LoRaWAN payload encoding

Proposal of April 30th, 2016. Thierry Schneider, Tetraedre Sarl

The maximum size of the LoRaWAN payload is 64 bytes. This limited size is the main reason to develop a flexible and compact data encoding. 

Tetraedre is active in the Internet Of Things technology since many years and has developped efficient and versatile data encoding.

The goal is not to be the most size efficient but to find a good trade-off between size, versatility and portability.

Header_main

The first byte of the payload is named header_main. This is the only fixed part of the payload. The description of this parameter is given below. The rest of the payload is a concatenation of data chunks. Chunks are either of type A, type B or type C. There might be any mix of chunks, in any order.

The two must significant bits of header_main are reserved and must be set to 0. Allowed values for header_main are thus 0 to 63. 

Chunk type distinction

Each chunk start with a header byte. This byte is used for data identification but also to know the size of the chunk.

Type A Chunk

The type A chunk is 3 bytes long. The first byte is the chunk header. The header is followed by 2 bytes

 byte8 is MSB

Type B Chunk

The type B chunk is 5 bytes long. The first byte is the chunk header. The header is followed by 4 bytes

byte24 is MSB

When data are transmitted in floating point the IEEE754 format must be used

Type C Chunk

The type C chunk is a variable size block. The first byte is the chunk header. The header is followed the size parameter which indicates the size of the data field

Type D Chunk

The type D chunk is 2 bytes long. The first byte is the chunk header. The header is followed by 1 byte

Content definition

 The content is defined with a combination of header_main and of header. The following table is a proposal of standard values.

if (value>=81)
    Volt = 4.2 + (value-80) * 0.1
else
    Volt = 1.8 + value * 0.03

If you intend to use this specification or if you have comments, please inform me.

Type definition

float16 type

The goal of the float16 type is to encode small and large values using only 16-bit. Of course this is not as efficient as IEEE754 but it already gives interesting possibilites. Only positive values are accepted. Values in range 0 to 181930 are converted.

The two most significant bits define a decimal range (0.001, 0.2, 1, 5) (exponent), while the 14 less significant bits defines the value (mantisse). Use the following code to decode the values:

left = (value >> 14) & 0x03;
value0 = value & 0x3FFF;
result = (float32) value0;

switch(left)
{
 case 0:
         result = result * 0.001;
         break;
 case 1:
         result = result * 0.02;
         result = result + 16.38;
         break;
 case 3:
         result = result * 5;
         result = result + 16725;
         break;
 case 2:
         default:
         result = result + 344;
         break;
}

Examples

Profile decoding

 Example index without delta:

01 80 5b 6d 63 b0 82 00 12 d6 87 ca 0b 00 43 2a 00 00 ff ff ff ff ff ff

01 main header

80 timestamp

5b 6d 63 b0 UNIX timestamp = 2018-08-10 12:06:40 GMT+2

82 serial number

00 12 d6 87 hexa =1234567 decimal

ca header = uncorrected gas meter

0b size = 11. Unit: bytes

00 conf = 0000 0000 = acq interval of 3600 seconds (1 hour) & no battery error & 0 = no other type of error

43 2a 00 00 index IEE754 Floating point MSB first = 170 decimal. Unit: m3

ff ff delta 1

ff ff delta 2

ff ff delta 3

Example with delta:

01 80 5b 6d 68 68 82 00 12 d6 87 ca 0b 00 43 34 00 00 02 58 01 2c 00 64

01 main header

80 timestamp

5b 6d 68 68 2018-08-10 12:26:48 GMT+2

82

00 12 d6 87 hexa =1234567 decimal

ca header

0b size = 11

00 conf = 0000 0000 = acq interval of 3600 seconds (1 hour) & no battery error & 0 = no other type of error

43 34 00 00 = index IEE754 Floating point MSB first = 180 decimal. Unit: m3

02 58 delta 1 hex value = 0000 0010 0101 1000 binary

left = (value >> 14) & 0x03;

this means: (0000 0010 0101 1000 >> 14) == 00

then apply the "and" mask

00

&11

---------------

result: 00

value0 = value & 0x3FFF;

0000 0010 0101 1000

& 0011 1111 1111 1111

result: 0000 0010 0101 1000

result = 0b0000001001011000 = 0x258 =0d600

result = (float32) value0;

result = (float32) 0d600;

result = 600.00

switch (left)

case 0

result = result / 100.0;

=> result = 600.00/100.00 == 6.00

Delta 1 = 6.00 m3

As last index = 180 m3

The previous index can be calculated as last index – delta 1 = 180 – 6 = 174 m3and it was the index reas 1 hour ago as conf == 00