no not cleaner

Hello, alexeenko there they simply selected the original sample itself to make it sound better - but in general the sound is worse. I immediately looked through the sources - who saw my article about sound on ZX will also look and understand that there are shortcomings there - cycles on registers (DEC C JR NZ, $-1) and so on. My algorithm gives a very clean sound, despite the fact that there are such terrible there are no slow constructions at all, there the cycle clock is 1 nop (i.e. 4 tact). And yet, in this demo it is not indicated how this thing was pressed for speck, it is quite possible that they used the flawed Error Diffusion, which however somehow it works ;) Luckily I'm working on a terribly fast algorithm... but here the memory is simply limited zero is missing... I don’t know what to do... I’ll post samples like a thread...