A simple project

The Project

This project will use the pulse-width modulator (PWM) to ramp an LED on and off every two seconds. An AT90S2313 processor will be used as the controller. The circuit for this demonstration is shown in the schematic diagram. If you have a development kit, you should be able to use it, rather than build the circuit, for this project.

Note:

Meanwhile, the AT90S2313 became obsolete. Either use its successor, the (pin-compatible) ATtiny2313 for the project, or perhaps the ATmega8 or one of its successors (ATmega48/88/168) which have become quite popular since the original demo project had been established. For all these more modern devices, it is no longer necessary to use an external crystal for clocking as they ship with the internal 1 MHz oscillator enabled, so C1, C2, and Q1 can be omitted. Normally, for this experiment, the external circuitry on /RESET (R1, C3) can be omitted as well, leaving only the AVR, the LED, the bypass capacitor C4, and perhaps R2. For the ATmega8/48/88/168, use PB1 (pin 15 at the DIP-28 package) to connect the LED to. Additionally, this demo has been ported to many different other AVRs. The location of the respective OC pin varies between different AVRs, and it is mandated by the AVR hardware.

Figure 1. Schematic of circuit for demo project

The source code is given in demo.c. For the sake of this example, create a file called demo.c containing this source code. Some of the more important parts of the code are:

Note [1]:: As the AVR microcontroller series has been developed during the past years, new features have been added over time. Even though the basic concepts of the timer/counter1 are still the same as they used to be back in early 2001 when this simple demo was written initially, the names of registers and bits have been changed slightly to reflect the new features. Also, the port and pin mapping of the output compare match 1A (or 1 for older devices) pin which is used to control the LED varies between different AVRs. The file iocompat.h tries to abstract between all this differences using some preprocessor #ifdef statements, so the actual program itself can operate on a common set of symbolic names. The macros defined by that file are:

OCR the name of the OCR register used to control the PWM (usually either OCR1 or OCR1A)
DDROC the name of the DDR (data direction register) for the OC output
OC1 the pin number of the OC1[A] output within its port
TIMER1_TOP the TOP value of the timer used for the PWM (1023 for 10-bit PWMs, 255 for devices that can only handle an 8-bit PWM)
TIMER1_PWM_INIT the initialization bits to be set into control register 1A in order to setup 10-bit (or 8-bit) phase and frequency correct PWM mode
TIMER1_CLOCKSOURCE the clock bits to set in the respective control register to start the PWM timer; usually the timer runs at full CPU clock for 10-bit PWMs, while it runs on a prescaled clock for 8-bit PWMs

Note [2]:: ISR() is a macro that marks the function as an interrupt routine. In this case, the function will get called when timer 1 overflows. Setting up interrupts is explained in greater detail in <avr/interrupt.h>: Interrupts.

Note [3]:: The PWM is being used in 10-bit mode, so we need a 16-bit variable to remember the current value.

Note [4]:: This section determines the new value of the PWM.

Note [5]:: Here's where the newly computed value is loaded into the PWM register. Since we are in an interrupt routine, it is safe to use a 16-bit assignment to the register. Outside of an interrupt, the assignment should only be performed with interrupts disabled if there's a chance that an interrupt routine could also access this register (or another register that uses TEMP), see the appropriate FAQ entry.

Note [6]:: This routine gets called after a reset. It initializes the PWM and enables interrupts.

Note [7]:: The main loop of the program does nothing -- all the work is done by the interrupt routine! The sleep_mode() puts the processor on sleep until the next interrupt, to conserve power. Of course, that probably won't be noticable as we are still driving a LED, it is merely mentioned here to demonstrate the basic principle.

Note [8]:: Early AVR devices saturate their outputs at rather low currents when sourcing current, so the LED can be connected directly, the resulting current through the LED will be about 15 mA. For modern parts (at least for the ATmega 128), however Atmel has drastically increased the IO source capability, so when operating at 5 V Vcc, R2 is needed. Its value should be about 150 Ohms. When operating the circuit at 3 V, it can still be omitted though.

