Could you explain in a little more detail exactly what you mean by this? It sounds very interesting.
S
Shoot an image at ISO 100, but take 8 frames (for example - I've done stacks with 100 frames before). Expose for the highlights - you can't rescue those once they're clipped.
Import all the images as layers into Photoshop, align them perfectly and convert them into a Smart Object.
From here, you can set various stacking modes, but the two relevant ones are Mean and Median (Maximum and Minimum are also often useful - say, if you want to enhance or eliminate light trails along a stretch of road - but don't come into play here). Mean stacking averages the RGB values in each pixel, while Median stacking selects the one in the middle.
Whether mean or median is better in a given situation depends on movement in the scene and what you are trying to do with it. Mean stacking averages out all the frames, essentially making it into one long exposure, while median stacking is good for eliminating moving subjects from the scene, since most subjects moving through an image will only appear in one spot in a single frame (although you can run into issues if someone stays in the same spot for multiple frames). Sometimes, you'll want to do both a Mean stack and a Median stack (or even a Maximum or Minimum stack) of the same images, then manually blend them.
In either case, by taking 8 frames, you've collected 8 times as much light, or 3 stops more. Either by averaging them or just selecting the median R, G, and B value for each pixel, you reduce random noise by 3 stops. Read noise remains about the same (barring any variability) but can be further reduced by shooting a few dark frames and subtracting them. You then end up with a much less noisy image (equivalent to shooting at a 3 stop slower ISO) with more flexibility to recover shadow detail without getting too much noise.
When not using dark frames, my preferred technique is to push the shadows and pull the highlights in RAW conversion first, before stacking in Photoshop That way, I know I've pulled every possible detail out of the highlights and shadows that would otherwise have been clipped in the TIFF file. This doesn't work when using dark frames, though (you can still pull highlights in RAW conversion, since read noise contribution there is negligible, but not push shadows).
Sometimes, when there are small areas with extreme differences in illumination (say, street lights or candles in an otherwise dark scene) I'll shoot an extra frame, exposed specifically for the brightest highlights, then manually blend it in. This is because, in such situations, the highlights are going to be blown in any reasonable exposure - they're only very small areas points, but often 8 stops brighter, or even more, than the 'highlights' in the rest of the image, and exposing for them and pushing the shadows would otherwise require hundreds of frames.
This isn't just for ultra-low ISOs or ultra high DR situations. I frequently combine this with high-ISO (640 or higher on A7r3) shooting where motion is an issue, aligning them before stacking to produce the equivalent of a low-ISO shot. This is particularly useful in telephoto shots, where golden-or blue-hour lighting would often call for exposures of 1/4s or more at landscape apertures, but wind- or ground-related camera shake, even on a tripod would turn such exposures into a blur. But shoot at 1/60s and each frame will have minimal blur; align and stack 30 such frames and you'll have the equivalent of a 1/2s exposure.
