What does reciprocity failure in color negatives look like?

@Romanko I understand your pragmatist approach and to an extent I can agree, but there are several things you state that I feel (know) are false, such as your statements about wavelength vs. color, or that in practice can work out differently than you seem to realize. If I photograph the dimly lit (with natural light) cloisters of a medieval monastery and I run into reciprocity territory of the film, the effects can be counterintuitive and the test that Les has posted so far wouldn't help me in managing expectations and/or compensating to the extent possible, since the scanner plays too dominant a role. That's a purely practical observation, which of course ultimately is rooted in a deeper theoretical basis. We don't have to go through all the theory though to conclude that there are methodological issues with the test as presented now, and that those issues really do have practical implications, even if they might not become an actual (experienced) problem in all cases.

My understanding of color science is very limited so take anything I say with a grain of salt. I like Les's images though and this is the most important thing!

I can't comment on Les's tests as I do not understand them. I only looked at the landscape images that Les posted and I do not see any signs of severe underexposure in them. We don't know of course to what extend the scans were manipulated by the scanner software and post-processing.

A rigorous reciprocity test should probably include a color checker card and some means of estimating negative density. A weak but stable source of light in a darkened room might be all that is required for the experiment (and possibly some adjustable neutral density filters). Is this what Les did in his tests? Taking several hundred density measurements is not something many people consider doing for a hobby, I must admit.

To illustrate the problems with frame-by-frame scanning, consider the following sets of frames:

These are four different negatives of two identical scenes shot under identical conditions. The scans were done frame-by-frame with Epson Scan, with the scanning software automatically balancing colors. Note that there are distinct differences in color between both sets of frames, but they're still fairly close, I'd say.

Here's how the same negatives look like, but scanned as a positive, inverted manually and then the same adjustment curves applied to the respective pairs:


Note how the differences in reality are far less subtle...
If the negatives were truly as similar as the initial examples above suggested, then the second set of examples would have come out the same. But they don't. Ergo: the negatives are quite dissimilar, but the scanning software has managed to hide most of the dissimilarities from us.

I've played with negatives from this same test series under the enlarger and from a practical, entirely non-theoretical viewpoint the filtration and exposures needed to get sort of comparable prints are "all over the f* place" and the final outcome when trying to match the prints is again, "all over the F* place". That's real life.

Scanners throw us for a loop much of the time, for better or worse. Better = you can get away with massive differences in development, exposure etc. and still end up with quite comparable results. Worse = you can absolutely make no claims whatsoever about the behavior of the film if you scan on a frame-by-frame basis and let a scanner handle the color balancing. All you can say is how well the scanner manages to match your preference. That's also important from a practical viewpoint, but it's not a relevant outcome from a film reciprocity test.

Romanko said:

I only looked at the landscape images that Les posted and I do not see any signs of severe underexposure in them.

Click to expand...

I look at images that may or may not have underexposure going on and that may or may not have shifted in colors, and that may or may not have been scanned from printable negatives. Basically, I'm looking at pleasing pictures, but it doesn't tell me anything about the negatives other than that they could be scanned.

To repeat myself, anyone who has taken a look at @Les Sarile's first post would see that he actually does compensate (whether he's aware of that or not) for film's reciprocity failure and that, shockingly, his film behaves exactly like manufacturer describes in the spec sheet in regard to the reciprocity failure.

So, it's perfectly plausible that his night shots presented later in the thread are all pretty well exposed negatives and that we do not necessarily have to get sidetracked by what all scanning can do to the image to explain why his pictures don't look underexposed. They don't look underexposed because they are not.

brbo said:

he actually does compensate

Click to expand...

His exposure is variable, that's the only thing I can make out of his data.

Vertical axis: +stops exposure compensation
Horizontal: seconds of exposure (note: log scale)
Orange = Fuji Superia 100 datasheet suggested exposure compensation
Blue = stops of overexposure as calculated from the meter readings in the first image in this thread
grey = sort of average of the blue line to get a feeling for how it relates to the orange line...

Note how Fuji's recommendation is a neatly log/log progression. Les' exposure range does not show a clear logic to me.

If the lighting was kept constant and if the meter readings are consistent with the lighting conditions of those shots and if the exposures were timed as indicated in the overlayed text, the average density of the negatives would have varied. The fact that they remain fairly similar (with some differences in contrast) suggests to me that the scanner is polishing away the real variations in the negatives.

brbo said:

we do not necessarily have to get sidetracked by what all scanning can do

Click to expand...

I think it's the dominant factor in explaining the conclusion presented in the first post.

brbo said:

They don't look underexposed because they are not.

Click to expand...

