Basic Interrupts and Io
Basic Interrupts and Io
TABLE OF CONTENTS
Lets' get physical ....................................................................................................................................2 Basic interrupts.......................................................................................................................................2 What's in a name?..................................................................................................................................3 Structure of an interrupt-driven program on the AVR ..........................................................................4 Real code part 1 .....................................................................................................................................7 The timer overflow interrupt.................................................................................................................11 Real code part 2 ...................................................................................................................................13
The program will be written in assembly, because: assembly is very "machine-near" and provides a very educative approach to what goes on inside the processor during our program high-level languages and different compilers all have different notations and routines for doing the same thing. Learning a compiler and the respective C-style (e.g.) is a story of itself. The code for this project is something we found among leftovers from O'Guru Sean Ellis; which we brutally and without due respect ripped apart. Shame on us.
Basic interrupts
An interrupt is a flow control mechanism that is implemented on most controllers, among them the AVR. In an MCU application interacting with the outside world, many things are happening at the same time, i.e. not in a synchronized manner, that are to be handled by the microcontroller. Examples: a switch pressed by the user, a data read on the UART (serial port), a sample taken by the ADC, or a timer calling to say that "time is up!". All these events neeeds to be handled by the MCU.
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 2
Instead of polling each instance round-Robin style to ask whether they are in need of a service, we can have them call out themselves when they need attention. This is called "interrupts", since the peripheral device (e.g. a switch pressed) interrupts the main program execution. The processor then takes time out of the normal program execution to examine the source of the interrupt and take the necessary action. Afterwards, normal program execution is resumed. An interrupt service in other words is just like a subroutine; except that it is not anticipated by the processor to occur at a particular time, since there are no explicitly placed calls to it in the program.
What's in a name?
When reading this article you will from time to time get the feeling that you are confronting a term possibly denoting an actual physical entity or an entity in some sense relevant to the current activity; namely playing around or building serious applications with the AVR...: INT0, INT1, GIMSK, PORTB, PB7 etc... You are sure to come across such names in any assembly code, Atmel appnote, AVRfreaks Design Note or any posting in the AVRforum. One might think these are just common names used by individuals accustomed to the jargon, but we will try to use them consciously - in the sense that these names actually denote actual memory locations in the AVR you will be programming. The mapping of these name to actual memory locations is in the part's def file (*def.inc). Have a look at the included def file for the 8515 each time you come across an unknown term: 8515def.inc
Example snippet; only a few lines are shown
;***** I/O Register Definitions .equ SREG =$3f .equ SPH =$3e .equ SPL =$3d (~6kB) File included in zip .equ GIMSK =$3b .. .. .. .. .. ..
When including this file in the assembly program file, all I/O register names and I/O register bit names appearing in the data book will be known to the assembler and can be used in the program. Note that some high-level language compilers may use proprietary terms other than these. But they will have files similar to this def file, defining the memory space of the AVRs. As previously stated; this is another story. Another document that will prove very useful to anyone working with the AVR, is this document: 8515 datasheet
The datasheet. The datasheet is the ultimate reference for any AVR microcontroller. It even includes an instruction set summary; look up every instruction you don't know when you come across it!
(~2MB) In this article, we will be using the 8515. Keep this .pdf close for File included in zip reference.
The datasheet. The datasheet is the ultimate reference for any AVR microcontroller. It even includes an instruction set summary; look up every instruction you don't know when you come across it!
Now you know where to look when anything unknown pops up. Let's move on! (~2MB) In this article, we will be using the 8515. Keep this .pdf close for
Structure of an interrupt-driven program on the AVR You may even want to print it, but think twice. It is long .
Take a deep breath. This is the heaviest part. We are going to write an "interrupt-driven" program where the main loop simply does nothing but wait for interrupts to occur. What interrupts?
