This is an early modelling digital synth (docs I've seen from 1985 actually so I'm sure the design wasn't started long after the DX appeared.) Each patch takes almost 1K to describe the parameters of in the dump, compared to I think it was 67 bytes/K3 patch or like that. 63 harmonics per 'half' in each patch. Pan setting for the halves. And each halve has it's own DFG (digital frequency generator) each with it's own 6 seg envelope, which forms the oscillator control base for the DHG (dig. harmonic generator) and is 'whole' mode rather than two 63 harmonic sets running side by side, you get one longer 126 harmonic set. But each half has 4 busses anyway, each with it's own 6 seg envelope and modulation independent of the global LFO. Then each half can be routed through a DDF (dynamic digital filter) which has dynamic slope and cutoff and a flat level setting....and accompanying 6 seg envelope. And then before they mix each half has a final 7 seg DDA. Then the mix goes through a DFT (digital formant) which behaves like an 11 band EQ sort of. 15 sounds can be run simultaneously in multi mode with their own channel or..all layered! Voice allocation can be flexible or a max voices can be allotted. Controller filters, etc. A very well laid out synth architecture but..too much for the typical user. I wrote a DOS software piece to control the synth after extensive sampling so that I could write for instance things like real time entry for the envelopes rather than the cryptic and unpredictable 'rates'. I'd love to pull it out some day and work it over to run in windows. Well...maybe 'love' isn't the right word. :-) HEY, I found an old link that has me listed at my old address on it...nice pic of the architecture and stuff though! k5info
Displays often go out. I have one inverter left. I think Kawai is out of them. But I think the display backlight element fails more commonly as was the case in mine. I've never bothered to troubleshoot it. Telesis in Orange County, CA does backlights for a lot of things including this one I believe. HEY here's a great link where someone researched this problem at K5 Hackpage. These machines were pretty reliable otherwise to my knowledge.