If the exposure data I gleaned from the first post are correct, and those or similar data are applicable to the actual night scenes posted, then sure. However, I also note that the night scenes presented are shot on at least partly different materials. IDK what kind of compensation may or may not be applied there.

The main conclusion I draw from this is that Nikon's scanning solutions were really nice. You could do pretty much *whatever* and it would still yield a decent result.

@koraks, Les explicitly stated that he kept lighting fixed when going through aperture f2.0 to f22. Only after that did he start playing with light intensity to get to longer exposure times.

Let us assume that at f2.0 his camera's predetermined exposure time and actual camera's exposure time and the time that Les recorded with his stopwatch are all 1s. He then progressively closed down the aperture with lighting of the scene remaining constant and timed camera's exposure times (Les T). This is what the data looks like.



My wording that Les' off-the-film camera metering being in perfect agreement with Fuji's exposure compensation might have been too strong, but the Pentax LX actual exposure times track Fuji's exposure compensation pretty well, I certainly wouldn't call it "not showing any logic".

@brbo, sorry, I goofed there - in part due to silly mistakes, partly due to being thrown off by the Sekonic readings also in the picture. I admit it's mostly the former though.

Still - it looks to me like all kinds of funny things have been happening here. A camera light meter apparently decided for itself to start applying a reciprocity correction. A sekonic light meter all of a sudden veers totally off course at some point (f/11 read 50 sec, then f/16 reads...16 minutes!?) - funny that; I have the same meter but never noticed it doing this... And in the scans, the color and contrast wobble from frame to frame.

There's too much going on to make much sense of this. I don't see firm support conclusion that the film behaves as it should in terms of reciprocity, nor that it behaves better. I do see support for methodological objections related to the light source and/or metering - and still scanning.

So much of this is unfortunately biassed not only by the lack or real experimental control, or even a proper understanding of it, but even more, due to inexperience judging color itself. To a trained color eye like myself (and I've even trained color matching pros), all the given Crayola shot examples are miserably off.

One needs iron out all the variables on at a time, or you end up with such a tangled up ball of yarn that it becomes quite difficult to sort anything out. And to do that, as I hinted earlier, you need to bag a correct standard normal exposure to begin with. Otherwise, you can't really appreciate the extent of drift and hue crossover as things start falling apart beyond the acceptable zone of exposure.

If it were me, I'd scrap the whole alleged test and learn some basics first. All we have so far is a rabbit chase down any number of dark burrows.

DREW WILEY said:

all the given Crayola shot examples are miserably off.

Click to expand...

DREW WILEY said:

All we have so far is a rabbit chase down any number of dark burrows.

Click to expand...

I agree on both counts...

It’s clear that the real problem started years ago, and can all be blamed on the poor soul who first coined the term “reciprocity failure”. More specifically, it was the use of the term “failure”. That was a very unfortunate choice of words.

The term “failure” implies a catastrophic turn of events. A shutter failure means you camera probably doesn’t work. If your engine fails, you know you won’t be driving anywhere soon. If you have an exposure failure, you’ve made a hot mess of your image.

When you hear the term “reciprocity failure”, you naturally key on the term “failure”, and all the normal responses to that kick in. The natural expectation is that you’ll see some obviously catastrophic result.

If we could go back in time, we could fix this, and call it something like reciprocity deviation, or reciprocity adjustment. But it’s too late now, the damage has been done.

From a scientific perspective, we have an exposure law E=Intensity x Time. E is Exposure and Intensity is the quantity of light. If you double the intensity of the light, and half the time, then the exposure remains the same. The intensity is controlled by your f stop. We all know this.

Unfortunately, the exposure law E=Intensity x Time starts to become inaccurate when the time factor becomes very long, or very short. Some knucklehead decided to call this inaccuracy a “failure”, and started us all on this long torturous path of misinterpretation and false assumptions. He/she needs to be spanked!

When film manufacturers speak of “reciprocity failure” it’s usually in the context of “for critical work”. There are cases when unexpected exposure deviations are important. For example, film was used in astrophysics for measuring the light intensity of distant stars. They’d measure the grain size on the film. The grain size is related to the light intensity. The grain size will be related to the exposure, so the light intensity can be calculated since the time is known. They’d need to know that the law is inaccurate at long exposure times, and that a correction factor needs to be applied.

For the rest of us, who aren’t doing “critical work” with long exposures, this is good information to be aware of, but probably isn’t going to be of huge importance.

At the end of the day, it’s all in the name. So the lesson here is: take the advice of Johnny Cash, and don’t name your son Sue.

koraks said:

There's too much going on to make much sense of this.

Click to expand...

Well, I did say that I wouldn't do a film reciprocity failure test in that way.

I just wanted to point out that Les' conclusion that he didn't detect any reciprocity failure in his test is plain wrong. He most certainly DID. In fact it's almost a textbook display of film reciprocity failure. As to why Les thinks there is no reciprocity failure... I don't have an explanation.

Sharktooth said:

Some knucklehead decided to call this inaccuracy a “failure”, and started us all on this long torturous path of misinterpretation and false assumptions. He/she needs to be spanked!

When film manufacturers speak of “reciprocity failure” it’s usually in the context of “for critical work”.

Click to expand...

Well, if anyone should get spanked, it should be us that haven't updated out terminology. Manufacturers use "long exposure compensation/adjustment" in their spec sheets pretty consistently.

brbo said:

Well, if anyone should get spanked, it should be us that haven't updated out terminology. Manufacturers use "long exposure compensation/adjustment" in their spec sheets pretty consistently.

Click to expand...

Ouch, baby!

Does anyone else have long exposure compensation/adjustment testing results to offer???

koraks said:

His exposure is variable, that's the only thing I can make out of his data.
View attachment 375463
Vertical axis: +stops exposure compensation
Horizontal: seconds of exposure (note: log scale)
Orange = Fuji Superia 100 datasheet suggested exposure compensation
Blue = stops of overexposure as calculated from the meter readings in the first image in this thread
grey = sort of average of the blue line to get a feeling for how it relates to the orange line...

Note how Fuji's recommendation is a neatly log/log progression. Les' exposure range does not show a clear logic to me.

If the lighting was kept constant and if the meter readings are consistent with the lighting conditions of those shots and if the exposures were timed as indicated in the overlayed text, the average density of the negatives would have varied. The fact that they remain fairly similar (with some differences in contrast) suggests to me that the scanner is polishing away the real variations in the negatives.



I think it's the dominant factor in explaining the conclusion presented in the first post.



If the exposure data I gleaned from the first post are correct, and those or similar data are applicable to the actual night scenes posted, then sure. However, I also note that the night scenes presented are shot on at least partly different materials. IDK what kind of compensation may or may not be applied there.

The main conclusion I draw from this is that Nikon's scanning solutions were really nice. You could do pretty much *whatever* and it would still yield a decent result.

Click to expand...

brbo said:

@koraks, Les explicitly stated that he kept lighting fixed when going through aperture f2.0 to f22. Only after that did he start playing with light intensity to get to longer exposure times.

Let us assume that at f2.0 his camera's predetermined exposure time and actual camera's exposure time and the time that Les recorded with his stopwatch are all 1s. He then progressively closed down the aperture with lighting of the scene remaining constant and timed camera's exposure times (Les T). This is what the data looks like.

View attachment 375466

My wording that Les' off-the-film camera metering being in perfect agreement with Fuji's exposure compensation might have been too strong, but the Pentax LX actual exposure times track Fuji's exposure compensation pretty well, I certainly wouldn't call it "not showing any logic".

Click to expand...

First of all, I appreciate the interest in the topic regardless of whether or not you find my test clear or unclear.

For clarity, the test shots of the Fuji 100 long exposure sequence, I referenced the Sekonic and the Pentax LX meters and they do appear to match up to 50 seconds when the Sekonic started to timeout and the display no longer matches. I also timed those exposure themselves using my stopwatch to confirm the LX exposure expired at about those exact times plus my response time to stopping the watch. So to be clear, no additional exposure compensations were added.

However, one other variables that I may not have accounted for is the fact that I was using cheap daylight CFL lights at that time when they were still relatively new. Having used many of them since, I now know they are far from being able to provide a constant reliable source of light. By how much of a factor I cannot even guess.

I will definitely find those frames and do a scan and/or shot of all at once as there seems to be enough interest in that analysis of this phenomenon and that it might be helpful for others.

Les Sarile said:

I am showing individual scans of each frame without pre or post adjustment. How else would you suggest to do this? Been awhile since I conducted these tests but I'll see if I can find them.

Click to expand...

Les, What scanner are you using? How can the colors appear normal without any pre or post-adjustments? Something is doing it, maybe the scanner unbeknownst to you is automatically adjusting colors.

Alan Edward Klein said:

Les, What scanner are you using? How can the colors appear normal without any pre or post-adjustments? Something is doing it, maybe the scanner unbeknownst to you is automatically adjusting colors.

Click to expand...

I do use autoexposure settiing on normally which will influence the results to the degree shown in the over & under exposure latitude test. I suppose unless the failure exceeds that threshold, it could be masked or minimized.

_T_ said:

Well you can draw exactly and only the conclusion that if you did exactly what op did then you might also find that the effects of reciprocity failure on the look of the images produced would be minimal.

Which is still interesting in and of itself and might be taken as encouragement to explore the topic further

Click to expand...

My sentiments exactly. Whatever the theory says it is clear that with Les' workflow he is able to produce pretty well the same picture over a massive range of exposures

My only concern is whether optical enlarging could get anywhere close to being able to do the same

A digital photo(s) of the negatives might give us a better idea as to how much the negatives change over the range of exposures and then someone who dóes much more optical printing might be able to say what limitations the darkroom printer will experience

pentaxuser

brbo said:

Well, if anyone should get spanked, it should be us that haven't updated out terminology. Manufacturers use "long exposure compensation/adjustment" in their spec sheets pretty consistently.

Click to expand...

Speaking simple English as they are doing now is always easier to understand than Reciprocity failure which barely no one can interpret unless you work for Kodak.

Les Sarile said:

I do use autoexposure settiing on normally which will influence the results to the degree shown in the over & under exposure latitude test. I suppose unless the failure exceeds that threshold, it could be masked or minimized.

Click to expand...

Like everything else about negative color film, you never know what you really got. It's like a box of chocolates.

One of the reasons I like shooting chrome film even though exposure is harder to get right, is that you can hold the film up to the window and know immediately if you nailed the exposure before wasting time on a scan.

With negative color, there is a lot of play when adjusting levels, and other variables before or after the scan. So you really don't know easily if you nailed the exposure. I think what your test does prove, that even with some or even a lot of exposure or reciprocity error, negative color film scanning allows you to adjust a lot to get reasonable results. I think that's what your test proves and it is helpful to see it. Thanks.

Les, thanks for adding some more explanation on the exposures and my apologies for having misread that part in your posts. I admit I still find it puzzling that your exposures do increase as @brbo highlighted. Maybe it's the light source as you pointed out. CFL

Les Sarile said:

I was using cheap daylight CFL lights at that time when they were still relatively new. Having used many of them since, I now know they are far from being able to provide a constant reliable source of light. By how much of a factor I cannot even guess.

Click to expand...

Yeah, who knows, this may be one contributing factor. CFL's are notorious for their startup/warmup behavior. Output will then drop gradually as the bulb heats up, until it approaches a stable plateau. Maybe these effects played a role here as well.

koraks said:

Les, thanks for adding some more explanation on the exposures and my apologies for having misread that part in your posts. I admit I still find it puzzling that your exposures do increase as @brbo highlighted. Maybe it's the light source as you pointed out. CFL


Yeah, who knows, this may be one contributing factor. CFL's are notorious for their startup/warmup behavior. Output will then drop gradually as the bulb heats up, until it approaches a stable plateau. Maybe these effects played a role here as well.

Click to expand...

My expertise is not in this field so I am sure there may be many other factors I have not properly accounted for.

Interestingly, I only considered testing after seeing my results. I didn't do an extensive search, but can't say that I've seen these kinds of tests done. So, it might actually require much more then I had anticipated doing it on the cheap by the seat of my pants . . .

Oh, such things always get frustratingly complicated very fast - tell me about it!

Les Sarile said:

I didn't do an extensive search, but can't say that I've seen these kinds of tests done. So, it might actually require much more then I had anticipated doing it on the cheap by the seat of my pants . . .

Click to expand...

There seem to be far more sites and forum threads that describe the phenomenon, or criticize the experience of people who actually investigate, or repeat the vendor suggestions than there are "tests".

Hree is one that I found interesting:

Alan Edward Klein said:

Speaking simple English as they are doing now is always easier to understand than Reciprocity failure which barely no one can interpret unless you work for Kodak.

Click to expand...

Well, I've always understood it pretty clearly.
Within a certain range of light intensities, reciprocity works perfectly - if you use a series of exposures incorporating variations where the shutter speeds are doubled or halved, accompanied by apertures that are increased or decreased a stop at a time, each reciprocal pair gives negatives of the same density.
Outside that range of light intensities reciprocity fails - the negatives come out less dense, even though the shutter speed and aperture pair remain in the same reciprocal arrangement.
The "misleading" comes from when one tries to put "long exposure" into the term. The "failure" phenomena actually has nothing to do with long exposures - it arises because the light intensity is so low. Discussions about reciprocity failure reference long exposure, because that is the tool we tend to use to try to compensate for the low light levels.
The other type of reciprocity failure - the one that usually occurs only with extremely short exposure times associated with some types of electronic flash - does relate to speed, rather than light intensity