 Hello, this is Ran Kumar and I'm the product marketing engineer for STM32 and STMA product line and I'm working out of the Santa Clara ST site. Today I take this opportunity to go over our new product line and the topic is Explore Graphics, AI and YCUI applications with STM32 H7A3, 7B3, 7B0 product line. Before going to the STM32 H7A3, B3 and B0 product line, I just wanted to give you a recap of the entire portfolio of STM32 and its use. As you can see on the left hand side, there are five categories. The first category of products that we produce are MPUs. This was introduced last year. It's called STM32 MP1 and the second category of the MCUs are high performance MCUs, in which you have STM32 F2, STM32 F4, STM32 H7 and STM32 F7. And then the mainstream MCUs and then we have many of the ultra-low-power MCUs and we also have wireless MCUs on the last row. So today the discussion we will continue is on the STM32 H7 product line. And we have 10 years of longevity commitment on all our STM32 MCUs and MPUs. And so does the newer product line that we are introducing this year in Q3, which is STM32 H7A3, B3 and H7B0 product line. Here is the product portfolio of STM32 H7 and its roadmap. So with the introduction of this new product, which is H7A3, H7B3 and H7B0, we will increase the total number of part variants to be more than 170 part numbers within the STM32 H7. As you can see here on the dual-core lines, these are already in production and they are H745, H747 product lines, these are dual-core product lines. And then on the single-core, the H742, H743 and the H750, which is a value line, all of these products are in production from last years in 2019. They are already in designing in multiple places. So those are already in production. What we are introducing now is the STM32 H7A3 and H7B3 and H7B0, which is marked in the yellow here. So these are single-core value lines and single-core line products. Here the single-core line products can run up to 280 megahertz. And then you have a RAM of 1.4 megabytes on this H7A3 and H7B3. And then you have a flash up to 2 megabytes. These are all internal flashes in the internal RAMs here. And then the value line to give cost effectiveness, we have reduced the flash to 128 kilobytes. And then the RAM still remains at 1.4 megabytes. As you can see on the right-hand side, you see the longevity commitment. And this is again 10 years. And this applies to this product line as well. Again, longevity commitment of 10 years. Here you can see a graph of performance of STM32 H7A3, H7B3, H7B0 product line. We have achieved 1,414 four-mark number with Cortex-M7 running up to 280 megahertz. We have built in highly coupled memories for instruction and as well as data. And this is used for executing critical code or keeping the critical data for easy access at zero weight states. And this is guaranteed. And then we have DMAs that are capable of moving the data over 2.2 gigabytes per second between the cheap resources. And then we have built in L1 instruction data and caches for internal and external code accelerations. So we have designed the product line with an architecture for performance in mind. So what is H7A3, H7B3, H7B0 in a nutshell? There are four main pillars on which these product lines were designed. The first one is the high memory integration. The second one is to have accelerated graphics on this product line. The third one is about the power efficiency that this product line can support. And the fourth one is the embedded security. Before going into the depths of the product features, I wanted to show you a few of the application use cases that can be achieved using this product lines. The first use case I like to show you is how to create a smartphone like GUI from an embedded device. The first one is the graphics hardware accelerations we have built in into the product line, which will bring a better effects transition and fluidity. We have built in large internal SRAMs to bring cost effectiveness for a single shift solution. And the third one is we have TouchEFX, which is a free graphics tool suite for developing stunning applications in the HMI and also simplifies the development. And then the fourth feature is the graphic support is possible even on a 64-point QFP package. And this is improving the cost effectiveness. Here is a simplified block diagram to achieve a use case that we just discussed in the previous slide. In this block diagram, we are showing a Quotex M7 core that's handling an audio and rich HMI with certain real-time control tasks. In this block diagram, you see the RAM. The size of the RAM is 1.4 megabytes. And out of this 1.4 megabytes, one megabyte is continuously addressable and these can be used for frame buffers. And then you have JPEC Quotex Accelerator, which is hardware built in into the IP. And then we have Chromot Accelerator for graphics accelerations. And then we have CryptoHash for doing all the cryptology. And we have the DMA built in inside the processor. And then we have the Flash going up to 2 megabytes on this one. And we have the Display Interfaces, which is an LCD TFT interface, talking