The Source Code

Compiling and Linking

This first thing that needs to be done is compile the source. When compiling, the compiler needs to know the processor type so the -mmcu option is specified. The -Os option will tell the compiler to optimize the code for efficient space usage (at the possible expense of code execution speed). The -g is used to embed debug info. The debug info is useful for disassemblies and doesn't end up in the .hex files, so I usually specify it. Finally, the -c tells the compiler to compile and stop -- don't link. This demo is small enough that we could compile and link in one step. However, real-world projects will have several modules and will typically need to break up the building of the project into several compiles and one link.

The compilation will create a demo.o file. Next we link it into a binary called demo.elf.

It is important to specify the MCU type when linking. The compiler uses the -mmcu option to choose start-up files and run-time libraries that get linked together. If this option isn't specified, the compiler defaults to the 8515 processor environment, which is most certainly what you didn't want.

Examining the Object File

Now we have a binary file. Can we do anything useful with it (besides put it into the processor?) The GNU Binutils suite is made up of many useful tools for manipulating object files that get generated. One tool is avr-objdump, which takes information from the object file and displays it in many useful ways. Typing the command by itself will cause it to list out its options.

For instance, to get a feel of the application's size, the -h option can be used. The output of this option shows how much space is used in each of the sections (the .stab and .stabstr sections hold the debugging information and won't make it into the ROM file).

An even more useful option is -S. This option disassembles the binary file and intersperses the source code in the output! This method is much better, in my opinion, than using the -S with the compiler because this listing includes routines from the libraries and the vector table contents. Also, all the "fix-ups" have been satisfied. In other words, the listing generated by this option reflects the actual code that the processor will run.

Here's the output as saved in the demo.lst file:

Linker Map Files

avr-objdump is very useful, but sometimes it's necessary to see information about the link that can only be generated by the linker. A map file contains this information. A map file is useful for monitoring the sizes of your code and data. It also shows where modules are loaded and which modules were loaded from libraries. It is yet another view of your application. To get a map file, I usually add -Wl,-Map,demo.map to my link command. Relink the application using the following command to generate demo.map (a portion of which is shown below).

Some points of interest in the demo.map file are:

The .text segment (where program instructions are stored) starts at location 0x0.

The last address in the .text segment is location 0x114 ( denoted by _etext ), so the instructions use up 276 bytes of FLASH.

The .data segment (where initialized static variables are stored) starts at location 0x60, which is the first address after the register bank on an ATmega8 processor.

The next available address in the .data segment is also location 0x60, so the application has no initialized data.

The .bss segment (where uninitialized data is stored) starts at location 0x60.

The next available address in the .bss segment is location 0x63, so the application uses 3 bytes of uninitialized data.

The .eeprom segment (where EEPROM variables are stored) starts at location 0x0.

The next available address in the .eeprom segment is also location 0x0, so there aren't any EEPROM variables.

Generating Intel Hex Files

We have a binary of the application, but how do we get it into the processor? Most (if not all) programmers will not accept a GNU executable as an input file, so we need to do a little more processing. The next step is to extract portions of the binary and save the information into .hex files. The GNU utility that does this is called avr-objcopy.

The ROM contents can be pulled from our project's binary and put into the file demo.hex using the following command:

The resulting demo.hex file contains:

The -j option indicates that we want the information from the .text and .data segment extracted. If we specify the EEPROM segment, we can generate a .hex file that can be used to program the EEPROM:

There is no demo_eeprom.hex file written, as that file would be empty.

Starting with version 2.17 of the GNU binutils, the avr-objcopy command that used to generate the empty EEPROM files now aborts because of the empty input section .eeprom, so these empty files are not generated. It also signals an error to the Makefile which will be caught there, and makes it print a message about the empty file not being generated.

Letting Make Build the Project

Rather than type these commands over and over, they can all be placed in a make file. To build the demo project using make, save the following in a file called Makefile.