Next question: What is Knee.....

[Mark Thorpe]

.....and what is its importance / relevance / role? when considering filming with RED. What does knee basically translate as?

Cheers in advance,
Mark.

[Graeme Nattress]

Knee, is the name given to a portion of a curve that is used to map a high dynamic range into a lower dynamic range, for recording or display, in such a way that the image is brightened without the highlight detail being crushed totally out of existance.

Graeme

[Mark Thorpe]

Hi Graeme,
So this is something that is addressed more in post? Or is this a consideration when prepping a shot? Would this be something that would be dialled in (?) as one works with the histogram for example?

Cheers,
Mark.

[Emmanuel Decarpentrie]

Although, at first, it might not sound like a better explanation to you (unless you know a little bit about audio-music-mixing like I do), the knee's function in video is pretty much comparable to what a compressor does to an audio signal. The compressor is one of the least understood device in the audio world. In very simple terms, a compressor is a device that can automatically control the level of an audio signal. Imagine an engineer using the volume fader to keep a singer’s performance at a constant level, making fine adjustments at the right time – well a compressor does this automatically and a lot more accurately!

The two most important parameters for setting up a compressor are:

Threshold – This sets the threshold level, measured in dB (decibels). When a signal exceeds this level, compression will be applied.

Ratio – This is the amount that the signal is reduced by the compressor. A compression ratio of 5:1 means that if the signal exceeds the threshold level by 5dB, the output signal only exceeds the threshold by 1dB.

Well, it is very similar with the "Knee" in the video world, except the compressor's "Threshold" is called the "Knee Point" and the compression's ratio is called "Knee Slope". Some cameras (typically "high end", broadcast professional cameras) allow you to set up both of these parameters for you to adjust, to play around with the compression of your highlights... In order for you to capture more details in the brightest parts of your pictures...

[Mark Thorpe]

arkham,
Thanks for a straightforward and easy to follow response. So would there for example be a standard limit one would set their knee point to or I would take it that this would differ depending on the 'look' people are after for specific scenes etc?

I'm actually quite knowledgeable about audio too, I know I like rock music and I know I like it LOUD!!! haha,

Thanks again.
Mark.

[Chris Gearhart]

I dunno, but if you get a shrubbery, things should be cool.



Seriously, will Red need knee?

[Emmanuel Decarpentrie]

So would there for example be a standard limit one would set their knee point to or I would take it that this would differ depending on the 'look' people are after for specific scenes etc?

Low end cameras typically have a factory set-up (not really "standard") knee point. Some, like the DVX100, have different (gamma) presets, some of which don't really have a "knee" at all, as far as I understand them correctly. Indeed, you have to know that compressing highlights by the means of a "knee" to artificially increase the dynamic range of your picture, has an important drawback: the compressed highlight areas can look "washed out" and "pale". This is especially true when you use higher "compression ratios" or, to use the correct term, with higher "knee slope" values. This makes the picture look "like video", hence the supression of the "knee" in those famous DVX100 or HVX200 "cinelike" gamma settings.

I would take it that this would differ depending on the 'look' people are after for specific scenes etc?

Of course!

Seriously, will Red need knee?

As long as you can disable it, I don't see any reason why you shouldn't get one. It's all about having more tools to achieve the desired look.

[Graeme Nattress]

Overuse of a typical video knee does indeed contribute to the "video look". But if you're shooting RAW, then no bother at all - just dial in the "right" curve in REDCINE and you're laughing.

A compressor in audio is similar, but it has a temporal dimension that the knee in video does not, and also the compressor will be adaptive, whereas in video it's not. An equivalent in video would be some kind of contrast masking or like "shadows / highlights" in Photoshop.

Graeme

[Chris Gearhart]

. . . if you're shooting RAW, then no bother at all - just dial in the "right" curve in REDCINE and you're laughing.

Graeme

So then, . . . we are no longer the knights who say knee.

[Stuart English]

If you want to vizualize this - draw a graph with a 45 degree straight line. Mark the vertical axis from 0 to 4000 in units of 500. Mark the horizontal axis with light intensity values from 0 to 12 in units of 2.

For a sensor that has 12 stops of dynamic range, this graph is a pretty good representation of the numeric values of its digitized output. If you have a 12 bit recording stsystem, life is good as everything can be recorded.

Now consider that you have only a 10 bit recording system, what do you do with all the values above 1,000 - throw them away? No, enter the knee.

For a 10 bit system, choose a point starting at say 800, then draw the rest of the curve as a straight line so it finishes at the intersection of 1000 / 12.

The knee is the point where the slope of the graph changes... if there were a corresponding slope change near the shadows it would be called a toe.

Now consider that in an 8 bit system (HDCAM / DVCPRO HD) you only have 250 values to work with ... hence knees, and more recently toes get used a lot - and the term "film gamma curves" implies their use. But note that as Graeme has said, using a knee does not change the sensor's dynamic range, it just allows the native dynamic range of the sensor to be recorded (or viewed) on a recording system (or monitor) with a lower dynamic range. The trade off is lower signal to noise and additional compression artifacts.

As REDCODE RAW supports 12 bit recording, but any downstream NLE or video tape systems do not, the best place to use a Knee and Toe - or any kind of Tonal Response Curve - is in REDCINE as part of the "one-light" color correction process.

The second best choice is to map 12 bits down to 10 in the camera pre-recording, using a logarithmic coding which is how REDCINE RGB recordings and HD-SDI can be set up (i.e SMPTE style DPX format)