The interrupts are handled in turn, and a return to the main program is performed at the end of each interrupt service (that's what I call it; "service"). This is a rather wide topic with many pitfalls. But we need somewhere to start and will mainly discuss aspects concerning elements of our little example application. The main important thing that constitutes such elements in a program is: 1. 2. 3. 4. Setting the interrupt vector jump locations: .org Setting the correct interrupt mask to enable desired interrupts: GIMSK Make necessary settings in control registers: MCUCR Globally enable all interrupts: SREG
The lowest part of the AVR program memory, starting at address $0000, is sometimes referred to as the "Program memory vector table", and the actual program should start beyond this space. The vector table is reserved for storing interrupt vectors; i.e. locations to jump to when this or that interrupt is calling. This means that each interrupt has a reserved memory location, and when a particular interrupt comes in, the MCU looks in this location to find the address where code that handles this interrupt resides. 8515 Vector table
Program memory Vector address $0000 Reset Comment Start address of Reset handler is stored here Start address of code to handle external INT0 is stored here Start address of code to handle external INT1 is stored here ...
$0001
INT0
$0002 etc...
INT1 ...
In assembly code, the .org directive is used to set vector jump locations. This
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 4
assembler directive (or "command", if you like) tells the assembler to set the location counter to an absolute value. It is not part of the AVR instruction set, it is just a command that the assembler needs to make sure the program code is mapped correctly when making a binary for the AVR. Example: Sample Code ; Interrupt service vectors ; Handles reset and external interrupt vectors INT0 and INT1 .org $0000 rjmp Reset .org INT0addr rjmp IntV0 .org INT1addr rjmp IntV1 ; Reset vector (when the MCU is reset) ; INT0 vector (ext. interrupt from pin PD2) ; INT1 vector (ext. interrupt from pin PD3)
; - Reset vector - (THIS LINE IS A COMMENT) Reset: ldi TEMP,low(RAMEND) ; Set initial stack ptr location at ram end out SPL,TEMP ldi TEMP, high(RAMEND) out SPH, TEMP ... ... Note that labels are used instead of absolute numbers to designate addresses in assembly code - The assembler stitches it all together in the end. All we need to do is tell the assembler where to jump when e.g. the reset vector is calling, by using the name of the code block meant for handling resets. A label denotes a block of code, or function if you like; which is not terminated with a "}", an .endfunc or anything like that. The only thing that ends a code block definition, is it being released by another block name, followed by a colon (":"). This also implies, unlike with functions in e.g. C, that all blocks are run by the processor consecutively, unless the flow is broken up by un/conditional jumps, returns, interrupts etc. In assembly, the whole file is the main() function, and the flow control is more like Basic... Please also note the first lines of the reset handler. This is where the stack is set up. The stack is used to hold return addresses in the main program code when a sub- or interrupt routine is run; i.e. when a "digression" from the main program is made. For any interrupt service or subroutine to return to the main program properly; the stack must be placed outside their vector space. The SP is namely initialized with the value $0000, which is the same location as the reset vector. This goes for any program, especially such as this, where we are involving several interrupt vectors besides the reset vector. For AVRs with more than 256 bytes SRAM (i.e. none of the Tinys, nor 2343 and 4433), the Stack Pointer register is two bytes wide and divided into SPL and SPH (low and high bytes).
The GIMSK register is used to enable and disable individual external interrupts. GIMSK
Bit Read/write Init. value 7 INT1 R/W 0 6 INT0 R/W 0 5 R 0 4 R 0 3 R 0 2 R 0 1 R 0
We are going to use the external interrupts INT0 and INT1 for the switches on the STK500. These interrupts are enabled by setting INT0 and INT1 in GIMSK; i.e. bits 6 and 7.
MCUCR
Bit
We will be using bits 0,1,2 and 3 in this register to control the interrupt from
INT0 and INT1. These bits control how to sense the external interrupts; either by level, falling edge on pin, or rising edge of pin: ISCx1 0 0 1 1 ISCx0 0 1 0 1 Description Low level on INTx pin generates interrupt Reserved Falling edge on INTx pin generates interrupt Rising edge on INTx pin generates interrupt
We will use the rising edge of the switches on the STK500 to trig the interrupt; so the 8515 must be programmed to trig external interrupts on rising edges of each pin PD2 and PD3. Hence; all the ISCx bits must, for our program, be set to "1". You can see on the diagram to the right how pushing the switch will close the lower branch and pull the pin low. Hence; releasing the switch causes a rising edge when the branch is re-opened and the pin is pulled high.
