fenarinarsa

Amazing that a simple technical question still fires the old Amiga/ST war The original Amiga has a 4 voices chip with independant volume/frequency setting. The mixing is done in hardware and all is DMA so it actually takes 0% CPU unless you want to manage a partition, like mod players do. One of the best thing on the Amiga is the sound quality : it sounds good even if you heavily change the frequency of a sound sample, which is I think because Paula is mostly an analog chip. Drawback : it's not really stereo, more like dual mono : channels 1 and 4 are on the left, 2 and 3 on the right, and you cannot change that. MOD is one of the music format created for Amiga, so it actually sticks a lot to this hardware : 4 channels, and the frequency data stored in the file are the values to write directly to Paula registers. The original Atari ST has an YM2149 chipset which cannot natively replay samples, but you could do it in software by changing the volume of 1 to 3 analog channels x times per second, x = the frequency of the sound played, and so the final quality of the sound. Let's be honest, it sounds like crap, but you can : - play a sample on 3 channels (best quality) - play a sample on 1 channel for instance, keeping the others with regular chip sounds (used in digidrums, or sound effects in some games like Wings of Death) Because you have to push data x times/second to the chipset, it can take a lot of CPU. You can also mix in software many sounds in one final sample and play it the same way. So you can actually mimic the Amiga by doing soft mixing (for example to replay a MOD), getting a final sample to be replayed on the YM2149. The higher final frequency you want, the best quality you get, and the more CPU you need. I think (not sure) the higher you can get is around 25kHz by using various optimization techniques. You can also mimic two or more channels, mix them and play them on 1 YM channel instead of 3. That keeps 2 channel for the chipsounds and 1 to play multiple digital sounds. Or 1 digital sound per channel, etc. But if you use less channels to play a digital sound, the quality is lowered. Example : Reset screen of The Froggies over the Fence, digi sounds start at 7:40 The STE has two additionnal sound chips : - A standard DAC chip that can replay 8-bits PCM digi sounds, mono or stereo, at 6/12/25/50kHz, with DMA - A mixer chip that allows to set the volume between YM and DAC, and to adjust treble/bass The treble/bass settings are quite useless IMO since changing them distort the YM output. It gives good results if you're using the PCM output only. The volume settings can be useful if you mix YM & PCM, and some MOD players use it as an optimization trick to avoid applying software volume on 1 or many channels. So it's not made to have an Amiga-like chipset but to output a single PCM sample, as it's the case on computers nowadays. The mixing should be done in software if you want to replay multiple channels. My guess is that Atari did that because they didn't have the R&D to make a custom chip, and also it was cheaper. Since this PCM chip is DMA, the CPU is relieved of all low-level replay task, and it takes near 0% of the memory bus (like on Amiga). But if you want to replay a 4-channels MOD, you need to mix in software, so you end up with an inferior quality than on the Amiga because the mixing is digital and not analog, unless you use high-quality samples. On the other hand with replay at 25 or 50kHz you can get crystal-clear sounds with high quality sources. You mentioned Talk Talk 2, this demo does NOT use a looping sample. It's a multichannel song with one part digital, on part analog (like maybe 2 channels digital, 3 channels analog). Note that when doing this you must keep the YM sound level quite low because the PCM sound level is really lower than the native YM output, surely because the YM outputs a really raw and distorted square signal. But that's true a lot of STE demos use looping samples to avoid software mixing and then keep CPU for the visual effects. Since they are often made to fit on floppy disks or in 4MB RAM without streaming them from HDD, the samples are often low quality and loops a lot. ​ And finally, the MOD software mixing technique developed on ST also had a use on Amiga. Jochen Hippel (Mad Max) gave his ST mixing routine to Chris Huelsbeck which used it for TFMX7 : 3 regular Amiga channels + 4 extra channels mixed in software But as it eats CPU like it does on ST, Huelsbeck limited the use of 7 channels to game intros.