PSG samples
You should try out PSG Sampler from Michel Shuqair 
nf, would you please introduce yourself in the who=who forum? I'm sure I'm not the only one interested in knowing more about you
Actually, our samples were 8-bit or even 12-bit. We used this method someone came up with the Atari ST: you've got three channels with 4 bit volumes, so each value of an 8-bit sample is mapped with a different combination of the three channels. I got the table straight from some Atari guy, so don't know how it works exactly.
Hi, nf! Long time no C Actually I asked this table from you few years ago, but I got only the replay routine. I anyway think, that I already figured that out (Correct me, if I'm wrong) as I think this is what the article in map.tni.nl/?p=articles/psg_sample.html is telling about the 5.5bit samples...
The volume controll of PSG seems to be linear, as you can directly play 4bit samples on it. If you use all 3 channels, they are just added to each other, so you get the range of 0-45. (It may be anyway more usefull to use these channels to play 3 samples at a same time.) Here is, example of the table, where N is the sample value (0-45) and A, B and C are the volume controls to each channel of PSG:
N: A B C
0-15 N 0 0
16-30 15 N-15 0
31-45 15 15 N-30
If you use this table, it is recommended to output the values to PSG in order of A,B,C or C,B,A to get minimum amount of unwanted click sounds.
As 0-45 is not any kind of vice table it is usefull to expand the table to 0-255, like it was made in your demos. This takes a bit more memory, but you can easyly use RAW samples created on any sample program without modifying the sample data.
If you would add also "Key click sound output" to the sample playing routine, you could get even more accurate sample playing, but I don't know if volume between PSG and key click is standard, so if it is not, then that might cause the sample to sound dissorted on some MSX models.
On SCC you have two different possibilitys... You can try to write the 32 byte sample RAM in same speed as it is playing, but I think, that it is VERY hard to get it in sync... or then you can use a bit modifyed version of the abowe mentioned method.
In SCC you have to fill the sample first with only one byte (#7F) to get the sound output up. Then you can choose sample rate. I don't know if the sample rate affects to volume setting in anyway, so pick just something.
SCC volume is not linear, so makeing a table for that will need oscilloscope (Or maybe PC with a good soundcard can do the trick). Anyway using different sample bytes on different channels, you can probably get the whole 8bit sample out of SCC, or maybe even more...
I have also heard very good quality samples played with FM-PAC, but I don't know details about that. I think, that the theory behind is about same, there is just more channels to use...
Note: All of this is just theory and I have not tried any of these methods.
~NYYRIKKI
I just checked out, that Amiga volume can be converted to dB using formula:
dB=20*log10(Volume/64)
I think, that this can be quite a close also with SCC:
dB=20*log10(Volume/16)
Don't shoot me, if I'm wrong
~NYYRIKKI
About the FM-PAC, apparantly you can set a bit somewhere in the test register, and that will enable you to write 6-bit sample data directly to the DAC. It is an undocumented feature of the FM-PAC. Although I don't know the exact details yet we intend to use this in our upcoming game aswell.
And thanks for the explaining text about '5.5-bit' samples and samples on the SCC. I'll add it to the MAP sometime (when I have more time ).
~Grauw