Choices for Industrial Designs
As MCU performance and functionality improve, the boundaries between MCUs and microprocessor units (MPUs) have gotten murky. In this article, Jacko examines the changing landscape in MPU vs. MCU capabilities, OS implications and the specifics of new SiP and SOM approaches for simplifying higher-performance computing requirements in industrial applications.
As semiconductor suppliers ramp up efforts to satisfy the needs of the Industrial Internet of Things (IIoT) or Industry 4.0, the choices of computing alternatives have broadened. However, as microcontroller (MCU) performance and functionality improve, the traditional boundaries between MCUs and microprocessor units (MPUs) become less clear. Understanding today’s options, including the use of System in Package (SiP) and System on Module (SOM) solutions, can help you make the right decision for a specific application.
Traditionally, the perception of an MPU is a powerful processing unit with the requirement to have several external memory and power management devices. In contrast, an MCU has some or all of these items integrated within a single chip and package. Meanwhile, software and programming were markedly different between the two categories. The hardware and software differences make the transition from one design approach to the other quite difficult. Figure 1 shows the current differences between the integrated hardware aspects for today’s embedded MPUs and MCUs.
By applying the multi-chip SiP technology that has been used in high-volume consumer products for many years to embedded industrial applications, MPUs can have some of the missing system blocks integrated into a semiconductor-type package. Further system aspects can be added using System On Module technology. A SOM is essentially a small PCB containing the SiP and additional components. That SOM can in turn be mounted directly onto another PCB.
Today, with increased integration possible at both the chip and the packaging level, SiPs and SOMs remove most—if not all—of the hardware design and programming differences between 32-bit MPU-based SiP or SOM and a 32-bit MCU designs. Table 1 shows an updated comparison of these design decision attributes.
|Time to market||Good||Best||Better||Best|
|Packaging example||10 mm × 10 mm BGA||2.84 mm × 2.84 mmWLCSP||14 mm × 14 mmBGA||38 mm × 40 mmPCB|
TABLE 1 – The comparison of key attributes of MPU, MCU, SiP and SOM shows how the landscape has changed.
For industrial applications that could use either an MCU or an MPU, embedded system designers typically fall into three categories: (1) designers with an MCU background; (2) designers with an MPU background; and (3) designers who are new or have no preference for either. The designer’s background can play a major role regarding the design choice. However, SiP and SOM-based MPUs can sway the decision.
The first category is perhaps the largest. MCU-background system designers have to consider an MPU if they need to run Linux and/or they need external memory. Adding that external memory to an MCU sometimes makes the total solution cost even higher than an MPU-based system. This approach also has less flexibility because the memory choices supported on MCUs limit the amount of code and data space available to the firmware designer. The SOM provides many, and in some instances all, of the functions the designer was used to in the MCU world with more memory and improved performance.
In the second category, an existing MPU designer is interested in a SOM for simplicity of design, fast time to market and/or faster time to cash flow. This type of user can also benefit from having hardware that is easy to design-in due to its integrated power management and memory, potential to be preprogrammed and ease of installation/soldering. As an alternative to the SOM, the SiP gives designers in both categories more freedom to implement their hardware while still benefiting from some of the reduced complexity with DDR high speed memory interface being integrated in the same package with the MPU. While not committed to one choice over the other, designers in the third category just need to be aware of the new design options available to them.
In any case, one of the tradeoffs that must be made by the technical design maker—hardware developer and software developer, project manager or even the CEO—is the use of Linux. Moving to Linux is a substantial investment but based on the extent of connectivity in a new project and the company’s plans to implement connectivity more extensively in the future, the move to Linux and an MPU gets easier.
LINUX FOR CONNECTIVITY
For the IoT and especially the IIoT, designers need to add connectivity to systems. Network connectivity requires protocol stacks, security layers and potentially even a web browser, to exchange and forward information and communicate data. This is typically software that is readily available in the Linux platform without any additional licensing fees. With the Linux OS comes the largest selection of device drivers, which removes barriers to the choice of peripheral devices and their suppliers. However, unless experienced Linux designers exist within the organization, proper training, education of software engineers, is a must for successful MPU implementation. The big advantage in the Linux community is an abundance of qualified engineers or consultants.
Embedded system developers moving to Linux will be successful if they recognize that they need to bring in expertise that they are lacking. This is typically the case of an MCU customer transitioning to an MPU. In contrast, an existing MPU customer/ Linux user knows the benefits of expertise—so no convincing is required. Another key decision factor is support of Linux distros that are used in industrial applications—such as Buildroot or Yocto—where the MCU to MPU transition and subsequent need for SiP and or SOM would be considered.
