video upload by Joe Landers
B52: 303 Growler
B00: Space Adventure
https://github.com/joelanders/profligacy

"This is a low-level emulation of the hardware in the Korg Prophecy.
I'm not affiliated with or endorsed by Korg or anyone else.
You bring your own ROM dumps of the original firmware (2 chips), and this thing runs it.
I spent some months probing the NEC V55PI, Hitachi H8/3003, and (3x) TI TMS57002 chips. I got all the inter-chip UART and DSP host interface signals (that I observed) matching in time and content exactly.
For example, the end-to-end signals are like: MIDI -> UART -> DSP host interface -> digital audio.
Demonstrating accuracy of the digital audio is a more complicated story because of the free-running oscillators. I had to do things piece by piece: render some long sustaining oscillator waveforms through the emulator and hardware, line them up by a phase offset. They still won't be sample exact, but the hardware-to-hardware error is consistent with the hardware-to-emulator error.
DSP2 and DSP3 were easier: I could record the inputs and outputs of the hardware chips (4 channels in and 4 channels out), play back the inputs through my emulated chips, and ensure that the outputs matched.
Additionally I could run the original firmware's DSP programs on a standalone DSP board, which let me control the phase of the oscillators. I don't know if this is an easier claim to believe than the in situ logic probing. In reality, I was coming at this from a lot of directions, and eventually I fixed all the discrepancies.
I'm not modelling the post-DAC analog parts at all because the service manual says the frequency response is supposed to be flat to within 1 dB across the audible band. (Annoyingly, I found my hardware unit isn't flat, so I might have to replace some 30-y.o. capacitors or something. I'll let you know if it turns out to be the magic analog warmth.)
...
The main chips are an NEC V55PI, a Hitachi H8/3003, and three TI TMS57002s. The H8 was already in MAME. The V55 was not (though the related V20 and V30 were), and even finding a TRM was a bit difficult. I could only find some copies in Japanese at first (here and here), which I couldn't read at all, but which was no problem for my LLM agents. Some months in, a fellow dev told me they had found an English copy (thanks, Giulio)."

























No comments:
Post a Comment
Note: To reduce spam, comments for posts older than 7 days are not displayed until approved (usually same day).