Oh, just beeps, but I think I could get it to play notes pretty easily now. Before, I did it on a very old version of Grammer that had really crappy sound support. It might actually go well for back ground music in my checkers game since a lot of it is waiting for user input o.O

Can't we just bind a timer to the port so that the program doesn't freeze every time a note has to be played? Also, I made this a while ago:
www.youtube.com/watch?v=7kK_vRbiibc (no comments about my voice please )
It plays the sfx on another calc so that they won't cause the game to freeze for a while.

Can't we just bind a timer to the port so that the program doesn't freeze every time a note has to be played? Also, I made this a while ago:
www.youtube.com/watch?v=7kK_vRbiibc (no comments about my voice please )
It plays the sfx on another calc so that they won't cause the game to freeze for a while.

Yeah I remember that and it was great. Of course that wouldn't work for everyone, though, since not everyone can afford to buy a second calc, let alone carry two calcs to play a game.

Can't we just bind a timer to the port so that the program doesn't freeze every time a note has to be played?

Sure. The the calculator totally has a southbridge that can handle DMA. It even has a PCM wave buffer.

OK, sarcasm aside, sound synthesis is difficult because it has to be done via pure bit banging. You could set up a crystal timer interrupt to handle changing the port status at a fixed rate, but it will sound terribly out of tune. The reason is that the maximum frequency that you can get interrupts to fire at is 32768 Hz. As a result, you have to quantize the period of each note to intervals of ¹⁄₁₆₃₈₄ of a second. The human ear can hear frequency distortions of less than 1%. With such a synthesis technique, the error from the correct frequency will begin to exceed 1% around middle C.

If you're wondering, the quantized frequency of a frequency f_n quantized to a half-frequency of f_s (here, 16384) is f_q(f_n, f_s) = f_s/round(f_s/f_n)

The ideal frequency of piano key m is f_m(m) = 440 * 2^((m-49)/12)

The percent distortion for a particular note n quantized to a half-frequency f_s is f_e(n, f_s) = (f_q(f_m(n), f_s) - f_m(n)) / f_m(n)

Go ahead and graph this on your calculator and you'll see the graph cross the 1% line for the first time at G#3.

Incidentally, this is why even simple delay-loop based music sounds slightly distorted. It's not the fact that the frequencies will differ from the ideal by a small percentage, it's the fact that those percentages are all different that makes the music sound out-of-tune. (Also, the square waves take some of the blame.)

Actually, it might be better if I make this compatible with the MuseInc format >.>