Tag Archives: programming

A (Horrible) Method of Hard Resetting an Arduino

Leave a reply

Ethernet ArduinoLast year I worked on a little home project involving an Arduino Ethernet board. It all worked perfectly except that my program would hang after a few days operation. The time to failure wasn’t consistent, sometimes it failed after only a couple of hours, often it took a week. But sooner or later it would stop sending data across the network and require a hard reset. Adding logging told me only that it was the network code that was stuck.

It turns out that this is a common problem with the Arduino Ethernet hardware. The fancy pants W5100 TCP/IP chip just goes on strike and needs to be reset. Using the Arduino’s watchdog interrupt to software reset the CPU isn’t enough, the W5100 needs to reset as well.

Reading the schematic provided a clue. The W5100′s reset line is connected to the normal Arduino reset line (via a small reset controller IC), so resetting the Arduino by pulling the RESET line low resets the network chip as well.

There is a right way to do this that involves a bunch of extra circuitry, and a wrong way that involves a piece of wire.

Photo of Arduino showing a wire soldered to the reset pin
Connecting the reset pin to one of the Arduino’s digital IO pins is not recommended. The CPU requires that the reset line is help low for a certain amount of time to reset correctly and it can’t hold the line low and reset simultaneously. That wouldn’t make sense. Except that when I tried it the CPU reset every time and took the network chip with it. There is a capacitor between the RESET pin and GRD that I think helps out here, but my electronics is very weak.

I am still making use of the watchdog interrupt but instead of letting it reset the CPU, I handle the interrupt and bring the RESET line low using the digital IO pin.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

#include <avr/wdt.h>
 
#define RESET_IO_PIN 3
 
void setup()
{
	//  Some people recommend setting the IO pin to HIGH on start up
	//  but it is better to leave it floating as INPUT. Pulling it HIGH interferes 
	//  with the normal reset mechanism, and the pin is held HIGH by a pull-up   						  
	//  resister anyway
	//  digitalWrite( RESET_IO_PIN, HIGH );
	//  pinMode( RESET_IO_PIN, OUTPUT );
 
	noInterrupts();
	wdt_reset();
	// these cryptic lines set the watchdog timer control register
	// to trigger an interrupt (not reset) after 8 seconds. The normal
	// Arduino function assumes you want to reset, so we can't use it here.
	// See the ATmega328P Datasheet section 10.9.2 for the gory details
	MCUSR &= ~(1<<WDRF);
	WDTCSR |= (1<<WDCE) | (1<<WDE);
	WDTCSR = (1<<WDP0) | (1<<WDP3) | (1<<WDIE); /* 8.0 seconds */
	interrupts();
}
 
ISR(WDT_vect)
{
	// The watchdog timer has fired, pull the pin low
	digitalWrite( RESET_IO_PIN, LOW );
	pinMode( RESET_IO_PIN, OUTPUT );
}
 
 
void loop()
{
	wdt_reset();
	// do some stuff here, if you don't call wdt_reset() more frequently
	// than 8 seconds the Arduino will reset
}

I spent weeks trying to figure out what was going on, hopefully this post will help somebody in the same situation.

Jumping Frogs – Using Python to Solve Puzzles

A few months ago I came across the following puzzle in a video game I was playing:

The starting position for the siz frogs puzzleThree frogs are happily hopping along a narrow board together when they meet another group of three frogs traveling in the opposite direction. These frogs can only move in the direction they are facing, and only if there is a space directly in front of them. Additionally, a frog can jump over the frog in front but only if there is clear space on the other side to land in.

How can the frogs (moving one at a time) pass each other and continue on their way?

Of course, this is a hoary old puzzle that most people come across and solve as children. It should be only a couple of minutes work with a pen and paper to confirm that it is possible to exchange both sets of frogs but I wouldn’t be much of a programmer if I used a piece of paper where hundreds of dollars of computer equipment would do just as well.

To solve a puzzle like this programatically requires three things: a representation of the current state of the problem, a way of generating every possibly legal move from a given position, and a way of figuring out when is a good time to stop.

