I must be missing something.

First do we mean the same thing by "repeat"? You hold the key pressed for a longer time and want to get the same behavior as the original remote would have produced for a long press.

I based my entire understanding of how your original remote does long presses on the third learn you posted for each of Power and Enter. In each case, near the end of the signal (beginning of last line as viewed in my web browser) it has the pattern 0164 0059 0016 0EF5, which is very close to the NEC1 repeat pattern, so I assumed NEC1 repeat was correct.

In my .irp file I left the repeat timing from NEC1.IRP, which generated 0168 005A 0016 0F47 twice per signal as the repeat.

The trivial differences in value shouldn't matter, but there is a long shot chance that they do. You might try replacing the two copies of 0168 005A 0016 0F47 at the end of one signal with two copies of 0164 0059 0016 0EF5 and see whether that makes a difference.

The more likely theory is that your original remote uses two different systems for long presses. For keys like Power and Enter, where long presses shouldn't repeat, it uses NEC1. But for keys like VOL+ etc. where a long press should repeat it uses NEC2. In that case, something closer to what you did before I started helping would be correct.


I've seen other devices mixing NEC1 and NEC2 this way, but they are rare enough that I decided not to mention that complication earlier.

It should be possible in a few tries to get a bad enough learn of one of those keys that it isn't recognised and translated, but good enough that the repeat rule can be determined.

It is easy to modify the .irp file to generate exactly the repeat structure of your first successful test, or generate whatever we think the repeat structure really should be once we see some learned signals.