AudioStreamPlayer3D: Correct amplitude panning

[berarma]

The effective number of speakers was being calculated on uninitialized parameters. It worked by chance for 2 channels, but it fails for more channels.

Before this change the front/center speakers would have less volume and the side speakers would be unbalanced.

I don't have a multiple speaker setup, so my testing is limited to hearing only two channels at a time.

Effective number of speakers calculation for 7.1 before this PR:

channel 0: 4.000000 (front left)
channel 1: 4.000000 ( front right)
channel 2: 4.707107 (center)
channel 3: 3.146447 (rear left)
channel 4: 3.146446 (rear right)
channel 5: 4.000000 (side left)
channel 6: 3.500000 (side right)

And after:

channel 0: 3.853553 (front left)
channel 1: 3.853553 (front right)
channel 2: 4.000000 (center)
channel 3: 3.146447 (rear left)
channel 4: 3.146446 (rear right)
channel 5: 3.500000 (side left)
channel 6: 3.500000 (side right)

[goatchurchprime]

Holy cr#p! That's exactly the bug I found yesterday and was going to do a PR for today.

Were you using #103655 to force the other channels or do you have another method?

The reason it worked in stereo was that it had perpendicular speaker vectors whose dot product was zero.

[berarma]

Holy cr#p! That's exactly the bug I found yesterday and was going to do a PR for today.

Were you using #103655 to force the other channels or do you have another method?

The reason it worked in stereo was that it had perpendicular speaker vectors whose dot product was zero.

Wow, that's a coincidence. I came across this accidentally while reading code to understand how multispeaker setups were handled after doing some audio work for my theora seek PR. This code just jumped to sight immediately as a possible obvious mistake.

I had seen your PR when you put it in the chat, and I'm interested, but I didn't follow it. I don't know if it would have helped here. I've just modified the code to output the channels I wanted. But in the end, I relied more on the numbers than the output, just for confirmation that it didn't break.

I had sat on this because I couldn't find any reported issue that I could relate it to. Finally, I could do some tests today, and it confirmed that this was most probably not intended.

I'm definitely interested in some solutions to test different speaker setups. I'll test your PR and comment on it when I can.

[goatchurchprime]

Follow up in Rocket-Chat

[akien-mga]

Superseded by #104853. Thanks for the contribution!