to the display hardware. And then we have memory interfaces that can talk to any types of memory, such as the NOR memory, or the PSRAM memory, OptoSpy, or an EMMC, or an SVRAM, or an SD card. And this is a block diagram that is supposedly you can achieve and create a rich human machine interface. Here is STS, HMI Design TouchEFX Designer Tool that can be used on any STM32 MCU, including the STM32 H7A3, H7B3, H7B0 product line. This designer is an easy to use GUI builder that lets you create the visual appearance of your end application in an instant. The Designer Tool cuts hours of your development as it automatically generates and flashes the code to your target board without any compromise to the quality of your end application. And this tool is free of charge. And this is an advanced UI development suite, which is part of the STM32 ecosystem. Here is another GUI solution by our partner, Tara Systems. And the name of this GUI solution is Embedded Wizard. And this is an embedded GUI technology, and it enables creation of a platform independent and a high-performance GUI. On a wide range of embedded systems, this includes all STM32 NCUs, and they have produced a demo on the STM32 H7B3 Discovery board here. And the main features of the development tool chain is one, the resources that it uses are lean. And then it is versatile, which means that it can be used in multiple or different market segments. And then they are scalable because they can run on any platform, which is an MCU or an MPU. And they are fast. They have a VCV editor. They have instant prototyping and debugging capabilities. And then they have ready to use and fully customizable widgets. And all of these are possible on the STM32 platforms. And then the demos are up and running with the Embedded Wizard tool chain. And if you need additional information, the email and the website are provided on this slide. Here is a second use case I want to discuss here. How do I build a wireless module? That's also all my challenges. The first challenge is the smallest package size. On the H7A3, B3, and B0 product line, we have packages going down to 4.57 millimeter by 4.37 millimeters. And the second challenge is most of the module makers require higher RAMs and flashes. In this, in the H7A3 and B3, we have 2 megabytes of flash and 1.4 megabytes of S-RAM. These are all internal memories. And then we have, this is an OTA ready product, and the flash sector sizes are very small, which is like 8 kilobytes of Sector sizes. And the third challenge is about the power efficiency. In this case, we are able to achieve 32 microamps in stop mode with all RAM retention. This is a very important factor requirement for module makers. And the fourth one is security. There is secure firmware upgrade and secure firmware install procedures, which are part of the RSS, the security services that are available on the H7 platform. And this applies to H7A3 and B3 as well. Here is another use case that I like to show you. This is adding natural language, cloud-based voice UI to your product. And for this, the first one that you will require is an all necessary memory should be embedded. You have 2 megabytes of flash and 1.4 megabytes of S-RAM on this one. The second one is the performance. The M7 core can run up to 280 megabytes. And the third one is the security. We have security services built into the H7 platform to do the secure firmware install or secure firmware updates whenever required on the flag. And then the next one should be a one chip solution with all the internal memories and the performance. And then it has to be a cheap PCB solution, which means it is a 2-layer PCB solution can be achieved using this MCU. And then the most important one that in the last one that you can see here is we can connect up to 8 digital microphones in this case because we have 8 channels with 8 filters. This is because of the 2 DFS-DM hardware IP built into the STM32H7A3, B3, and B0 platforms. And this is essentially used as a voice acquisition front-end and then the B-forming and other solutions can be done upstream. Here is another use case that I like to show you on STM32H7A3, H7B3, H7B0. This product line is ready for AI. The first important factor for AI is processing performance for advanced neural networks. In this case, on this product line, we have the Cortex-M7 core running up to 280 MHz, which means you can achieve a core mark of 1,014. That's the first factor. The second one, if you require additional memories, we do have external memory interfaces for complex neural networks. And then these memory interfaces are the octal spy memory interfaces. There are two of those. And then we have on-the-fly decrypt on each of those interfaces. And then we have camera and digital microphone interfaces on this product line. And then we have the power efficiency, which is 120 microamps per megahertz in the run mode. And as you saw before, it is 30 microamps in the stop mode. Here is an overview of STM32H3 extension AI conversion tool. As you see in the slide, you have three hexagons. And the leftmost one is a trained neural network