The W5100 bug(s) – and how to fix it (them)

Bug 1- The SPI bug

On a recent project in which I shared the SPI bus of a 3V3 8MHz ProMini, with a W5100 Ethernet module and another SPI slave, I experienced some problems: sometimes it would work perfectly, sometimes not.
I had hooked up the ethernetmodule with Pin D10 as Chipselect and the other slave (an RFM69) with Pin D8 as Chipselect. Pretty standard and it should work, shouldnt it?
Well as it is, it didnt. Oddly enough, I had build it before with another Ethernet module and that worked fine.
When I connected my EthernetShield in the same way, the issue was gone. Time to throw suspicion on my Funduino W5100 Ethernetmodule. Well on inspection it is immediately clear that eventhough the functionality is supposed to be the same, the ethernetshield and also my previous module have more chips than my Funduino module. Why for instance does my shield have a 74LVC14 and my Funduino module just the W5100 chip and nothing more?

When looking at the circuit of the Ethernetshield it is clear that only one gate of the 74LVC14 (a SchmittTrigger inverter) is used: it takes the SS signal (the ‘chipselect’) inverts that and sends that inverted signal to Pin 31 of the W5100. Pin 31 is the SEN pin.

In the Funduino module that was different. The ‘SEN’ pin is just tied high to 3V3 with a 10k resistor. My shield has that pull-up resister as well, but still is controlled by the inverted chipselect signal.

Time for the datasheet of the W5100

Hmm.. that doesnt say much, other than that in my module the SPI mode is enabled, however, judging from the 74LVC14, it seems that this pin needs to be driven to LOW when the Chip is not selected. In other words, it is apparently necessary to disable the SPI mode in order to release the bus to another slave.

Time for some soldering: apart from the Vcc and ground, one needs two signals: the SS pin and the SEN pin. As the W5100 is an 80 LQTF chip, at my age (eyes) I was not even going to try soldering on the chip, but fortunately the Pullup resistor gives an entry point to the SEND signal and the connector is where we find the SS signal

Pullup resistor on SEN pin
Pullup resistor on SEN pin
Pullup resistor on SEN pin, close-up
Pullup resistor on SEN pin, close-up

I used a 74LCV14  (Low Voltage Inverting SchmittTrigger) for my modification, but I am sure an 74LC04 or 74LVT04 (an inverter) would do just as well. If you vcannot find the 74LCV14, try a 74LVT14 or 74HC(T)14 (it accepts  2 Volt Vcc). John Crouchley who also describes this problem uses a CMOS CD4011. He feeds that from 5Volt, which I did not want to do. It should be posible though, given the fact that the W5100 pins apear to be 5 V tolerant. Perhaps even a simple 1 transistor inverter is possible. Ideal would be an 74AHC1G04 as that is only a single inverter chip that measures only 2x3mm and can easily find a plce on the module.
Currently my modification is breadboarded. I will think of a neat solution and then add some more pictures.

Bug 2- The ‘510’ bug

There is another problem with some of the ‘asian webstore’ W5100 based Ethernet shields which isnt really a bug with the W5100 but more a problem with bad sourcing.
Some of those shields refuse to make contact with the internet because of some wrong components (resistors)In the picture here of the top of an ethernet shield, directly right of the Ethernet jack there is a “spider” resistor that is labelled with “49R9” that are actually 4 resistors in one  package with each a value of 49.9 Ohm. Apparently that was hrd to source at some time and  clone manufacturers decided to use 51 Ohm resitors. That is not a problem if indeed they had used 51 Ohm, but by mistake 510 ohm resistors were used (labelled ‘511’). Resistors with a value of 51 ohm should have been labelled with ‘510’.
So if you have such a board you need to replace the 510 ohm resistors. If you can’t find the proper replacement resistorpack then apparently it is also OK to solder  a 100 ohm resistor between pin 1-2 (Tx+/Tx-) and a 100 ohm resistor between pin 3 and 6 (Rx+/Rx-) as explained here. (So 2×100 Ohm resistors in total).

