we can move around. So what is this? I can come back to this. So this is the main page of our ST Wiki. So the address is wiki.st.com. We put a lot of effort here to put many documentation that will help our customer to move from the existing MCU to our MCU. And the other thing, the one known as the STM32 Cube and Mix Tool was enriched with new features in order to easily be able to allocate our peripheral whether to the Cortex A7 for Linux or to the Cortex M4 for STM32 Cube. Nice. So is this platform a new launch? Yes, it's absolutely a launch. It's been launched just last week. But basically we did the launch in a way that we are right now in the mass production. So customers can get the samples starting from March. Can get real ones? Exactly. And the Discovery kits will be available from our distributors from April. But around here you already have a lot of partners working on this? Absolutely, because it's not only about the product, right? What is important for us that our customers have very easy and quick development time. And to do so we need a rich set of ecosystem partners both on hardware and software part. So here together we have in total 20 of such partners. Yeah. And we can check some of them right now. Let's do. All right. So let's go in there. Cool. Let me check this one. Hi. Hello. So I'm Simon. I'm from Kontron. What is Kontron? Kontron, we're making customer electronics and customer demand. We usually don't have our own products, but with the ST MP1 we started our… What is this? Yeah, this is our first MP1 sum in the size of a stamp. You do a sum in the size of a what? Of a stamp. Just a stamp. Just a stamp, yeah. A little tiny stamp. That size? Oh, a big stamp. It's one… Yeah. Isn't it that big? It's very small. It's one by one inch. Cool. So it's a little module you put on… What board do you put it on? What is this? All you need. We have a DRAM. We have a NAND flash. We have an SPI NOR flash. And we also have an Ethernet file. So you can only connect your voltage and the Ethernet plug and the web service ready. And what is this demo here? Well, we try to demonstrate the benefits of the MP1. The MP1 has a dual cortex A7 inside, which is doing all the communication and the Linux. So we have… This is connected over Wi-Fi to this demo. And what you see is that we have this connectivity of Linux. But also the Cortex M4 core inside is doing all the real-time stuff to control the LEDs. Because you can't really do real-time with Linux. And for that purpose, you can use the M4 to do motor control and all that. Is it easy to do one and the other? Yeah, it's really easy. In this case, we have some kind of shared memory where the Linux and the Cortex M4 can access at the same time. So we can put the image into the shared memory and just read it out by the M4. So it's really easy. How soon is available this solution? This is pre-order. Yeah, we will have the next songs available in April. April? Yeah, we are already building some new. Is there a price? I don't know exactly, but I think it will be about $35 or below. Nice. Thanks a lot. You're welcome. Thank you. Thank you, Simon. So you see Contron is one of the partners, but we have many others all around. Yeah. They are partners linked also to the software. Cute. Yeah. Do some optimization. This is right here. So that you can try for something. And get ready to work with MP1. Yeah. Okay. We have some other partners and solutions. So we have seen the system on module, but very interesting solution for customer. It could be also a SIP where you can integrate everything into one chip. And this is something that we have here with Octavo Systems. Hi. Hi. I'm Greg Sheridan. I'm with Octavo Systems. And we're doing a system in package around their new MP1. Basically what we're doing is we're taking the MP1 processor, the power management IC, DDR, passives, and EEPROM, and all the other components that would typically go around the system and integrating it into a single BGA package that's only 18 millimeters on a side. So everything's in there? Everything's in here. So basically that means you don't have to worry about DDR routing. You don't have to worry about power sequencing, anything like that. You just hook up power and your EMMC and you're often running with the new MP1 system. So basically making it as easy as possible for people to develop with the MP1, allowing them to easily make low-cost four-layer boards, simplifying the supply chain, everything like that. Really our goal is to make adopting the new MP1 look just like a microcontroller. Is this your invention, this kind of thing? So system in package has been around for a very long time. What we're really doing is we're bringing it to the masses. So it's really been just used for people like Apple or military applications and stuff like that. And what we're doing is we're bringing this technology to innovative companies and allowing anybody to take advantage of it. Like the Apple Watch uses this kind of technology. But again, we're bringing it and making it mass market available. And what does it say here? So this is basically what we've integrated inside. So you've got your MP1 processor, the PMIC, one gigabyte of DDR memory, EEPROM, oscillator, all the passive components. And then what's really cool is... Can I go just one step back? So this is also showing the... Yeah, this is all the integrations. So here's