Pulse Sounds..

[andym00]

Well, actually more about synthesising lots of software voices, but with little CPU overhead, well much less than traditional means, and the bonus is the voices are playing back at 15.6K.. Really

It's an idea from a long time ago, and after a few silly ideas ideas ended up with something useful..

So, without further ado, here's a demo of it for the unexpanded VIC..

In the pacfollin.zip at https://sites.google.com/site/andym00/files

There's the original development of the idea here:
http://formatwar.net/forums/viewtopic.php?f=5&t=265
Although it started off on the Atari A8 since I really didn't think the ~1MHz machines would have the power for it, but a revelation in being able to (essentially) compress the output samples so we have way less interrupts than usual due to the nature of the pulse waves we're synthesising made for a game changer..

In this demo, the colours bars off the irq at the bottom of the screen do nothing other than..

Code: Select all

IRQ:
			lda $9124
			dec $900f
			ldy #$0
-
			inc $900f
			inc $900f
			inc $900f
			dec $900f
			dec $900f
			dec $900f
			dey
			bne -
			inc $900f
			jmp	$eb18

It's simply there to burn cycles and show how much time left in the main loop for synthesis is required, and how much the interrupts slow down the code..

Also this version doesn't due the full NSF driver emulation.. It's simply playback a cached version of the registers, even so, this pacfollin tune still only comes to 1.5K.. Huge I know for what it is, but there's no way I can get the NSF driver working on the VIC due to the memory map.. But since using the NSF driver was merely a stepping stone to writing a proper music driver for this it did the job nicely..

I've also done a 64 version, which is exactly the same bar the VIAs make it some cycles quicker (and less memory) to setup the time to the next sample, it also show you more how much cpu time it actually uses..
http://noname.c64.org/csdb/release/?id=106466

And the code is there as well for the core of it..
http://noname.c64.org/csdb/forums/?room ... icid=90296

It's a lot less CPU time than you might imagine, and doesn't sound too bad, although the VICs volume register doesn't sound as nice as the SIDs..

[FD22]

Impressive!

[Witzo]

I ran

https://sites.google.com/site/andym00/f ... ects=0&d=1

and it sounds remarkably un-VIC-like; that these things are even possible.

Where is the music from?

[andym00]

Well the whole thing was initially just an idea of a way to synthesis some additional voices without the normal massive cpu-overhead required..
To test it out, rather than writing a full blown music driver in parallel, I simply grabbed so NES NSF music files, and then wrote a later to convert the 3 main channels to my driver..
Unfortunately on the VIC there's not enough memory to do this, so I just captured the tunes APU register output, and stored that, simply to get it working..
The code itself is tiny for the playback system, and the CPU overhead would really surprise you
And once it has a proper music driver in place, rather than the NES NSF player, things will get brighter, and less memory intensive..

[Kweepa]

Pretty cool!
I don't quite understand. Are the borders indicative of how much CPU is free? If so is it the blue bit or the multicolour bit that is the free time?

[andym00]

Erm, it's not clear is it, but I didn't have the memory for a fancy CPU load like the 64 version..

The multicoloured bit is just a fixed loop (of ~10000 cycles) run from a per frame interrupt.. It's there to show how much variance there is from the interrupts running in the background, hence why it varies in time..

The blue bit is just how much CPU is free after the interrupts fixed timing thing has finished.. The actual synthesis doesn't take all of that time, although the load varies depending upon the frequencies of the notes being synthesised..

I should have put in some better cpu load display, but for some reason I really wanted to get this into an unexpanded VIC, for reasons I'm not entirely sure of myself

[Vic20-Ian]

Sounds good, tried it on Vice and get some fuzzy bars top and bottom and a short loop of what sounds like a Pop remix of Pacman!

I agree it is very unlike a Vic.

Try datapop on this forum also.

http://sleepingelephant.com/ipw-web/bul ... ht=datapop

[matsondawson]

Here's the JS Vic20 link for those to lazy to download and vice it.

http://www.mdawson.net/vic20chrome/vic2 ... eType=ntsc

Took me ages to figure out why it didnt' work in NTSC turned out to be the branch not taken interrupt ignoring feature of the 6502. Oddly enough it really sucks the cpu emulating it. But I have things in motion for a 5+ times speed up. It should run 100% speed on Android when I'm done.

[Jeff-20]

I just love that online VIC script so much!