Calculating exposure for extremely long exposures

I'm thinking of trying my hand at extremely long exposures, in the order of days, weeks, months and maybe even years.

The plan is to exploit the reciprocity failure of some films to get the long times I'm looking for, especially Kodak 400TX since it's the worst of the of "standard" films.

My question is this, what to do with varying light conditions. For example, as a test, I intend to shoot a shot of my apartment for couple days. For maybe 12 hours per day, the apartment will be lit by sunlight/artificial light, and during the night it will be dark. In my head, I'd figure that if I meter for the light part of the day, and then overexpose one stop, it should make up for the darkness half of the time. Is this right, close to the truth, or completely off?

And how accurate must I be when doing these exposures? I mean, if I expose a piece of film for say a month, how much does it alter the exposure if I cut the time short by a couple days, or forget about it and stop the exposure a week too late?

If you calculate the correct aperture for a month long exposure, and leave it for an extra week, I'd call that about 1/4 stop overexposed. I think the math still works pretty much the same again once you've gone that deep into the reciprocity failure end of things.

Who was that guy who took the years-long exposures of building projects and stuff? Oh yeah, Michael Wesely:

http://www.unfinishedman.com/the-long-exposures-of-michael-wesely

Must be a heck of a stack of ND filters involved

Duncan

Okay, so if it's dark half the time of the day, that would mean a 50% decrease in light, which would require a one stop increase in exposure from the computed exposure at daylight?

I guess, but most places even when it's dark it's not really DARK... and across a bunch of days you might need to figure out more precisely what percentage of the day it's really dark, etc. I hope you have a lot of patience, to zero in on the proper exposure, it may take quite a bit of "bracketing"!

Duncan

It seems to me that the idea of metering goes against the grain of extremely long exposures. If it were me, I would pick an arbitrary exposure time, develop the film, and see if I had blank film, some image, or a completely opaque negative, then adjust exposure of the next negative based on the outcome.

I am curious as to how you are going to restrict the light to permit such long exposures.

If you have an old box camera you don't care about you might want to try setting it up at the same film, aperture, etc. for a day and develop it and see what you get and go from there. Heck, you might even want to start with an hour of exposure during daylight to make sure it isn't already completely white and if it comes out then do a 4 hour exposure to see if it actually captures anything more than 1 hour.

The plan is to load up my Canon FTb with Tri-X, lock the shutter open, and stack ND-filters infront. My calculations have resulted in that if the average light is around EV6 I would need a -16 stop decrease to reach a 347 day exposure with 400TX after compensating for reciprocity failure (if the effect computed with the formula T=Tm+a*Tm^b still works for these insane times, i.e. 2048 minutes of metered exposure)

With that tiny amount of light my main concern is that the reciprocity failure will be so severe that you'll get absolutely nothing on the film.

Are you going to do any shorter tests first? It sounds like an interesting project but I think it would be a bit disappointing to get a completely black or completely white frame after the year because no real world tests were done first. I guess you want to get started but I'd use one camera for the year test and set up some more for shorter tests to develop early in case the calculations were off.

frobozz said:

If you calculate the correct aperture for a month long exposure, and leave it for an extra week, I'd call that about 1/4 stop overexposed. ...

Click to expand...

To be more precise, and assuming reciprocity, the f/stop error is calculated as:

f/stop error = log(actualTime/targetTime)/log(2)

which gives us an overexposure of 0.32 stops for a time increase of 25%.

However, this example is far into reciprocity failure, and therefore, all bets are off.

My intentions are to start off at a couple of days, then do more tests until I think I got it right. The only problem is how to test for 1 year exposures, if the calculations are off, then it takes a year until you notice, and if the formulas for reciprocity failure are off, then how am I supposed to go from there? One way would be to just set up a bunch of pin-hole cameras, and do 10-15 simultaneous tests, but then it becomes problematic to make sure that all the cameras are identical, and that the apertures actually are the size used for calculations.

How about setting up a few cameras and leaving them out for as long as is practical to do a test at, let's say, 1, 2, 4, 8, 16, 32 days? Then use the achieved densities and extrapolate them out for up to a year on a 2^x scale. That should get you pretty close, and in one year you can adjust the curve to be more perfect next time.