In addition to setting up the interrupts individually, the SREG (Status Register) bit 7 must also be set to globally enable all (i.e. any) interrupts. SREG
Bit 7 I 6 T 5 H 4 S 3 V 2 N 1 Z 0 C
Status register The bits in SREG indicate the current state of the processor.
Init. value
All these bits are cleared on reset and can be read or written by a program. Bit7 (I) is the one we are currently interested in; as setting this bit enables all interrupts. Vice versa, resetting it disables all interrupts. In AVR code, we have an instruction of its own to set this flag; sei: sei ; lots and lots of initialisation, and then... ; this instruction enables all interrupts. ;...and off we go!
1. 2. 3. 4.
A switch is pressed -> ext. INT0 generated The vector for INT0 is found Code at the according location is run, and jumps to a common subroutine The common subroutine returns to the main loop by reti instruction
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 7
This is what our code will do. Nothing more. Besides initialization, the short routine for handling the other switch (generating INT1) and a few directives for the assembler, that's it all. 8515def.inc Just to make sure I'm still not kidding you; have a look in the 8515def.inc file and search for "INT0addr" and "INT1addr". Lo and behold; they are real addresses. Reset is placed at $0000.
(~6kB)
File included in zip
OK, here is the entire program code, with some excessive comments removed (these are still left in the available file). Look up any unknown instruction for full understanding while you read through it. You can click each code block label to jump to their respective comments below. INTs_1.asm Source for first part of program ;----------------------------------------------------------------; Name: int0.asm ; Title: Simple AVR Interrupt Verification Program ;----------------------------------------------------------------.include "8515def.inc" ; Interrupt service vectors .org $0000 rjmp Reset .org INT0addr rjmp IntV0 pin D2) .org INT1addr rjmp IntV1 pin D3) ; Reset vector ; INT0 vector (ext. interrupt from ; INT1 vector (ext. interrupt from
(~3kB)
File included in zip
;----------------------------------------------------------------; ; Register defines for main loop .def .def .def TIME=r16 TEMP=r17 BEEP=r18
;----------------------------------------------------------------; ; Reset vector - just sets up interrupts and service routines and ; then loops forever. Reset: ldi ram end out ldi out SPL,TEMP TEMP, high(RAMEND) SPH, TEMP TEMP,low(RAMEND) ; Set stackptr to
; Set TEMP to $FF to... ; ...set data direction to ; ...all lights off! ; ...all high for pullup on
ldi TEMP,(1<<DDD6) ; bit D6 only configured as output, out DDRD,TEMP ; ...output for piezo buzzer on pin D6 ; set up int0 and int1 ldi and 1 set out ldi rising edge only out ldi sei we go! loop: rjmp terminates loop ; Infinite loop - never TEMP,(1<<INT0)+(1<<INT1) ; int masks 0 GIMSK,TEMP TEMP,$0f MCUCR,TEMP TIME,$00 ; interrupt t0 and t1 on ; Start from 0 ; enable interrupts and off
;----------------------------------------------------------------; ; Int0 vector - decrease count IntV0: dec rjmp new count TIME Int01 ; jump to common code to display
;----------------------------------------------------------------; ; Int1 vector - increase count IntV1: inc new count Int01: mov com out reti r0,TIME r0 PORTB,r0 ; display on LEDs TIME ; drop to common code to display
OK, lets go through the code step by step, though at a pace. It may be easier if
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 9
you have the source printed out next to you while reading the following comments: The first lines includes the define file for the 8515; thus making all register and I/O names known to the assembler. What happens next is the Interrupt vector table is defined. At $0000, the reset vector is set up. This is where the 8515 wakes up in the morning - everything is supposed to start from here. Also, the INT0 and INT1 vectors are set up, and their handling routines named IntV0 and IntV1, respectively. Look up their labels down the code, and you can see where they are declared. Following this, registers r16, r17 and r18 have labels put onto them. This is a way to make variables in assembly - only we also get to decide where they are placed in memory. Where? In registers r16, r17 and r18... hence; they are all one byte wide.
