I currently working to recreate a quad-copter controller.
I am working on getting data from my gyro sensor, and to do that, I'm using
an ISR with an interuption.
My problem is, when I call my function "gyro.getX" on the main program, it work.
But, when I call this function from my ISR, it doesn't work.
I thing that I find the reson of the bug, the function I'm using is provided by the "Adafruit_LSM9DS0" library (from ST), and it used a "timestamp".
I thing that the current time from my ISR is different that the current time from my Main program, but i don't know how to fi it.
Here a shortcut of my program:
void loop(){
/*main prog*/
}
/*
*Reserve interrupt routine service (ISR) by Arduino
*/
ISR(TIMER2_OVF_vect)
{
TCNT2 = 256 - 250; // 250 x 16 µS = 4 ms
if (varCompteur++ > 25)// 25 * 4 ms = 100 ms (half-period)
{
varCompteur = 0;
SensorGet(pX, pY);//Feed gyro circular buffers
}
}
void SensorGet(float * pRollX, float * pPitchY)
{
lsm.getEvent(&accel, &mag, &gyro, &temp);
GiroX_Feed(pX, gyro.gyro.x);
GiroY_Feed(pPitchY, gyro.gyro.y);
}
bool Adafruit_LSM9DS0::getEvent(sensors_event_t *accelEvent,
sensors_event_t *magEvent,
sensors_event_t *gyroEvent,
sensors_event_t *tempEvent)
{
/* Grab new sensor reading and timestamp. */
read();
uint32_t timestamp = millis();
/* Update appropriate sensor events. */
if (accelEvent) getAccelEvent(accelEvent, timestamp);
if (magEvent) getMagEvent(magEvent, timestamp);
if (gyroEvent) getGyroEvent(gyroEvent, timestamp);
if (tempEvent) getTempEvent(tempEvent, timestamp);
return true;
}
The problem isn't the time. The problem is likely that your sensor uses the I2C and it is disabled during an interrupt routine, or it's some other communication protocol that relies on interrupts to function and is therefore disabled during your ISR.
You are really abusing the interrupt. This is not the kind of thing interrupts are for. Interrupt should be super fast, no time for communications there. So the real question is why do you think you need an interrupt for this?
Related
I am using a Remote Control from FlySky. For my robotics project, I want to read PWM from the receiver on an Arduino. I came across 2 options:
pulseIn() arduino function
ISR(PCINTx_vect) (interrupt)
I cant use the first option of pulseIn() because I want my robot to continue with the operation if receiver signal are not coming (Tx not available etc.) So I used ISR.
Most reliable source : Mr. Brookings channel on YouTube.
Here is what I did (Only the required part for 1 axis):
// [R] where R is defined as 0 => [R] == [0]
volatile long CH[4]; //4 pwms to read so array of 4
float IN[3]={0,0,0}; // throttle is directly written
unsigned long timer[4],curr_time;
byte last[4];
void setup(){
PCICR |= (1 << PCIE0);
PCMSK0 |= (1 << PCINT0);
PCMSK0 |= (1 << PCINT1);
PCMSK0 |= (1 << PCINT2);
PCMSK0 |= (1 << PCINT3);
/* There is some more code here */
Serial.begin(115200);
}
void loop(){
/* There is some more code here */
IN[R] = ((CH[ROLL] - (1500 + R_TRIM))/11.0); // eg.: (1200 - (1500 + 8))/11.0 = -28 (interpreted as setpoint of -28° by the robot)
Serial.println(IN[R]);
}
ISR(PCINT0_vect){
curr_time = micros();
//channel 1 roll
if(PINB & B00000001){
if(last[ROLL] == 0){
last[ROLL] = 1;
timer[ROLL] = curr_time;
}
}
else if(last[ROLL] == 1){
last[ROLL] = 0;
CH[ROLL] = ((curr_time - timer[ROLL]));
}
}
I can read the PWM actually, but the robot keeps showing random twitches in its control at a given set point. I managed to trace the reason and found out that the PWM is insanely ridden by noise. Its not stable like it should be - steady. I have a MATLAB plot I used for analysis:
Signal (IN[R]):
Close up (when Tx stick was in the middle w/o movement) :
There are such spikes coming which is adding up to the control signal eventually making my robot to twitch. I tried some filtering techniques like 'moving average' and '1st and 2nd order exponential filters'. Also checked if it was due to power supplied to it - tried putting a capacitor or an iron core to the power lines but in vain. I can figure out how to remove them as their some constrains :
platform is Arduino Uno (slower in heavy computation)
Control loop shall not go below 100Hz (Currently its at 108Hz exponential filters on 4 axes took it to
~85Hz)
I would appreciate some guidance!
