I have discovered a few other conditions that make a big difference.

1. Getting a proper color match (material-vs-keyer). A large 'swatch' from one place will not give the same results as will a large 'swatch' from another place on the same material. I find it best to work out what the *color* of my background actually is (as if in mixing paint), and enter the R=41, G=244, B=24 values in directly. I'm using eefx.com's foam-backed material. These HD cams can see the individual fuzz pieces on the surface of the chroma material, which is why we want a shallow depth of field (subject as far from the chroma as possible, camera as far from the subject as possible).

2. Keeping the High Threshold high seems to assist. I've found that if the color sample requires High Threshold to fall below .739 then I need to improve the color sample. Bring the Low Threshold up to about .1 below High.

3. Carefully tweek Chroma Blur. I get the best edge detection with Hor=.500 and Ver=.1.500 - but then I shot this subject breaking all the rules, apparently (grin). It is a 1:12 scale model Monster Truck. I anchored it to a thin metal post vertically (nose straight up) 1ft above the chroma material that was laid on the floor, and I boomed the Z1U pointed straight down directly above the truck a mere 2ft. I focused at .8m. Lighting was natural shaded sunlight indoors (in a glass room with bright outdoor sun). The chroma and truck were situated in a non-window corner of the room so that both were uniformly and identically lit from all directions (super-soft lighting). f-stop was 2.8 at tight zoom. I set a 6-second pre-set A-B zoom-in from widest to tightest; 60fps. I shot the truck with all the wheels spinning wildly and with the truck bouncing on the steel post (Monster Truck, remember?). The objective was to create the illusion that the truck was rapidly approaching, and about to run over, the viewer (the final frames being a death-defying closeup of the front bashbar (slightly out of focus) and the huge right wildly-spinned monster tire and it's suspension system (also out of focus).

The effect works: bright driving lights, spinning tire tread, bouncing frame and axles. All we need is the engine and road noise, plus the booming airhorn. I am very impressed with Vegas that it could hold the key with clean sharp edges *even when those edges went back out of focus* and the edges were spinning wildly. Remember, I had to key on the subject when it was the smallest image, and Vegas had to hold the key while that image grew to about 8-times the actual frame size.

It helped immensely that I can preview in Best High on a 20.1 in widescreen, letting me *see* the individual pixels. I threw a cookie cutter around the image to make sure I didn't have any extraneous not-quite-keyed areas. I was using a mere scrap of the chroma material, and at wide angle (the truck appearing to be about 50ft from the camera in the first frame) I was capturing tripod legs, my boot, my hand operating the remote control, some very nice oak wood floor, etc.

Y'alls still with me? One success story leads to another problem.
This 11-second segment is about as perfect a key as I could hope for on all frames. But when I render it to an HDV intermediary the black comes out "solid" not transparent, making the intermediary useless on another timeline. Ah, V6.0d. Anyway...

Q: How do I render so that only the subject is in the clip (the black keyed area representing 100% transparency)?