0

I am now hopelessly stuck and need help please.

I am building an automated observatory and now at the stage of trying to control the observatory dome rotation. The dome is rotated using a step motor and it works fine. The rotation of the dome also turns a rotary encoder to give an independent verification that the dome is moving. Although the RE is a quadrature type I am simply counting RE ticks from both channels as I don't need direction information.

What I am trying to do now is to count how many motor driving pulses occur between each pair of RE ticks. The reasoning behind this is to monitor the dome movements and detect jamming. The actual motor pulses will be used for all dome position calculations.

To achieve this, I have built a small board with a ULN2803A inverting buffer and an Arduino Nano. The Nano uses Pin Change interrupts on pins D11 and D12 to count the RE ticks. Pin D7 will also be used to monitor a Home Position sensor although this doesn't exist at present. The motor pulses are sent via the buffer to Pin D2 (INT0) and the motor direction is sent via the buffer to pin D3 (INT1). The external interrupts are then used to count the motor pulses, up or down depending on the motor direction.

For information, there is another board present, a Velleman VM110 USB breakout board, that also receives the rotary encoder ticks. This gives me a simple way to check that my board gives the same answer when I rotate the rotary encoder shaft by hand.

OK - sorry this is so long winded but now it gets weird.

When I rotate the RE without the motor running everything is fine. I get four times as many pulses as the Velleman board as I am using Pin Changes, triggering on each rise and fall, not on full cycles of the encoder. This does happen as predicted and everything is nice and stable.

BUT - when I start the motor running, I get hundreds of RE ticks every time I move the encoder and sometimes I get increasing RE ticks without touching the encoder at all.

The obvious answer is some form of RFI but I have done everything I can to eliminate this. The motor driver and the Nano board that powers it are in a shielded enclosure, the cable to the motor is a shielded cable and I have 10uH chokes on each of the four motor leads. Finally, I have fitted a filter to the incoming power supply to the motor box to minimise any RFI going back down the power lines.

The use of the ULN2803A buffer was my latest attempt to make this work. Previously the signals went direct to the Nano pins. With the buffer, I have used 20k pull-ups on the input side and 10k pull-ups on the outputs. This was a direct copy of the Velleman board input circuit that I knew worked without problems.

I have looked at the motor pulses at the input pin to the Nano on my scope and they are nice looking, sharp edged pulses of duration 70 uS and a frequency of 497 Hz. Not bad as I set the pulse rate using the Accelstepper library to 500Hz.

I now suspect that this is a software problem. This wouldn't surprise me as I am very new to all this, just trying to learn enough at each stage to do what I need. I have attached the code - this is a stripped down version without a lot of I2C stuff that is not relevant to my problems.

Again, for information. I have tried using the attachInterrupt() function and by directly setting the relevant registers - no difference!

Anyway, I really really hope someone can help me sort this out.

Regards, Hugh

/*          
ABoard3  
ROTATION MONITORING AND POSITION
Version AB3_stripped
*****************************PURPOSE*****************************************
This sketch is used on an Arduino Nano to count the motor pulses from ABoard1
and the Rotary Encoder ticks. The motor pulse count between encoder ticks is
used to detect dome jamming.
****************************INPUTS*******************************************
PIN CHANGE INTERRUPTS
**********ROTARY ENCODER INPUT*********
The rotary encoder is a Bourns ENA1J-B28 L00064L 1139M MEX 64 cycle per turn
optical encoder. This is connected to ABoard3 Pins D11 and D12. These pins
connect to Channel A and Channel B respectively. Pin change interrupts are used
to receive the rotary encoder output.
(The output pulses from the rotary encoder is also sent to the Velleman board
for use by LesveDomeNet for finding dome position)
*********HOME POSITION INPUT*********
The home position sensor is an Omron EESX671 Optical Sensor.
The sensor output is connected to ABoard3 Pin D7.
Pin change interrupt PCINT21 records activation/deactivation of the sensor.
EXTERNAL INTERRUPTS
*********MOTOR PULSE INPUT***********
The pulses sent to the G251X stepper driver are also sent to Aboard3 Pin D2.
This pin is the External Interrupt pin, vector INT0
*********MOTOR DIRECTION INPUT********
Motor direction is input to ABoard3 from ABoard2. ABoard2 pin, pnVmInRotMotor
(AB2:A0{54}) is connected to Velleman pins DI4 and DO2 and also to AB3:D3.
AB3:D3 is an External Interrupt pin, vector INT1.
*/

