Red Epic HDRx in Action

Image from fxguide.com

RED came out of the gate strong with a message of the importance of spatial resolution. We werre told that the RED One was an important camera because it “shot 4K,” and 4K is better. A more-is-more argument that I agree with only in part.

In the stills world, the obsession with resolution became the “megapixel race,” and only in the last couple of years has some sanity been brought to that conversation. Canon’s 14.7 megapixel PowerShot G10 was assailed for being a victim of too much superfluous resolution at the expense of the kind of performance that really matters, and Canon backpedaled, succeeding it with the G11 at 10 megapixels.

Why is more not always more? First, there’s the simple matter that there is such a thing as “enough” resolution, although folks are happy to debate just how much that is. But there’s also an issue of physics. Only so much light hits a sensor. If you dice up the surface into smaller receptors, each one gets less light. Higher-resolution sensors have to work harder to make an image because each pixel gets less light. This is why mega megapixels is a particularly disastrous conceit in tiny cameras, and why the original Canon 5D, with it’s full-frame sensor at a modest 12.8 megapixels, made such sumptuous images.

All things being equal, resolution comes at the expense of light sensitivity. Light sensitivity is crucial for achieving the thing most lacking in digital imaging: latitude.

What we lament most about shooting on digital formats is how quickly and harshly they blow out. Film, glorious film, will keep trying to accumulate more and more negative density the more photons you pound into it. This creates a gradual, soft rolloff into highlights that film people call the shoulder. It’s the top of that famous s-curve. You know, the one that film has, and digital don’t.

You know how sometimes you drag a story out when you know you have a good punchline?

When RED started talking about successors to their first camera, it was all about resolution. Who ever said 4K was good enough? We need 5K and beyond! Of course the Epic would be have more resolution. But would it have more latitude?

As the stills world’s megapixel race became the high-ISO race (now that’s something worth fighting for!), so too did the digital cinema world get a dose of sanity in the form of cameras celebrating increased latitude. Arri’s Alexa championed its highlight handling. And RED started swapping its new MX sensor into RED One bodies, touting its improved low-light performance and commensurate highlight handling.

Life was good.

And then Jim Jannard started hinting at some kind of HDR mode for the Epic. HDR, as in High Dynamic Range, as in more latitude.

The first footage they posted seemed to hint at a segmented exposure technique. It looked like the Epic was using two frames two build each final frame, and Jim later corroborated this. The hero exposure, or A Track, would be exposed as normal (let’s just say 1/48 second for 24p at 180º shutter). The X Track would be exposed immediately afterward beforehand (see update below) at a shorter shutter interval. Just how much shorter would determine how many stops of additional latitude you’d gain. So if you want four additional stops, the X track interval would be four stops shorter than 1/48, or 1/768 (11.25º).

The A Track and the X Track are recorded as individual, complete media files (.R3D), so you burn through media twice as fast, and cut your overcrank ability in half. Reasonable enough.

But could this actually work? You’d be merging two different shutter intervals. Two different moments in time (again, see comments). Would there be motion artifacting? Would your eye accept highlights with weird motion blur, or vise versa? Would the cumulative shutter interval (say, 180º plus 11.25º) add up to the dreaded “long shutter look” that strips digital cinema of all cinematicality?

RED’s examples looked amazing. But when the guys at fxguide and fxphd got their hands on an Epic, they decided to put it to the real test. The messy test. The spinning helicopter blades, bumpy roads, hanging upside down by wires test. In New Zealand. For some reason.

Thankfully, they invited me along to help.

But before I’d even landed in Middle Earth, Mike Seymour had teamed up with Jason Wingrove and Tom Gleeson to shoot a little test of HDRx. They called it, just for laughs, The Impossible Shot.

This is not what HDRx was designed to do. It was designed to make highlights nicer. To take one last “curse” off digital cinema acquisition. This is not that. This is “stunt HDRx.”

And it works. Perfectly.

Sure, dig in, get picky. Notice the sharper shutter on the latter half of the shot. Notice the dip in contrast during the transition. The lit signs flickering.

Then notice that there’s not another camera on the planet today that could make this shot.

I guess Mike should really have called it “The Formerly Impossible Shot.”

Read more at fxguide, and stay tuned to fxphd for details on their new courses, coming April 1.

