My idea was, that a sprite unit would work this way: it has two "data structures" for a sprite. First, there is the "shape information". That is, controls /EXTC, this is 1 bit info per sprite pixel. Can be also thought as the "alpha channel" in modern gfx things, just we have only on/off, no "semi-transparency" etc. And then there is 4 bits of colour info for the given pixel, which does not count at all if /EXTC is not (low) active. Now it's interesting that it _seems_ I was right, in the "palette-less" 256 colour mode, the external colour still uses the palette. That makes sense, as this external colour feature is created for "sprites" keeping in mind in the future, this it would be a bit odd, that sprites can have only fixed colours for the first 16 EP palette colours. And anyway, in 256 colour mode the palette by Nick internal pixel generation is not used, so it's kinda handy to use it for something at least

Surely, for a more simple sprite unit, some can say, that external colour value 0 means the transparency, then no need for shape/colour info, but only a simple logic, to drive /EXTC for low active if 4 bits of EC are all zero. This would mean - if I am correct - that 0 would mean transparent, 1-15 would select the corresponding internal Nick palette (8 LPB palette values + FIXBIAS for the second

. Thus we lost one colour with this technique. The price for not having separated sprite "shape" and "colour" info in the external sprite unit ...