Knee setup

[ColorBalance]

Hello all. New to the site.

Been playing aroud with the EX1. Wonderful camera except that can't figure out what to do with the battery when not shooting. Keep it plugged in, and it loses the juice. Very weird, I would say.

Anyway, the question I have is, has anyone played around with the Knee settings? On Sony's site, there is something on it. However, haven't been able to convince myself that it works.

From what I understand, you can use knee settings to reduce the brightness of the bright areas while keeping the dark areas at the same level. This can be a useful control when shooting outdoors on a sunny day.

Any verified setup / guidelines that someone has worked with?

[PerroneFord]

The battery drain problem is a known issue. Call Sony and get that fixed. As to the knee and picture profile stuff, this is the best resource I've found thus far:

http://www.dvinfo.net/conf/showthread.php?t=110902

[basspig]

Sony service in San Jose, CA will fix the battery problem with a new PCB installed at no charge, under warranty. Stongly-recommended.

The Knee settings control the compression of highlights in the image. It's not avaiilable with the CINE gammas, only with STD gammas. The slope controls how early, on the range from mid-tones to highlights, the compression should start. The threshold is the actual point where this compression begins.

The image sensor has a dynamic range that's maybe 600% more than can be fit into the recorded stream, according to the technical notes on Sony's web site. The knee and gamma curves can help the shooter to optimize the range of a particular scene to fit better into the constrained TV broadcast brightness range.

Don't forget to look at Knee Saturation settings too, as these control how much color is conveyed above the knee point.

[Gogu]

What do you mean by: "The image sensor has a dynamic range that's maybe 600% more than can be fit into the recorded stream " that the XDCAM codec cannot capture the full data that de sensor relays to it ?

[ColorBalance]

I am not sure if I can explain it in a technically correct way, but what it means is that the image sensor has a much wider range than the video we record. So, in order to put more in the video, you compress the image a bit (in terms of brightness). This is what the Knee setup does. By lowering the threshold and changing the slope, you can suqeeze in more brightness than you normally would. In the end, the result is that you can see the details in the bright areas as well as in the the dark areas.

[chagchag]

just som people won't panic and ucse sony of robbing them of dynamic range....

It's the same with film. There is a vast deal of information below and above blacks and highlits in what you actually see on screen.

[basspig]

What do you mean by: "The image sensor has a dynamic range that's maybe 600% more than can be fit into the recorded stream " that the XDCAM codec cannot capture the full data that de sensor relays to it ?

That's essentially what I understand it to be, based on Sony's documents on their web site. They're saying the dynamic range of the sensor is about 600% greater than what can be recorded, so the Knee (and black gamma) adjustments enable you to choose what portions of the brightness range to fit into the recorded stream.

In the old days of CCD cameras, you were lucky if the imager had enough dynamic range to make use of all 8 bits per color. Nowadays, with CMOS imagers and 14 bits per color, you have to pick and choose the best luminosity levels, based on your scene, to get even better results than what the default settings will render (both very excellent).

[Gogu]

Does an umcopressed capture solve this problem. Like the RAW data from a RED camera for example ?

[csundbom]

I really don't think this is a problem at all. The concept of dynamic range is meaningless when we talk codecs. 16 is black and 255 if peak white, end of story. It has absolutely nothing to do with the actual brightness of the captured picture.

I'm guessing the sensor has more bits than the codec, so the sensor levels have to be re-mapped to fit into the codec's 8 bits. The user has an option to manipulate that transform, and knee is one of the parameters controlling it.

[basspig]

To use another analogy, if you recall early DOS PCs, the way to get more than 640K RAM was to use "Expanded Memory Specification", and since the system couldn't see all that RAM at the same time, a method of accessing "pages" of expanded memory was devised. The result was a window, big enough to fit in conventional memory, but able to be moved around in expanded memory, to access chunks of memory that could not normally fit in to DOS conventional memory.
What the EX1 does is enable, as csundbom said, remapping of luminance levels to fit within the more constrained dynamic range of the output format.
Red One is a different situation, as RedCode RAW format contains all the information the sensor captures, and once again, this process of choosing gamma and white point knees is needed to get that footage into a destination or distribution format, but it is done in post, whereas the EX1 has to do it in-camera.

[csundbom]

Yep, that's what I meant to say. It's a good thing, allowing us to squeeze the real world dynamic range into the limitations of a 8-bit codec. Doing in-camera or in-post makes less of a difference. You just have to be careful to set these parameters for the situation and look you want.

[adamr316]

Indeed, this is much like taking a RAW formatted digital picture vs. a JPEG. The RAW file contains a lot more information (and as such has a larger file size). You can manipulate it a lot easier in post-production and do things with it that you can't to a JPEG.

Digital video, be it MiniDV, XDCAM EX, XDCAM SR, DVCPro HD, etc. is like JPEG. You set it so that it captures what will be a lot closer to your final output (or will be your final output in the case of instant turn-arounds) but at the expense of the ability to REALLY change parameters after capture.

If you saw the "true" dynamic range cameras could capture it appears very washed out and dull. It's a good thing that cameras crush the dynamic range to how we perceive things with our eyes.

[basspig]

And in reality, the camera is only part of the loop. When you think about it, the monitor is a huge limiting factor.
In the real world, light levels go from 100,000 lux to zero. But no monitor has 100,000 lux brightness, so we have to crush dynamic range, lest the sun be exposed properly but everything else is too dark.
With all the focus on imagers and CODECs, there is little thought or consideration ever given to the display technology.
When someone develops a monitor with 14 f-stops of dynamic range, a 10,000,000:1 contrast ratio and a 100,000 lumen backlight, we will see interesting things happen.