#include <Wire.h>                       
#include "streaming.h"                        
#include "I2C_AnythingHEG.h"   
#include <EEPROMex.h>                    

//CONSTANTS                       
//PIN ASSIGNMENTS For ABOARD3
const uint8_t pnMotorPulse = 2;      //Port PD2, INT0, ISR(INT0_vect){}
const uint8_t pnMotorDirection = 3;  //Port PD3, INT1, ISR(INT1_vect){}
const uint8_t pnDomeAtHome = 7;      //Port PD7, PCINT23,ISR(PCINT2_vect){}
const uint8_t pnREChanA = 11;        //Port PB3, PCINT3, ISR(PCINT0_vect){}
const uint8_t pnREChanB = 12;        //Port PB4, PCINT4, ISR(PCINT0_vect){}

//*****************EXPERIMENTAL STUFF FOR PULSE COUNTING*******************************                  
uint16_t volatile myTest = 0;
int32_t volatile ratioCount = 0L;
int32_t volatile totalCount = 0L;
int32_t volatile tickCount = 0L;
uint8_t volatile halftickCount = 0;
int32_t volatile addPulse = 0L; 

void setup() {
  //**********************************SERIAL FOR DEBUG ONLY************************
  Serial.begin(115200);
  //*************************INPUT PIN MODE SETUP**********************************
  //NOTE Set all UNUSED PCINT0:5 pins (D8, D9, D10, D11) to OUTPUT.
  //and set the value to LOW
  //The actual pins used to receive interrupts have external 10k pull-ups.
  //This is to reduce susceptibility to interference causing false triggering.
  pinMode(pnMotorPulse, INPUT); //D2
  pinMode(pnMotorDirection, INPUT); //D3
  pinMode(pnDomeAtHome, INPUT); //D7
  pinMode(pnREChanA, INPUT); //D11
  pinMode(pnREChanB, INPUT); //D12
  pinMode(4, OUTPUT); //D4
  digitalWrite(4, LOW);
  pinMode(8, OUTPUT); //D8
  digitalWrite(8, LOW);
  pinMode(9, OUTPUT); //D9
  digitalWrite(9, LOW);
  pinMode(10, OUTPUT); //D10
  digitalWrite(10, LOW);
  //******************SET UP AddPulse According to Motor Direction******************
  //This is needed to make sure the pulse counting starts in the correct direction
  //as the direction is only checked in the ISR after a change has occurred.
  //If Motor Direction is AntiClockwise, change to Subtract a pulse
  if( digitalRead(pnMotorDirection)) {
    addPulse = 1L;
  } else {
    addPulse = -1L;
  }
  //**************************SET UP PIN CHANGE INTERRUPTS**************************
  //Set the Pin Change Interrupt Masks
  //Clear all bits to start with
  PCMSK0 &= 0b00000000; //Clear all bits
  PCMSK1 &= 0b00000000; //Clear all bits
  PCMSK2 &= 0b00000000; //Clear all bits
  //Mask for PCINTs 0 through 7 is PCMSK0 (Rotary Encoder Pulses)
  //Need to allow interrupts on PCINT3 and PCINT4, so set bits 3 and 4
  //PCINT3 is Nano  pin D11 and PCINT4 is Nano pin D12
  //Use a compound bitwise OR to set the bits
  PCMSK0 |= 0b00011000; //Enable PCINT3 ONLY (bit 3)
  //Interrupt on pins 11 and 12, RE Ticks
  //Mask for PCINTs 16through 23 is PCMSK2 (Home Position)
  //Need to allow interrupts on PCINT23 so set bit 7
  PCMSK2 |= 0b10000000; //Interrupt on pin 7, Home Position Sensor activated
  //Now enable the interrupts (TURN THEM ON) by setting bits in PCICR
  //PCICR is Pin Change Interupt Control Register.Set bit 0 (PCIE0)  
  //to enable interrupts PCINT0:7 This catches PCINT3 and 4 - RE Ticks
  //Set bit 2 (PCIE2) to enable interrupts PCINT16:23. Catches PCINT21 - Home Position Sensor
  PCICR &= 0b00000000; //Clear PCICR register 
  PCICR |= 0b00000001; //Set bit 0 - Catches PCINT3 & PCINT4 - RE Ticks
  PCICR |= 0b00000100; //Set bit 2 - Catch PCINT21 - Home Position Sensor
  //**************************SET UP 'EXTERNAL INTERRUPTS'**************************
  //Interupts on External Interrupt Pins D2 (INT0) and D3 (INT1).
  //INT0 (Nano pin D2)occurs when a stepper motor driver pulse is received
  //INT1 (Nano pin D3)occurs when the ABoard2 pin, pnVmInRotMotor toggles 
  //indicating a motor direction change.
  //To use the 'External Interrupts' the following registers need to be set-up:         
  //EICRA External Interrupt Control Register A       
  //Need to set Interrupt Sense Control bits ISC11 .. ISC00 (bits 3:0 in EICRA)
  //ISC01     ISC00 (INT0)    Interrupt
  //ISC11     ISC01 (INT1)    Generated by
  //  0         0             Low level on Pin
  //  0         1             Any Logical Change
  //  1         0             Falling Edge
  //  1         1             Rising Edge
  //First clear all bits, then set as follows:  
  //For INT1 - Motor Direction - Generate on ANY LOGICAL CHANGE     
  //bit 3 ISC11 0     
  //bit 2 ISC10 1 This combination = Any logical change causes interrupt 
  //For INT0 - Stepper Motor Pulses  - Generate on RISING EDGE      
  //bit 1 ISC01 1     
  //bit 0 ISC00 1 This combination = Rising edge of pulse causes interrupt
  //NOTE: To provide some immunity to RFI, Aboard3:Pins 2 & 3 are pulled high
  //using 10k resistors. 
  //So, code is
  EICRA &= 0b00000000; //Clear EICRA register
  EICRA |= 0b00000111; // Set bits 0,1 and 2 in EICRA register
  //EIMSK External Interrupt Mask Register        
  //Need to set External Interrupt Request Enables INT1 & INT0  (bits 1:0)          
  //First clear both bits, then set as follows: 
  //bit 1  INT1  1 External interrupt pin (D3) enabled   
  //bit 0  INT0  1 External interrupt pin (D2) enabled   
  //So, code is
  EIMSK &= 0b00000000; //Clear EIMSK register
  EIMSK |= 0b00000011; // Set bits 0 and 1 in EIMSK register
  //NOTE: Setting EIMSK TURNS ON THE EXTERNAL INTERRUPTS
  //************VARIABLE INITIALISATION*********
  myCommand = 0;
  myTest = 0;
  tickCount = 0L;
  totalCount = 0L;
} //END of setup