The RJ45 Jack, seen from component side

Bug 3 – The Funduino Reset Bug

Screenshot_2017-04-19_22-53-51The “Funduino” W5100 module as is a cheap, but qualitatively good W5100 board. However, apart from the ‘problem’ of not being able to share the SPI bus in unmodified state (adressed above), some  people have a lot of trouble getting it to work in the first place. The problem is then most likely in the Reset of the module. Modern versions of the Ethernetshield have a seperate Reset controller, that is triggered by the RST of the Arduino. The Funduino Module, just has its RESET tied to 3.3V via a 10 k resistor. As a result the Module’s reset state is a bit unpredictable. Some people get it to work by powering the module up and down a few times, but that remains random luck. What I found that works immediately is to add a slight delay in the setup, before the Ethernet connection is initialized. For me 250mS was enough, but you may need a bit more or a bit less.
So my Setup looks as follows:

void setup() {
// setup ethernet communication using DHCP

Also, when you use the shield, the Reset pins of the Arduino and the shield are connected, so the Arduino can reset the shield. With this module you cannot do that: if you would connect both resets, you will notice that you cannot upload sketches. Perhaps it will work if you only connect the RST after you upload a sketch, I did not try that, too much hassle. I found the 250mS delay to work for me.
This ‘bug’ does not happen with all Funduino W5100 modules. Supposedly some batches are ‘OK’. There are 2 identification numbers: one on the PCB and one on the RJ45 jack. Supposedly if that says “15/10”  it is OK, “14/10” is not. Mine said “16/38” and needed the delay.
The only conenctions (other than Vcc and ground)  that I make are:
D10  -> nss
D11  -> miso
D12  -> mosi
D13  -> sck


Adding an MCP23017 16 port IO expander to Arduino or Esp8266 or Attiny85 or……..

After I made this expander module, a ready made module
with this chip has become available. So I actually would advice anybody needing a 16 bit expander, to buy that one rather than build it. The module will cost you abt 1.50 euro, while the individual chip may set you back a euro or so.

I am not claiming that what I am describing here is earth shattering or trailblazing, because in fact it is very simple and no doubt has been done by many already. But sometimes what is simple for the one, is still a question mark for the other, so here is quick ‘how-to’ of adding 16 I/O ports to your microprocessor. This is especially handy when working with a chip like the ESP8266 that has only limited I/O
The MCP23017 is an I2C enabled 16 I/O port chip. That means that you only need 2 pins (yes with Vcc and ground it makes 4) to control the chip and the added advantage is that you can share I2C with various other devices as well.

The 16 I/O lines are divided into an 8 I/O PORT A and an 8 I/O PORT B. Both can be used as input as well as output. The chip also has 2 configurable interrupts (that I will not be using). The physical layout of the chip makes it quite easy to use it on a piece of strip board.

The circuit (at right) is rather simple. At a last moment I decided to leave out the pull up resistors so it would be more flexible to use together with other I/O devices. The 3 Address pins A0-A2 determine the I2C address that ranges from 0x20 (all pins on ground) to 0x27 (all pins on Vcc).
The chip  can take a Vcc from 2.7V to 5V and this is perfect for 3.3 Volt devices as  the modern arduino’s and the ESP8266 range of boards.

Using the chip in a program is fairly easy. There are good libraries available, but it might help if you know how to program the chip without a library.
In my case I have all  address lines tied to ground and therefore my I2C address is 0x20. Suppose I want to use all PORT A lines as outputs. I do that  as follows:

Wire.write(0x00); // IODIRA register
Wire.write(0x00); // set entire PORT A to output

For PORT B that  is rather similar:

Wire.write(0x01); // IODIRB register
Wire.write(0x00); // set entire PORT B to output

If we then want to send a specific value ‘X’ to that PORT A, we do that as follows

Wire.write(0x12); // address port A
Wire.write(X);  // value to send

‘X’ ofcourse is a byte value that determines whether we set a specific port HIGH or LOW.
If for instance ‘X’is ‘0’ that means we write a LOW to all PORT A outputs. If it is 255 that means we write a HIGH to all PORT A outputs.
To determine what value to send, consider the 8 I/O lines of PORT A as a byte in which the individual bits determine HIGH or LOW.
So if we only want to make PORTA.0 HIGH and the rest LOW, we write a binary value of 0b00000001 =1 to the A register. If we want to make PORTA.0 and PORTA.2 HIGH and the rest LOW we write a binary value of 0b00000101 = 5.
For PORT B it is similar:

