Miscalculations in Volume and Modulation Envelopes about fluidsynth

[derselbst]

Original comment author: Mr_Bumpy

I couldn't attach the file because it was too large, but you can find it here.

Original comment by: elementgreen

[derselbst]

Original comment author: diwic

Hello Christian,

sorry that nobody has taken the time to investigate this bug further, now that you have put so much effort in creating good test cases. I'm new to that part of FluidSynth, but I thought I would give it a go and see if I could fix the bugs you are experiencing. I started by going through the math behind the first one:

-2400 cents = two octaves down

Sustain of 13% means 13% off the top level, 100% - 13% = 87% remaining.

87% of -2400 = -2088

Then the desired behaivour is an additional 7 semitones (i e a perfect fifth up), so total adjustment should be 27 semitones and 88 cents, which is what you have in the "Soundtrack FS Fix" instrument.

I have a hard time seeing what FluidSynth does wrong, or if my calculation above are wrong in some way, or if there is something else affecting the pitch...? Can you help me out?

Original comment by: elementgreen

[derselbst]

Original comment author: Mr_Bumpy

My apologies... it appears as though it is the Audigy that is doing the miscalculation; all of my measured tests show FluidSynth to be 100% what the values should be.

In the Soundtrack example, I have the modenv->pitch set to -2400 cents. The Audigy, however, seems to not go any lower than -1200 cents, so it uses it's actual limited value for performing the calculations. So you end up with:

FluidSynth:

set value: -2400
calculation: 13% of -2400 cents

Audigy:

set value: -2400
calculation: 13% of -1200 cents

Setting the modenv->pitch to -1200 in the SoundFont provides identical playback for both Audigy and FluidSynth.

For the sweep pad, I discovered two factors that create the large difference in playback between the Audigy and FluidSynth:

1. The Audigy's mod envelope does not take into account the difference between actual and audible linear scaling. For example, 100 to 200 hz is an octave, but 200 to 300 hz is only half an octave. The Audigy scales numerically linear, and the audible result is that the pitch/filter moves very quickly at low frequencies and very slowly at high frequencies. FluidSynth scales based on the perceived sound, so it covers 100 to 200 hz in the same amount of time it takes to cover 200 to 400 hz. FluidSynth's method is preferred, IMO, but it does change the percieved filter/pitch of the sound compared to the Audigy.

2. The Audigy just plain seems to miscalculate some things, and I haven't completely figured out why. For example:

FluidSynth:

instrument-level attack: 0.1 sec
preset-level attack: x60
resulting attack: 6 sec

Audigy:

instrument-level attack: 0.1 sec
preset-level attack: x60
resulting attack: 10.5 sec (estimated)

I don't know why the Audigy gets its results, but I can't see any reason why it would be correct in doing so.

Original comment by: elementgreen

[derselbst]

• status changed from new to closed
• resolution set to invalid

Original comment by: elementgreen