Update on Tuesday, April 19, 2011 at 2:30PM by Stu

Graeme Natrress confirmed for me that the X track is not sampled out of the A Track interval, but is in fact a seperate, additional exposure. There is no gap between the X and A exposures, but they don’t overlap.

The just-posted first draft of the Red Epic Operation Guide has a few nice deatils about HDRx as well.

Epic Movie

I’m in New Zealand with Mike Seymour and John Montgomery shooting with their new Epic M. Anything more I could say about that is probably better expressed by this video, shot by John:

Mike has an ongoing thread going on reduser about his first Epic experiences, and is updating his Dean’s Blog over at fxphd. I’ll have some thoughts on the camera soon, but for now, suffice it to say that RED picked the right name for it.

I thought Mike needed a behind-the-scenes shot as nice as the one John got of me, so I grabbed this on our afternoon run: 

The Shot You Can Make Gallery

Many people commented on The Shot You Can Make that they’d love to see this as an iPhone app. Great idea, but I can’t make that happen quickly, so I looked around for other options. It turns out the photo gallery feature of Apple’s MobileMe service (AKA the litmus test of whether you’re totally Steve Jobs’s bitch) has some impressive features that work pretty well for presenting whichever camera and lens combinations I’ve pre-made using the simulator. It’s not the same as a dedicated app that would allow you to enter arbitrary camera backs, focal lengths, and f-stops, but it does do some cool stuff.

From your browser, visit gallery.me.com/prolost. You can browse the pre-rendered images of course. You can also subscribe to an RSS feed to be notified of updates. Images you have not yet viewed are marked with a blue dot.

SYCM Gallery on iPhone

But perhaps the coolest feature is for those who have an iPhone, iPad, or iPod Touch. Download the free MobileMe Gallery app if you don’t have it already, and add “prolost” as a friend (you don’t need a MobileMe account of your own for this). You can now browse the SYCM gallery using a lovely graphical interface. You’ll see the familiar blue dot next to albums with new images, so it’s easy to see when I’ve made an update. Better still, the app keeps a local cache of images you’ve viewed, so you can continue to browse them even if you’re not connected to a network.

SYCM Gallery on iPad

I’m not ruling out the possibility that this could be a real app someday, but in the meantime, this gives you a taste of SYCM on the go, and it’s free. Do check it out, as I’ve added some of the most requested cameras, including the RED fixed-lens Scarlet and the Sony EX1. I’ve also created a special gallery reserved for images that perfectly match the baseline shot’s AOV, so you can get a more direct comparison of DOF performance on this setup.

Enjoy, and keep those requests coming for cameras and lenses.

The Shot You Can Make

There are some exciting new camera options out there these days, ranging from inexpensive, large-sensor hybrids from the likes of Panasonic and Canon to groundbreaking high-end digital cinema rigs from folks like RED and Arri. When contemplating buying or using these cameras, one has many resources to evaluate things like focal length equivalencies and depth of field. This chart from Barry Green is a great example — it’s extremely helpful and the kind of thing you could refer to again and again.

In fact, if you’re like me, you’ll need to refer to it again and again, because there’s nothing really intuitive about it. It’s a bunch of numbers. You might find yourself staring at it and thinking, “Yeah, but what kind of shot can I actually make with these camera/lens combinations?”

Read down the thread on DVXuser and you’ll see one member exclaim with glee that with only three Panasonic zoom lenses, he can cover every focal length from 7mm to 300mm on his Micro Four-Thirds (MFT) rig. Barry quickly points out that this is true, but at undesirably slow apertures. The poster is, I’m sure, left scratching his head, trying to grok just how these slower lenses are going to affect the cinematic look and feel of his shots.

You can feel my own frustration with this in my article on the Panasonic AF100. I discussed there the dilemma of MFT — is it “big enough” for cinema? Yes, but if you want cinematic DOF, you’ll need fast lenses, faster than most of what Panasonic makes.

Another example — you might be looking at the recent demonstrations of the fixed-lens RED Scarlet and wondering just how much DOF control you’ll have with an ƒ2.4 lens on a 2/3” sensor.

All the charts and sensor-size comparison images and spec sheets won’t answer the question: “Yeah, but what shot can I make?”

