I noticed in 8bit Power Music Final and kirby's adventure that the triangle channel is softer than the other channels. I am using a powerpak to play them.
I don't personally know about "8bit Power Music Final", but Kirby's Adventure uses DPCM samples that tend to leave the DAC level quite high, which reduces the volume of the triangle and noise channels. Maybe this was intentional because the game tends to use the triangle for non-bassline parts.
Quote:
DPCM samples that tend to leave the DAC level quite high, which reduces the volume of the triangle and noise channels
I don't fully understand this effect, and the wiki doesn't seem to address it.
Does anyone have a link to the technical details of reducing triangle volume with DMC level?
http://wiki.nesdev.com/w/index.php/APU_Mixer
Basically the three channels tri/noise/DMC share a single digital sum output, but the DAC that's connected to is a nonlinear curve. Low output values are farther apart, high values are closer together. This means that if the DMC output value is high, the effective volume of noise and triangle is lowered. (SMB and Startropics both intentionally use this as a crude way to make the triangle a little quieter. Many homebrew NSF composers have done the same.)
The pair of squares have a similar problem on the other DAC.
The exact nature of the curve I don't think has been accurately documented. Blargg's old approximation doesn't seem quite right to me, but it's widely used. I'm planning to start measuring it thoroughly in the near future, actually.
Basically the three channels tri/noise/DMC share a single digital sum output, but the DAC that's connected to is a nonlinear curve. Low output values are farther apart, high values are closer together. This means that if the DMC output value is high, the effective volume of noise and triangle is lowered. (SMB and Startropics both intentionally use this as a crude way to make the triangle a little quieter. Many homebrew NSF composers have done the same.)
The pair of squares have a similar problem on the other DAC.
The exact nature of the curve I don't think has been accurately documented. Blargg's old approximation doesn't seem quite right to me, but it's widely used. I'm planning to start measuring it thoroughly in the near future, actually.
I always knew that using dmc affected some channel but didn't know why. Now I know, learning new things everyday 
I guess the next obvious question.
Should I set $4011 to zero to make triangle notes louder?
Will this cause a popping sound?
Should I set $4011 to zero to make triangle notes louder?
Will this cause a popping sound?
Decreasing $4011 will make triangle and noise louder. An immediate write to $4011 (such as using the Zxx effect of FamiTracker) will make a popping sound, but ramping $4011 down to 0 over time, such as by playing an all-$00 DPCM sample, won't.
Well, any change in the DMC output causes an audible sound, but if you do it slowly the volume is low. You can kinda hide the DPCM sound a little if you pick a samplerate that's harmonically in line with your music.
There was a thread about that: https://forums.nesdev.com/viewtopic.php?f=6&t=17302
There was a thread about that: https://forums.nesdev.com/viewtopic.php?f=6&t=17302
So I'm not crazy then. I have a pretty good ear from playing classical music and from form analysis.