The second and third 4-digit numbers in Pronto Hex give lengths (in pairs) of the two sections that follow. The two sections are "one time" and "repeating".
The 0026 in your bad learn means 26 hex (38 decimal) pairs in what the Pronto (incorrectly) thought was the one time part. Those 38 pairs are:
0081 0041 0010 0010 0010 0031 0010 0010 0010 0031 0010 0010 0010 0031 0010 0010 0010 0031 0010 0010 0010 0031 0010 0010 0010 0031 0010 0031 0010 0010 0010 0031 0010 0010 0010 0031 0010 0031 0010 0031 0010 0031 0010 0010 0010 0010 0010 0031 0010 0031 0010 0010 0010 0031 0010 0010 0010 0010 0010 0031 0010 0010 0010 0010 0010 0010 0010 0031 0010 0010 0010 0031 0010 0031 0010 0010
The 0004 in your bad learn means 4 pairs are in what the Pronto thought was the repeating part. Those four pairs are:
0010 0031 0010 0010 0010 0010 0010 002B
If I had trusted that the learning error was that simple, I could have constructed the repaired signal from just that information. I recognise Kaseikyo protocol and know that there are actually 50 pairs in the correct signal. Your Pronto learned it as 38 pairs followed by 4 pairs repeating. So the repaired signal was those 38 pairs followed by three copies of the 4 pairs, and with the header adjusted to indicate a one time part of 50 pairs (0032 hex) and a repeat part of 0000. I also adjusted the final value (second number in the 50'th pair). Most Pronto's ignore that number when the repeat part is 0000, but just in case, I wanted to make sure it wasn't too small.
At first look, that reconstructed signal had a value I didn't expect in the one 4-bit section of Sharp Kaseikyo whose meaning I don't know (but which usually has the same value all the time). That made me guess incorrectly that the learning error was more complex. Also, there are several air conditioner IR systems that take a long protocol, such as Kaseikyo and extend it to be even longer. So I thought the correct length was 50 pairs, but I wasn't confident of that.
The two good learns you posted have exactly the same value in that 4-bit field as the reconstructed signal, and they are 50 pairs (0032 hex), which made me decide the reconstruction was probably correct.
Last edited by johnsfine
on August 15, 2006 14:21.