September 30, 2014

The crappy digital look: Demystifying the lie about how sensors vs. film handle light sensitivity

In this installment of an ongoing series (search "digital film" for earlier entries), we'll explore another case of digital photography offering supposedly greater convenience at the cost of compromised image quality. The end result is pictures that are too harshly contrasting, more pixelated, and color-distorted where it should be white, black, or shades of gray.

This time we'll look into the properties of the light-sensitive medium that records the visual information being gathered into the camera by the lens. I have only read one off-hand comment about the true nature of the differences between digital and film at this stage of capture, and mountains of misinformation. The only good article I've read is this one from Digital Photo Pro, "The Truth About Digital ISO," although it is aimed at readers who are already fairly familiar with photographic technology.

Given how inherent the difference is between the two media, and how much it influences the final look, this topic is in sore need of demystifying. So hold on, this post will go into great detail, although all of it is easy to understand. By the end we will see that, contrary to the claims about digital's versatility in setting the light-sensitivity parameter, it can do no such thing, and that its attempts to mimic this simple film process amounts to what used to be last-resort surgery at the post-processing stage. One of the sweetest and most alluring selling points of digital photography turns out to be a lie that has corrupted our visual culture, both high and low.

To capture an image that is neither too dark nor too bright, three inter-related elements play a role in both film and digital photography.

The aperture of the lens determines how much light is gathered in the first place: when it is wide open, more light passes through, when it is closed down like a squint, less light passes through. But light is not continually being allowed in.

The shutter speed regulates how long the light strikes the light-sensitive medium during capture: a faster shutter speed closes faster after opening, letting in less light than a shutter speed that is slower to close, which lets in more light.

Last but not least in importance, the light-sensitive medium may vary in sensitivity: higher sensitivity reacts faster and makes brighter images, lower senstivity reacts slower and makes dimmer images, other things being equal. This variable is sometimes labeled "ISO," referring to the name of a set of standards governing its measurement. But think of it as the sensitivity of the light-reactive material that captures an image. This scale increases multiplicatively, so that going from 100 to 200 to 400 to 800 is 3 steps up from the original. Confusingly, the jargon for "steps" is "stops."

A proper exposure requires all three of these to be in balance — not letting in too much or too little light through the lens aperture, not keeping the shutter open too long or too briefly, and not using a medium that is over-sensitive or under-sensitive. If you want to change one setting, you must change one or both of the other settings to keep it all in balance. For example, opening up the lens aperture lets in more light, and must be compensated for by a change that limits exposure — a faster closing of the shutter, and/or using a medium that is less light-sensitive.

Between digital and film, there are no major differences in two of those factors. The lenses can be opened up and closed down to the same degree, whether they are attached to camera bodies meant for one format or the other. And the shutter technology follows the same principles, whether it is opening up in front of a digital or film recording medium. (Digital cameras may offer slightly faster maximum shutter speeds because they are more recent and incorporate improvements in shutter technology, not because of digital properties per se.)

However, the two formats could not be more different regarding light-sensitivity of the recording medium.

Film cameras use rolls of film, which are loaded into and out of the camera on a regular basis. Load a roll, take however-many pictures, then unload it and send it off to the lab for development. The next set of pictures will require a new roll to be loaded. Digital cameras have a light-sensitive digital sensor which sends its readings to a memory card for later development and archiving. The sensor is hardwired into the camera body, while the memory card is removable.

Thus, no matter how many pictures you take with a digital camera, it is always the exact same light-sensitive piece of material that captures the visual information. With a film camera, every image is made on a new frame of film.

A digital sensor is like an Etch-a-Sketch that is wiped clean after each image is made, and used over and over again, while frames and rolls of film are like sheets of sketching paper that are never erased to be re-used for future drawings. The digital Etch-a-Sketch is just hooked up to a separate medium for storing its images, i.e. memory cards. Frames of film are both an image-capturing and an image-storage medium wrapped up into one.

Whether the light-sensitive material is always fresh or fixed once and for all has dramatic consequences for how it can be made more or less reactive to light — the third crucial element of proper exposure.

Film manufacturers can make a roll of film more reactive to light by making the light-sensitive silver halide crystals larger, and less reactive by making the crystals smaller. Hence slow films produce fine grain, and fast films large grain. What's so great is that you can choose which variety of film you want to use for any given occasion. If you're worried about too much light (outdoors on a sunny summer afternoon), you can load a slowly reacting film. If you're worried about not getting enough light (indoors in the evening), you can load a fast reacting film.

