Vibrato effect messing up note detection about gbsplay HOT 5 CLOSED

The NOTE macro basically takes a sound generator frequency and calculates the corresponding note to it. The GameBoy has no notion of a note, a sound generator can be only set to a frequency which will then make a sound.
If you set the proper frequency, you can play a certain note.
If you add vibrato to your sound, the frequency will wobble up and down.

In order for the MIDI plugout or the on-screen display to detect a note, we can only take the current frequency, put it in the NOTE macro and see which note is nearest to that frequency. If the vibrato has a very high 'spread', the NOTE macro will oscillate between different notes. Because the algorithm is 'dumb', it will recognize this as different notes. (Also, if it were a trill instead of a vibrato, recognizing different notes would be correct.)

Other problems with note detection are:

• As you have no notes, you also can't really detect where a note starts or ends. Currently we guess based on volume changes and envelope triggers. While that works surprisingly well, it is by no means perfect.
• Sound effects: Nothing distinguishes sound effects from music. In the worst case we will try to map bleeps, blops and sweeps to notes because they also might be music.
• If somebody would make a sound routine that is not pitch-perfect (eg every note is played a quarter-tone too low, which the casual listener probably wouldn't even notice), the NOTE macro would fail miserably.

I currently don't have enough time to look at the IO dumps (I'd have to relearn everything, last time I worked with that part was some years ago). I don't want to rule out that our detection can be made better, but I think all of the easy parts are already done ;-)

But I have had a look at the file. You mean track 1 of subsong 1, right? I think I can see the problem in the console output.
Perhaps this is indeed a case of an out-of-tune track? You could try to add an frequency offset to the NOTE macro and play around until it fits like this:

#define FREQUENCY_OFFSET 123
#define NOTE(x) ((long)((log(FREQ(x)+FREQUENCY_OFFSET)/LN2 - MAGIC)*12 + .2))

You could wrap this in a script that iterates over different offsets and rebuilds gbsplay with -DFREQUENCY_OFFSET=xyz as needed ;-)

from gbsplay.

Thank you very much for taking the time to write such an elaborate answer! I fully understand. In the meanwhile, I looked a bit into the topic of musical temperaments and realized, this is much more complicated than I thought.

So maybe, after some time, I will understand the magic NOTE() macro and can tell, if that track is out-of-tune, it is trill or just vibrato.

Let's close this issue for now and should I ever get to the bottom of the problem, I'll add it then :-)

Oh and btw: Thank you for creating this software, it is not trivial at all and I think you made some people very happy! 👍

from gbsplay.

Have you tried the altmidi output plugin? It might give better results than the old midi plugout.

from gbsplay.

Yes I have, same result.

I believe the problem is with the NOTE() macro (which I don't understand unfortunately): It returns that a different note has started, while in reality it's the same note, just a little pitch-bend...

from gbsplay.

btw: I've just pushed 2a48a12, now there is only one NOTE macro (in common.h).
Up until now, the macro was defined in three different places…

from gbsplay.