Leave it to you, Ralph, to suggest a "stop" solution to his problem. This is why I think your book is so useful.
Dan

hpulley said:

With that tiny amount of light my main concern is that the reciprocity failure will be so severe that you'll get absolutely nothing on the film ...

Click to expand...

Not so. There's a long history of extra long exposure photography. Reciprocity doesn't take away all the exposure, just a huge fraction. If you get a blank. you used way too much neutral density.

A few misc. thoughts on very long exposures...

Metering and calculations are pretty useless other than for providing a start for exposure guesstimates.

To shoot your apartment with exposures ranging from months to years, you don't have to worry about a few rainy days or whatever, since it will all average out.

How would I come up with a "film speed" for, say, a one month exposure? Well, I wouldn't even worry about the math or trying to rate a speed other than starting with nothing more than a rough guess. I think I'd end up with an empirical / rote formula that would look something like - Tri-X f/8, 2 months, two stacked ND such and such zillion factor fiter.

Since over the course of a month or whatever, having enough light is not a problem to put it mildly, why not choose whatever is the optimum fstop for sharpness for your lens, probably f8 or so.

Length of day might be a variable, since winter days are shorter than summer, and in high latitudes probably average less bright. That might be a one or two stop seasonal factor.

I'd start with a very, very wide range of exposure for a test. And I'd rely only on a test since there are no published tables of reciprocity losses for days, weeks, and months...

Were cost no object, to make the test process quicker I'd probably just buy perhaps 5 cheepo 35mm cameras off craigslist or such for $10.00 each, load them with the 400TX or whatever you plan to use, and set them up side by side on a table with whatever neutral density filters (and that's the expensive part because ND filters just aren't cheap, especially the really dense ones) you plan to use, and have very wide ranges of exposure - a middle exposure, a second exposure of a hundred times less exposure than the middle exposure, a third with hundred times less than that, then the same sequence on the plus side; a plus one hundred times more, and a hundred times that. All the exposure variation would be through through neutral density.

Why such a huge exposure range? Since, pardon the pun, effective film speed for months and years exposure is a complete shot in the dark, the first thing is to just get somewhere in the rough neighborhood of the right exposure. If you get within a couple of stops in the first run through, you are doing very well. The next test can refine that value.

If you get a total blank, up your exposure a thousand fold. If the film is totally blackened, drop exposure by a thousand fold. there's little point in muddling around with half stops. You'll spend the rest of your life sneaking up on the right exposure if you take little steps. Each step will set you back a week or a month or a year. If you establish right off the bat that three stacked 1000x ND filters cut too much exposure, and two stacked 1000x NDs blow you away with a totally black negative, finding the middle ground won't take that much longer.

Dan Henderson said:

Leave it to you, Ralph, to suggest a "stop" solution to his problem. This is why I think your book is so useful.
Dan

Click to expand...

Let's not hijack this interesting thread, but it should be said that many processes only show significant change if variables are at least halved or doubled. One exception may be a film test, where development time is increased by square root of 2.

In the OP's example, it makes no sense to increase the exposure time by one day at a a time. Always double the exposure to create meaningful data relatively quickly.

A couple of examples to consider.

Adding just 1/3 stop to a 365 day exposure means adding 12 weeks, and that doesn't count the additional reciprocity failure introduced by the longer time.

Adding 1/3 stop to a 4 week exposure means adding 6.5 days, also not accounting for additional reciprocity speed loss.

If you expect to get any meaningful data from testing, you'd better have a good densitometer and working habits.

Manufacturers state that reciprocity failure may vary significantly among different emulsion batches of the same film. Some of that is just denial of responsibility for using the film outside design parameters, but even a 1/3 stop variation, as can be seen above, will be significant to you in adjusting time.

Lee

Lee L said:

... but even a 1/3 stop variation, as can be seen above, will be significant to you in adjusting time.

Lee

Click to expand...

On the bright side, a day, a week, or a month more or less won't matter much.

Hmm, this does give some interesting problems. I'll think I'll start of with my intended plan, since that doesn't require me to invest a pile of money in a stack of ND-filters, and see where the 1-2 week exposures leave me. That's fast enough for me to alter parameters, while still giving me the effect I'm looking for. It also let's me see if reciprocity failure calculations are even in the ballpark for these exposures.

