Friday, March 25, 2016

CCD versus CMOS Sensor - Magic Skin Tones or not?

I had an interesting discussion on a photography forum some time ago. In particular, there was someone who absolutely insisted that Nikon's older CCD sensors give much better skin tones compared to the modern CMOS sensors. In other words, this person's argument was that a Nikon D40 gives you better "out of the camera" images compared to a D3300, to use a model of the same category.

Is there any truth to that? Are CCD sensors magically better for skin tones than CMOS? Let's try to find out in this article. Like with all Amateur Nikon articles, I won't be using technical language, diagrams, or anything of that sort. We're photographers, not physicists; I speak with pictures, and so do you. Furthermore, the idea of "good skin tones" is subjective, and therefore not quantifiable.

First of all,we need to at least try to define what we mean by "good skin tones". As I said a moment ago, this is something entirely subjective - a bit like talking about the quality of bokeh. Still, and as with bokeh, a good number of us will probably be able to agree on what "good skin tones" means. At the very least, although defining it or quantifying it ("skin tones in that pic are three times better than in this pic") might be hard, we all know it when we see it.

So, good skin tones images are images that should have these two characteristics:
  1. Natural, true color (that is, the color you see in the photo is the true color that person's skin looks like)
  2. Good tonality (that is, there is a smooth, natural progression from one shade to the next).
Already we're facing some problems. Let's try to detect and talk about them.
  •  the word natural appears in both of the above statements. The problem is, how do you define even that? Is natural the skin you see in daylight? Indoors? In overcast? The answer of course must be related to that: true, natural color on skin tones means that it should look as that person's skin really looked in those given conditions. This is something many photographers don't realize when they start talking about "good skin tones", or a sensor (more of that later) being particularly good for red tints.
  • There are as many skin tones as there are humans. OK, this is a gross overstatement, but you know what I mean. When we're talking about "white people", or "black people", or "Asian people", we cannot possibly imply that every person in that group has the exact same (or even similar) skin tone. This is something you really need to consider when thinking about skin color. Natural skin tone does not mean you can apply a certain color or use a certain filter based on some preset (there are such plug-ins).
  • Limitations related to vision. The way John sees colors might be different than the way Jill does. For John, red might have a slight magenta tint to it, that Jill doesn't see. Oh, and is your monitor properly calibrated?

And now that we know what we're talking about, let's talk about sensors. As stated earlier, the argument affirms that CCD sensors are somehow "magically" better for skin tones than CMOS ones. That means, or so the argument goes, that straight-out-of-the-camera jpegs look more natural. At this point, let's take a look at an unedited, out-of-the-camera image from a modern, CMOS sensor.

Taken with a (CMOS) Nikon D700.
The photo above was taken with a Nikon D700 and a Samyang 85mm f/1.4 lens. I was so pleased with the out-of-the-camera result that I didn't even bother processing the raw file. In my experienced, professional opinion, the skin tones in this image are "better". Why? Because they are more natural; they correspond in a better way to reality as I remember it. Perhaps the skin tones are less warm (due to the overcast), with less deep reds in them.

Here's an image taken in sunlight (with the same D700 CMOS sensor)

Can we reach any grand conclusion merely from examining three images (which, since they are selected among untold thousands, inevitably contain some bias)? No. But I challenge you to pick any images you'd like from your personal collection, and make similar comparisons. If your comparison is a fair one (you can't blame the sensor or camera if your white balance setting is wrong), you should reach similar conclusions.

  1. No, it makes no difference whatsoever whether you're using a CCD or a CMOS sensor. CCD sensors aren't magically better for skin tones.
  2. It's different to speak of "skin tones I like" and of "good skin tones". What you like might be some weird "retro-like" Instagram filter with a green tint. It doesn't mean that's what we mean when we say "good skin tones". For most knowledgeable, experienced photographers, good skin tones means natural-looking, and not with a particular tint.
  3. Perhaps the most important conclusion. Even if you like a certain look - for example, if you like skin tones that are quite warm, it is much easier to begin from a natural skin tone and add to it, than to begin from something that can occasionally be too red and subtract from it. Quite often, entry-level cameras can have too intense reds out-of-the-camera. This is because intended audiences for these cameras like vibrant colors and high contrast. Conversely, higher-end cameras are usually tuned in a way that produces more natural results (that beginners might often find too neutral out-of-the-camera). The bottom line, however, is this: with any camera - whether entry-level or pro, CCD or CMOS - you can take the RAW file and apply every kind of adjustment you wish. If you want vibrant reds, you can do it. If you want dull, neutral reds, you can do that too. 
So, basically, the whole idea of CCD versus CMOS is in my opinion irrelevant. After all, even if we assume you are someone who never uses anything but jpegs out-of-the-camera, and you like that particular look from your D40, would you stick with it over a modern sensor that gives you much better results at anything beyond base ISO? That's hardly sense-making for someone so obsessed with results that s/he would pick cameras based on such arcane (and largely inexistent differentiation).

No comments:

Post a Comment