Wire.write(0x13); // address PORT B
Wire.write(X);  // value to send

If we want to use PORT B (or PORT A for that matter) as input, we do that as follows:

Wire.write(0x13); // address PORT B
Wire.requestFrom(0x20, 1); // request one byte of data
byte; // store incoming byte into "input"

The byte “input” will vary between 0 and 255, in which the individual bits determine the input on the corresponding IO line. So if ‘input’  reads ‘3’  which in binary is 0b00000011, that means that both IO line 0 and 1  were HIGH and the rest LOW

#include <Wire.h> // Wire.h
byte input=0;
void setup()
  Wire.begin(); // wake up I2C bus
  Wire.write(0x00); // IODIRA register
  Wire.write(0x00); // set entire PORT A as output
void loop()
  // read the inputs of bank B
  Wire.requestFrom(0x20, 1);;
  // now send the input data to bank A
  Wire.write(0x12); // address PORT A
  Wire.write(input);    // PORT A
  delay(100); // for debounce

That’s basically it if you want to do the adressing yourself. Using a library, such as the one from Adafruit, makes it much easier though as it has commands to write and read from individual IO lines. One of the example programs to read a single button, looks  for instance like this:

#include <Wire.h> // Wire.h
#include "Adafruit_MCP23017.h"

// Basic pin reading and pullup test for the MCP23017 I/O expander
// public domain!
// Connect pin #12 of the expander to Analog 5 (i2c clock)
// Connect pin #13 of the expander to Analog 4 (i2c data)
// Connect pins #15, 16 and 17 of the expander to ground (address selection)
// Connect pin #9 of the expander to 5V (power)
// Connect pin #10 of the expander to ground (common ground)
// Connect pin #18 through a ~10kohm resistor to 5V (reset pin, active low)
// Input #0 is on pin 21 so connect a button or switch from there to ground

Adafruit_MCP23017 mcp;

void setup() 
mcp.begin();      // use default address 0
mcp.pinMode(0, INPUT);
mcp.pullUp(0, HIGH);  // turn on a 100K pullup internally
pinMode(13, OUTPUT);  // use the p13 LED as debugging

void loop() {
// The LED will 'echo' the button
digitalWrite(13, mcp.digitalRead(0));

If you want to use more than one MCP23017 do that as follows:

#define addr1 0 //addr1 =A2 low , A1 low , A0 low =000
#define addr2 1 //addr 2 = A2 low , A1 low , A0 high =001 

Mind you that “0” is in fact 0x20 and ‘1’ is in fact 0x21

MQTT with the ENC28J60 Ethershield

My old ENC28J60 based shield surely has gathered some dust
My old ENC28J60 based shield surely has gathered some dust

While playing with MQTT on various ESP8266’s, I started to wonder if maybe I could do something with an old ENC28j60 shield and module I still had laying around and actually hardly ever used.

For the youngsters: The ENC28J60 Ethernetshield was the first shield to connect the Arduino with the internet. The major drawback of the chip was that it was lacking a stack, that subsequently had to be constructed in software. As a result it was kinda memory hungry. The initial version was without an SD card slot, the later version had an SD card slot. There were two libraries available: The Ethercard and the Ethershield library. The UIPEthernet library came later.
As far as I could figure out, the PubSubClient library that is needed for MQTT doesnt work with the original Ethershield/EtherCard libraries. It does work with the UIPEthernet library though (extended fork here). The UIPEthernet library is a smart piece of coding that made programs written for the W5100 Ethernetshield suitable to be used with the ENC28J60 Shield, simply by changing the included library. Ofcourse there is a price to pay for this, namely more memory consumption. There is also an MQTT client for the ENC28J60/Atmega328 based Nanode. The Ethercard seems to work with the EthercardMQTT library.

Anyway, there still is enough memory to read a DHT11 sensor an analog port and some switches. In my case those were 3 door contacts. The PubSubClient is the original from Knolleary, though normally I am more a fan of the fork by Imroy. For this example I have used the Mosquitto public broker but ofcourse any broker can be used. I have installed the Mosquitto broker on a local raspberry and I found that a call to (“raspberrypi.local”,1883) does not work, using the  local ip number does work though

MQTT-Output ‘Dicht’ means ‘Closed’

The program is rather ‘spartan’, to save space. Print statements have been removed or commented out after initial testing.

