2007-09-09   The rise of high-speed 8-bit network MCU
  By Jason Wang and Harvey Jan, ASIX Electronics Corp. [2007-09-09]

The evolution of the Internet began in the early of 1970's with the ARPANET project by the US Department of Defense to make the communication between different computer systems possible. In the 1980's, the academic world and universities began to participate in the development, using LAN (local area network) as a means to share data and files between different computers. By the 1990's, corporations and businesses began to operate under the World Wide Web (WWW), and online users increased dramatically with email and e-commerce as significant driving forces. The ensuing Internet era will augment the computer-to-computer networks of the 1980's and the user-to-user networks of the 1990's with Internet capable devices, resulting in the rise of M2M communication and the so-called "Ubiquitous Network Society."

The microcontroller used in embedded system design is primarily focused on control capabilities, where integration is the primary objective. Based on the target application, chip designers will incorporate different degrees of memory, input/output interface, and computational ability into a single controller.

Generally, microcontrollers integrate a variety of serial or parallel peripheral interfaces with GPIO (general purpose input/output) capabilities. Serial interfaces include UART, SPI , I²C, Microwire and 1-wire.

The parallel interface is linked to other chips using an external memory interface (EMI), local bus interface (LBI), or PCI bus interface. A microcontroller can be configured to be a Slave host, at which point any other controller configured as Master host can communicate to the Slave using a conventional 8/16-bit parallel interface. The timing functions can be achieved through a watchdog timer, timers/counters, programmable counter array (PCA), or a real-time clock(RTC).

The inclusion of other interfaces depends on specific applications. For example, speech channel applications build in I2S, AC'97, SP/DIF or other speech codecs. Consumer electronics build in USB, USB OTG, LCD controllers/drivers, and possibly include MMC or SD memory card interfaces. Battery or module controllers build in SMBus, industrial automation applications build in CAN controllers, automobile data communications build in CAN/LIN controllers. LAN applications use network controllers with built-in 10/100Mbits Ethernet MAC/PHY. Figure-1 depicts the 8-bit network microcontroller discussed in this article and it's I/O interface and applications.

Figure-1 Application diagram of High-speed 8-bit network MCU AX110xx

Based on its ease of use, low cost, high bandwidth, stability, security, and compatibility across devices, Ethernet has become the de facto standard of network access. Today, Ethernet has surpassed the use of both SOHO and enterprise networks and expanded into consumer electronics, gradually becoming the most attractive solution for embedded systems to access the network. With the growth of home networking, media sharing, and the gradual prevalence of hi-definition content, the expansion of high bandwidth Ethernet connections to other non-PC devices in the home is also on the rise. In addition, the stability and security of Ethernet makes it an attractive solution for more industrial applications as well.

The proliferation of M2M communications anticipates a rise in Ethernet-capable microcontroller market.

Ethernet can be applied to include a broad range of products, from household appliances, factory and industrial automation, security systems, remote surveillance and management, environmental observation, remote data accumulation, and other applications.

In recent years, the market for microcontrollers has shifted away from consumer electronics to data/telecommunications, resulting in increased demand of high-powered 32-bit microcontrollers for advanced products.

However, as the prices of 8-bit microcontrollers have plateaued and begun to decrease, they have been singled out as the low cost solution for M2M and networking devices. How does one increase the performance of the 8-bit controller? How does one increase the network bandwidth? The question of how to achieve a high level of integration yet maintaining the cost-cutting and miniaturization demands of the market are the key issues facing today's chip designers.

Embedded Ethernet Solution

There are four primary forms of embedded Ethernet designs (see Figure-2). Scenario (I) and Scenario (II) feature a microcontroller without an integrated network controller, and are able to interface with an external network controller through a serial (i.e. USB host) or parallel (i.e. PCI or non-PCI local bus) interface. Scenario (III) and (IV) feature an integrated Ethernet/network controller and microcontroller; in particular, Scenario (IV) involves the highest degree of chip integration. Also, the form factor for the embedded Ethernet system design is getting smaller with each successive generation. This contributes to lowering the total cost of the system (i.e. eliminating PCB components costs, chip material costs) and power consumption.

