YM2151 Fine detuning

[kliepatsch]

Hi Folks,

I've got another pesky question about the YM2151. I have played around a little with the parameters and I would like an opinion if the following is the correct/normal behavior of the chip.

DT1 (fine detuning) is a 3 bit value, i.e. there should be 8 levels of detuning. However, it seems there are only four, so the bottom 4 and the top 4 levels have the exact same effect. The most significant bit is ignored.

Moreover, the detuning is quite underwhelming. At max setting (which is 3 or 7 for me, doesn't matter), the phase difference between two operators is about 1 wave duration per 3 to 4 seconds. In other words, if two sine waves are added, and one of them is detuned with DT1=7, those two waves cancel each other out once per 3 to 4 seconds.

Edited July 31 by kliepatsch

[Elektron72]

DT1 uses sign-magnitude representation to allow for both positive and negative detunes. Values 1-3 are positive detunes, while 5-7 are negative. Setting one operator's DT1 to 3 and the other to 7 will produce a more powerful effect.

[ZeroByte]

It’s using 1’s compliment so 4 is the same as zero (negative zero).

The effect is going to vary a lot depending on which ALG you’re using and what pitch is being played.

[ZeroByte]

For anyone interested, when adding two tones, the “warble” you hear is the two waves periodically being in phase or 180deg out of phase. The frequency of this warble (how fast it warbles) is the difference in the two “carrier” frequencies.

So if you have two tones at 440hz and 443hz, the warble is going to be 3hz - 3 warbles per second.

The easiest con alg for hearing this is alg 7 which simply outputs all 4 oscillators directly with no FM effect at all.

[kliepatsch]

55 minutes ago, Elektron72 said:

DT1 uses sign-magnitude representation to allow for both positive and negative detunes

Thank you!

[kliepatsch]

1 hour ago, Elektron72 said:

Setting one operator's DT1 to 3 and the other to 7 will produce a more powerful effect.

Well, I think it should go all the way up to 11

Edited July 31 by kliepatsch

[ZeroByte]

But… why can’t you just make ten the top number, and then go up to that when you need a little more?