1. Capture original LD signal from high-quality player, as exactly as possible (at least 9-bit ADC, to lossless codec).
2. IVTC, not in real time, with best possible algorithm in AviSynth (free, and testable/tunable) to get full-resolution progressive frames.
3. Scale, not in real time, with best possible scaling algorithm in AviSynth (free, and testable/tunable).
Wouldn't that be a radical idea?