white questions

[Richard Herd]

That’s an obscure question. 🙂

I encourage you to research the following:
— Sampling frequency
— discrete cosine transformations (DCT)
— Fourier transformations
— logarithmic scale
— wavelet transformations

It sounds like the general problem you’re describing is the mathematics of how photons are digitized and quantized, by these fancy gadgets we call cameras.

Another very interesting Web site is https://lurkertech.com on how all the stuff mentioned above becomes editable video.

I assume your camera is 8-bit DCT. It’s very interesting to read how 8-bit differs from 10 bit log, and how DCT differs from wavelet.

In sum, it’s absolutely amazing that these cameras work at all! Kudos to the geniuses who build these tools.

[Bob Cole]

Thanks Richard, although I try to avoid those kinds of topics. I was simply wondering whether this is something that a shooter really needs to worry about, given the in-camera processing as well as legalizing filters in the NLE.

Bob

MacPro 2 x 3GHz dualcore; 10 GB 667MHz
Kona LHe
Sony HDV Z1
Sony HDV M25U
HD-Connect MI
Betacam UVW1800
DVCPro AJ-D650

[Richard Herd]

Probably there’s nothing to worry about. I’m a curious kind of guy. Now, I think about all this in terms of computer-data rather than film-chemistry. Somewhere on the cow is an explanation of the difference between 10-bit log and 8 bit DCT. Can’t remember who wrote it, but it’s quite good.

Here’s how I think about image capture (I use a panny ag-hvx200 but recently shot a movie on an XL2), and it’s why I use my incident light meter, and I set the zebra pattern to 100%.

For example, let’s suppose your camera’s minimum luminance is 100 lux (about 10 Foot Candles) at 60i (like most of the 8-bit DCT cameras), and let’s assume the camera can capture 5 f/stops.

Anyway, the f/stops would correspond as follows (@ 320 ISO):
f/1.4 (10 fc), f/2 (20 fc), f/2.8 (40 fc), f/4 (80 fc), f/5.6 (160 fc)

A “perfect” image in the NLE, then, could be set so the f/1.4 (10 fc) corresponds to 7.5 IRE and the 5.6 (160 fc) = 100 IRE. This makes a very nice image, but requires a studio environment.

When runnin’-an’-gunnin’ exterior day, I’ve seen f/stops way up there, like f/32 (500,000 FCs!—and sometimes hotter!) (Hey, I live in South Lake Tahoe and a noon day sun over the beautiful lake is really quite bright—obviously, split ND filters and a polarizer is essential).

If we assume the same 5 f/stop contrast, then the f/stops are as follows: 8, 11, 16, 22, 32.

That means the f/32 corresponds to 100 IRE and the f/8 = 7.5 IRE.

In other words, all the available luminance that’s less than f/8 is blacked-out, like under the brim of a golfer’s hat. Or shadow in the trees, etc.

Another technique to fake extra contrast is to use gels–again, that’s impossible during external day covering a golf tournament.

[Jeff Greenberg]

Addressing #2:

Look, it’s critical you have the people in your shots exposed properly. On the variety of cameras you have, you should be using a meter or zebra striping (at 70-80%) on the highlights of someone’s face.

Yes, it’s good not to blow out your sky (or anything else for that matter…) but getting exposure right so people are visible is always more important…especially in run and gun.

Best,

Jeff G
Principal Instructor
Future Media Concepts

FMCTraining.com
Editorsretreat.com

[John Sharaf]

Setting the face at 70 ire is old school. With the new HD cameras that are so good at reaching into the blacks, we set faces at 45-55ire dependind on whether high key or low key. This automatically allows more detail in the highlights because you’re stopped down a little more.

As far as whites, it’s really a matter of on one hand not clipping the video so that it interferes with the auto circuits of home televisions and/or distorts the audio and on the other of holding detail in the white areas. When shooting white limbo’s (like the eHarmony ads or the Apple PC vs. Mac commercials) the whites are at 80 ire, yet they look really white. Where you want to hold detail in whites, like a chef’s jacket, so you can see the wrinkles and textures, this is where you have to be careful with the exposure.

I agree with the previous reply that zebras can be very useful in determining the exposure and where the clip occurs, but only if set to 100 ire. In the case of a face in front of a bright object, obviously the face is more important and unless the background is critical in the visual context it’s ok to let it go “detail-less” by overexposure, as long as you’re camera or NLE will clip it to make it “legal”.

[Bob Cole]

[john sharaf] “With the new HD cameras that are so good at reaching into the blacks, we set faces at 45-55ire dependind on whether high key or low key.”

Thanks John. That is really good information. My Sony Z1 HDV camera still seems to look best with the 70 IRE on the highlights. Thanks too for the info about going over 100 IRE and making 80 IRE “white.”

When you say 45-55 IRE, is that for the highlights? I use a “triangle” on the brightest part of the face when I meter 70 IRE.

Still curious about whether anybody heard that old truism about including some white in the shot.

Bob C

MacPro 2 x 3GHz dualcore; 10 GB 667MHz
Kona LHe
Sony HDV Z1
Sony HDV M25U
HD-Connect MI
Betacam UVW1800
DVCPro AJ-D650