model. You see that in this block, you have many of the off-the-shelf, cafe, TensorFlow, Onyx, PyTorch. These are the tools that you use to train your neural network model. Once you have your neural network model, then the next hexagon that you have to look at is the STM32Cube.AI tool. This tool can convert your pre-trained neural network into an optimized code for an MCU. And this is the code generated to run on the third hexagon, which is process and analyze the new data using trained neural network on the STM32. So the STM32H7A3 can run neural networks that are intended for such devices. Here are a few AI examples that are running on the STM32 platforms today. The three examples that we show you here are today running on different cores of STM32. The leftmost one is an audio scene classification. This is coming out as an example in terms of the function pack that we deliver it as a function pack package. And this is today running on an STM32L4, which is a Cortex-M4-based product. So this example could be converted back and run on the STM32H7, which is a Cortex-M7-based. It could be easily ported. This is for audio scene classification. That is whether you're inside or outside or in a noisy environment, you can find this. And then the second use case that we can think about is a human activity recognition, whether you're sitting, standing or leaning towards right-hand side or left-hand side. Some of these examples can be recognized using the sensor technologies around the microcontroller. And we have a function pack that is today running on the STM32L4, which is a Cortex-M4-based core. This could be ported and reused on the STM32H7A3B3 platforms. And the third one that you're seeing is a full recognition example. This is what we have demonstrated using the dual core STM32H757. This function pack of AI vision package could be used on the STM32H7A3, but we have to look at the performances required for the end solution. So this is a set of solutions that you can reuse on Cortex-M7-based core running up to 280MHz with the increased RAM, which is 1.4 MB, and then the flash is up to 2 MB with external memories that are available on the STM32H7 platform. This makes it much usable in the AI space as well. After those use cases, let's look at the lineup of STM32H7A3, H7B3, H7B0 product lines. Here, let's look at the top features and details of each of the features. The first one is the performance. As we discussed above and until now, it's a Cortex-M7 can run to 280MHz. And then we have zero weight state TCM memories included in the hardware IP. We have 2D graphics and graphics MMUs and JPEG hardware accelerators included in the hardware inside this processor. These are all contributing to the higher performance of this product. The second feature of this product is the efficiency of power. Then in this product, we are able to achieve 120 microamps for microhertz in the run mode, and we are able to reach 32 microamps in the stop mode with SRAM retention. And we have 2 different power domains which we can discuss later. And we have an embedded SMPS on this part. The third main feature is the high memory integration. We have 2 MB of dual bank flash and 1 MB of contiguous RAM out of the total 1.4 MB of SRAM. And we have TCMs included in this one. And then we have 128 kilobytes of contiguous SRAM on an AHP bus. And this is for the high memory section, integration of the STM32H7A3B3B0 platforms. And then the final one is the security. We have included on the flight decrypt hardware on the OctaSpy IP. We have 2 of the OctaSpy IPs inside the STM32H7A3B3. And then we have cryptographic and hashing hardware functions. And we have all the security services enabled on the services as part of the STM32H7 product line features. Here is a block diagram of STM32H7A3B3 and B0 product line. On the left hand side, you see the features that are common to all the 3 sublines. If you can see, this is Cortex M7 280 megahertz based product line here. And we have many interfaces in them. On the top line, you have external memory interfaces, which are 2 OctaSpy buses on the flight decrypts. And we do have 2 DFSDM channels to get it to 8 digital microphones. And we have 2 16-bit ADCs on this product line. And then we have built-in USB OTG, 2 KNFDs and multiple spies and I2Cs and many such peripherals. To differentiate between the product lines, we have on the STM32H7A3, we have 2 megabytes of flash and up to 1.4 megabytes of a strand with a DCDC and LDO integrated into the H7A3 product line. On the H7B3, we have added crypto hash on the flight decrypt along with the security services that are enabled on the H7B3 platform. And on the H7B0 platform, we have reduced the flash from 2 megabytes to 128 kilobytes. And that is the difference between the H7B3 and the B0 product line. H7B0 also has the crypto and the security services enabled on it. These are the 3 sublines of STM32H7A3B3 B0 product line. Here we are showing you the graph of the package lineup for the STM32H7A3, H7B3, H7B0. On the graph on the x-axis, you will notice the pin counts. On the y-axis, you will see the flash in the RAM sizes. And