Well, it is 512 KB for 30 sec of recording. I think it is doable. I remember a program I've found on a magazine (in its cassette) when I was young... ehm, younger.
It was able to direct play or record sounds (and then replay them) from the datassette. And it was damned good (well at least in my memories, maybe I was too excited for it at that time)
The technique is the one described by Grawu but that solution is limited by the unwanted volume transitions due to the fact the 3 channels cannot change at the same time. The solution here minimises the unwanted transitions by the search across all possible sequences of volume changes on the 3 channels.
hi, i didn't check. Did not the solution already use the viterbi optimization?
if not it would be exceptional to test the same song with the PCM Encoder, maybe one could get even better quality!
It's not a fair comparison, GX4000/CPC+ range have a DMA sound feature that allows to feed the PSG with near to zero CPU cycles wasted. An Atari ST can also get cool effects with the same PSG thanks to its powerful processor and architecture