The right panel in Figure-2 depicts two possible solutions for achieving a small form factor. Top right depicts a small chip package using a non-PCI Ethernet controller with 64-pin connector to achieve a form factor that is about one-fourth the size of a US dime. Bottom right achieves a high level of integration, featuring a network capable microcontroller combining 10/100Mbits Ethernet PHY, MAC, TCP/IP accelerator, and integrated flash, decreasing the physical size of the total package significantly.

Figure-2 Embedded Ethernet Solution

Embedded Ethernet Microcontroller technological advances

An alternative approach to embedded Ethernet development is to observe the trend of integrating networking microcontrollers into the so-called single chip SoC (System on Chip) solution. This progression can be divided into four levels (see Figure-3).

The first level has a relatively low degree of integration, requiring an external flash memory and a non-PCI local bus to connect to the Ethernet controller; this creates a three-chip solution. The second level integrates flash to the microcontroller, but still requires a non-PCI bus to connect to the Ethernet controller; this creates a two-chip solution. The third level combines the microcontroller, flash memory, and the Ethernet MAC layer into a single chip, which then connects to an external PHY; this creates also a two-chip solution. The fourth and ideal outcome combines the microcontroller, flash memory, and both the MAC and PHY layers in a single-chip solution, creating both the chip with the broadest and most comprehensive applications and the smallest form factor in the market.

Figure-3 Embedded Network Evolution

As the semiconductor process technology has advanced, the standards for microcontroller performance in operation timing and machine cycle have also increased. The Machine Cycle is defined as the process of executing a complete instruction within the microcontroller. Shortening the machine cycle speeds up the execution of the instruction. When Intel announced the first generation of the 8051 microcontroller in the 1980's, the standard operation timing was clocked at 12Mhz. Every twelve clock cycles constitutes a machine cycle, or so-called 12T, and the majority of 8051 instructions can be executed in one or two machine cycles. The operation timing of later generations of 8051-compatible microcontrollers has been upgraded to support 16/24/33/40/60MHz clock speeds, and machine cycles have been shortened to 4T/2T as well.

Breakthroughs in chip design technology today have made possible operation timing speeds of up to 100MHz, and the newest generation of the 8051 microcontroller performs at 1T, or Single Cycle instruction. The acceleration of both the operation timing and machine cycle creates a multiplicative effect that creates exponential increases in the total performance with speeds running up to 100MIPS. The microcontroller AX110xx in Fig-1 is an example of such a chip.

The program memory of the original 8051 microcontroller was 4kB; this was gradually upgraded to support up to 8/16/32kB and reached an upper limit of 64kB. Subsequently, using "Bank Select" technology to switch between two sets of 64kB, engineers were able to break the limit and support up to 128kB. The data memory of a conventional 8051 is 128B; this was also gradually upgraded to support up to 256/512/1k/2k/8kB. Upgrading both the program memory and data memory were essential as microcontroller became controllers for memory and resource intensive tasks such as TCP/IP and embedded web server.

In the past, the clients tended to select 32-bit high-end network controller SoC's due to higher density requirement on both program and data memory in their applications. This class of network controller usually supported program and data memory up to 256kB and 32kB respectively, and used an external bus interface to expand memory if necessary. Today, designers have the option to use a low-cost 8-bit network SoC controller. For example, the 8051 compatible network microcontroller 80390 has no limitation on the memory addressing and can support up to 16MB of program memory, large enough to support TCP/IP and embedded Web server applications. The central panel in Figure-1 depicts this design with an 8051/80390 processor core with an embedded flash size of 512kB and 32kB of data memory.

Also, unlike Mask ROM or One Time Programmable (OTP) ROM, whose firmware cannot be modified after shipment, the integration of flash in the network microcontroller makes field upgrades possible through Ethernet and UART. This not only speed up the development process, but makes it easier to patch or upgrade the firmware at a later time.

Increasing Network Throughput through Hardware

In addition to increasing the memory size, clock rate, and reducing machine cycle to improve the performance of the core processor of network controller, there are also hardware mechanisms to relieve the computing power of the processor such as TCP/IP Accelerator and DMA (Direct Memory Access). In the header of IP or TCP, there is a column labeled as header/packet data checksum. This checksum is a mathematical algorithm to ensure the integrity of packet during the transmission on the network. Network controllers normally use a software program to process TCP/IP protocol, but the computation of the checksum requires a great deal of computer power from the processor and may degrade the network performance (see Figure-4). The hardware enabled TCP/IP Accelerator can offload microcontroller computer power and improve the network performance. Also, the process of TCP/IP packets involves mass data transfer between hosts. The hardware-enabled DMA moves the data within memory blocks without going through the read-write cycle from the main processor to further improve the system performance.