then as you see here, we have variance in the 128 kilobytes flash size variance or 1 megabyte slash variance or 2 megabyte slash variance. And then you also see variance in dark blue, which are with hardware crypto and hash. And light blue is without hardware crypto and hash. And there are variance which have an internal LDO plus an internal SNPS which are shaded with the LO background. And there are some variance with only LDO in them, which is with an LO diamond on them. On the x-axis, if you see there is a 64 pin LKFP variant to a 28 pin TFBGA variant. As you can see, we have a CSP package with a pitch packaging of size of 0.35 millimeter of pitch. And this is 130 to pin WL CSP packaging. And we can go up to 225 pin packaging on a BGA. And then on the y-axis, you are seeing products at 128 kilobytes of flash. The second set is 1 megabyte of flash and 1.4 megabytes of RAM. And the third set is 2 megabytes of flash and 1.4 megabytes of RAM. And these are the variance that are available on the STM data H780, H7B3, H7B0 platforms. One of the main features of H7A3 in H7B3 product line is the graphics subsection inside this IP. Let's just take a look at this one right now. Here are subsections of STM data H7A3, H7B3 to realize powerful graphics. The first one is the chromat accelerator. Because of this, we can realize advanced GUIs with less intervention of the CPU load. And then we have hardware codec for JPEG. This is part of the hardware IP. And we have Chrome GRC, which is a patented IP, which will reduce the number of RAM resources required, especially for the round displays. You can see the saved memory in the part of the graphic picture that is shown on the right-hand side. And we have introduced octal spy interface, which is used to support octal spy PS RAMs if the internal RAMs are not sufficient. And as we have noted until now, this product contains larger internal RAM up to 1.4 megabytes of internal RAM. And without any external RAM or support, you can realize HVGA or WVGA resolutions. Thanks to various graphics peripherals inside the IP, we are able to achieve higher performances on the graphics side. Here in the slide we are showing you how you can realize low and medium resolutions using the embedded graphics engine of H7A3, H7B3, using the internal RAM. As you can see here, on the table below, we are able to realize resolution of XGA, which is 1024 by 768, up to 16 colors, which is 4 bits per pixel, which means the two frame buffers of 384 kilobytes are sitting in the internal RAM of the H7A3. Similarly, you can achieve 32 bits per pixel deep color with two frame buffers of 250K with internal RAM inside the H7A3. If you really need higher resolution than a 384 kilobytes buffer or more colors required in the XGA resolution, then you may have to use an external RAM. But in this case, what we are showing you here is that you can achieve up to 16 colors of XGA resolution with internal RAM of 1.4 megabytes. And then we are showing you here that our graphics support is starting from a 64th in package on this product line. And here there is no external RAM is required for the resolutions that you want to achieve in the pink color that you're seeing in the table here. Here is a block diagram to realize high resolutions of support using the graphics engine of STM that it had H7A3 and H7B3. In the table below, you can see a shaded gray color regions. For example, if the customer use case wants to build a graphic solution at 16 bits per pixel, which is called high color. And then if the customer wants to do an XGA resolution, which is 1024 by 768, the single frame buffer size would be 1,536 kilobytes, which is 1.5 megabytes. So for this, the customer may have to go with an external RAM parts. In this case, they have a choice to use an external serial PS RAMs or an external SD RAM. So in case the customer chooses to use external serial PS RAMs for low cost and low power, they could use the octal spy bus, which is shown in the block diagram above. They can put both the frame buffer 1 and frame buffer 2 in the external serial PS RAMs connected to the octal spy 1. And if in case they need external flash other than or more than the 2 megabytes we have, then that can be connected to the octal spy 2. And if it is needed, if the customer has chosen not to use the serial PS RAM and instead they want to use an external SD RAM, they can connect it through the FMC bus. So this is a semantic block diagram to support a use case if the customer requires high resolutions using the graphics engine of STM director H780, H7B3. One of the important features of the H780 and H7B3 that we have seen earlier is its power efficiency. Let's take a look at this one. Because of the flexible architecture for power efficiency in this product line, we're able to achieve a low power consumption going down to 30 nano amps in the VBAT mode. That is without RTC. And with RTC, we can go to 740 nano amps or 0.74 micro amps. And if you're looking at the runtime, you're going to run at 280 megahertz when the peripherals are off. We're able to realize 120 micro amps per megahertz. And this is