Py-haa! The reset label contains all initialization code; this block is run at startup. The first 4 lines sets up the stack pointer, as mentioned earlier. Note how the ldi(load immediate) instruction is used to hold any value temporarily before writing to the actual location by out. low() and high() are macros returning the immediate values of their arguments, which are memory locations defined in the .def file. The next six lines sets the Data Direction Registers of ports PORTB (used for LEDs) and PORTD (switches). Please check the datasheet under "I/O Ports" for functional descriptions of these registers. Now, notice this line: ldi TEMP,(1<<DDD6)
This line of code simply (!) means: "Load TEMP register with a byte value of 1 shifted DDD6 places leftwards". Ok. Then what is DDD6? From the .def file, we find that this value is 6, and it is meant to point to the 6th bit of the PORTD Data Direction Register. The value loaded into TEMP and then into DDRB, becomes 01000000 in binary. Hence, the bit in this position in the DDRB register is set. So what happens? That pin (pin PD6) is to be used for a special twist in the next stage of the program, so that particular pin is set as an output; the others will be inputs For now, just notice the notation. You can probably imagine what happens if you combine such notation in an addition? Well, this is what happens next, when the GIMSK register is loaded, and then the MCUCR. Please refer to the previous section or the datasheet for a description of these registers and why they are set this way. Only thing remaining in the reset block now, is to call our friend the sei instruction for enabling the interrupts we have just set up.
The loop label simply contains nothing but a call to itself. It's an equivalent of writing while(1); in C. After reset is run, the program pointer falls through to the loop block and it will run forever only interrupted by - interrupts.
This label comtains the handling code for the INT0 interrupt. Whenever that interrupt calls, this code will be run. It will simply decrement the TIME register. Then it just jumps to a common block called...:
This block consists of common code that displays the value of TIME (r16) on the
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 10
LEDs connected to PORTB. Note that the value is inverted by 1's complement (com instruction) before written to PORTB, since a low value means no light and vice versa. This block then performs a return to wherever it was called from through the reti instruction - which was from the loop label.
Timer overflow 0
The 8515 has two timer/counters; one 8 bits wide and one 16 bits wide. This means that they are capable of counting from any value you set, until they reach their limit which is determined by the number of bits available (256 or 65535, respectively). Then they will issue an interrupt, if you have set it up to do so. Upon overflow; the Timer/Counter just keeps counting "around" the range... so if you have set the timer to start from some special value and want it to start from there again; you will have to reset it to that value. What we need to do in the code, is to add three little blocks of code more. These are (could you guess them?): 1. Another interrupt vector, for the TimerOverflow 0 interrupt: OVF0addr 2. Initialization code for timer/counter 0: TIMSK, TCCR0,TCNT0 3. The interrupt handling subroutine.
OVF0addr
This is the name set in the 8515def.inc file for the location where this interrupt vector should reside (check it, I may be pulling your leg). We add these two lines of code to the vector block:
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 11
; -- new interrupt vector .org OVF0addr rjmp TimerV0 ; T/C0 overflow vector You are very able to read this now, and realize that it is just like the previous .org's in this program. Let's move on!
TIMSK, TCCR0,TCNT0
Together, these 3 registers are all we need consider to have timing interrupts in an application on the AVR. TCCR0 controls the operation of Timer/counter 0. The count is incremented for every clock signal at the input of the timer. But the clock input can be selected, and prescaled by N. We'll just consider the 3 lowest bits of this register: TCCR0
Bit 7 Read/write R Init. value 0 6 R 0 5 R 0 4 R 0 3 R 0 2 1 0
Timer/Counter0 register Note that bits 7-3 are reserved, and always read as zero
TIMSK; the Timer/Counter Interrupt Mask register is simply a "mask" register for enabling/disabling interrupts just like you have already seen with the GIMSK register: TIMSK Timer/Counter Interrupt Mask register
7 6 5 4 R 0 3 TICIE1 R/W 0 2 R 0 1 TOIE0 R/W 0 0 R 0
Bit
Note that bits 4,2 and 0 are reserved, and always read as zero
...and again; the only thing you really need to know for this little tutorial; is the
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 12
position of one special little bit: this one is called "Timer/Counter0 Overflow Interrupt enable", abbreviated "TOIE0" and found in bit postition 1 of this register. To enable our Timer interrupt; set this bit (to "1"). TCNT0 is the actual "Timer/Counter" register. This is where the timing and counting is done, in accordance with the settings in TCCR0. This is simply a register for storing a counter value; there are no special bits in it. It is entirely readable/writable; so you can load it with any desired starting value for your counting if you like. Note that it does not reset itself automatically, even if an interrupt is issued. This is already becoming old news to you now, since it's just more or less another instance of registers controlling similar functions that you have already heard about regarding the external interrupts...