your DDR, the processor, the power management. Sorry. Power management, all the passives, a couple of oscillators as well. So how long does it take you before it's ready? So we will have samples in Q3 and in production in Q4. And the price will be good? We think so. The price should be competitive with doing a discrete solution yourself. What might be the products that are going to be using this? Anything that can use the MP1 really. We're expecting it to be very... Really work for any application. Really size constrained applications are really key. Are people coming up from the microcontroller space? So it's the most compact way of using it? We believe so. We're 18 millimeters by 18 millimeters. So we're typically going to be about 50 to 60 percent smaller than a discrete solution today. Cool. Thanks a lot. Yeah, no problem. Thanks a lot. Good job. Thank you. Have a good one. Thanks. So let's move on a bit. One of our partners is Shiratek. He's just finishing a call, probably calling a customer. It's all right, no worries. Sorry. No problem. So Shiratek can you introduce? Yes, of course. So Shiratek is here together with us since the beginning. We're going to talk about an important element of our ecosystem following the 96 boards platform. Hi. So you just start with the 96 boards you have right here. So which one is this? Okay. So we have here 96 boards IOT edition extended, the big one. We have here the 96 board IOT edition extended. What's the chipset on this? The chipset is the new STM32MP157 MCU from ST. We also have this model for QuickTill, the BG96 for an narrowband IOT model. So you have narrowband IOT together with MP1 here. Yes, exactly. And this is already, it looks like it's ready. It's finished. It's ready. It's working. It's finished. Soon it will be available on Arrow website. And for how much? Not known. I don't know yet. Okay. And what is this? Okay. And this is our new IOT cube. Actually what we did is we took this new motherboard that we just saw a minute ago. And we designed another mezzanine, 96 board mezzanine that fits on top of it. Okay. So you have the 96 boards in there? We have the 96 boards in there and we have a mezzanine also. And on the mezzanine we have a few sensors like gas sensors, microphone. We've also added Wi-Fi and Bluetooth. And we have some terminal blocks for connecting general purpose IOs. And another connector for connecting, actually giving the option to connect to IcePercy, SPI. You also can connect any sensor that you want. This unit is actually an IOT gateway. It's a basic IOT gateway with a nice chip set, the new chip set from ST. And it allows you to connect any sensor you want to this unit and send its data to the cloud. IOT gateway. Yeah. So you call it the box cube? The IOT box cube. It's an IOT gateway based on the new STM32 MP157 MCU from ST. Along with the Quicktel BG96 for neuroband IOT communication. With the mezzanine that allows you the ability to connect to Wi-Fi, Bluetooth with a couple of more sensors and the terminal block for connecting any sensor you like. And send its data to the cloud. How soon? Also very soon. It's in the final stages of design. Cool. Alright. Thank you. Bye. Very good. So of course together with the hardware we have also some firmware and software related partners. Yeah. TimeSys is one of them. Okay. So they are on the Linux embedded site of the development. Combining and bringing the system together with things like security and embedded Linux in general. I'll let Atul tell you more. Hi. So who are you? I'm Atul Bansal from TimeSys. And what we are showing here is the security solution for embedded Linux. One of the problems because we have in the industry right now is what is known as vulnerabilities. And they are coming at the rate of 350 vulnerabilities a week. And we have built a vulnerability scanner that scans the embedded Linux. And it can actually find out the vulnerability specific to your application, your embedded Linux. It's running on this. It's scanning this. It's scanning against this. With the MP1? Yeah. So this is the MP1 BSP. And it is scanning the MP1 BSP and it's finding out all the vulnerabilities. And then it's finding out in this thing where the patches are, what the configs are. And it allows the filtration of the CVEs based on the CVS test score, based on the kernel and provide the patches. So that's one of the things we are showing. The second thing we are showing is a concept of a board farm cloud, on-premise board farm cloud. This is helpful for remote debugging, helpful for doing the test automation. For example, this board we can control from anywhere, anywhere in the world. We can be there and we can control this board by, like I can show you, I can just power off the board. And now the board is off. I can go and turn on the board and you'll see the board is booting here and it's connected. It's somewhere in the cloud, it's connected and we can access the board basically. Wow, that's cool. So this helps with the remote debugging and test automation. Thanks a lot. Thank you. Thank you, Agil. So you see, in a short way, this is where we really come with the MP1, a new product line to our customers. It's the first step, we want to grow this family more further, same like we did with STM32 in past. So stay tuned for future products and solutions. But now I would suggest let's look also on the other things we