DVI-16-Hotplug Detect <>
That leaves hot-plug detect, DVI-16, as the only wire still hanging at this end. Unfortunately, I can't be sure whether it should connect to VGA-4 (marked as RESERVED on VESA DDC2, ID2 on the original VGA) or VGA-11 (ID0 on original VGA).
OK, some more info on DVI-16-Hotplug-Detect.
The DVI spec can be downloaded for free from the Digital Display Working Group (DDWG) website (~1.2MB PDF, you'll have to select and paste in the address bar as it won't let guests post links, and who knows if I'll be here again, so I'm not going to register): ddwg.org/lib/dvi_10.pdf
Starting on ~pg 16 of that document, it lists the requirements for Hotplug-Detect (HPD). Briefly, it comes down to this:
HPD is a Monitor to Computer signal. < .4V monitor side, .8V computer side, is to be interpreted as no EDID/DDC compliant display present. > 2.4V Monitor side, 2.0V computer side, is to mean an EDID/DDC capable display is present. Monitor controls the level. 5V signaling levels are appropriate.
The spec didn't specify fully what action software should take to the hotplug event (HPD switching levels, indicating the plug or unplug of a compliant monitor), but did to the extent that if a compliant monitor comes online, the signal should be activated, and if it goes offline, the signal must be terminated within 1 minute. It didn't say, but this would appear to refer to the digital signal.
Both of the monitors I have handy here ground both VGA pins 4 and 11. With pin 12 open, that would be interpreted on the original VGA spec as Color monitor that can display 1024x768. Of course, as mentioned upthread, DDC2 compliant connectors have changed that. VGA-4 is DDC-Reserved, and I found one mention (not well supported) that VGA-11 is DDC-ClkReturn.
While I still can't say for sure where that DVI-16 connected in the broken adapter of mine, the upshot of the above appears to be that for VGA adapter purposes, DVI-16-HPD should be grounded, as the whole purpose is to use the analog signal from the DVI-I, not the digital signal. It would further appear that today's monitors apparently ground both VGA-4 and 11, so that grounding will appear in either case.
OTOH, from my googling, it was obvious that various video cards react to DVI-HPD in different ways. Most seem to ignore it, while some do indeed keep the digital signal off if it's not high. Also note that here, the xorg log (Linux) indicated that it could get DDC data from the monitor connected to the real VGA connector, but not the one connected thru the DVI/VGA adapter. This is on a Radeon 9250 using the xorg native open source drivers, not ATI's proprietaryware closed drivers. I've not experimented yet, but it's possible that forcing DVI-16-HPD high would allow the DDC detection to proceed, and that it would then detect an analog monitor on the DVI-I connector and activate that, deactivating the digital. Forcing it high could be accomplished by hard-wiring the upthread mentioned +5V to HPD, preferably thru a current-limiting resistor.
So... it should be safe to experiment with HPD forced high, or grounded, either one. VGA monitors appear to ground it, regardless of whether it's supposed to connect to VGA-4 or VGA-11. However, forcing it high could mean better automated detection of what's there. OTOH, it may have undesirable side effects such as disabling the analog signal. Effects are likely to be somewhat video card dependent, variable from model to model.