01. introduction to embedded systems
- 1. Introduction to Embedded System Design Babak Kia Adjunct Professor Boston University College of Engineering Email: bkia -at- bu.edu ENG SC757 - Advanced Microprocessor Design Microcontrollers A Microcontroller is essentially a small and self- sufficient computer on a chip, used to control devices It has all the memory and I/O it needs on board Is not expandable – no external bus interface Characteristics of a Microcontroller • Low cost, on the order of $1 • Low speed, on the order of 10 KHz – 20 MHz • Low Power, extremely low power in sleep mode • Small architecture, usually an 8-bit architecture • Small memory size, but usually enough for the type of application it is intended for. Onboard Flash. • Limited I/O, but again, enough for the type of application intended for Microprocessors A Microprocessor is fundamentally a collection of on/off switches laid out over silicon in order to perform computations Characteristics of a Microprocessor • High cost, anywhere between $20 - $200 or more! • High speed, on the order of 100 MHz – 4 GHz • High Power consumption, lots of heat • Large architecture, 32-bit, and recently 64-bit architecture • Large memory size, onboard flash and cache, with an external bus interface for greater memory usage • Lots of I/O and peripherals, though Microprocessors tend to be short on General purpose I/O 1
- 2. Harvard Architecture Harvard Architecture refers to a memory structure where the processor is connected to two different memory banks via two sets of buses This is to provide the processor with two distinct data paths, one for instruction and one for data Through this scheme, the CPU can read both an instruction and data from the respective memory banks at the same time This inherent independence increases the throughput of the machine by enabling it to always prefetch the next instruction The cost of such a system is complexity in hardware Commonly used in DSPs Von-Neumann Machine A Von-Neumann Machine, in contrast to the Harvard Architecture provides one data path (bus) for both instruction and data As a result, the CPU can either be fetching an instruction from memory, or read/writing data to it Other than less complexity of hardware, it allows for using a single, sequential memory. Today’s processing speeds vastly outpace memory access times, and we employ a very fast but small amount of memory (cache) local to the processor Modern processors employ a Harvard Architecture to read from two instruction and data caches, when at the same time using a Von-Neumann Architecture to access external memory Little vs. Big Endian Although numbers are always displayed in the same way, they are not stored in the same way in memory Big-Endian machines store the most significant byte of data in the lowest memory address Little-Endian machines on the other hand, store the least significant byte of data in the lowest memory address A Big-Endian machine A Little-Endian machine stores 0x12345678 as: stores 0x12345678 as: ADD+0: 0x12 ADD+0: 0x78 ADD+1: 0x34 ADD+1: 0x56 ADD+2: 0x56 ADD+2: 0x34 ADD+3: 0x78 ADD+3: 0x12 2
- 3. Little vs. Big Endian The Intel family of Microprocessors and processors from Digital Equipment Corporation use Little-Endian mode Whereas Architectures from Sun, IBM, and Motorola are Big-Endian Architectures such as PowerPC, MIPS, and Intel’s IA- 64 are Bi-Endian, supporting either mode Unfortunately both methods are in prevalent use today, and neither method is superior to the other Interfacing between Big and Little Endian machines can be a headache, but this is generally only a concern for TCP/IP stack and other interface developers Program Counter (PC) The Program Counter is a 16 or 32 bit register which contains the address of the next instruction to be executed The PC automatically increments to the next sequential memory location every time an instruction is fetched Branch, jump, and interrupt operations load the Program Counter with an address other than the next sequential location During reset, the PC is loaded from a pre-defined memory location to signify the starting address of the code Reset Vector The significance of the reset vector is that it points the processor to the memory address which contains the firmware’s first instruction Without the Reset Vector, the processor would not know where to begin execution Upon reset, the processor loads the Program Counter (PC) with the reset vector value from a pre-defined memory location On CPU08 architecture, this is at location $FFFE:$FFFF A common mistake which occurs during the debug phase – when reset vector is not necessary – the developer takes it for granted and doesn’t program into the final image. As a result, the processor doesn’t start up on the final product. 3