are showcasing at Embedded World. Alright, let's check it out. One of them is the solution link to AI, the STM32Q.AI. This is new. It's pretty new, yes. It has been announced recently at CES, a show in the beginning of January. And here we are showing it in live action together with our colleagues. Hi, hello. Hello. So who are you? Danilo Pau, STMicroelectronics. So you are able to do the AI on the microphone? Yes. Through STM32Q.AI, that is the tool, it's a software extension of Cubemax and allows to automatically convert any pre-training neural network into optimized code that can run on any STM32 Cortex-M microcontroller. Any? Any. M4, M7, even the latest M33. So which one is running here? For example, this is image classification. It's a mobile net on 18 classes running on STM32Q H7 at 400 megahertz. So it's just algorithms or how do you make it possible to do AI on just any STM32? Because usually the customers use popular deep learning tools, such as Keras, TensorFlow, and so on. And after that they export the results of the training, the neural network with topology and weights, and then the tool does the conversion in optimized code. So these Cortex-M devices don't have hardware acceleration of AI, right? These are standard microcontrollers. So they don't do all the AI only on the Cortex-M? So it happens somewhere else? Well, the training happened on cloud, on powerful GPU, but then the inference, the deployment happens on standard STM32 microcontrollers. How much can you do in a small STM32? I can go down to a few kilobytes. I can demonstrate neural networks even on STM32F3 with 12 kilobytes of RAM and a few kilobytes of flash. Can you do some inference, some AI neural network, simple like that? Yes, absolutely. Can you demonstrate? What can you show over here, around here? What is this one? Hi. So what are you showing here? So we are showing here, we are working with a partner called Carteziam. It's a startup in the south of France. And they have developed Nado-H AI, which is a technology that is allowed to do the learning of the normal set of motors and to detect anomalies from it. So it will connect, it will start the learning phase when it's connected just to display the result to the tablet here. Here it's starting the learning. So on just this little, on this center-tide. Not this. Not this, on this. This is a motor control kit. This one is controlling the motors, rotation. And here you have the learning process. It's based on vibration analysis. And here your system has been learned and you see that it's detected not any anomalies. And then any kind of small anomalies will be detected. This kind of anomalies will be detected. Or typical use guess is also unbalanced. It's going to start playing. And balanced. That's nice. So you managed to put the AI on all your small microcontrollers somehow? Well yes, of course, it depends on the neural network model how much performance you need. But this is exactly the one of the benefits of the Cube MX AI helping you to understand what can run on which microcontroller. So if you move on, I'd like to also present you one of our latest, very interesting technologies. And I'd be happy if Steven will give a short word about the 60 gigahertz RF communication technology. Hello. Hi, nice to meet you. I'm Steven. So what is the relation here? So what we show here is this very, very tiny chip here. It's two millimeter by two millimeter, meant for short-range transmission of very high data rates at a very low form. Is it in here? Yeah, that's the same demo here and there, not the same form factor. So the idea is that basically we're having this box, the Schrodinger's cat, filmed by a camera, sent to this nuclear board over a single UART link to this receiving board there, but through a wireless link. So we just get the wire and replace it by our wireless solution in minutes and it just works. Not a single line of code to update, not a software stack to grow or anything, just out of the box. Almost zero latency in a matter of picoseconds and you got a 27 megs UART link here running between two MCUs, enabling very fast data transmission for a very... How fast? Here it's 27 megs, but this chip is able to do up to 6 gigabit per second. 6 gigabit? Yeah, 6 gigabit per second. Of course, not suited for MCU applications, this 6 gig, but our load at a rate goes up to 100 megs, which is not so low. Alright, and what's the distance that can be done? It can be... Well, it depends on the antenna, but with those nice whole antennas, it can go up to 10 centimeters, so meant for short range, but not so short either, depending on the application you're targeting. And the application could be... Yeah, imagine that you have a sealed box like that. You can fetch data out of the first sealed box with quite nice distance. You can go and fetch data out of it and go in to establish it. It's point to point, no pairing and no interference. You can have thousands of it in an industrial environment because there is no cross talk and no interference between systems. Nice. So you are one of the first ones to come with this kind of solution? Absolutely. So this is a unique solution today on the market, and I think it will open very nice use cases. Because today, if you want to have, let's say, RF communication, you can have NFC, you can have other protocols, but you will never reach such a speed like with this one.