Closed LeonardoDemartino closed 3 years ago
So actually, this is a bit subtler than you suspected. An LFO value of 0 does not really disable the LFO. According to the YM2151 application manual (see excerpt below), an LFO speed of 0 is 0.0008Hz, or a period of ~20.8minutes. What you're hearing is totally what I would expect, unless something else is causing the LFO to be reset.
Which does seem to be the case. The YM2151 test register is known to reset the LFO when bit 1 is set. I had implemented that, but only as an instantaneous effect when the register was written. Your VGM writes to the test register and sets the 1 bit once at the start, but clearly expects that bit to be "sticky", so that it should disable the LFO until it is cleared. Implementing this gets rid of the LFO.
I discovered that even if you set the LFO speed to 0 (disable LFO), you actually get a very slow LFO on ym2151. Which means that notes will begin to detune after a while.
Doesn't happen on other emulators: https://deflemask.com/ym2151_lfo_bug.vgm
Check this .wav: https://deflemask.com/ym2151_lfo_bug.wav (If you scroll around the detuning will be obvious)