Hi!

You mentioned that [they were] curious about resolution comparison. Below, I’m going to give you the requested comparison, but before I do, I’d like to make a pre-amble to say why the usual way that people talk about and compare resolution is at best misleading if subtleties are not understood and at worst completely meaningless.

Here are quick looks at the three main reasons that the usual resolution comparisons are misleading (though there are others reasons too):

1.The usual comparison of “megapixels" assumes that the pixel-count is the resolution limiter. But it is not necessarily so (and in my opinion it is not so for most high resolution imaging) that this is the limiter. By “limiter," I mean “the attribute that makes the image go bad if you enlarge it too much.” Other limiters could be camera optics (the lens), sensor optics (the hot mirror or Opical Low Pass Filter mounted on the sensor), digital compression, and (most important in my opinion) sensor noise. Sensor noise is the really sticky one because in most cases, MORE “resolution” in a sensor also means more noise. But the noise problem appears larger than single pixels, so it’s worse as a resolution limit. The higher the amplitude that noise has, the more it appears much larger than single pixel size (even if it isn’t actually larger) because of a perceptual phenomenon called Poisson Clumping — so more “resolution” can actually be worse for cropping than less resolution in many actual cases.

2.The numeric comparisons are usually made indiscriminately interchanging RGB pixels and Bayer pixels, which are not equivalent. It’d be like comparing the time in which two different cars go zero to 60 but failing to mention that for one car it was 60 mph and for the other it was 60 km/h. But we do it all the time [with resolution]. Remember “K” just means “thousand,” it’s not a unit. So, “2k,” just means “two thousand,” it doesn’t tell you WHAT you have two thousand of. Camera raw is (usually) measured in Bayer while post files and non-raw camera formats are usually measured in RGB. Again, these two (RGB and Bayer) are not equivalent but people always talk as though they are. (There is no literal conversion like there is from miles to kilometers, but one true RGB pixel is certainly worth more than one Bayer pixel for resolution, by some unquantifiable margin).

3.The resolution rhetoric assumes that true resolution (fidelity to the original at very small sizes) is the same as perceived sharpness by the viewer. This is not so and has been unquestionably proven not so by science. This is unquestioned by actual impartial scientists as opposed to “scientists” who, say, work for companies trying to sell 4k TVs to consumers. The fact that perceived sharpness and actual resolution are not the same thing is why the numeric resolution of an image is not really a great measure of how much it can be “cropped” or enlarged in the final movie.

Okay, those are my objections to the usual way of comparing resolution, but I will now give you the comparison anyway. Below is the info for [our current job].

The thing to remember is that, because we’re shooting anamorphic, you can’t just look at the horizontal resolution because, when it gets de-squeezed, vertical resolution BECOMES horizontal resolution. So you have to just look at the total number of pixels making up the image. Here we go:

We are shooting 2.39 aspect ratio. So for [the final HD master], the image will be:

1920 x 804 = 1.54 Million Pixels (that’s RGB pixels, not Bayer)

For the anamorphic shots that we’re doing in ArriRaw, we’re shooting:

2454 x 2054 = 5.05 Million Pixels (that’s Bayer pixels, not RGB)

For the anamorphic shots that we’re doing in ProRes, we’re shooting:

1744 x 1460 = 2.55 Million Pixels (RGB)

For our [special cropped] shot:

1986 x 830 = 1.65 Million Pixels (Bayer)

So, ignoring the difference between Bayer and RGB (which I don’t like doing, but people do all the time whenever they talk about this stuff):

Our ArriRaw stuff is oversampled for HD broadcast by a factor of 3.3-to-1

Our ProRes4444 stuff is oversampled for HD broadcast by a factor of 1.7-to-1

Even our [special cropped] shot which is extracted out of the center of the screen (so you’d think it wold be understampled, not oversampled) is still oversampled, by a factor of 1.07-to-1.

Okay, hope that helps!

-Steve