#include <UIPEthernet.h>
#include "PubSubClient.h"
#include "DHT.h"

#define CLIENT_ID       "UnoMQTT"
#define INTERVAL        3000 // 3 sec delay between publishing
#define DHTPIN          3
#define DHTTYPE         DHT11
bool statusKD=HIGH;//living room door
bool statusBD=HIGH;//front door
bool statusGD=HIGH;//garage door
int lichtstatus;
uint8_t mac[6] = {0x00,0x01,0x02,0x03,0x04,0x05};

EthernetClient ethClient;
PubSubClient mqttClient;

long previousMillis;

void setup() {
  // setup serial communication
  // setup ethernet communication using DHCP
  if(Ethernet.begin(mac) == 0) {
    //Serial.println(F("Ethernet configuration using DHCP failed"));
  // setup mqtt client
  //mqttClient.setServer("",1883); //for using local broker
  //Serial.println(F("MQTT client configured"));

  // setup DHT sensor
  previousMillis = millis();

void loop() {
  lichtstatus = analogRead(A0);
  // check interval
  if(millis() - previousMillis > INTERVAL) {
    previousMillis = millis();

void sendData() {
  char msgBuffer[20];
  float h=dht.readHumidity();
  float t = dht.readTemperature();
  if(mqttClient.connect(CLIENT_ID)) {
   mqttClient.publish("hal/temp", dtostrf(t, 6, 2, msgBuffer));
   mqttClient.publish("hal/humid", dtostrf(h, 6, 2, msgBuffer));
   mqttClient.publish("hal/door", (statusBD == HIGH) ? "OPEN" : "DICHT");
   mqttClient.publish("hal/garage",(statusGD == HIGH) ? "OPEN" : "DICHT");
   mqttClient.publish("hal/kamer",(statusKD == HIGH) ? "OPEN" : "DICHT");
   mqttClient.publish("hal/licht", dtostrf(lichtstatus, 4, 0, msgBuffer));
 //hal=hallway, DICHT=Closed, kamer=room, licht=light
The old ENC28J60 shield at work
The old ENC28J60 shield at work

Just for completeness sake, this sketch takes about 77% of memory. This same sketch for the WIZ5100 based Ethernet shield, with the  Ethernet.h library takes about 52% of memory. If you decide to adapt the sketch, be careful with altering the character strings. For instance, using the string “OPEN” 3 times is likely less memory consuming than having 3 different strings of the same or even shorter length. Obviously the ‘topic’ strings take a lot of space as well and if you were to shorten them to less meaningful names, you could add more sensors such as for instance a PIR sensor.

Should you copy the program from this  blog page, you may copy ‘stray characters’  that you have to delete. You can also download it here.

Freeing up memory

Should you really be pressed for memory, there is a way to free up about 5K of Flash: Go to your /../sketchfolder/libraries/UIPEthernet-master/utility/uipethernet-conf.h  and open the uipethernet-conf.h file.
in that file you will see the following section:

If you set UIP_CONF_UDP to ‘0’ you will save 5kB flash, by disabling UDP. However, if you use DHCP to connect to your router, you cannot disable UDP as the DHCP connection requires UDP. In that case you still can gain a bit of memory by reducing the UIP_UDP_CONNS.
An example of using a fixed address is this:

#include <UIPEthernet.h>
byte mac[] = { 0xDE, 0xAD, 0xBE, 0xEF, 0xFE, 0xED }; 
//the IP address for the shield:
byte ip[] = { 192, 168, 1, 120 };

void setup()
 Ethernet.begin(mac, ip);
void loop() {}

A final warning… after I updated my libraries, including the Adafruit DHT library, I received an error on compiling this sketch. That disappeared when I returned it to version 1.2.1. I thought I was not using the Adafruit library but the Tillaert library, but DHT libraries come a dime a dozen so maybe my compiler linked in another than I thought.
(The main sketch is my adaptation of work I found on internet, but I think the original source is from Luca Dentella)

Reset a program from a ‘freeze’