void loop() {
  //******************************COMMAND ACTIONS******************************
  if (myTest == 3) (
    //RE tick
    Serial << "Tick Count = " << tickCount << "  totalCount = " << totalCount << "\n";
    myTest = 0;
  }
}
//*************************FUNCTIONS / ISRs FROM HEREON*************************
ISR(INT0_vect) {
  //Triggered by stepper motor drive pulse from ABoard1
  totalCount = totalCount + addPulse;
}
ISR(INT1_vect) {
  //Triggered by a change in motor direction
  if(digitalRead(pnMotorDirection)) {
    addPulse = 1L;
  } else {
    addPulse = -1L;
  }
}
ISR(PCINT0_vect) {
  //Triggered by a ROTARY ENCODER TICK
  halftickCount++;
  if (halftickCount == 2) {
    //Make count same as Velleman
    tickCount++;
    halftickCount = 0;
    myTest = 3;
  }
}
ISR(PCINT2_vect) {
  //Triggered by activation of Home Position Sensor
  myTest = 4;
}
6
  • 1
    PCMSK0 &= 0b00000000; //Clear all bits - wouldn't PCMSK0 = 0; be a lot clearer? Meanwhile I agree with Gerben's answer. Try "guarding" your access to multi-byte variables set by interrupts. See Interrupts.
    – Nick Gammon
    Commented Mar 14, 2016 at 19:44
  • 1
    1) You have a scope, you used it to look at the motor pulses. Did you look at the encoder pulses also? How do they look like? 2) On a quadrature encoder, I would expect at most one channel to be “bouncy” at any given time. Then, if you use the encoder in the usual way, you should get a sequence of +1, −1, +1, −1... adding up to zero. 3) The standard idiom for setting the AVR I/O registers is PCMSK0 = _BV(PCINT3) | _BV(PCINT4);, which looks clearer than PCMSK0 &= 0b00000000; PCMSK0 |= 0b00011000;, at least to those familiar with AVR programming. Commented Mar 16, 2016 at 16:55
  • Do you have a proper shared ground between quadrature encoder outputs and the Arduino? If so, then @EdgarBonet is right. Quadrature encoders can bounce like crazy, so you really do need to read the encoder properly, determining the step direction and all from each and every read. This way, if it bounces +1 and then -1, or any other sequence, 100 times in a row, you end up knowing what the new location is, and thereby you've debounced it too. Try adding a 2.2uF cap on each quadrature output too, to help a little bit with hardware debouncing. Commented Aug 12, 2016 at 0:30
  • 1
    Properly reading the encoder is the software debouncing you really need. Commented Aug 12, 2016 at 0:32
  • Is your rotary encoder electrical or optical? I ask as I wonder if you have issues with switch bounce or possibly the optical receiver swithing on and off at a boundary. If this is the case, one way that may help solve it is to properly decode the quad encoded pulses. This blog post shows a reliable way of decoding quad encoded pulses: thewanderingengineer.com/2014/08/11/… Regads, Commented Sep 12, 2016 at 10:35