running from the flash and the cache is on and the peripherals are off and SMPS is on. And this is a state when we have a typical VDD of 3 volts at 25 degrees centigrade. So at a run mode at 280 megahertz when the peripherals are off, we are able to achieve 120 micro amps per megahertz. And in the stop mode, we are able to achieve 32 micro amps. And in the standby mode, we're able to go down to 2.2 micro amps. And this is in the standby mode. And on the left hand side, you can also see the time it takes to wake up to the run mode. It goes to 10 microseconds going from an autonomous mode to the run mode. Or a stop mode to the run mode, it takes 38 microseconds. And from standby to the run mode, it takes 257 micro seconds. And if you see, there are parts that are embedded with SMPS inside. So if you compare the power consumption in the run mode, it will be drastically low when you're using SMPS based part. The third important feature of the STM32H7BX family is the embedded security that we have built in. And let's take a look at this one. Here, you can see a full set of security enablement on the STM32H7B platform. On the left hand side, you see the encryption, decryption, authentication algorithms that we built into the product. On the right hand side, you see the memory and the IP protection. On the left hand side, you can see the AES crypto engine that can go up to 256 bits, or 128 or 192. We have a SHA-256, that is the authentication engine that we have. And we have a certified crypto library that's part of the STM32 ecosystem. And then we have a true random number generator that is built into this product line. And then you have a unique ID and a key provisioning from STM32 authentication services. These are all part of the encryption and decryption authentication enablements on the STM32H7B platform. Coming to the right hand side, how do we protect the memory? This is the most important question, how do we protect the IP? This is another question. So for this, we have tamper, and as you know, we have backup registers on each of the STM32s. Tamper protection is available. MPU is memory protection unit. And the secure boot that is added into this platform. And then we have read-write protection that is available on every STM32. The important part here is the secure user area, which is also known as a hide protect. So this is a memory area of the two megabytes of flash that you're getting in the STM32H7. A part of that could be converted into a secure user area, depending upon the size of the secure boot code that you are going to develop for your customers. And then we have PCRock, which is also a proprietary code read-only protection. That's another hardware feature. And finally, we have introduced the octal spy on the fly decrypt on the external north flash. So this is another important feature that you can execute in place from the external memories and also with a decrypt on the fly on top of it. So these are the few features of this H7B platform from a hardware perspective. And let's take a look at the features that are enabling the security services coming up. Here you're seeing STM32 trust. STM32 trust is a brand name for STM32 security services that are built on each product line. For example, in this case, STM32H7 series. In this case, you go from bottom to the top. We go from the bottom. We have SP, SFU, which is essentially a reference software package for the firmware update, which is a secure boot. And it also checks for the root of trust for a particular MCU that it boots from a known set background. And you have SFI, which is a secure installer for the first time. This is almost an important requirement for many of our customers where they need to program the code into the MCU for the very first time in a premises that they are not sure whether it is secure or not secure. So this is one of the services that are coming out, which is called SFI, secure firmware install or secure first install. And then the next block that you see is the STM32 concept of support of customers for secure boot and root of trust. This is part of the STM32 platform. And all of these things combine and we call it as a global security, which is ecosystem and the services behind this are many. So these are all the building blocks of the STM32 trust, which is a branding of security services on the STM32 platforms. Here is one of the security services and this is called SFI, which is secure firmware install. In this case, we are able to program the STM32 in an uninterested environment. And this is the message we are trying to pass here. So either a third party premises or a customer premises will actually use something called a trusted package creator to create two outputs. One is an encrypted firmware, which is the code that the customer has developed. And then they will also inject the keys that are required into the HSM. And these two are sent to an interested environment. In this case, it could be a CM or a programming house that is receiving the HSM and also they will receive the encrypted firmware. And once they receive these two informations, they will also receive the