From analog to digital: How to make the transition smoothly

The world we live in is analog. The analogue realm is where we find the beauty we admire. A rose; a sundown; a pretty face: It is difficult to intuitively comprehend how these images’ essence can be recorded and reproduced as a string of discrete, rigid values.

However, at a very high level, we digitally record, process, and reproduce the natural world. The basic idea is well-known to most of us: There is nothing lost “in the gaps” because there is nothing significant to lose if you sample frequently enough and with sufficient depth per sample. Put another way, If you sample well, you’ll get everything you need for an accurate and faithful reproduction.

This is unquestionably undisputed. This is not the same as the debate over CDs versus vinyl. This is about how far you can go to capture images using the most advanced technology available. In addition, the answer is very far away.

I wanted to find out what Graeme Nattress, RED Digital Cinema’s “Problem Solver,” thought about the process of switching from analog to digital.

What are the problems and how can they be solved?

(The “Problem Solver”) Graeme probably deserves a title like that. However, it provides little insight into the kinds of issues that he resolves. His primary title would probably be “Chief Image Scientist” in any other company.

RedShark Media’s (DS) Editor-in-Chief David Shapton:

The concept of sampling, or taking measurements at predetermined intervals to capture a scene, appears to be antagonistic to the idea of a smooth analog world.

RED’s Graeme Nattress (GN): Because we have a shutter speed and discrete time samples in the sense of one frame following another, we have always sampled in the temporal (or time) domain. To create a video, we take a predetermined number of pictures each second. Sample rate is shutter speed: digital sampling We’ve always carried it out. It’s a part of making a movie. Therefore, imagining that we also sample in the spatial domain is less of a conceptual leap. This means that, on the one hand, you have frames, which are your temporal sampling, and, on the other, you have to sample vertically and horizontally what’s in a frame. Each of these spatial samples is a pixel with a number that represents a color. We can record and represent more colors with more bits per pixel. Naturally, as well as colors and dynamic range.

We can very accurately reproduce analog images with enough frames per second, enough pixels, and enough depth per pixel.

DS:

Additionally, we do receive analog artifacts that we are accustomed to in the digital realm, don’t we? Is the reverse movement, like on wagon wheels, simply temporal aliasing? Also, moire: Occasionally, a fence can be seen through another fence. I assume that the sensor’s grid-like structure takes the place of one of these fences in digital?

GN:

Temporal aliasing can take many forms, including reverse wagon wheels and “stutter” on pans. To avoid temporal aliasing on movement, use the speed indicated in the ASC data tables on panning speed.

When I was a young child, the aliasing patterns that appeared on highway bridges when the metal grille on one side of the bridge interfered with the metal grille on the other side fascinated me. The moire would get bigger and more angular as the car drove toward the bridge. I was intrigued even though I had no idea what the effect was or what it was called.

Although aliasing is clearly visible as a moire pattern when two repeating patterns are overlaid, it is not necessary for aliasing to occur. When you look at a screen door, you can see that aliasing can occur in any sampled system—even though it is most obvious in uniform sampling, randomizing the samples does not eliminate aliasing; rather, it only disguises it. Aliasing will occur throughout the scene, but you will be able to see it most clearly on sharp edges. A moire pattern will appear as a result of aliasing if you have another screen door.

As a result of the sensor’s grid-like sampling structure in digital video, aliasing must be taken into account.

DS:

It is essential, in my opinion, to keep in mind that sensors are analog devices. They are there to produce a “signal” that represents the image; however, before it can be processed digitally, the output from each photosite must be fed into an analogue to digital converter. In the analog domain, what aspects of a sensor are crucial?

GN:

In the analog domain, dynamic range matters. The dynamic range of the various sensors that RED has produced over time has generally increased: sometimes in the shadows and other times in the highlights. There is a change in the noise characteristic as you descend into the shadows, as well as how dark the detail can be while still being visible. The noise has a textural quality to it. Take a look at the newest HELIUM and MONSTRO sensors. Because MONSTRO is seeing deeper into the dark and the mid-tones are less noisy, you need to think about noise across the entire luminance range rather than just the noise in the dark.

If you know your math, our perception of noise is like integrating under a curve. When working with raw data and without a predetermined rendering intent, we must be concerned about noise at all brightness levels because we have not yet determined which levels in the resulting image will be important or displayed in what contrast.

This is where comparing sensors and cameras of different generations can be challenging: They are not contrasting similar things. And if the noise performance at the end points improves while it does not at the middle, the improved dynamic range might not actually be noticeable.