1 Answer 1

1

There is no need to use int32_t for some of the variables. The problem with using variables that are more than 8 bits on a 8-bit processor, is that the processor needs 4 reads to get the value. In the middle of those reads an interrupt might occur, resulting in a value that has some bits of the old value, and some bits of the new.

Some variable only need 8 bits.

uint8_t volatile myTest = 0;
int32_t volatile totalCount = 0L;
int32_t volatile tickCount = 0L;
uint8_t volatile halftickCount = 0;
int8_t volatile addPulse = 0L; 

The next thing is to make the reading of the value atomic.

void loop()                                             
{                                               
//******************************COMMAND ACTIONS******************************  

if (myTest == 3)   //RE tick
{
    noInterrupts();
    int32_t tickCountCopy = tickCount;
    int32_t totalCountCopy = totalCount;
    interrupts();
    Serial << "Tick Count = " << tickCountCopy << "  totalCount = " << totalCountCopy << "\n"; 
    myTest = 0; 
}
}   

PS. this is not necessarily an answer, but it didn't fit in the comments.

5
  • Hi, and thanks Gerben and Nick. I made the changes you recommended but they actually seemed to make things worse, not better, in that even without the step motor running I was getting multiple triggering on each of the RE channels. Commented Mar 15, 2016 at 17:48
  • Oops - wanted a new line and pressed return but that sent the comment before I had finished. So, excuse the lack of formatting from now on. I have now reached the conclusion that I need to debounce the Rotary Encoder. I had ASSUMED that as it's an optical encoder it wouldn't require debouncing but I now think my assumption was wrong. I am going to make a small hardware debouncer using RC timing and a 74HC14 hex inverting Schmitt trigger IC. Hopefully hat will solve the problem. The good news is that the actual step motor pulses are counted without any problems. Thanks, Hugh. Commented Mar 15, 2016 at 17:54
  • There's no real bounce here, but probably not as clear a transition as you'd expect. My guess is a schmitt trigger alone will work. Even though my answer doesn't solve your problem, it still point to a problem you might account with your code.
    – Gerben
    Commented Mar 15, 2016 at 19:31
  • 1
    Hi, I tried to use my scope to look at the encoder outputs but I don't know how to use it properly and couldn't get meaningful results. I have made the debouncer board, using ca 5 mS delay on each encoder channel and, finally, it all works as it is supposed to. I first had this problem in early December so it has taken all this time to get it solved. Actually bought my scope to help with this specific problem. Thanks again to everyone who answered and commented. Regards Hugh Commented Mar 17, 2016 at 14:33
  • @HughGilhespie, if the problem is solved now, please post an answer to your own question and mark it as solved. Commented Aug 12, 2016 at 0:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.