There's no way of telling from this if the input is noisy, or if your code is reading the PWM wrong, of if something else is going on, like external noise on the line, the Arduino's clock jitter, or other interrupts taking time. Also note that micros() on an Arduino Uno only has a resolution of 4µs, not 1µs.
You should check the input for jitter and noise, and try fast code that isn't influenced by other interrupts.
A fairly simple and fast way of getting the PWM pulse width is something like this, preferably without using anything else that uses interrupts:
volatile int pwmPulseWidth = 0;
volatile unsigned long int previousTime = 0;
void setup() {
attachInterrupt(0, rising, RISING);
}
void loop() {
// pwmPulseWidth is available here.
}
void rising() {
attachInterrupt(0, falling, FALLING);
previousTime = micros();
}
void falling() {
attachInterrupt(0, rising, RISING);
pwmPulseWidth = micros() - previousTime;
}
Untested, but it should give you an idea. This will return the width of the PWM pulse.
There are other ways of doing this, of course, like using a timer in capture mode.
Knowing the PWM frequency and the width of the PWM pulse is enough to reconstruct the PWM signal, should you want to.
I try original O-Scope project (PigOScope) without touchscreen based on STM32F103C8T6 Bluepill board, but got some problem:
I used newest rogerclarkmelbourne/Arduino_STM32 Core and downloaded
pingumacpenguin/STM32-O-Scope sketch. I compiled and uploaded it to the device via UART from 0x08000000 address. Then I started the device. The grid and coordinate lines were displayed on the screen. Also on the screen were displayed inscriptions below 0.0 uS/Sample etc... But any noise or pulse signal from PB1 on my Probe. Why the chart is not drawn?
Also I tried to log my steps in Usart in DMA activation code function:
void takeSamples()
{
// This loop uses dual interleaved mode to get the best performance out of
the ADCs
Serial.println("Init DMA");
dma_init(DMA1);
dma_attach_interrupt(DMA1, DMA_CH1, DMA1_CH1_Event);
Serial.println("Enable DMA for ADC");
adc_dma_enable(ADC1);
dma_setup_transfer(DMA1, DMA_CH1, &ADC1->regs->DR, DMA_SIZE_32BITS,
dataPoints32, DMA_SIZE_32BITS, (DMA_MINC_MODE |
DMA_TRNS_CMPLT));// Receive buffer
Serial.println("Set DMA transfer");
Serial.println(maxSamples / 2);
dma_set_num_transfers(DMA1, DMA_CH1, maxSamples / 2);
dma1_ch1_Active = 1;
Serial.println("Enable the channel and start the transfer");
dma_enable(DMA1, DMA_CH1); // Enable the channel and start the transfer.
samplingTime = micros();
Serial.println(samplingTime);
while (dma1_ch1_Active); // SOME BUG OR WHAT.... PROGRAM STOP HERE!!!
samplingTime = (micros() - samplingTime);
Serial.println("Disable DMA");
dma_disable(DMA1, DMA_CH1); //End of trasfer, disable DMA and Continuous
mode.
}
Event handler for stop interrupt
static void DMA1_CH1_Event()
{
dma1_ch1_Active = 0;
}
Volatile flag to stop routine
volatile static bool dma1_ch1_Active = 0;
Program keep crushing on while loop i think... And program does't work beyond takeSamples() function.
Why program does't exit the loop?
I am using an Arduino Uno, connected to a USB shield, a RFID shield(adafruit PN532), an LCD, EEPROM(24AA256) and a RTC module(DS1307). I will not post my code here because it is too large and it is separated in multiple files.
In my program, I realize that if my programs enters a certain functions, after entering function after function, if I use a delay() at the end of the function I am currently in, the arduino resets. An example of what I mean is below.
void a() { b(); }
void b() { c(); }
void c() { d(); }
void d()
{
lcd_string("Testing", 0x80);
delay(2000); <---- Arduino resets at the delay here
}
At first, I thought it was because my dynamic memory was at 80%, and when I compiled, they said the Arduino might have some stability issues. So I modified my code such that my dynamic memory is now 57%. Problem still exist.
I thought maybe the delay() function has some overflow or something, so I tried replacing the delay with the following code.
unsigned long timing;
timing = millis();
timing += 2000;
while(millis() < timing);
The Arduino still resets.
Next, I thought maybe because my arduino is connected to my PC, some serial pin might have been causing the reset, so I used an external Power to power up the arduino and disconnected the USB. The arduino still resets.
