Okay, so I've made the vortex generator array for my roof, but haven't tested it due to lack of additional people to take photos of tufts and/or run stopwatches(coast down testing)...
So, I basically paraphrased the details from the Mitsubishi Lancer EVO: 8 fins(plus one in the center that was the antenna on the EVO), 5mm thick(the coro I used was 4mm), 30x60mm(I used 1"x2" because I was using a measuring tape), 100mm apart(I used 4"), at a 15* angle relative to airflow(I didn't have a protractor handy, so used 1/4"width for 1" length and it looked close), and 100mm from the point of separation... I could have tested that but instead assumed it to be the top of the back glass and went 4" forward from that... I can't tell any FE gain, but I did this at basically the same time as finishing my wheel skirts, so I might need to remove it and actually do some ABA testing, and/or tuft testing...
What I found was, that before, when the back glass was wet, the middle 50-60% would take forever to clear, without running the defroster... After installing the vortex generator array, the top half of the middle 50-60% clears much faster, but the bottom half still takes a while...
Yeah, I know, I still need to tuft test to show any attached flow on the glass, but it seems promising... However, I'm also aware that in ideal conditions, with probably millions in R&D budget, the above mentioned Mistubishi engineers only got a .006 Cd reduction, which basically amounts to nearly no fuel savings for the effort...