Firstly, the representation of the board is a simple python list:

1

start = [1, 1, 1, 0, -1, -1, -1]

Frogs traveling right or left are represented by “1″ and “-1″ respectively. Empty spaces that frog can move into are represented by “0″. The advantage of this representation is that you can calculate the new position of a frog by:

1

newPos = pos + (representation * distance)

where pos is the current index in the array, distance is the size of the hop (either 1 or 2) and representation is either 1 or -1.

Next, we need a way of generating legal moves for a given position:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

def legalMoves(board):
	moves = []
	for pos, piece in enumerate( board ):
		jumpmove = pos + (piece * 2)
		move = pos + (piece)
		if ( piece == 0 ):
			continue
		if (not (( jumpmove < 0 ) or ( jumpmove >= len(board)))):
			if (board[jumpmove] == 0):
				t = list(board)
				t[pos] = 0
				t[jumpmove] = piece
				moves.append(t)
		if (not ((move < 0) or ( move >= len(board)))):
			if ( board[move] == 0):
				t = list(board)
				t[pos] = 0
				t[move] = piece
				moves.append(t)
	return moves

Now we need a way of keeping track of all board positions we have seen, so once we find the target we can print the states that led to the solution:

1
2
3
4
5
6
7
8
9
10
11

def evalAll( current, target ):
	next = []
	for a in current:
		n = legalMoves(a[-1])
		for q in n:
			t = list(a)
			t.append(q)
			if ( q == target ):
				return t
			next.append(t)
	return next

This code keeps a list of lists, each sublist being it’s own list of the sequence of moves investigated so far. For each sequence of moves, the next legal moves are discovered and new sequences are added to be investigated the next time this function is called. Technically this is called a breadth-first search because at all of the current legal moves are investigated before moving on the next stage. This is a very simplistic way of doing the job, but in this case the puzzle is small enough that it works well enough.

Finally, a simple wrapper that we can use to set things up and return the final result.

1
2
3
4
5
6
7

def solve(start):
	temp=[[start]]
	end = list(start)
	end.reverse()
	while(temp[-1] != end):
		temp = evalAll(temp, end)
	return temp

So now we can do this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

print solve([1, 1, 1, 0, -1, -1, -1])
 
[[1, 1, 1, 0, -1, -1, -1],
 [1, 1, 0, 1, -1, -1, -1],
 [1, 1, -1, 1, 0, -1, -1],
 [1, 1, -1, 1, -1, 0, -1],
 [1, 1, -1, 0, -1, 1, -1],
 [1, 0, -1, 1, -1, 1, -1],
 [0, 1, -1, 1, -1, 1, -1],
 [-1, 1, 0, 1, -1, 1, -1],
 [-1, 1, -1, 1, 0, 1, -1],
 [-1, 1, -1, 1, -1, 1, 0],
 [-1, 1, -1, 1, -1, 0, 1],
 [-1, 1, -1, 0, -1, 1, 1],
 [-1, 0, -1, 1, -1, 1, 1],
 [-1, -1, 0, 1, -1, 1, 1],
 [-1, -1, -1, 1, 0, 1, 1],
 [-1, -1, -1, 0, 1, 1, 1]]

Success!

You might say this is a waste of time since you figured out the problem in your head. Good for you, but try this on for size:

1

[1, 1, 1, 1, 1, 1, 1, 1, 0, -1, -1, -1, -1, -1, -1, -1, -1]

Arduino POV Device Part II

Last week I wrote about starting with digital electronics with the Arduino prototyping platform. Getting something working was easy enough but I was dissatisfied with ugly result, especially when my friend Lloyd showed up with his extremely tidy version of the same idea. The piece of cardboard had to go!

My first version used a breadboard and uncut LED leads because I thought that I would want to dismantle the components for reuse. However I have since realised that I am a grown man with a career and can easily afford the $4.50 material cost. Behold POV version 2:



I won’t be entering this in any soldering competitions but it works OK.

