I will answer [part of] my own questions in detail (corrections may be necessary), for further future reference:
While reading this thread:
http://www.amiga.org/forums/showthread.php?t=18247I stumbled upon MskoDestny's comments which are the exact answers I was looking for:
Those digital signals only give you 4-bit color (for attaching to a CGA monitor). Only way to get the full color resolution is to use an ADC on the analog RGB signals, make a video slot scandoubler or make something that clips on to the appropriate chips.
The added ADC of the analog solution pushes you over what makes sense for a $100 product unless you're willing to order the parts in sufficient quantity (which of course requires capital). A digital solution is complicated by the fact that no one manufacturers 5V DRAM chips anymore (well at least none that I've come across) and SRAM is kind of expensive for the quantities you need. Since AGA has 24-bits worth of video signal it's not a trivial number of signals to convert (though the fact that it's not a bidirectional signal probably allows for cheasier solutions).
So the digital RGB outputs (on the DB23 RGB port) are only 4bit (16 colors), and thus pretty much useless
As mentioned before the options are:
1) Internal Video Slot solution (occupies slot, much like CyberVision 64/3D, has access to best quality/full OCS+ECS/AGA 12/24bit DIGITAL signals, is problematic since not all video slots are the same, and some Amiga models don't even have one, also a pain since the machine has to be opened up)
2) Internal Chip Hijack solution (attaches on top of Alice or appropriate chip and thus allows an empty video slot, much like some DCE solutions, has access to best quality/full OCS+ECS/AGA 12/24bit DIGITAL signals, is problematic since different Amiga models have different video chips, also a pain since the machine has to be opened up and furthermore one has to much around with the delicate circuitboard components)
3) External Analog To Digital solution (easily attaches externally to the RGB port and is thus the most "universal" solution, has access to the "not so best" quality OCS+ECS/AGA 12/24bit *ANALOG* signals)
Clearly for the best quality solutions 1 and 2 are the right choices. Clearly for the most trouble-free and most compatibility solution 3 is the right choice.
One thought that I had in mind is that even though the 3rd solution doesn't give the best results, an improvement would be to use the digital RGB signals to "dynamically calibrate" the ADC circuits. In other words, when the ADC samples a single RGB pixel value as 23,45,9 it can then compare the lowest 4 bits of each component to the digital RGB values, and do some fancy "auto-calibration" or even "auto-error-correction" (this would only work with component values less than 16 due to only having 4bits per component from the digital RGB lines).
EDIT:
Now that I read the quote again, I'm not 100% sure if the 4bits digital output pertains to each RGB component. Anybody have a link to someplace documenting the digital RGB outputs?
Also, when AlexH said that the ADC is only 16bits, I think he means that the ADC might sample at 10bits per component, but will only output 5 or 6 bits per component (depending on if it's 5:6:5 or 5:5:6, etc). Can you verify this AlexH? Thanks
EDIT2:
And now I will fully answer my questions.
Looking at the following schematic by Ian Stedman:
http://www.ianstedman.co.uk/Downloads/scandoubler.png and looking at the AL250 chip, one can see what AlexH is referring to as the "16bit limit": it's those pins named VDIN[0-15]. The input to the scandoubler piece (AL250) is limited to 16bit. BLAH! Hopefully a replacement for the AL250 is easy to find.