Next, I thought maybe Timer1 might have been crashing with the delay() function, although the delay function uses Timer0 so I disabled my Timer1 . The arduino still resets.
Is there any other possibilities that I am missing out? My program storage space is at 69% which I believe shouldn't be an issue.
Edit
Here is my code for Timer1 ISR
ISR(TIMER1_OVF_vect)
{
TCCR1A = 0;
TCCR1B = 0;
TCNT1 = 0;
OCR1A = 34286;// = (16*10^6) / (1*1024) - 1 (must be <65536)
TCCR1B |= (1 << CS12);
// enable timer compare interrupt
TIMSK1 |= (1 << TOIE1);
triggered = 1;
}
Any other interrupt of flags used are in the library header files.
I am using the following external libraries
USB Host shield library 2.0
Adafruit PN532 master
A little sample to come close to RAM corruption ...
#define MEM_PER_LEVEL 50
#define TRY_TO_SURVIVE 10
void KillMe(int level) {
byte dummy[MEM_PER_LEVEL];
for ( byte i = 0; i < MEM_PER_LEVEL; i++)
dummy[i]= i;
Serial.println(level);
delay(1000); // not sure why this would hurt more than others
if (level < TRY_TO_SURVIVE) KillMe(level+1);
for ( byte i = 0; i < MEM_PER_LEVEL; i++) {
if (dummy[i] != i) {
Serial.println(F("corruption happened"));
while(1) {} // HALT
}
}
if (level == 0)
Serial.println(F("survived"));
}
void setup() {
Serial.begin(9600);
KillMe(0);
}
void loop() { }
I had the same problem - wherever I put a delay in my setup function the Arduino would restart.
For me, the problem was an instance of SoftwareSerial with invalid pin numbers.
SoftwareSerial mySerial(30, 31);
Anyone else landing on this question should check their pin numbers are appropriate for the board they're targeting. Not sure why the crash only happens if a delay is called, would be interested if anyone has insight into this!
Im trying to figure out how to dim a led over time(time is defined by the user lets call it rampUp). Im using arduino with the adafruit breakout PWM-Servo-Driver (http://www.adafruit.com/products/815) with the library : https://github.com/adafruit/Adafruit-PWM-Servo-Driver-Library
this breakout has 4095 steps (0-4095) so now my problem:
I want to be able to take a variable ( int minutes) and send that to a method that will dim a LED from 0 to 4095 in a equal light intensity increase for the period minutes. I want the increase to be incremented by 1 each time it increases.
So how do I write the method without using delay() ?
void dimLights(int rampUp){
// some code to regulate the increase of the third value in setPWM one step at a time over a period of rampUp(int in minutes)
led.setPWM(0,0,4095);
}
the reason for not wanting to use delay() is because it will pause/stop the rest of the program.
I actually implemented something close recently:
void loop() {
/* ... */
if (rampUp != 0)
led.setPWM(0,0,rampUp);
}
void led_update() {
// here you can prescale even more by using something like
// if (++global_led_idx == limit) {
// global_led_idx = 0;
++rampUp;
}
void start() {
TCCR1B |= _BV(WGM12);
// set up timer with prescaler = FCPU/1024 = 16MHz/1024 ⇒ 60ms
TCCR1B |= _BV(CS12) | _BV(CS10);
TCCR1B &= ~_BV(CS11);
// initialize counter
OCR1A = 0x0000;
// enable global interrupts
sei();
// enable overflow interrupt
TIMSK1 |= _BV(OCIE1A);
}
void TAGByKO_LED::stop() {
TIMSK1 &= ~_BV(OCIE1A);
rampUp = 0;
}
ISR(TIMER1_COMPA_vect, ISR_NOBLOCK) {
led_update();
}
then you can call start() to start the timer, or stop() to stop it. You can read more documentation about the registers I've been using here and the ISR statement. Beware that it's one of the most tricky things of AVRs to really understand, and even then you'll always need to have a cheatsheet or the datasheet close by.
You may also use #sr-richie's solution with timers, but do only setup variables in the dimLights() function, and call led.setPWM() only within the loop().
I have one Arduino with 4 Pots. The other Arduino receives these 4 values via i2c and prints them on a Display. The problem is that I don't know how to send these 4 values that the Slave is able to know which value belongs to which Pot.