So I’m introducing a new Prolost feature designed to help answer that question. It’s called the Shot You Can Make (SYCM) Simulator, and it’s sort of a 3D “Marcie” for focal length and depth of field. Here’s what it looks like:

I started with a shot from a movie called 12 Rounds. Directed by Renny Harlin, 12 Rounds is an action flick set in New Orleans and starring WWE’s John Cena. It’s the kind of movie you can rely on to contain the kinds of shots you see in many movies — in this case, a guy (Cena) with a gun. The focus is on Cena’s face, the gun is slightly soft, and there are a few big, boke-liscious out-of-focus lights in the background.

12 Rounds was shot with Panavision cameras on Super 35mm film. My guess is that this particular shot was made with a 125mm Panavision Primo at ƒ2.0. Based on this estimation, I simplified the image into cartoon-shaded layers and split them out in 3D in Adobe After Effects, essentially recreating a simplified model of what was in front of the lens that night. Using a lens blur plug-in rigged with expressions, each 3D layer gets the correct amount of defocus for its distance from camera. The result is a simulation of the shot with accurate angle of view and depth of field.*

The shot contains a number of recognizable things, like a man, a hand, a gun, and some distant lights. If you’re familiar with your camera — any camera — you can probably easily imagine what kind of shot of this setup you could make. But what about a camera that you’re not familiar with?

The Shot You Can Make Simulator allows me to place any camera I want, with any lens, at any stop, into this same scene, and re-photograph the virtual scene with that rig. In this way it provides a real-world-ish benchmark for the kind of lens performance that matters most to filmmakers: 

• What’s the angle of view, i.e. how wide or telephoto is this lens?
• What kind of depth of field performance can I expect? I.e., what will be my ability to isolate my subject from the background using focus?

Here’s an example. Simulating a Canon 5D Mark II with a 50mm lens at ƒ1.4, you can see that I’ve had to move closer to my virtual Cena to achieve similar framing. His hands appear larger, and much softer. The lights in the background are still blooming, but not as much. Although we’ve opened up about a stop and focused closer, we’ve also gone wider in AOV, so our ability to make big circles in the background has diminished.

Another example, this time simulating the shot you might make with a Canon 5D Mark II using the Canon 70–200 F2.8 II IS, at maximum zoom:

Note that we now have a slightly narrower Angle Of View than the Super 35 125mm baseline, so we’ve stepped back a bit to maintain the framing. But despite focusing longer and stopping down, we have a much larger image sensor on the 5D, so we maintain the same “feel” in terms of the softness of the gun barrel and the size of the background boke.

Now, if someone tells you that 200mm on a 5D Mark II is a decent rough match for 125mm on Super 35, that’s useful information. But unless you’re a Cybog Killer from the Future, you might have a hard time getting a sense of how all the other factors will balance out — slower lens, but larger sensor, but longer focus, but more money left over for tacos.**

I was trying to explain to someone the other day why I felt that the MFT zooms Panasonic has on offer are not very sexy. Maybe this will help — here’s the same shot at 140mm ƒ5.8, the max zoom of a popular “do it all” MFT lens from Panasonic:

Even though we’re zoomed way in, tighter than the baseline shot, we can just barely soften the background, and the gun is razor sharp. Interestingly, the Shot You Can Make with the Panasonic 14–140mm is not unlike the Shot You Can Make with a Canon HV20, which has a tiny sensor by comparison, but a surprisingly fast lens. Here’s the HV20 making the shot at its max zoom of 61mm, ƒ3.0:

The similarity between these images shows that you can very easily slap a lens on a MFT camera that will completely undo any perceived DOF advantage of the large sensor. To me, this is useful information. If you agree, I’ll use the SYCM Simulator to profile lenses and cameras I discuss here.

* My respect for you the reader demands that I generally avoid disclaimers, but in this case I would like to point out that this is all guesswork on my part, from the original lens used to the dimensions of the set. And I could very likely have my math wrong on any of the DOF calculations too. Please let me know if anything jumps out at you as wrong.

** Tacos have not yet been integrated into the SYCM Simulator.

Update on Thursday, January 13, 2011 at 10:26AM by Stu

Be sure to also check out this camera/lens comparison tool from Abel Cine. I’d been looking for the link when writing this, but luckily Jason Wingrove helped me out on Twitter.

Update on Wednesday, January 19, 2011 at 9:34PM by Stu

Those of you asking about a mobile app version of the SYCM Simulator, be sure to check out this follow up post.