physical STM32 device because they are the programming house. They will receive the STM32. So they have a package within the tool set that are required to authenticate that the STM32 has actually come from ST. And this is managed through the HSM that is going to find a relevant certificate on the STM32 H7. That's how we maintain that this is the right STM32 and this is not a fake part. And then once this is decided, then we will identify that then the interested environment or the CM or the programming house will use the tools provided to them to decrypt the firmware and then securely load that firmware back into the MCU, which is an STM32 device that will be programmed. And this is the first time there is securely programming the STM32 device in an interested environment with the means of the HSM and the certificate enabled STM32 H7 device. Here is a show of our secure boot and secure firmware update block diagram. What we provide here is a reference library with source code. This reference library demonstrates software modules that are required for a number one and the secure boot. Number two, the secure engine. This is a consortium of software or hardware crypto libraries. And the third one is firmware update image management. This is the three modules that we deliver as part of the library and the source code. And we have reference implementations of the SB SFU, basing upon the end products. For example, in this case, STM32 H7's hardware memory protections that are enabled on its platform. Here is a show and tell of an optical spy encryption and decryption with on-the-flight decrypt engine. In this, we are showing you STM32 H7 in the block diagram on the left-hand side. In the middle, you're seeing an external flash device. We have subgraded this into, you know, in this block diagram for regions, region one, two, PN4. And on the H7, we have a write function to write it through the octal spy, which is going and writing into the external flash with an encrypted code or data in the region one. And when you want to execute it, then you will bring it back. The octal spy is going to on-the-flight decrypt and the decrypt engine is going to put this back as a plain text code or data onto the STM32 main flash device. And for writing function, you could either use AES crypto mode, AES mode, or there is an additional mode that we have, which is called proprietary mode. More details are available on the reference manual, but this is a generic block diagram of how you can implement multiple regions using the octal spy and on-the-flight decrypt engine. We have used AES CTR 128 bit mode for the code and data management on the on-the-flight decrypt. So this is how we are using the system on the STM32 H7B. Many details we can discuss it offline. Talk to your local PMAs on this topic. Now that we have seen all the features of the STM32 H7A and H7B platforms, let's just take a look at the ecosystem for this product line. The ecosystem consists of one, the software, the second one is hardware, and the third one is customer support. On the software side, we have the STM32 cube supporting this platform now, and then we have the STM32 cube programmer supporting this new product line. And we also have a nuclear board, which is the simplest board that you can see. And then we have a discovery board with a display. And then we have a full evaluation board with all the peripherals mounted on the board. And this is a bigger board. And these are the three different types of hardware solutions that we have for our customers to be used. And then we have, on the customer support, you have local FAEs, the PMAs, and the worldwide support we have in the community that's supporting the STM32 platform. And we have many books that are coming online on this one. And this completes our entire ecosystem support for our STM32 H7A and H7B platforms. So far we have seen all the features of STM32 H7A3, H7B3 platform. This is a new product line that we are coming out in the Q3 2020. And what are the key takeaways from this session? Let's just take a look at these. Here is a summary of STM32 H7A and H7B products. We are bringing a unique balance of features in this product line. The first one is the high performance. We are able to achieve 1,414 core mark or 599 drystone maps using a Cortex-M7 core at 280 megahertz of top speed. And the second and the important one is the high integration of memories. This is up to 2 megabytes of embedded flash and 1.4 megabytes of embedded RAM memory. And then the third key takeaway or feature is the power efficiency, which is at one mode, you are at 120 microamps per megabytes. And at stop mode, you are going down all the way to 32 microamps with full RAM retention. And this is a key takeaway, which is the power efficiency. And the third one is the advanced security features, including the enablement of external memory interfaces to do an on-the-flight decrypt. And we have native security services, including the secure firmware install, secure firmware updates. And this is our fourth key takeaway. And these are the four key takeaways of STM3 to the H7B product line.