- 4. Stack Pointer (SP) The Stack Pointer (SP), much like the reset vector, is required at boot time for many processors Some processors, in particular the 8-bit microcontrollers automatically provide the stack pointer by resetting it to a predefined value On a higher end processor, the stack pointer is usually read from a non-volatile memory location, much like the reset vector For example on a ColdFire microprocessor, the first sixteen bytes of memory location must be programmed as follows: 0x00000000: Reset Vector 0x00000008: Stack Pointer COP Watchdog Timer The Computer Operating Properly (COP) module is a component of modern processors which provides a mechanism to help software recover from runaway code The COP, also known as the Watchdog Timer, is a free- running counter that generates a reset if it runs up to a pre-defined value and overflows In order to prevent a watchdog reset, the user code must clear the COP counter periodically. COP can be disabled through register settings, and even though this is not good practice for final firmware release, it is a prudent strategy through the course of debug The Infinite Loop Embedded Systems, unlike a PC, never “exit” an application They idle through an Infinite Loop waiting for an event to happen in the form of an interrupt, or a pre-scheduled task In order to save power, some processors enter special sleep or wait modes instead of idling through an Infinite Loop, but they will come out of this mode upon either a timer or an External Interrupt 4
- 5. Interrupts Interrupts are mostly hardware mechanisms which tell the program an event has occurred They happen at any time, and are therefore asynchronous to program flow They require special handling by the processor, and are ultimately handled by a corresponding Interrupt Service Routine (ISR) Need to be handled quickly. Take too much time servicing an interrupt, and you may miss another interrupt. Designing an Embedded System Proposal Definition Technology Selection Budgeting (Time, Human, Financial) Material and Development tool purchase Schematic Capture & PCB board design Firmware Development & Debug Hardware Manufacturing Testing: In-Situ, Environmental Certification: CE Firmware Release Documentation Ongoing Support System Design Cycle Writing code conjures up visions of sleepless nights and stacked up boxes of pizza And if not done correctly, that is exactly what the outcome will be! The purpose of the design cycle is to remind and guide the developer to step within a framework proven to keep you on track and on budget 5
- 6. System Design Cycle There are numerous design cycle methodologies, of which the following are most popular • The Spaghetti Model • The Waterfall Model • Top-down versus Bottom-up • Spiral Model • GANTT charts The Spaghetti Model Not in this course, thank you. The Waterfall Model Waterfall is a software development model in which development is seen flowing steadily through the phases of • Requirement Analysis • Design • Implementation • Testing • Integration • Maintenance Advantages are good progress tracking due to clear milestones Disadvantages are its inflexibility, by making it difficult to respond to changing customer needs / market conditions 6
- 7. Top-down versus Bottom-up The Top-Down Model analyses the overall functionality of a system, without going into details • Each successive iteration of this process then designs individual pieces of the system in greater detail The Bottom-Up Model in contrast defines the individual pieces of the system in great detail • These individual components are then interfaced together to form a larger system The Spiral Model Modern software design practices such as the Spiral Model employ both top-down and bottom-up techniques Widely used in the industry today For a GUI application, for example, the Spiral Model would contend that • You first start off with a rough-sketch of user interface (simple buttons & icons) • Make the underlying application work • Only then start adding features and in a final stage spruce up the buttons & icons The Spiral Model In this course, we won’t focus on the aesthetic qualities of the application - only the underlying technology 7
- 8. GANTT Chart GANTT Chart is simply a type of bar chart which shows the interrelationships of how projects and schedules progress over time Design Metrics Metrics to consider in designing an Embedded System • Unit Cost: Can be a combination of cost to manufacture hardware + licensing fees • NRE Costs: Non Recurring Engineering costs • Size: The physical dimensions of the system • Power Consumption: Battery, power supply, wattage, current consumption, etc. • Performance: The throughput of the system, its response time, and computation power • Safety, fault-tolerance, field-upgradeability, ruggedness, maintenance, ease of use, ease of installation, etc. etc. Schematic Capture 8
- 9. PCB Layout PCB Board Portions of this power point presentation may have been taken from relevant users and technical manuals. Original content Copyright © 2005 – Babak Kia 9