It's like buying different types of sketching paper depending on how much response you want there to be to the pencil lead — smooth and frictionless or bumpy and movement-dampening. Depending on the purpose, you're able to buy sketchpads of either type.

What was so bad about the good old way? The complaints boil down to:

"Ugh, sooo inconvenient to be STUCK WITH a given light sensitivity for the ENTIRE ROLL of film, unable to change the sensitivity frame-by-frame. What if I want to shoot half a roll indoors, and the other half outdoors?"

Well, you can just buy and carry two rolls of film instead of one — not much more expensive, and not much more to be lugging around. And that's only if you couldn't compensate for changes in location through the other two variables of aperture size and shutter speed. For the most part, these were not big problems in the film days, and served only as spastic rationalizations for why we absolutely need to shift to a medium that can alter the light-sensitivity variable on a frame-by-frame basis, just as aperture size and shutter speed can be.

That was the promise of digital sensors, which turns out to be a fraud that the overly eager majority have swallowed whole, while enriching the fraudsters handsomely.

Digital cameras do offer a means for making the image look as though it had been captured by a material that was more sensitive or less sensitive to light, and this variable can be changed on a frame-by-frame basis. But unlike film rolls that may have larger or smaller light-sensitive crystals, the photodiodes on the digital sensor have only one level of sensitivity, inherent to the material it is made from.

Because this sensitivity is baked into the materials, it certainly cannot be altered by the user, let alone on a frame-by-frame basis. And because the sensor is not removable, the user also has no recourse to swap it out for another with a different level of sensitivity.

How then do digital cameras attempt to re-create the many degrees of sensitivity that film offers? They choose a "native" sensitivity level for the photodiodes, which can never be changed, but whose electronic output signal can be amplified or dampened to mimic being more or less sensitive in the first place. In practice, they set the native (i.e. sole) sensitivity to be low, and amplify the signal to reach higher degrees, because dampening a highly sensitive "native" level leads to even lower quality.

Most digital cameras have a native (sole) sensitivity of ISO 100 or 160, meant to evoke the slowly reacting less sensitive kinds of film, and allow you to amplify that signal frame-by-frame, say to ISO 800, 3200, and beyond. But remember: it is never changing the "ISO" or sensitivity of the light-reactive material in the sensor, only amplifying its output signal to the memory card.

It is like always recording sound at a low volume, and then using a dial on an amplifier to make it louder for the final listening, rather than record at different volume levels in the initial stage. And we all know how high-quality our music sounds when it's cranked up to 11. It does not sound "the same only louder" — it is now corrupted by distortions.

We should expect nothing less from digital images whose "ISO" was dialed up far beyond the native (sole) sensitivity of 100 or 160.

Below are some online digital test shots taken with the lens cap fully in place, blocking out most light, with higher and higher settings for the faux-sensitivity ISO setting. Now, these images should have remained black or gray the whole way through. The only change that would have occurred if they were shot on more and more highly sensitive film material is a grainier texture, owing to the larger film crystals that make film more sensitive, and an increase in brightness, since what little light was sneaking in past the lens cap would have produced a stronger reaction.

And yet look at the outcome of a digital sensor trying to see in darkness:


Not only does the texture get grainier, and the light level brighter, when the native (sole) sensitivity is amplified, there are now obvious color distortions, with a harsh blue cast emerging at higher levels of sensor amplification.

What's worse is that different cameras may produce different kinds of color distortions, requiring photographers to run "noise tests" on each camera they use, rather than know beforehand what effects will be produced by changing some variable, independent of what particular camera they're using.

The test shots above were from a Canon camera. Here's another set from a Pentax, showing a different pattern of color distortions.


Now it's red instead of blue that emerges at higher levels of amplification. Red and blue are at opposite ends of the color spectrum, so that shooting a digital camera without test shots is like ordering a pizza, and maybe it'll show up vegetarian and maybe it'll show up meat lover's. Unpredictable obstacles — just what a craft needs more of.

These distortions can be manipulated in Photoshop back toward normal-ish, but now you've added an obligatory extra layer of corrections in "post" just because you want to be able to fiddle with light-sensitivity frame-by-frame, which you're not really doing anyways. Convenience proves elusive yet again.

So, if amplification of the native (sole) light sensitivity is not like using film rolls of different sensitivities, what is it like? As it turns out, it is almost exactly like a treatment from the film era called push-processing, which was a last-ditch rescue effort in the developing stage after shooting within severe limitations in the capturing stage.