The code has also changed. Compiling in what amounts to a 2D bitmap was a quick way to get something up and running but it wasn’t very flexible, especially since in the future I want to display long strings of text. This means storing glyphs (letter shapes) for every letter of the alphabet (plus digits and symbols) – what I needed was a proper font; what I got was this:

I then turned to Python to generate the program data to embed in the new Arduino code. This script slices the image vertically, generating one byte (actually only 7 bits) for each slice of pixels. What comes out is a long list of byte values that can be output directly to the Arduino’s IO pins to display each section of the text, along with a set of indexes that map letters to the start and end positions of individual glyphs in the array.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52

import sys
 
def GetSlice(i, pos):
	t = 0
	for q in range( pos, len(i), len(i) / 7 ):
		t = t * 2
		if ( i[q] != '\x00' ):
			t = t + 1
	return t
 
def GetGlyph( i, pos ):
	end = pos
	start = pos
	data = []
	while (end < len(i) ):
		t = GetSlice( i, end )
		if ( t == 0 ):
			break;
		data.append( t )
		end = end + 1
	return (start, end, data)
 
def GetFont( d ):
    result = []
    i = 0
    currentIndex = 0
    indices = []
    while ( i < len(d) / 7 ):
	(start, end, data) = GetGlyph( d, i )
	if ( len(data) == 0 ):
		break
	i = end + 1
	indices.append( currentIndex )
	currentIndex = currentIndex + len(data)
	result.extend( data )
    return ( result, indices )
 
if __name__ == '__main__':
    if ( len(sys.argv ) != 3 ):
        print "USAGE: program <input> <output>"
        exit()
    data = []
    i = file( sys.argv[1], "rb" )
    data = i.read()
    i.close()
    ( d, indices ) = GetFont( data )
    o = file( sys.argv[2], "w" )
    o.write( "static const int indices[] = { " + ",".join(map(str, indices ) ) + " }; ")
    o.write("\n")
    o.write( "static const unsigned char fontdata[] = { " + ",".join(map(str, d)) + "};")
    o.write("\n")
    o.close()

This scheme is more complex that the original bitmap but allows arbitrary text to be display without lots of editing. The modified Arduino code now looks like this. The first two lines contain the new font data, and you can see that the text to display is simply declared on line 5.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55

static const int indices[] = { 0,4,8,12,16,19,22,26,30,33,37,42,45,52,59,63,67,71,75,79,82,86,91,96,101,106,111,116,121,126,131,136,141,146,151,156,161,163,165,167 }; 
static const unsigned char fontdata[] = { 63,72,72,63,127,73,73,54,62,65,65,34,127,65,65,62,127,73,65,127,72,64,62,65,69,39,127,8,8,127,65,127,65,2,1,1,126,127,8,20,34,65,127,1,1,63,64,64,60,64,64,63,127,32,16,8,4,2,127,127,65,65,127,127,72,72,48,62,65,71,62,127,72,76,51,50,73,73,38,64,127,64,126,1,1,126,112,12,3,12,112,126,1,15,1,126,65,54,8,54,65,96,16,15,16,96,67,69,73,81,97,62,69,73,81,62,17,33,127,1,1,71,73,73,73,49,65,65,73,73,54,24,40,72,127,8,121,73,73,73,6,62,73,73,73,6,65,66,68,72,112,54,73,73,73,54,48,73,73,73,62,3,3,123,123,27,27,127};
 
static int outputPins[] = { 12, 11, 10, 9, 8, 7, 6 };
static char text[] = "ANDREW";
 
void setup()
{
  for (int i = 6; i<=12; ++i)
  {
    pinMode(i, OUTPUT);     
    digitalWrite(i, HIGH);
  }
  pinMode(13, OUTPUT);
  digitalWrite(13, LOW);
  delay(2000);
}
 
void loop()
{
  byte row = 0;
  byte endRow = 0;
  char* currentChar = text;
 
  while (1)
  {
    if (*currentChar == 0)
      currentChar = text;
    if ( row == endRow )
    {
        char t = *currentChar;
        currentChar++;
        t = t - 'A';
        row = indices[t];
        endRow = indices[t+1];
    }
    while (row != endRow)
    {
      byte mask = 64;
      for (byte pin = 0; pin < (sizeof( outputPins) / sizeof(outputPins[0])); ++pin )
      {
        digitalWrite( outputPins[pin], (fontdata[row]&mask)? HIGH : LOW  );
        mask = mask / 2; // divide mask by two, moving it down one bit
      }
      row++;
      delay(1);
    }
    for (byte pin = 0; pin < (sizeof( outputPins) / sizeof(outputPins[0])); ++pin )
    {
      digitalWrite( outputPins[pin], LOW);
    }
 
    delay(1);
  }
}

Still to do: This code is not very efficient. It uses digitalWrite() in a loop to turn on the LEDs. digitalWrite() is very easy to use but if you look at the source code you will see it does all sorts of unnecessary stuff for this job. I plan to replace it with direct writes to the required Arduino IO registers.