Slave Code:
#include <Wire.h>
#include <LiquidCrystal.h>
LiquidCrystal lcd(12, 11, 5, 4, 3, 2);
void setup()
{
Wire.begin(5);
Wire.onReceive(receiveEvent);
Serial.begin(9600);
lcd.begin(16,2);
}
void loop()
{
}
void receiveEvent(int)
{
while(Wire.available())
{
//How to create this part? How does the Slave know which value belongs to which pot?
}
}
Master Code:
#include <Wire.h>
void setup()
{
Serial.begin(9600);
Wire.begin();
delay(2000);
}
void loop()
{
int sensor1 = analogRead(A1);
Wire.beginTransmission(5);
Wire.write(sensor1);
Serial.print(sensor1);
Wire.endTransmission();
delay(100);
int sensor2 = analogRead(A2);
Wire.beginTransmission(5);
Wire.write(sensor2);
Serial.print(sensor2);
Wire.endTransmission();
delay(500);
}
Ahh what we have here is a basic question on how to design I2C communication. Unfortunately Examples for I2C master and slave included in Arduino IDE are IMO too limited to provide clear guidance on this matter.
First of all in your examples the master and slaves roles are exchanged and should be switched. Slave should read values from analog inputs and master should request them. Why? Because it's master which should decide when to request values and properly decode the request. Slave should provide proper answer to a given request eliminating the problem of data interpretation.
I2C communication is based on requestFunction-(wait)-requestResponse sequence controlled by the master.
Plese refer to the range finder example on arduino page. In a nutshell:
First: master requests a function to measure distance:
// step 3: instruct sensor to return a particular echo reading
Wire.beginTransmission(112); // transmit to device #112
Wire.write(byte(0x02)); // sets register pointer to echo #1 register (0x02)
Wire.endTransmission(); // stop transmitting
(sometimes slaves need some time e.g. 10 - 50 ms to process requests but in the example I'm refering to master doesn't delay read)
Second: master requests response:
// step 4: request reading from sensor
Wire.requestFrom(112, 2); // request 2 bytes from slave device #112
Third: master tries to read and analyze response.
You should design reliable I2C communication in a similar way.
Here is how I do it; you can follow my pattern and get extensible slave implementation which will support one function: read analog inputs but can be easily extended by adding additional function codes and required processing implementation to the slave main loop
Initial remarks
some kind of a simple protocol is needed to control slave - e.g. it should support requesting functions. Supporting functions requests is not absolutely needed in such simmple scenario as reading four analog inputs but what I'm describing is a more general pattern you may use in other projects.
Slave should not perform any additional actions (like reading inputs) on request response as I2C communication may break (due to delays) and you will get partial responses etc. This is very important requirement which affect the slave design.
response (and also request if needed) can contain CRC as if master waits not long enough it may get empty response. If nobody else is going to use your code such countermeasures are not needed and will not be described here. Other important thing is Wire library buffer limitation which is 32 bytes and implementing CRC checksum without modifying the buffer length limits the available data length by two bytes (if crc16 is used).
slave:
#include <WSWire.h> // look on the web for an improved wire library which improves reliability by performing re-init on lockups
// >> put this into a header file you include at the beginning for better clarity
enum {
I2C_CMD_GET_ANALOGS = 1
};
enum {
I2C_MSG_ARGS_MAX = 32,
I2C_RESP_LEN_MAX = 32
};
#define I2C_ADDR 0
#define TWI_FREQ_SETTING 400000L // 400KHz for I2C
#define CPU_FREQ 16000000L // 16MHz
extern const byte supportedI2Ccmd[] = {
1
};
// << put this into a header file you include at the beginning for better clarity
int argsCnt = 0; // how many arguments were passed with given command
int requestedCmd = 0; // which command was requested (if any)
byte i2cArgs[I2C_MSG_ARGS_MAX]; // array to store args received from master
int i2cArgsLen = 0; // how many args passed by master to given command
uint8_t i2cResponse[I2C_RESP_LEN_MAX]; // array to store response
int i2cResponseLen = 0; // response length
void setup()
{
// >> starting i2c
TWBR = ((CPU_FREQ / TWI_FREQ_SETTING) - 16) / 2;
Wire.begin(I2C_ADDR); // join i2c bus
Wire.onRequest(requestEvent); // register event
Wire.onReceive(receiveEvent);
// << starting i2c
}
void loop()
{
if(requestedCmd == I2C_CMD_GET_ANALOGS){
