The last days I’ve been playing around with my JeeLink and a JeeNode including three sensors. I made a JeePlug with the TMP36, an LDR and a TSOP infrared receiver, which I can plug into my JeeNode. So I don’t have to plug everything in a breadboard and connect the wires, just place the JeePlug on top of my JeeNode in ports 1 and 4. I use port 1 for the temperature (TMP36) on the analog pin and I use the analog pin of port 4 for the LDR. The TSOP is connected to the digital pin of port 4 which leave the digital pin of port 1 empty to be used in the future.
The goal is to create a WSN based on several JeeNode and a Jeelink as central node. The JeeLink/JeeNode have a RF12Demo sketch with which it is possible to easily create a mesh network. The JeeLink is connected to a computer through USB. JeeNodes can communicate wirelessm using the on board RF12 chipset. All you have to do is configure the two units s that they can ‘find’ each other. See POF03 @ JeeLabs for an instruction on how to do that.
After I made the setup of my nodes and got the sketch started I saw data being displayed in my serial monitor. The JeeNode (sensor node) displayed information about the light and temperature on screen and sent the data to the JeeLink. I saw valid messages in the serial monitor of the JeeLink. The data shown by the JeeLink was something like:
OK 33 103 113 175 65
This line contains the measured temperature sent by the JeeNode. I knew that the value measured and sent was related to 21.93 degrees Centigrade. So somewhere in the sequence could be found. When you take a look at the Arduino Reference page about float (floating-point variables) you find that a float is represented by 4 bytes. So 4 of the 5 numbers in the message will together form the floating point number.
I wanted to know what 4 bytes were sent by the JeeNode so I did put in some extra Serial.print statements to display not only the measured temperature of (in this case) 21.93 degrees but I also the seperate bytes. To do this I had to ‘convert’ the float into an array of bytes. The lines below show you how:
float temperature; byte *ArrayOfFourBytes; temperature = 21.93; ArrayOfFourBytes = (byte*)&temperature; Serial.println(temperature); // print the temperature Serial.println(ArrayOfFourBytes,BIN); // print the 'first' byte in binary notation Serial.println(ArrayOfFourBytes,BIN); // print the 'first' byte in binary notation Serial.println(ArrayOfFourBytes,BIN); // print the 'first' byte in binary notation Serial.println(ArrayOfFourBytes,BIN); // print the 'first' byte in binary notation
The printed binary numbers were:
In decimal notation equal to 103, 113, 165 and 65 which are the last 4 numbers of the received message. This still doesn’t look like 21.93 but we’re getting there.
A float is represented by 4 bytes which are equal to 32 bits. The printed binaries above are not all 8-bits but three of them are 7-bits. This is because the preceding zero’s are left out and you have to place them yourself in front of the 7-bits.
Al together these 32-bits give you three pieces of the puzzle; a sign, a exponent and a mantisse.
The formula for a IEEE 754 float is: sign * 2exponent * mantissa , where sign is +1 or -1
There are many of pages that explain more about floating-point numbers and their representation in bits, together with formulas. You can also find websites with converters that let you enter a number in the different representations and show that number in a binary, hexadecimal, float or decimal representation. I used this website.
I entered 21.93 in the converter and got 01000001 10101111 01110000 10100100 (in bytes this is equal to 65 175 112 164). This isn’t equal to 103, 113, 165 and 65; except for the 65 which might be a coincidence or not.
When I entered the original bit sequence with preceding zeros 01100111 01110001 10101111 01000001 I got the float number 1.1413232E24 and not 21.93.
I know that there are more possibilities of ordering and combining bytes to get e.g. a float. You have big-endian and little-endian (see Wikipedia about Endianness). This has a relation with the hardware platform you are using.
So I tried the sequence: 01000001 10101111 01110001 01100111 (which takes the ‘last’ byte first, then the third, second and first byte). The converter now gave me the floating-point number 21.930372 which is not exactly 21.93 but close enough to continue.
The explanation for the difference between 21.93 and 21.930372 can be found on the Arduino Reference page about Serial.print. Where the second example shows: Serial.print(1.23456) gives “1.23”, in other words the float is rounded or capped to two decimals.
This difference also explains the difference in bit sequence returned by the converter.
Knowing this I can continue sending, receiving and translating data to build my own WSN solution with central reporting functionality.