The picture shows how to connect the speaker. We have chosen pin PD6 for no particular reason.
Huh? ?
Why is that the formula for the base frequency? The Timer/counter counts one up from some value every clock cycle until it overflows. Then we reset it, to repeat the drill. Let's say the timer can only count to 1 before overflow. Flipping the pin every time, will give us one cycle of a square-wave like waveform every 2 flips, right? (up a while, then down a while, repeat...). Hence, the base frequency would be: CLK/2 Now; the Timer/Counter register is 8 bits wide, and can count from any value it is set (TIME) to 255. The formula becomes: CLK/2*(256-TIME) Besides; we have a prescaler which, when set to N, makes the Timer count just every Nth cycle... CLK/2*N*(256-TIME)
These are the three snippets of code to insert. Please consult the complete source code (INTs_2.asm, available below) to where the snippets are inserted: INTs_2.asm Source snippets for second part of program
;---------- CODE SNIPPET #1 - OVF0addr vector -----------------; inserted BELOW the existing vector defs (~3kB) ;--------------------------------------------------------------File included in .org OVF0addr zip rjmp TimerV0 ; T/C0 overflow vector . . ;---------- CODE SNIPPET #2 - Initializing TIMSK,TCCR0,TCNT0 ---; inserted in the Reset: label, right before the 'sei' call ;--------------------------------------------------------------ldi TIME,$80 ; Start from 128. NB! out TCNT0,TIME ; set Timer/counter also. ldi enable 0 out ldi out TEMP,(1<<TOIE0) ; timer overflow interrupt TIMSK,TEMP TEMP,$02 TCCR0,TEMP ; clock prescaler = clk/8
. . ;---------- CODE SNIPPET #3 - handling the Timer overflow int. -; new subroutine label, inserted at the end of the file ;--------------------------------------------------------------TimerV0: out TCNT0,TIME ; reset time com ori out reti BEEP BEEP,$BF PORTD,BEEP ; bit 6 only ; important!
CODE SNIPPET #1
This part simply declares the Timer Overflow vector address. NB! It is imperative that you declare the interrupt vectors in the same order as they appear in the .inc file. If not, you will get segment overlap errors. That is; the .org directives must be placed according to the memory locations they represent, in ascending order. So an easy rule to use is to just enter them in the same order as in the .inc file!
CODE SNIPPET #2
First, we set the TIME register to a higher value (0x80 = 128 decimal) for
AN INTRODUCTION TO INTERRUPTS AND I/O, AVR EIVIND - AVRFREAKS.NET 14
starters, and load it into the Timer/Counter register. It's just a more fitting start value if you run the 8515 on a low CLK freq. Then the relevant interrupt enable bit is set in the TIMSK register, and the prescaling bits in the Timer Control register are set to Prescale=8. This way, if the 8515 runs @ 1.23MHz; the speaker will buzz with a base frequency equal to 1.23E6/2*8*127 = 605.3 Hz
CODE SNIPPET #3
The important issues in handling this interrupt is: Resetting the Timer/Counter - it won't do that itself! Flipping the beep pin Returning to the program
Resetting Timer/Counter is obviously done by loading the value of TIME (r16) into TCNT0, and returning from the interrupt routine is done by issuing a reti instruction. Flipping the beep pin (PD6) is a little curious, however: This is done by inverting every bit in BEEP (r18) with the com instruction, and then OR'ing it with this value 0xbf = 10111111 b (note the 6th position is '0'). Follow the sequence below:
BEEP after com: 'OR' with: Result: after com: 'OR' with: Result: etc... 00000000 11111111 10111111 11111111 00000000 10111111 10111111 ...
As you may see; whichever value is in BEEP, the 6th bit of it will flip every time... So, the pin will toggle up and down, and the speaker beeps this little song: "...10101010101010...". Haha. Now, these were the very basic basics of interrupts and I/O. Feel free to experiment with what you have learnt in this article; use other prescaler settings, try other flanks of external interrupt triggering, write programs that use switches to make flow control decisions, whatever... Good luck!