Michael, who has already pulled single-bit audio from an Apple II and an IBM PC, turned his pulse-code modulation experiments on the ZX Spectrum. The result was a partial failure: PWM, which should outperform PCM in audio quality, did not work. Unlike the Apple II and IBM PC, the Spectrum has no dedicated hardware timer driving its buzzer. Every timing cycle runs in software, and that constraint is where this attempt breaks down.
What makes the writeup worth reading is not the outcome but the teardown of why it failed. Michael documents the multichannel music algorithm he did get running, and the gap between what worked and what did not reveals exactly how much the Spectrum's all-software architecture costs you in audio fidelity. The failure is specific and technical, not vague.
Michael states outright that more attempts are coming. If you want the full arc, his earlier Apple II and PC sessions are linked from his blog and show the same methodology applied to hardware that cooperated. The Spectrum piece is the most instructive installment precisely because it is unfinished.
[READ ORIGINAL →]