Finally, some cloud computing companies’ products—for example, Amazon Web Services (AWS) Greengrass—are built entirely on top of Linux. An MPU running Linux simplifies the design of any product that interfaces to these types of services.
SiP AND SOM SOLUTIONS
Using a two-step approach to obtain more system functionality and solve common system design problems, Microchip Technology developed a SiP product based on the Arm Cortex-A5 core and a SOM that uses the SiP. The availability of both from a major MCU/MPU supplier and choice of either one by a system designer simplifies both the design process and sourcing and inventory in the factory.
The first step, the SAMA5D2 SiP, a Smart ARM-based Microprocessor (SAM), integrates DDR2 memory and the SAMA5D2 MPU. Three memory levels (up to 1 Gbit) are offered to address different system requirements. In addition to addressing space constraints, the integration removes the high-speed memory interface constraints from the PCB. With the impedance matching performed in the package, not manually during development, the system will function properly at normal and low-speed operation.
Power is an additional design consideration in the SiP. The use of DDR2 instead of DDR3(L) memory keeps the power management simple. DDR3(L) would add a 1.5 V or 1.35 V power supply requirement to the system. In addition, at the SAMA5D2’s maximum operating frequency DDR2 uses less power than DDR3 – without compromising performance. Figure 2 shows the additional memory as well as other integrated blocks for connectivity, user interface, control and security in the SAMA5D2 SiP.
The second step is the SOM. A SOM typically includes the processor, DRAM and boot memory, power management and support chips so that system designers only need to add non-volatile memory, and application-specific peripheral support on the host PCB. This integration of multiple external components eliminates key design challenges around electromagnetic interference (EMI), electrostatic discharge (ESD) and signal integrity.
The 38 mm x 40 mm SAMA5D2 SOM simplifies system design by integrating power management and non-volatile boot memory which offers single supply and flexible memory implementations similar to a classic microcontroller. The SOM can be readily used in a standard surface mount manufacturing process like a QFP package or can even be hand soldered during initial prototyping. Figure 3 shows the key additional aspects of the SOM MPU.
CASE STUDY EXAMPLE
Power distribution units (PDUs) that provide the electric power for computing, especially in server arrays, and networking equipment use either local and/or remote monitoring. Initially, system designers used a small MCU to obtain network connectivity. However, for cloud computing access they struggled to get all the protocol stacks and correctly address the security aspects. With an MPU with build-in hardware security and its easy access to Linux support tools, both of these issues were easily resolved. This example identifies two final system requirements and benefits of using Linux: long-term software support and security.
Source code is important to ensure support in the future. While users can provide the support themselves and they may very well have to with solutions from smaller providers, Microchip’s MPU, SIP and SOM products and their mainlined Linux distributions are a unique offering where longevity is guaranteed. The longevity aspect is all part of a corporate philosophy of customer-driven obsolescence. The hardware components and the software (based on Buildroot or Yocto distros for Linux), that goes with them follow these principles. Like its other products, Microchip will continue to manufacture and support its MPUs, SiPs and SOMs for 10, 15 years or more.
System designers recognize the need for security but often they do not feel comfortable implementing it. With integrated security in the SAMA5D2 MPU, either the SiP or SOM that uses it provides encryption, tamperproof access and all the building blocks to implement secure systems.
Once a decision to implement an MPU-based SOM has been made, the next step—or another factor to help make that decision—is the availability of an evaluation kit. The ATSAMA5D27-SOM1-EK1 evaluation kit (Figure 4) includes a camera interface, an LCD interface, other standard extension board connectors (such as PMOD or mikroBUS), standard communications interfaces (including Ethernet, USB, CAN) and SD card interfaces (both standard and micro). In addition, a reference design along with the free schematics, design and Gerber files and complete bill of materials are available online. A final design-in touch comes from Linux with the availability of the largest set of device drivers, middleware, and application layers for the embedded market free of charge.
Because project volumes in the industrial market are smaller than in consumer, communications and automotive markets, many smaller industrial system designs are ideal candidates for the SiP or SOM-based MPUs, especially when Linux connectivity is required. Medical, point of sale (POS) terminals, power distribution units in servers, building automation as well as factory automation and many other applications are included in the broad industrial marketplace. While MCUs may continue to be the right approach for many of these applications, the new options allow design decisions and tradeoffs to be determined based on each embedded system developers’ goals and future plans.
For detailed article references and additional resources go to:
Microchip Technology | www.microchip.com
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • JANUARY 2019 #342 – Get a PDF of the issue