I also want to make the whole thing interrupt driven to free up the CPU. Why would I want the extra CPU time? I have plans for that as well…

Starting with Arduino

I have a new hobby – digital electronics! Last week I bought an Arduino Uno on a whim from JayCar. Being a typical programmer, I have very little idea how computers actually work on the physical level so this gives me an easy introduction to lower-level coding and electronics in general.

The Arduino is a small board with a fairly capable 8 bit microprocessor (an ATmega328 to be precise) with the IO pins hooked up to convenient headers ready for attaching external devices. The CPU has its own RAM and programmable flash memory, and it comes with a bootloader that can reprogram the flash via the handy USB connection (which can also power the whole board). It is hard to imagine a more plug-n-play device.

I decided that I would follow the well-worn path and create a persistence of vision device for my first project. Mine consists of 7 LEDs hooked up to IO 7 pins (via appropriate resistors), the idea is to make the LEDs blink in such a way that recognisable shapes appear as the LEDs are waved quickly in front of your eyes.

My prototype is a little rough, but works fine in a darkened room (I was too cheap to spring for super bright LEDs and the darkness hides my shoddy soldering job.) The breadboard is too fragile to wave around so the LEDs are mounted on cardboard and connected using a couple of metres of CAT5, which conveniently has 8 internal wires to supply the 7 LEDs with a common ground. The total cost is less than $10, excluding the breadboard and the Arduino itself.

The Arduino programming environment is pretty nifty. The language used is a limited form of C++ (no standard library or runtime support for much of anything) with some extra libraries for managing the CPU’s features. Compiling and flashing the CPU is as simple as pushing a button. Here is the code that generated the above picture:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54

// Array holding the graphic to display
static char* text[] = {
" XX   X     X  XX     XXXX   XXXX  X     X     ",
"X  X  XX    X  X  X   X   X  X     X     X     ",
"X  X  X X   X  X   X  X   X  X     X     X     ",
"XXXX  X  X  X  X   X  XXXX   XXX   X  X  X     ",
"X  X  X   X X  X   X  X X    X     X  X  X     ",
"X  X  X    XX  X  X   X  X   X     X  X  X     ",
"X  X  X     X  XXX    X   X  XXXX   XX XX      " };
 
static int outputPins[] = { 12, 11, 10, 9, 8, 7, 6 };
static int graphicLength;
 
 
void setup()
{
	// setup the initial state of the pins
  for (int i = 6; i<=12; ++i)
  {
    pinMode(i, OUTPUT);     
  }
	// pin 13 is hardwired to the onboard LED, turn it off
  pinMode(13, OUTPUT);
  digitalWrite(13, LOW);
 
  graphicLength = 0;
  for (char* t = text[0]; *t!=0; ++t)
  {
    ++graphicLength;
  }  
}
 
void loop()
{
  while (1)
  {
    for (int i = 0; i < graphicLength; ++i )
    {
		// find out if the pin is supposed to be on (HIGH) or off (LOW)
      for (int pin = 0; pin < (sizeof( outputPins) / sizeof(outputPins[0])); ++pin )
      {
        if (text[pin][i] == ' ')
        {
          digitalWrite( outputPins[pin], LOW );
        }
        else
        {
          digitalWrite( outputPins[pin], HIGH );
        }
      }
      delay( 1 );
    }
  }
}

Not very efficient but simple enough to get going. I am already working on a more functional version which will have its own font and interrupt driven display.

A Better Boost Book

Boost is a excellent resource for C++ programming, but suffers from inconsistent documentation and a daunting array of sub-projects. Trying to make sense of it all is a fairly serious undertaking. I tried to get my head around it by writing my occasional series of boost blog posts, but now I see that somebody has done a much better job.

The Boost C++ Libraries is a free book that clearly explains some of the more generally useful boost libraries, with lots of useful examples. It even covers advanced libraries like ASIO in an approachable way. I highly recommended bookmarking it if you do any C++ programming.