Figure-4 Ethernet packet format and the location of header checksum

The applications of Network MCU

The overall performance improvement in the 8-bit network microcontroller makes it an attractive cost down alternative to the 32-bit microcontroller for embedded applications. Applications that once required a 32-bit microcontroller can now be powered by the lower cost, highly integrated and higher capacity of 8-bit network microcontrollers. This includes a number of growing applications including home appliances, factory/building automation, industrial equipments, security systems, remote control/monitoring, and streaming media applications such as POS terminals, vending machines, IP camera, Internet radio, automatic meter reading, environmental monitoring systems, network sensors, networked UPS, Serial to Ethernet adapter, and Ethernet to ZigBee bridge, etc.

The following section details three of the applications mentioned above: network cameras, serial servers, and Ethernet to ZigBee bridges.

Network (IP) Cameras

Network, or IP (Internet Protocol) cameras are rapidly becoming a common gadget in today's digital home or business. A network camera offers users the ability to conduct remote monitoring and management at any given time and place through the simplicity of the Internet and a web browser. The two most common formats used today for image compression are Motion JPEG and MPEG4. Due to the bandwidth limitation in broadband services today, there is trade-off between the image resolution and transmission speed. The advantage of Motion JPEG solution is its low cost; due to its low compression rate, it requires less power from the hardware. The budget network camera is an ideal give-away promotion item for both broadband and mobile phone service providers. Typical uses of network cameras include surveillance, remote monitoring of children, the in/outflow of shop entrances, no-man factory/warehouse, etc. With an 8-bit high performance network microcontroller and Motion JPEG codec, one can design and build a low cost network camera with video and audio as well as Pan/Tilt functionalities (see Figure-5).

Figure-5 Dual mode Ethernet/USB network camera:AX11015+STV0684

Serial Servers

As information technology advances, more and more automation equipments in the factory are connected into the Internet. The main purpose of serial server is to transmit data between serial ports (RS-232/485/422) and Ethernet, allowing users to access, monitor and manage the equipment or device with these serial interfaces through the Internet remotely. The applications cover both industrial and commercial uses including medical equipments, electronics billboards, PLC, Machine-Human Interface and barcode scanners, etc. With an 8-bit high performance network microcontroller, one can design and build a serial server with an appropriate serial interface as depicted in Figure-6. This can shrink the size of both the PCB and the system, making the design process easier and cheaper as well as replacing multi-chips solutions in the market today.

Figure-6 Single chip solution for RS-232 to Ethernet application

Ethernet to ZigBee bridges

Many vendors are promoting Wireless Sensor Network products with the strong emergence and industry acceptance of the 802.15.4/ZigBee standard. The applications of this new wireless network technology include both industrial and home automation such as digital home appliance control, security monitoring, goods control, home-care, etc. As ZigBee proliferates in the factory and the digital home, controlling and managing these equipments and devices embedded with ZigBee becomes an issue. With a 8-bit high performance network microcontroller with SPI interface and 802.15.4/ZigBee controller, one can design and build a compact Ethernet to ZigBee bridge as depicted in Figure-7. Once this compact Ethernet to Zigbee bridge is connected to a Home Gateway, the Home Gateway will become a dual role device (Home gateway and Zigbee Gateway), allowing the user to control and manage any ZigBee device through remote Internet access.

Figure-7 Ethernet to ZigBee Bridge:AX11001+UZ2400

The 8-bit network MCU's Future in the Market

The technological advances in 8-bit microcontrollers have allowed these highly integrated and low cost solutions to reach into the market once accessible only by 32-bit controllers. With the growth of Machine-to-Machine communication and a rising number of Internet-ready devices, the market for high capacity network controllers are likely to experience heightened demand in the foreseeable future. The role of the 8-bit network microcontroller will likely be an important one in powering future embedded systems in the marketplace.

About the Authors

Jason Wang is director of Marketing at ASIX Electronics. He can be reached at: jasonw@asix.com.tw

此文章源自《 Embedded.com》网站: