decodeGeneric receives and dumps correct input, fails to decode a value

[lhk]

I am trying to implement my own protocol by using sendGeneric. Therefore I created a list of definitions on both sender and receiver side: const int numBits = 2; const unsigned int Head_Mark = 1500; const unsigned int Head_Space = 800; const unsigned int Mark_One = 2000; const unsigned int Mark_Zero = 1300; const unsigned int Space_One = Head_Space; const unsigned int Space_Zero = 700; //ignored const unsigned int kHz = 56; const bool Use_Stop = true;

The signal uses varying marks. The invariant Space_One=Head_Space is fulfilled. It sends only two bits.

This code is used to send two signals. The first is the value 2 with the bits 1 and 0. Therefore the expected output should contain a head, then a mark of 2000, then a mark of 1300.

Here is the serial dump: Decoded Unknown(0): Value:0 (0 bits) Raw samples(8): Gap:3968 Head: m1400 s900 0:m1900 s900 1:m1200 s800 2:m1900 Extent=9000 Mark min:1200 max:1900 Space min:800 max:900

The next signal is 3 with two 1 bits. The expected output are two marks of 2000. Decoded Unknown(0): Value:0 (0 bits) Raw samples(8): Gap:42284 Head: m1400 s900 0:m1900 s900 1:m1900 s900 2:m1900 Extent=9800 Mark min:1900 max:1900 Space min:900 max:900

As you can see, both signals are transmitted correctly. There are small variations in the lengths of marks and spaces, but they seem negligible.

Nevertheless IRLib doesn't decode the value. In both cases the method decodeGeneric returns 0 and the value of the decoder is not changed.

How can I fix this ?

I am very sure that my setup is working correctly, since I can transmit values with the built-in protocol Panasonic_Old which appears to be the only builtin protocol which transmits at 56kHz. That's the frequency needed by my receiver.

[cyborg5]

When I first rewrote IRLib from IRremote one of the main features was the sendGeneric and decodeGeneric methods which eliminated a lot of duplicate code. It did however get a little bit complicated trying to handle variable marks and variable spaces in the same function. With much more experience I have discovered that Sony is the only protocol of about 50 I've seen that uses variable marks. In an upcoming version of IRLib the Sony protocol will no longer use the generic routines and the generic routines will only be able to use variable spaces. So for future reference you might want to either change your homemade protocol to use variable spaces or you will have to write your own send and decode routines.

In the interim I've noticed that you use a stop bit of true but Sony does not. All of the other protocols that use variable spaces use a stop bit. That may be your problem. Try turning off the stop bit.

Also try uncommenting IRLIB_TRACE in IRLib.h at approximately

39. That will turn on some debugging information that will tell you why the decoder rejected the data.

[lhk]

Thank you for the quick answer. I reworked my marks and spaces, now with different spaces. Didn't work. I turned on the debugging information and read your code again. Apparently the receivers RawCount is not necessarily the number of bits that was sent. For the 2 bits in my two test scenarios the decoder registered 8 samples. For 3 bits that became 10. And 4 bits create 12 raw samples.

Seems like a linear dependence. Is that correct ? RawCount = numBits*2 +4 ?

But after reading the code I think it should be +3: Head Mark and Space = 2 bits, mark and space for each bit = 2_numBits, stopBit = 1 bit => numBits_2 +3

With these changes the receiver can decode the transmission.

[cyborg5]

Yes you have one entry is the gap, 2 for the header, and one for the final bit means there are four extra. And it takes two counts per bit Mark and a Space. You actually compute bits from the raw count. In decodeGeneric look for bits= Mode however inside that method the count is a variable called "offset".