Embedded Exam 1 Flash Cards

What is a microcontroller?

A complete (or mostly complete) solution for embedded applications: one or more CPU cores, RAM, ROM, I/O ports, and timers

What is the Raspberry Pi 4 made up of?

Broadcom BCM2711 SoC processor, additional SDRAM, USB controller, Ethernet controller, WiFi controller, peripheral I/O

Why is the RP4 called an embedded subsystem?

It is a microcontroller plus additional support controllers

What important function is the RP4 missing?

An RTC (real time clock)

RP4 SoC key functions

4x ARM 64-bit Cortex-A72 cores, SDRAM controller, video decode (H.265/H.264), OpenGL ES 3.0 graphics, L1 and L2 cache, HDMI display interface, many I/O (DMA, USB, UART, SDIO, PWM, PCM/I2S, SPI, GPIO)

ESP32 compared to the RP4

Tensilica 32-bit core (1 or 2), onboard RTC, onboard ROM and RAM, onboard crypto engines, onboard RF logic, many I/O interfaces

Three things a CPU core does

1) Moves data 2) Manipulates data 3) Is controlled by instructions

Internal organization of a computer

CPU, memory, and I/O (input and output)

Memory capacity

The number of bits a memory can store (e.g. 128 Kbits, 256 Mbits)

Memory organization

How the locations are organized, e.g. 128 x 4 = 128 locations of 4 bits each

Memory access time

How long it takes to get data from memory; depends on location and interface speed

Access time: L1 cache vs L2 cache vs external RAM

L1: CPU clock speed. L2: multiple CPU clocks. External RAM: dozens to 100+ CPU clocks

RP4 clock speed

1.5 GHz (older), 1.8 GHz for the latest version. Study guide answer: 1.5 GHz

Three buses connecting CPU to memory and I/O

Address bus, data bus, control bus (read/write)

Memory control signals CS, WE, OE

Chip Select, Write Enable, Output Enable (read)

Port-mapped I/O

An IO/MEM select bit gives I/O its own independent address space

Memory-mapped I/O

I/O devices share the same address space as memory; a logic circuit (address decoder) enables CS for the right address range

Is the RP4 memory mapped or port mapped?

Everything is memory mapped, including I/O

Steps to design an address decoder

1) Write the address range in binary 2) Separate the fixed part of the address 3) Design a logic circuit whose output activates when the fixed bits appear

Address decoder for addresses 0 to 15 (8-bit address)

Fixed bits a7-a4 = 0000; CS (active low) asserted when a7..a4 are all 0

Address decoder for 300H to 3FFH

Fixed bits a11-a8 = 0011; decode those four bits to drive CS

Harvard architecture

Separate memory and buses for code and data

Von Neumann architecture

One shared memory and bus for code and data

Harvard pros and cons

More pins, power and area; more complex; faster with parallel/simultaneous instruction and data access; special purpose; used for registers and L1 cache

Von Neumann pros and cons

Fewer pins and less power; simpler; slower due to bottleneck; general purpose; used for L2, L3, main memory and storage

Program Counter (PC)

Special register inside the processor that stores the address of the next instruction

Stored program computer operation

Step 1: fetch the instruction at the address in the PC and increment the PC. Step 2: execute the instruction, repeat

Most powerful thing a computer can do

Repeat a set of operations until some condition is met

What is branching?

Setting the PC to a specified instruction location rather than the next instruction

Levels of program code

High-level language (C++, Rust, Python) -> assembly language (text form of instructions) -> hardware representation (binary opcodes)

What language does a processor actually run?

Binary machine code (opcodes)

What does a compiler do?

Converts high-level language to assembly/machine code for a specific processor architecture

RISC

Reduced Instruction Set Computing: operations as simple as possible, usually one instruction per clock. ARM, MIPS

CISC

Complex Instruction Set Computing: operations as powerful as possible, 1 to many clocks per instruction. x86

Hybrid (GPU) architecture

Vector or SIMD (single instruction multiple data); data-flow based. SSE, CUDA

Direct addressing

The operand is an address (a mailbox); the data is what is inside that location

Immediate addressing

The operand is a constant number, e.g. #3

Opcode

The fixed-size code in an instruction that identifies which operation to perform

Why do compilers/assemblers take multiple passes?

To resolve addresses of labels and variables that are not known on the first pass

RP4 pins you cannot control

3.3 V, 5 V and GND

Are the 5V and 3.3V PWR pins inputs?