Suppose you were shooting on film, and your only available rolls were of sensitivity ISO 100, which is a slowly reacting film best suited for outdoors in sunlight. Suppose you wanted to shoot an indoor or night-time scene, which might call for faster reacting film, say ISO 400. Could it still be done with such low-sensitivity film? You decide to shoot in the evening with a slow film, effectively under-exposing your film by 2 stops, worried the whole time that the images are going to come back way too dark.

Lab technicians to the rescue! ... kind of. If you let them know you under-exposed your whole roll of film by 2 stops, they can compensate for that by allowing your film to soak in the chemical developing bath for a longer time than normal, allowing more of those darkened details to turn brighter. (The film starts rather dark and the developing bath reveals areas of brightness over time.) Taking 100 film and trying to make it look as sensitive as 400 film is "pushing" its development by 2 stops.

But if that were all there were to it, nobody would've bothered using films of different sensitivities in the capturing stage — they would've let the lab techs worry about that in the developing stage. The costs of push-processing are various reductions in image quality, which Kodak's webpage on the topic summarizes in this way (click the link for fuller detail):

Push processing is not recommended as a means to increase photographic speed. Push processing produces contrast mismatches notably in the red and green sensitive layers (red most) compared to the blue. This produces reddish-yellow highlights, and cyan-blue shadows. Push processing also produces significant increases in film granularity. Push processing combined with under exposure produces a net loss in photographic speed, higher contrast, smoky shadows, yellow highlights and grainy images, with possible slight losses in sharpness.

Not a bad description of the signature elements of the digital look, is it? Blue shadows are exactly what the Canon test shots showed earlier.

Interestingly, they note that although push-processing produces less sharp images, they may subjectively appear to be normally sharp, given the increase in contrast. Sure, if a subject is wearing a normal red shirt and normal blue jeans, and you crank up the contrast parameter, the picture looks more defined — ultra-red juxtaposed against ultra-blue. But we're only fooling ourselves. Sharpness means how clear and crisp the details are, and push-processing and its obligatory counterpart in the digital world are actually losing details, while distracting us with more strongly contrasting colors.

Remember, this is what a digital camera is doing each time it takes a picture outside of its native (sole) sensitivity level of 100 or 160, i.e. when you shoot indoors, at night, or on cloudy days. In the digital world, every image is immediately rushed into emergency surgery.

Is there a way to compare side-by-side a film image that was processed both normally and with push-processing? Unfortunately, no, since developing the negative image from the latent image on the film cannot be undone, and then done a different way. I suppose you could take a shot of the same scene, with two identical cameras and two identical rolls of film, but with one camera set to the true sensitivity and the other set inaccurately, then develop the normal one normally and the under-exposed one with push-processing. That sounds like a bit too much just to make a perfect textbook comparison of normal vs. push-processed images, and I couldn't find any examples online.

But there are examples of film that has been push-processed. Although we can't compare them side-by-side with normally developed versions of the same film frame, at least we can pick up on some of the typical traits that push-processing introduces. Below is an example from this series at a photographer's website. The film is ISO 400, but was push-processed to look like ISO 3200. That is 3 stops of pushing, whereas Kodak and other photography guidebooks advise never pushing past 2 stops of over-development.


It's disturbing how digital this film photograph looks. It looks like someone opened a digital image in Photoshop and cranked up the contrast and saturation settings. Look for details on the man's shirt and pants, like folds and creases. They're hard to make out because push-processing renders the image less sharp. But we're distracted by how striking the contrast is between these overly rich yellows and reds and the cooler blues. It looks more defined, but is poorer in detail.

It's almost like a child drew an outline of pants and hit "fill" with yellow on MS Paint. Very little detail. The yellow pole also looks like a crude "fill" job. Even worse, these pictures were shot on medium-format film, which has a far higher resolution than the 35mm film we're all used to. It ought to have detail so fine that you could blow it up into a poster or banner without blurring of the details.

We also see the familiar blown-out sky from digital Earth, rather than the blue one we know and love. Other white areas look like intense spotlights, too. I can't tell if they have the red-yellow tint to them as Kodak warned, although they do look kind of bright pale yellow. There aren't many dark shadows to tell if they have the bluish tint warned about, although the asphalt on the road looks blue-gray. The color distortions might be more obvious if we had the same scene captured and developed normally, for comparison.

The ultra-contrasty, overly saturated, harshly blown-out bright areas are hard to miss, though. And they look like something straight from a digital camera plus Photoshop settings dialed up to 11.