While doing a test run on software for an incubator the software did fine, until suddenly on the ninth day weird things started to happen: the LCD showed some odd data and in spite of a low temperature the heating had not kicked in. I resetted the software, wondering what it could be but only an hour later again something off happened. This time in spite of the temperature being well over the upper limit, the heating was not switched off.
I checked my code, which really wasn’t so complicated: Read DSB1820, read DHT11, compare with high and low limit, switch a pin on or off and write the value to an LCD and I couldn’t find a single mistake and don’t forget, it had run flawlessly for 9 days. I suspected my display. This was a 20×4 LCD with I2C module with 4k7 pull up. Nevertheless there might still have been some ‘static’ on the SDA and SCL lines that supposedly can cause the Arduino to freeze.

Obviously that is not good if you are running an incubator as you dont want to check it and find your eggs boiled.
So, other than maybe straightening out the LCD cables a bit, I decided that I needed some software protection against ‘freezes’
The only (and possibly best) way to do this is with the watchdog timer. I dont want to go into the specifics and the background of the watchdog timer, but just keep it on a practical level.
What we do is to set-up the watchdog timer to initiate a system reset after say 4 seconds. Then in our loop we do a reset of the watchdog timer so it starts counting from zero again. So as long as the program tells the watchdog timer “I am still running” nothing will happen. Should the program freeze up, it will not reset the watchdog timer and then after 4 seconds the watchdog timer will reset the entire system.
It is very well possible to use the watchdog timer by manipulating the various registers yourself, but it is much simpler to use the watchdog libary that is part of the avr libraries.
We do this as follows:

#include <avr/wdt.h>

void setup()
	//wdt_enable(WDTO_1S);// 1 sec
	wdt_enable(WDTO_2S);// 2 sec
	//wdt_enable(WDTO_4S);// 4 sec
	//wdt_enable(WDTO_8S);// 8 sec

void loop()
	//  your program

1 Wire LCD interface: Part 1

The one wire LCD interface has been around on the internet for some time already and made popular by Roman Black, although he claims he is not the inventor of the principle. His article is mainly PIC processor oriented, but it gives a good explanation of the principle. Myke Predko is also a name that  should not be unmentioned as pioneer in 1 wire LCD driving. Also this article about  one wire driving of a shift register is very informative

The ShiftRegLCD123 website from Raron used to give a lot of information as well, along with circuits, but that (the Wiki) seems to have disappeared. But his library is in codebender.

In brief, the One Wire interface uses a latched shift register in which the clock and the latch are taken from the data signal through two RC networks that produce the necessary delay. Circuits are a plenty on the interweb, but finding the proper driver for it can be confusing sometimes.

Although the interface on the data, clock and latch lines is rather standard, the connection between the shift register and the LCD has thre major variants that for now i will just call the Roman Black, the LCD3Wire circuit (yes, it is 1 wire) and the ShiftRegLCD123 circuit. Also Elektor has published a 1 wire interface (Detlef Hanemann)(Published in issue 9/2015 on page 92) with yet again a different shift register to LCD connection. Their setup is said to have several advantages over the setup used by Roman Black, it uses the Q7′ direct output of the shift register to trigger the output latch register to carry over automatically the shifted data to the output pins. But for that it requires a monoflop built around a BS170 FET to generate the E-pulse. The setup to be used in Francisco Malpartida’s LCD library is yet again different from the other 3 and uses a diode to create an AND gate

Roman Black






Malpartida LCD library

. Roman Black LCD3Wire Shiftreglcd Elektuur Malpartida
RS  Qc  Qb  Qc  Qg  Qg
E  Qd  Qd  Qh  Qh’  Qh’ sort of
D4  Qe  Qh  Qd  Qa  Qe
D5  Qf  Qg  Qe  Qb  Qd
D6  Qg  Qf  Qf  Qc  Qc
D7  Qh  Qe  Qg  Qd  Qb