No. They power external devices and are not for input voltage

Default state of GPIO 0-8

High, with a pullup resistor

Default state of GPIO 9-27

Low, with a pulldown resistor

Default internal pullup/pulldown value

About 50 kohm (weak)

Pullup value on GPIO 2 and 3

1.8 kohm, to meet the I2C spec

Typical external pullup/pulldown value

About 10 kohm; a stronger external resistor overrides the internal one

50 kohm pullup in parallel with 10 kohm

8.33 kohm

Max current for one GPIO

16 mA

Max current per pin when using multiple GPIOs

3 mA per I/O

Total GPIO current limit

Stay below about 51 mA (17 outputs at 3 mA each)

What happens if you overdrive the RP4 GPIOs?

The RP4 may reboot due to voltage drop

Ways to debounce a switch

Analog RC filter (0.1 uF capacitor), buffering, digital logic (cross-coupled NANDs / SR latch), software

Output good practices: datasheet items to check

Current, voltage, and time constraints (setup, hold, period)

BJT LED driver: purpose of RB and RC

RB limits current drawn from the Pi; RC limits current through the LED

Python call to force a pulldown on an input

GPIO.setup(gpio_number, GPIO.IN, pull_up_down=GPIO.PUD_DOWN)

Extra capabilities some GPIOs have

PWM, I2C, serial ports, etc.

Primary function of the assembler

Convert readable assembly into binary machine code (.s files to .o files)

Why are assemblers multi-pass?

To resolve loop/label dependencies, variable name dependencies, register renaming and reordering

Linker

Combines multiple object files, resolves symbol references and relocates data sections

objcopy

Translates object output from one format to another

objdump

A disassembler that gives readable output of the final result

What does the compiler (GCC) take in?

A mix of assembly (.s), C files (.c), object files (.o) and library files to generate a combined executable with the linker

GCC

GNU C Compiler; common compiler for Linux and bare-metal systems

Bare metal

No operating system; the processor runs only the code the user wrote

EABI (Lecture 3)

Embedded Application Binary Interface; often used for bare-metal programming

ELF

Executable (and Linkable) Format; often used for assembly programs run in a Linux environment

What is the makefile?

A file read by the Make utility that describes how to build the program

What does Make do?

Controls generation of executables, uses the makefile as input, automatically uses the files it needs based on what changed, helps with file management, automates the build

Primary tools for assembly on the RP4 under Linux

GCC and GDB

GDB

GNU Project Debugger

Flag needed to debug with GDB

-g, e.g. as -g prog.s -o prog.o

GDB: set a breakpoint

b

GDB: run the program

r

GDB: show registers

i r

GDB: step

s (or s 100 for 100 steps)

Who coined "debugging"?

Admiral Grace Hopper, Harvard, 1940s, after a moth was found in the computer

Examples of IDEs

VSCode, Geany, Eclipse

What are flowcharts and pseudocode for?

Giving structure to a program: validate the algorithm, find key functions/macros, break a big job into chunks, divide work among teams, communicate with stakeholders

Flowchart characteristics

Symbolic/pictorial, too much detail is bad, hard to modify when large, fixed set of symbols, more straightforward to understand, best for smaller code jobs

Pseudocode characteristics

Natural text, flexible level of detail, easy to modify, open-world structure, can be more complex to follow, best for larger code jobs

Common flowchart symbols

Start/End, Input/Output, Process, Connector, Direction of flow, Decision/condition, Delay

When are flowcharts useful?

At the start of a project to think through architecture, and at the end to document the code

Advantage of pseudocode with real code

It can morph along with the code and become the comments

One hex digit equals how many bits?

4 bits (a nibble)

Convert from a base to base 10

Multiply each digit by base^position and add

Convert from base 10 to another base

Divide repeatedly by the base; remainders are the digits, least significant first

How many values can n bits encode?

2^n

Powers of 2 from 2^0 to 2^10

1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024

2^10, 2^20, 2^30 approximately

1 thousand, 1 million, 1 billion

2^32

4,294,967,296 (about 4.3 billion)

Bit

Can be 1 or 0

Nibble

4 bits

Byte

8 bits

Half word

16 bits

Word

32 bits

Kilobit (Kb) vs kilobyte (KB)

Kb = 1024 bits; KB = 1024 bytes = 8192 bits

Megabyte (MB)

2^20 bytes = 1,048,576 bytes

Binary: 1 + 1

10 (2)

Hex: 0x0A + 0x08

0x12 (18)

Prefixes for binary and hex

Binary: % or 0b. Hex: $ or 0x