questions on the dpcm channel

Hi,
I am quite a newbe at this, so please forgive me if I am asking stupid questions.
I have been playing with MCK, which is a nice little program to write instructions to an .nsf file, using a .mml file.
Using a nice little tutorial I found, I am able to program the two pulse, triangle and noise channels. While I can get the DPCM channel to work, I have three questions about its limitations and how to program it more than I already have, and one question about a playback problem that I am having.
First, suppose I sampled a short sound, a synth bass G. In the mml file, is it easy enough to say "okay, this nsf is a g," and then tell it to play an f, c, etc as needed? If so can you give me an example of how this would be done?
Second, I read that after using 8 kilobytes of memory, the bank is full so you have to switch banks. So if I had a dcm file that was 100 bytes, every time I used it I would be adding 1000 bytes to the bank?
Third, Let's say I have two wav files: a kick, and snare and have converted them to dcm. The dcm files have a sampling rate of 33.4KHZ. The kick sample is 353 bytes and the snare is 1153 bytes. They work fine, but are they too big? Something tells me they should not be this big and I should probably shorten the wavs and use a lower sample rate in DMCconv. Any guidelines on what I should and should not be doing?
Fourth a little off topic, but when playing my snf files with DPCM using the not so fat so plugin in Winamp, there is a lod of DC offset, which fortunately not so fat so is able to elliminate. I've heard that sometimes clicking problems come up with DPCM, I am having none of those at the moment. Are these circumstances normal? Should I try other players?
Any help you can give is greatly appreciated!

I know very little about MCK, but I do know about sequencers and trackers in general.

Yes, but you have to be careful. There are sixteen different valid sample playback rates, from 4.2 kHz to 33.5 kHz or so, and these represent various intervals relative to the original sample rate.
Say you have a C sampled at 4.2 kHz. You can play it as C, D, E, F, G, A, B in one octave, C, D, F, G, A in the next octave, C, E, G in the third octave, and finally C.
But if you record a B♭ sample, you can play it as B♭, C, D, E♭, F, G, A, then B♭, C, E♭, F, G, then B♭, D, F, B♭.
Sunsoft games use five bass samples to be able to play all notes.


For each sample that you use, the NSF contains one copy of each sample and then a couple bytes in the sequence every time you play it back.


The formula is one byte for each 8 samples. A 1153-byte snare has 9224 samples and thus 275 ms. Unless that snare sample has Phil Collins style gated reverb, it probably has quite a bit of decay that won't translate well to DPCM; you usually want to cut the original wave about when it hits -36 dB.


I wrote this program and ran it on a PowerPak to demonstrate that either my NES or my sound card also has a high-pass filter that kills DC. If your favorite NSF player doesn't, you can apply a high-pass filter in Audacity with a corner frequency around 30 Hz.


DPCM clicks happen when you abruptly change the DAC's center point. It only happens in those music engines that reset the DAC to center (like lda #$40 sta $4011) before each sample, and happens more often if

1. you play different samples with different center points,
2. you cut samples off before letting them go back to center, or
3. your music engine sets the center point back to the negative rail when NoteOff-ing a sample, like Famitracker.

Without specific experience in MCK, I can't help you much there.