Driving these OneWireLCD’s thus differs per configuration. The ShiftRegLCD123 library can do both the ShiftRegLCD and the LCD3Wire protocol. The Liquid Crystal Library of Francisco Malpartida  does have a One Wire protocol but make sure you have the latest version. However, it relies on connections different from the others mentioned:
// Bit #0 (QA) – not used
// Bit #1 (QB) – connects to LCD data input D7
// Bit #2 (QC) – connects to LCD data input D6
// Bit #3 (QD) – connects to LCD data input D5
// Bit #4 (QE) – connects to LCD data input D4
// Bit #5 (QF) – optional backlight control
// Bit #6 (QG) – connects to RS (Register Select) on the LCD
// Bit #7 (QH) – used for /CLR on the HW_CLEAR version (cannot be changed)
// (Q’H) – used for Latch/EN (via the diode AND “gate”) (cannot be changed)

// -----------------------------------------------
//                         74HC595     (VCC)
//                       +----u----+     |          2.2nF
// (LCD D7)------------1-|QB    VCC|-16--+      +----||----(GND)
// (LCD D6)------------2-|QC     QA|-15         |
// (LCD D5)------------3-|QD    SER|-14---------+--[ Resistor ]--+
// (LCD D4)------------4-|QE    /OE|-13--(GND)         1.5k      |
// (BL Circuit)--------5-|QF    RCK|-12---------+                |
//                       |         |             \               |
// (LCD RS)------------6-|QG    SCK|-11-----------)--------------+--(Seri
//                     7-|QH   /CLR|-10--(VCC)   /               |
//                  +--8-|GND   Q'H|--9---|<|---+--[ Resistor ]--+
//                  |    +---------+     diode  |      1.5k
//                  |                           |
//                  |      0.1uF                |
//                (GND)-----||----(VCC)         +----||----(GND)
//                                              |   2.2nF
// (LCD EN)-------------------------------------+
// (LCD RW)--(GND)

The Elektuur/Elektor configuration has its own program, or here (both direct download link)

A capacitive soil humidity sensor: Part 3

Initially I imagined the probe as a sleek device with a wire coming out that one would stick in the soil, if necessary completely. This would mean that the circuit pcb would be an integral part of the PCB that  formed the plates, but as always, things go different in practice. For one thing , the piece of PCB I intended to use just wasnt long enough. Second, my PCB etching possibilites are temporarily impeded. Third, I wanted to add an LDR, meaning that I had to have some sort of  translucent cover.

I constructed the circuit on a small piece of stripboard:

IMG_20160225_163125 i2cprobe

With regard to the LDR, sure that is nonsense. If I want to add an LDR I  could just as well add it to my base station. But I am not adding it because I have to, I am adding it because I can and because I wanted to get some experience in programming an I2C slave. An LDR might not be so usefull but in future I may want another sensor, e.g.a sensor that reads if there is really water flowing from the irrigation tube.

Also I had decided on putting the capacitor plates back to back, but as I did not have double sided PCB I just used two pieces glued and  soldered(!) together.
So My BOM was rather simple:

  • 2 equal size pieces of PCB  Size depends on what you have, but do not make them too small. I used 12×3 cm.
  • 1 piece of 0.5-1.0 cm plastic for a baseboard. I used an old cutting board.
  • 1 clear/translucent cover, I used the lid of a whipped cream spray can.
  • 1 piece of  thin 4 wire cable, length depending on your need

pcbprobe2I glued the two piecesof pcb back to back. drilled a hole in all 4 corners and through soldered a piece of wire through each hole, thus anchoring the plates together. I removed some copper around the solder so it would become an island isolated from the rest of the plate. (See picture).

Soldering the plates together in the corners may not be necessary if you decide to electrically isolate them from the soil with e.g. shrink tube.

Eventually I will place an NTC on the plate as well after it is covered with Shrink tube. Esthetically it might be better to put the NTC under the shrink tube, but that could create an air pocket.


baseplateI made the base plate from a o.5 cm thick piece of soft plastic. Cut out a round shape  with a 5.5 cm diameter to fit the base of my  clear dome and  made a slit 3 x0.3 cm into that in which the PCB  fits snug. made a round hole for the connecting cable.