// read inputs and save to response array; example (not tested) below
i2cResponseLen = 0;
// analog readings should be averaged and not read one-by-one to reduce noise which is not done in this example
i2cResponseLen++;
i2cResponse[i2cResponseLen -1] = analogRead(A0);
i2cResponseLen++;
i2cResponse[i2cResponseLen -1] = analogRead(A1);
i2cResponseLen++;
i2cResponse[i2cResponseLen -1] = analogRead(A2);
i2cResponseLen++;
i2cResponse[i2cResponseLen -1] = analogRead(A3);
// now slave is ready to send back four bytes each holding analog reading from a specific analog input; you can improve robustness of the protocol by including e.g. crc16 at the end or instead of returning just 4 bytes return 8 where odd bytes indicate analog input indexes and even bytes their values; change master implementation accordingly
requestedCmd = 0; // set requestd cmd to 0 disabling processing in next loop
}
else if (requestedCmd != 0){
// log the requested function is unsupported (e.g. by writing to serial port or soft serial
requestedCmd = 0; // set requestd cmd to 0 disabling processing in next loop
}
}
// function that executes whenever data is requested by master
// this function is registered as an event, see setup()
void requestEvent(){
Wire.write(i2cResponse, i2cResponseLen);
}
// function that executes when master sends data (begin-end transmission)
// this function is registered as an event, see setup()
void receiveEvent(int howMany)
{
//digitalWrite(13,HIGH);
int cmdRcvd = -1;
int argIndex = -1;
argsCnt = 0;
if (Wire.available()){
cmdRcvd = Wire.read(); // receive first byte - command assumed
while(Wire.available()){ // receive rest of tramsmission from master assuming arguments to the command
if (argIndex < I2C_MSG_ARGS_MAX){
argIndex++;
i2cArgs[argIndex] = Wire.read();
}
else{
; // implement logging error: "too many arguments"
}
argsCnt = argIndex+1;
}
}
else{
// implement logging error: "empty request"
return;
}
// validating command is supported by slave
int fcnt = -1;
for (int i = 0; i < sizeof(supportedI2Ccmd); i++) {
if (supportedI2Ccmd[i] == cmdRcvd) {
fcnt = i;
}
}
if (fcnt<0){
// implement logging error: "command not supported"
return;
}
requestedCmd = cmdRcvd;
// now main loop code should pick up a command to execute and prepare required response when master waits before requesting response
}
master:
#include <WSWire.h>
#define I2C_REQ_DELAY_MS 2 // used for IO reads - from node's memory (fast)
#define I2C_REQ_LONG_DELAY_MS 5 //used for configuration etc.
#define TWI_FREQ_SETTING 400000L
#define CPU_FREQ 16000000L
enum {
I2C_CMD_GET_ANALOGS = 1
};
int i2cSlaveAddr = 0;
void setup(){
// joining i2c as a master
TWBR = ((CPU_FREQ / TWI_FREQ_SETTING) - 16) / 2;
Wire.begin();
}
void loop(){
//requesting analogs read:
Wire.beginTransmission(i2cSlaveAddr);
Wire.write((uint8_t)I2C_CMD_GET_ANALOGS);
Wire.endTransmission();
delay(I2C_REQ_DELAY_MS);
// master knows slave should return 4 bytes to the I2C_CMD_GET_ANALOGS command
int respVals[4];
Wire.requestFrom(i2cSlaveAddr, 4);
uint8_t respIoIndex = 0;
if(Wire.available())
for (byte r = 0; r < 4; r++)
if(Wire.available()){
respVals[respIoIndex] = (uint8_t)Wire.read();
respIoIndex++;
}
else{
// log or handle error: "missing read"; if you are not going to do so use r index instead of respIoIndex and delete respoIoIndex from this for loop
break;
}
// now the respVals array should contain analog values for each analog input in the same order as defined in slave (respVals[0] - A0, respVals[1] - A1 ...)
}
I hope my example will help. It's based on code working for weeks making 40 reads a second from multiple slaves however I have not compiled it to test the function you require.
Please use WSWire library as the Wire (at least as for Arduino 1.0.3) may occasionally freeze your master if for some reason slave will not respond to request.
EDIT: The WSWire lib requires external pull-up resistors for I2C unless you modify the source and enable internal pull-ups like Wire does.
EDIT: instead of creating i2c slave implementation you may try the EasyTransfer library. I haven't tried it but it may be easier to use it if sending four bytes is everything you need.
EDIT[12.2017]: There is a new player on the block - PJON - a library suited for easy multi-master communication ideal to exchange pot values (and much more). It's been around for some time but gained a substantial development speed in recent months. I'm partially involved in its development and switched all field-level and local buses I've used so far (I2C, MODBUS RTU) to PJON over single wire, hardware serial or RF.
Check out GitHub-I2CBus, I've done the exact same thing. Hope it can help