You might object that, hey, this guy knows what he's doing, and he's using push-processing to give the pictures a flamingly dramatic style (he's gay). That misses the point: these kinds of distortions and reductions in image quality are built in with digital photography's light-sensitivity technology. They aren't going to be chosen purposefully for some intended artistic effect. They're just going to make ordinary people's pictures look cartoony and crappy because they don't know about them before buying a digital camera, and won't mind anyway because digital is all about convenience over quality.

Even Hollywood movies shot by pros will be subject to these digital distortions, although they'll have much better help cleaning them up in post — for a price. Good luck scrubbing your digital images that clean on your own with Photoshop.

In the end, is digital really more convenient, all things considered? All of these distortions require laborious and expensive corrections, which may well off-set the efficiency gains that were hoped for at the beginning. Or those corrections simply won't be done, and greater convenience will have been traded off against poorer quality. Either way, one of the fundamental promises of digital photography turns out to be a big fat lie.

4 comments:

  1. My father was an early adopter of digital cameras for two reasons: integration with computers and ability to preview images in the field, which was an unspoken tradeoff (as I see it in retrospect)

    A really interesting phenomenon results.

    When you have a limited number of shots, you conserve what you have left on the roll. Because of this, you are much more willing to stand in the rain, stand on a cold mountain, or some other equally uncomfortable place. The thought of wasting shots is much more unbearable than if you wait just a little longer and the weather breaks and you get the shot. As a kid, it was miserable- but god damn was it living, some of the memories that resulted are etched forever.

    With digital, the resulting mental laziness is like "eh, you get the picture i'm in the mountains." Fire the shot and go home. Push brightness and saturation in Pshop.

    Here's another aspect: when the shots ran out prematurely, out trip would get extended while he rush mailed more film to some remote town we would stop in. As a kid, that made the adventure longer. Or, if the weather was great the trip would get extended as well- better to stay and capitalize while its there.

    I haven't seen him do that since film.

    ReplyDelete
  2. Sigh. What's amazing is that if you look at your first sample, you'll see that ISO 100, 200, and 400 have almost the same exact noise pattern. Even 800 is close.

    Whereas if you'd pushed ISO 100 film to 400, you'd have all sorts of blocked shadows and wonderful color casts. And yep, the sensor on the Canon starts degrading pretty bad past 1600, and blows at 6400. But that's pushing your 100 film 4 stops and 6 stops, respectively. What would have Ecta E-100 have looked like pushed 4 stops? Or 6?

    At least with the digital camera you would have HAD a shot.

    But yeah, shoot 6400 on a basic P&S and blow the color and saturation out in PS and you're going to get a terrible image, quality wise.

    Who knew?

    ReplyDelete
  3. Film photographers rarely or never needed to push from 100 to 400 because 400 film was widely available to shoot on in the first place. Pushing is only ever useful when you're trying to go far beyond what is normally feasible. You get 800 film, but it's still night-time, indoors, and poor lighting, so try to push a few stops.

    Whatever would've happened pushing film from 100 to 400 is irrelevant -- never happened in practice. But what happens with higher sensitivities in digital is relevant -- that's the whole point of the amplification wars, and nerdgasms about "OMG, can you believe ISO 6400?!"

    "At least with the digital camera you would have HAD a shot."

    You don't get that shot for free. What allows digital to get a halfway-clear shot indoors, with poor lighting, at night makes its pictures look crappy when taken during the daytime, outdoors, or anywhere with halfway-decent lighting.

    Getting the shot in the dark is more of an abstract benefit, a hypothetical prize. Practically, those situations don't have much worth taking pictures of.

    And they are traded off against image quality in the situations that actually matter -- when people can see well enough to carry out their activities, and when the environment is visible enough to be worth exploring. Film wipes the floor with digital where it really counts.

    ReplyDelete
  4. Yes, but with film ISO 400 came with a cost: grain. Heck, some films had so much boulder-sized grain (Scotch 640T) that some photographers started snarfing it up and created their own "style" with it. (Sarah Moon)

    One might also mention that other films (Kodachrome 25, Velvia 50) were so contrasty that they tended to hand you the defects of your sample image out of the box, no push processing required.

    The only way to combat film grain was to start hauling out Hassie's and Sinars and start shooting 6x6 and 4x5 and 8x10, and that brought with it it's own set of problems, like needing 50,000 w/s worth of ProFoto lighting gear to get more than a few centimeters of DOF.

    Before I get further into this, just what "Canon" camera is your sample from?

    ReplyDelete

You MUST enter a nickname with the "Name/URL" option if you're not signed in. We can't follow who is saying what if everyone is "Anonymous."