It's also partly about the time constraints. It can be difficult to clear a set for enough time to shoot the best HDRI possible.

If you're shooting a 360 degree hdr environment map with a 180 degree (8mm fisheye) lens then I usually shoot 3 directions at 120 degrees apart to avoid vignetting at the edges/seam. With 9 exposures per direction - that's 27 images to shoot per hdri - which is sometimes not possible.

I usually shoot 5 exposures, 3 stops apart -6, -3, on-aim, +3, +6. This covers the same exposure range as 7 images at 2 stops apart but with fewer images. Doesn't suit all conditions but works as a general purpose setup for me.

It varies depending on the intended use. All the best plans and science fall apart when the first AD wants to move on, people in bright orange safety vests are milling around and the clouds are doing 200km/h. I think I can grab a sphere in about 3 minutes, at least that's the time I quote to the first and give them a lot of warning.

The method passed on to me by our lighters that I use (no complaints, yet) shooting with an 8mm/5dmkII on a nodal ninja goes like :

• Have the damn thing ready, cap off, remember to remove the extra cuff found on fishies.
• Get assistant/data to help level the ninja, check the measurements on the side so we have zero parallax and the sphere is somewhat aligned.
• assistant/data do notes for card/image number, time and any relevant easy survey.
• Point the camera at the sun or the fattest light, go to like f6-8 iso100 usually, best compromise for flare etc and the flare will hopefully be small and in one shot.
• Head straight to quickest shutter and bang one off, see if it's clipping if not, start clicking.
• Shoot with a 3 'click' gap unless the lighters have asked for different. I don't read the camera display, just count clicks and shoot as fast as possible.
• Keep shooting, when your shutter time is long enough to almost clip out the frame, it's time to rotate the ninja 120 degrees and start at the bottom.

I find the hardest part about that is convincing people to move out of the shot when you're on 8mm. "nope you're still in, yep still in, safest behind me, click"

Depends, largely, on two things.

1. The dynamic range of the scene you're shooting. Essentially, the shortest exposure should be short enough so that the brightest areas of the scene are not overexposed, and the longest should be long enough so that the darkest regions are not under exposed.

2. The response curve of the camera. This is a non linear function that maps the actual scene radiance to the pixel value recorded by the CCD. This function can be estimated in advance ( as it is dependent on the cmaera, and not the scene ), also from multiple exposures. In which case, a sequence 3 stops apart between the darkest and lightest image will be sufficient. If you don't have the response curve, try a sequence 1 stop apart.

For a full explanation, see Debevec and Malik; (Recovering high dynamic range radiance maps from photographs) or Mitsunaga and Nayar; (Radiometric self calibration) via ACM or IEEE respectively.

According to this site (which has lots of other good info on HDR shooting) http://www.hdrlabs.com/tutorials/index.html there is no definite answer as what makes "a good quality HDRI" is highly dependent on the desired use. However 7 exposures shot in raw at 2 exposure values apart seem to cover most bases.