The stop-system suggested by Ralph is also very interesting, by plotting the differences it shouldn't be impossible to calculate the effect, and if all the frames are shot on the same roll of film, it would take out variations in processing as well. Only problem is that I do not own a device for measuring film density (what are these called and how much do the cost anyhow?). An alternative would be using my epic excel sheet for reciprocity failure (based on measurements by another APUG:er who's name escapes me at the moment) to calculate the needed times as the time doubles, and then see how deviation it would get in the film densities.

BTW, would pulling affect reciprocity failure? I'm thinking of pulling the 400TX down to EI100 to decrease the amount of filters I would have to acquire. Or I could just use a ASA25 film instead and wing it, for example Rollei Ortho 25, but then again I'm not sure if there are any reciprocity failure data on that film at all...

If you want good reciprocity and 100 ISO, T-Max 100?

All bets are off

With super-long exposures, all bets are off. Suck-it-and-see is the only thing to do. I would choose a slow or medium speed pan film, because reciprocity effects are the least. (B.t.w., astronomical plates offer the lowest reciprocity failure and stand long exposure times.) I would use cut film in a plate camera, which allows a stepped exposure to be done, by withdrawing or inserting the darkslide "sheet", just like making a test strip in the darkroom.

Pulling isn't a good idea, nor is pushing, for that matter. With this project, pulling adds further uncertainties, but more importantly you loose shadow detail and contrast. All modern films, b&w and colour, are "minimum silver loaded", which means the D-max is set by the amount of silver in the emulsion, and development is taken to "unity". This isn't absolute, but compared to pre WW2 materials, speed and dynamic range is set by the manufacturer. That's why Ansel Adams's Zone system doesn't really work with present day materials.

There is no way to really tell with exposures that long until you just try something and see if it works. I'd start with negative film, and lean toward overexposure. Then let us know what happens. We look forward to your post on 1-12-12!

colourgeek said:

... That's why Ansel Adams's Zone system doesn't really work with present day materials.

Click to expand...

Really? Let's start a new thread about that!

I have to agree with the other Keith- the proof is in the pudding, you just have to try. If there are big doubts about exposure, my suggestion would be to develop test strips and determine the best dev conditions experimentally. Snip off a bit and develop and see how you're doing. Can you set up to expose a few extra pieces of film simultaneously? If so then this is going to be easy.

Ach now I see Ralph's post #11, sorry. My suggestion is almost identical to his except I am suggesting to "bracket" your development, not the exposure.

Okay, this goes a little against my nature, I'm the kind that likes to do a lot of calculations beforehand and then go and try it afterwards.

My original idea was to use a film that doesn't take well to reciprocity failure, i.e. Tri-X, since it allows for longer exposures with more light than for say T-Max (if the reciprocity failure calculations still apply at 72 hours of metered exposure, then T-Max need a little under a year while Tri-X needs over 3 years. )

Stuggi said:

Okay, this goes a little against my nature, I'm the kind that likes to do a lot of calculations beforehand and then go and try it afterwards.

My original idea was to use a film that doesn't take well to reciprocity failure, i.e. Tri-X, since it allows for longer exposures with more light than for say T-Max (if the reciprocity failure calculations still apply at 72 hours of metered exposure, then T-Max need a little under a year while Tri-X needs over 3 years. )

Click to expand...

If you're mathy try thinking in integrals - if your shot is light and dark in the same exposure then its unlike %99 of the still photos you'd normally take, where the light stays pretty much constant throughout the exposure (ignoring motion blur)...

Draw a graph of light intensity over time - your exposure is a sum total of the area under that graph...

Also remember long exposure reciprocity is a function of low light intensity not long exposure time, it just so happens that long exposures are required for low light. But just be aware of your cart and horse ordering, even if your horse is in front, it might be backwards, and your cart pointing up - (pedants lets ignore that reciprocity issue with short exposures also). Anyways, this will confuse your issue in that you're going to have more reciprocity 'lag' to deal with when its dark than when it is light... A side by side test over the same duration with two cameras, one shutter closed when it is dark would probably have a very similar image as the one that was open the full time

Its a bit of a mess mathematically - 2nd year university stuff ?

I'd do all the math I wanted, come up with a number - try it out, but have about zero faith it'd be nailed first time