Yes, I have a scan converter (RGB Scart) already, I was just curious if he built this particular implementation and if so, if it worked very well (things like timings and brown fix being the usual issues, as you noted). I was looking into the many variations with my friend (Adrian from ADB) and we kinda played with one board but it needed parts that were impossible to find so that was a wash (including finding the correct resistors which mostly means you need close resistors and some trimmer pots for tweaking) . He was also going to look at doing a programmed GAL and started to make up a board but got sidetracked on other projects.
I can't imagine "timing" being something that would affected by any of these DAC board designs, although there may be some black magic in the separate-to-composite sync circuitry in those that handle that. There is a lot of variation in how the actual RGB conversion is handled, which broadly split into "Diodes! Lots of Diodes!" and "Diodes? Why?" camps...
For laughs I just sat down to try to take a crack at GAL code for the equivalent of the SeeGeeLog converter linked earlier. In the process I also found the schematic to the three different generations there's been of the GGLabs board, and to broadly summarize the difference:
1: The GGLABs board sets the GAL up as an according-to-hoyle 6-bit DAC with two intensities of each color; the "intense" resistor is about twice the value as the "base" resistor; the programming of the GAL simply sets which 16 color subset of the 64 possible colors are used, which is how the "brown fix" is implemented. The SeeGeeLog one, on the other hand, had me scratching my head for a while in terms of how it uses the 74LS138, but after working out a truth table for it it looks like it uses a weird "pull down" arrangement to reduce the voltage on "G" under the one specific circumstance.
Frankly after looking at it if you have a GAL I'm inclined to say the GGLABS arrangement makes more sense, it actually requires fewer parts and is easier to grok. The only downside is it doesn't let you "tweak" the brown-ness. You essentially get a "brown" from the EGA palette, which I'm not sure *quite* matches CGA brown... but it also almost certainly hardly matters.
2: The later versions of the GGLABs board use a proper op-amp to provide the final drive of the analog pins with 75ohm impedance matching. If there's a solid quality difference between the current version of the CGA2RGB and a homebuilt one this may well be where it's coming from.
Something I'm not super clear on is why the *first* version the CGA2RGB, which doesn't have a proper op-amp, uses a 74AC541 line driver between its GAL16v8 and the resistor ladders. According to the datasheet the AC541 has an output source/sink of 24mA; the GAL20v8 datasheet I have handy claims it has a drive current of up to 64mA, so... not sure why they used it.
I think I'm tempted to basically build a GGLabs rev 1 without the '541 on a breadboard and see what I get. Worst case I'll blow up a $1.50 GAL, I guess.
So besides the SCART, I do have one of those GBS boards and honestly they really are crap. Adrian did some tests using the CGA2RGB with both the Scart box and the GBS and the GBS was almost unusable. I'm currently using the Scart box with my Atari ST and the appropriate cable and i'm pretty happy with the output. My 'gaming' is limited to things like Ultima, so i don't really know about timings and such.
I have a GBS lying around that I haven't tried yet so I'm thinking there's nothing to lose, but it won't surprise me if it's too terrible to bother with; the GBS is currently configured to work with an Apple IIgs and I sure can't say it does a great job there.
As for the CPLD ones, I personally would rather something that was a little simpler to build which is why i prefer the GAL option and mostly through-hole design, if at all possible, so that regular folks can build it if they are so inclined.
One potentially neat thing about the Raspberry Pi-powered dingus is *maaaybe* it might end up being possible to do capture with it.
I mentioned op-amps above, one thing that I think might be interesting if it turns out they really are necessary to get decent quality out of the first stage conversion is there are some schematics out there for doing RGB->YPbBr color space conversion using some fairly cheap video op-amps. (Which unfortunately only seem to come in surface mount packages). Those RGB-capable SCART converters are sort of hard to find; what might be a preferable strategy, at least for those in non-SCART-using regions, would be to make something like the GGLabs board but with built in component video conversion. Component to HDMI converters are only about $20 and I'd assume they probably have the effectively same guts as the SCART models, and it would also enable direct-plug to TVs and monitors that'll take component video. Component is starting to die out but it was common on TV LCDs until pretty recently.