multiple exposure on colour film -- order of exposures

Huh! How so?

 

As previously discussed, light and exposure are simply additive. And whether that's done in linear or logarithmic 'space' doesn't really matter.

 

(snip)

OK, first consider the digital case, as it avoids some non-linearity problems with film.

Say we take two very different scenes and take digital pictures of them.

Amazingly enough, they have the same histogram. Maybe even the

histogram is perfectly flat.

Now add them, as would happen with a double exposure, but we can just add the pixel values.

(Just to be sure, divide by two.)

Now what does the histogram look like? The light and dark parts of the two images are in

very different places. (That is, statistically, they are unlikely to be in the same place.)

The result is that the histogram gets sharper in the center, and the tails reduced.

For a case that everyone should know, what is the distribution of the total when you

roll two dice? Even though the distribution for each is uniform, the sum is peaked

at seven. It gets worse and worse as you roll more and more dice at a time.

So, given a sharply peaked in the center histogram, how do you change the exposure?

Maybe even the

histogram is perfectly flat.

In that case the combined histograms would also be flat. But greater in height and piled up at the righthand edge - unless the exposures were adjusted downward in proportion to the number of exposures.

Whatever. The shape of the resulting histogram is totally irrelevant to combining exposures.

An exposure reading is, in most cases, a weighted average that takes very little account of the distribution of light and shade in the image. And in the case of a multiple exposure the sum of any number of those weighted averages should remain roughly the same as one normal exposure. The shape of the resulting distribution of light and shade has no bearing on that.

Even if we take a million pictures and get a classic gaussian bell curve as a histogram; how would we know? If the film or sensor was so overexposed that the resulting positive image was pure white everywhere.

The 'shape' of the resulting image has no relevance to calculating the exposure needed for an individual exposure in a series of multiple exposures. Because exposure metering is already an 'average' of some kind.

Even if we use ETTR (Expose To The Right) as our exposure parameter, we need to compensate. We can't add two or more highlight values and not expect them to overexpose the film or sensor. The combined highlights would just crush up to the extreme right of the histogram. Likewise with the shadow values. Without compensation they'd also add together and end up as mid-tones.

Making multiple exposures with no compensation would be like rolling your dice several times and simply totalling their numbers. You need to divide by the number of throws to get an average!

Einstein said - "God does not play dice with the universe" - and neither is calculating multiple exposure values anything like rolling a set of dice.

Remember how highlight and shadow detail will be influenced by shooting two frames on top of each other.

-The shadows of your first frame will be filled in by your second frame

-The highlights of your first frame will be mostly lost in the second frame

It is equally true that the shadows of your second frame will be filled by your first...

It makes no difference which order they are shot in.

It's occurred to me that multiple exposures could be done on digital, in camera rather than in PP. Put the lens cap on, in full manual mode open the shutter on B, quickly remove and replace the lens cap, and repeat for subsequent exposures. Obviously the aperture would need to be small, and lighting not too bright, to avoid overexposure,

Perhaps someone has tried this. Or am I going crazy?

It's occurred to me that multiple exposures could be done on digital, in camera rather than in PP. Put the lens cap on, in full manual mode open the shutter on B, quickly remove and replace the lens cap, and repeat for subsequent exposures. Obviously the aperture would need to be small, and lighting not too bright, to avoid overexposure,

I tried it. With the Pentax KM on a tripod, I set it to F/16 and 30 seconds, I focused and refitted the lens cap. During the exposure I removed the cap and replaced it after around 1 second. Then repeated the procedure after repositioning the subject. Here's the result:

Err, can anyone think of a use for this?

Perhaps someone has tried this. Or am I going crazy?

Some Nikon DSLRs offer multiple exposure mode. And, yes, I've tried it.

Nikon's implementation is long-winded and not too useful, but has no restriction on the individual shutter speeds for each exposure. The strange restriction is that you have a total time of 5 minutes (IIRC) to complete the preset number of exposures.

I used it to experiment with 'pre-flashing' to see if shadow detail could usefully be lifted. The result was a bit inconclusive, and considering the time restriction, not practically useful.

Err, can anyone think of a use for this?

A stereoscopic third hand? Hell, yes! Just add robotic micro-motors and it'll be perfect.:)