I soldered two wires on the top of the PCB, one on each side. Soldered wires on an NTC, insulated those, attached the NTC to the bottom of the PCB with the wires leading to the top and then covered the PCB with shrink wrap.

probeEventually the probe looks like this (picture)probe3


A capacitive soil humidity sensor: Part 2

capacitive74HC14_attinyIn a previous article I presented a simple way to read a capacitive moisture sensor with a simple RC generator.
In this post I will present a sensor with some added functionality that can be read through I2C. The circuit (figure) doesnt need much explanation: the RC generator we saw in the previous article and the two variable resistors in a voltage divide read by analogue inputs. The Attiny 45 (NOTE, it is NOT an Attiny25) is the heart, or rather the brains of the sensor. As the Attiny  functions as an I2C slave we will need the TinyWireS library. The library comes with some examples and one example was quite easy to rework to what I needed. The code is as follows.

#define I2C_SLAVE_ADDRES 0x4
#include <TinyWireS.h>//
#define TWI_RX_BUFFER_SIZE ( 16 )

volatile uint8_t i2c_regs[] =
// Tracks the current register pointer position
volatile byte reg_position;
const byte reg_size = sizeof(i2c_regs);

byte LDRvalue;
byte NTCvalue;
int MoistPulse;

 * This is called for each read request we receive, never put more than one byte of data (with TinyWireS.send) to the
 * send-buffer when using this callback
void requestEvent()
	// Increment the reg position on each read, and loop back to zero
	if (reg_position >= reg_size)
		reg_position = 0;

 * The I2C data received -handler
 * This needs to complete before the next incoming transaction (start, data, restart/stop) on the bus does
 * so be quick, set flags for long running tasks to be called from the mainloop instead of running them directly,
void receiveEvent(uint8_t howMany)
	if (howMany < 1) { // Sanity-check return; } if (howMany > TWI_RX_BUFFER_SIZE)
		// Also insane number

	reg_position = TinyWireS.receive();
	if (!howMany)
		// This write was only to set the buffer for next read
		i2c_regs[reg_position] = TinyWireS.receive();
		if (reg_position >= reg_size)
			reg_position = 0;

void setup()
	pinMode(1, INPUT);

void loop()
	 * This is the only way we can detect stop condition (
	 * it needs to be called in a very tight loop in order not to miss any (REMINDER: Do *not* use delay() anywhere, use tws_delay() instead).
	 * It will call the function registered via TinyWireS.onReceive(); if there is data in the buffer on stop.

void readSensors()
	LDRvalue = (analogRead(A2)) >>2; // max value is 1023/4=255 is 1 byte (FF) physical pin3=PB4=A2
        i2c_regs[0] = LDRvalue;
	NTCvalue = (analogRead(A3)) >>2; // max value is 1023/4=255 is 1 byte (FF) pin2=PB3=A3
	i2c_regs[1] = NTCvalue;
	//pulsepin= pin6=PB1
	MoistPulse = pulseIn(1, HIGH);
	i2c_regs[2] = highByte(MoistPulse); // or use = MoistPulse >> 8;
	i2c_regs[3] = lowByte(MoistPulse); // or use = MoistPulse & 0xFF;

I have called the pulseIn function with an integer rather than with a long. This means that you will have to choose a value for R3 that gives a reasonable range for the type of soil that you use. A range of say 0-200 uS is very reasonable. Once you have done that you can also add a timeout to the pulseIn function. This should be abt 2 times the pulselength you expect (but depends on the duty cycle). With regard to the two variable resistors, they are in pull up so their value is Rntc = Rseries/((1023/ADC) – 1));  For the NTC this could be substituted in a Steinhart Hart approximation

The code to call the values (and that is loaded onto the master arduino) is even simpler:

#include <Wire.h>
void setup() {
  Wire.begin();        // join i2c bus (address optional for master)
  Serial.begin(9600);  // start serial for output

void loop() {
  for (byte i=0;i<4;++i){
  Serial.println(readRegister(i),HEX); // print the character
Serial.println(" ");

uint8_t readRegister(uint8_t regaddress)
	Wire.requestFrom(4, 1);


This code only prints out the values, you still need to combine the LSB and MSB from the cycle time and e.g. switch a pump based on the value. You could do that with this function:

int combine (byte lsbv, byte msbv)
	int value = msbv << 8;
	value = value | lsbv;
	//value= msbv<<8 | lsvb;//if you want to do it in one go
	return value;

Don’ t forget that the I2C lines need pull-up resistors of 4.7-10k.

In a next article I will present the construction of the probe itself.