Election 74

You can use any solid and (relatively) evenly lit colour to achieve a chroma-key. Blue and Green tend to be used because they work well and people generally don't have much blue or green in their skin tone. The BBC term for the technique 'Colour Separation Overlay, CSO' is quite a good one for actually describing what's going on - you're using a block of colour on the screen to separate out a background and generate a key signal for overlaying onto something else. I've not seen yellow used myself, and I'm sure it would cause no end of grief for someone with blonde hair, but as part of a set change any colour would do (not that it happens these days).

Just take any Jon Petwee Doctor Who episode to see how bad yellow is as the key colour.

Modern 'digital' chromakeyers are actually very good at differentiating between similar hues and saturations so the problem is not as potentially severe as you might think, even the latter generations of analogue keyers could cope quite well given some time and patience. Properly set up, the keyer should easiliy be able to differentiate between say a forecaster wearing a dark or medium green jacket against a bright green backcloth for example. Best not to push your luck though I would agree.

It's important to understand that Chromakey depends on colour difference signals and not the primary colour signals themselves, blue was pretty much the default standard choice in the days of analogue because the resultant colour difference signal was larger than any other colour could produce and that's always a good starting point in analogue television terms.

That was true in the days of analogue television - where you got equal bandwidth RGB signals out of cameras. However these days green is widely used for broadcast TV because we now use YCrCb 4:2:2 which gives you significantly more high frequency detail with a decent bit depth in the green channel than in the red, and particularly the blue.

In SD TV Y=0.59G+0.30R+0.11B, so you can see in the full bandwidth Y channel (the 4 of 4:2:2) you have a lot of information from green (and thus a lot of green information is carried with high bit depth and high frequency), with less from Red and Blue (and thus lower bit depth of the high frequency R and B information). The Cr (R-Y) and Cb (B-Y) signals are carried at half bandwidth (the 2 of 4:2:2) and thus most of the R and B information is there at lower resolution. This is one major reason that green is now massively favoured over blue for studio chroma keys - as you get a cleaner, sharper result, everything else being equal.

This is why older PAL studios which had separate RGB and Composite inputs from cameras (the composite was used to generate the Fill, but the RGB analogue - not usually bandwidth limited input was used to derive the chroma key) can often generate cleaner keys with sharper edges than basic 4:2:2 digital vision mixers. (Heck - you even had separate "Hue Supressor" modules you could switch into the composite path that would remove the chroma key colour from the fill, but not the RGB signals feeding the keyer!. Very odd when you see the camera output suddenly turn all the green or blue areas grey! Though given that you often had a B&W camera previews in the stack it wasn't as disconcerting as some may imagine)

As I recall the Grass Valley GV4000 SD mixer had an option board to allow you to feed sets of analogue RGB signals from the cameras into it! We didn't use it, and I don't know of any other people who did. Fill and key timing must have been an interesting prospect.

One reason I've heard is that digital sensors in cameras are traditionally more sensitive to green than red or blue, having twice the number of sensors. Is there any truth to this?

Traditionally lead oxide ("plumbicon") camera tubes were less sensitive to red, and so 'red' was the first colour to be separated out of the light splitting optics to maximise the signal. With the introduction of CCD cameras, blue was the less sensitive sensor and so 'blue' became the first colour to be split out of the block. I don't know the current position with Cmos sensors but I suspect it is 'blue' out first for the same reasons. You may have observed on some big LE shows where there is some deliberate shooting of the studio lamps that modern Cmos cameras seem to 'dislike' very high luminance blue specular highlights - the effect is more of a blue splodge than a sharply defined highlight. I did point this defect out to Sony when they introduced their HDC 1500/1000 camera range and although they acknowledged the problem and said they would look into it nothing happened. Sale already made I guess.

In a modern broadcast CCD/CMOS camera the sensors are identical, so the number light senstive cells and resolution is identical but that's not to say the semiconductor materials used in the sensors themselves are not uniform in the sensitivity to the visible light spectrum. The colour analysis is obviously performed between the back of the lens and the target area on the individual sensors. Complicated business. I'm hoping 'Noggin' may be add some more recent practice and experience to this topic.

In current 3CCD studio cameras that isn't the case, as they use three separate sensors of equal resolution. Ikegami made an SD 4 CCD camera with two offset green sensors to increase the green resolution - but there were also issues about avoiding the EU's 3CCD import tariffs ISTR (a 4CCD camera didn't count?)

However the 4:2:2 YCrCb sampling system used to get pictures from camera CCUs into vision mixers does favour green over red and blue.

Single-sensor cameras used for location filming (and high-end 4K OBs like the Sony F55) use a single large sensor (similar to DSLRs) that use a Beyer or similar filter pattern that usually has twice as many green sensor sites as Red and Blue (often a 2x2 block with 2 Green, 1 Red and 1 Blue)

Yep - I remember that option as well, as it was specifically there to mitigate the 4:2:2 chroma subsamplign ISTR?

Ah yes, If you remember back far enough, RCA used to import and 'flat pack' build broadcast kit in the Channel Islands to avoid some sort of European tax issues, so the principle in the tv market is well established! I think that's why Sony build their studio cameras in Wales too.

Taking your points about the Beyer sensor a bit further, my understanding is that a part of the popularity (as a sensor design) is that you can get the target (Beyer sensor) very close to the back of the 'taking' lens which means you can get very narrow depth of focus fields which you cannot get in a traditional optical splitter design, i .e a traditional studio camera cannot match it for depth of focus. I understand, for cinematographers, that the ability to get a narrow depth of field is an essential requirement for 'product' designed to be shown in a cinema.

Yes - large single sensors allow a shallow DOF which is beloved by drama directors and DoPs as it allows you to compose frames and direct attention. It also has huge advantages for art directors (as out of focus backgrounds need less attention...) However they are also usually CMOS based so then have to solve the "rolling shutter" issues (high-end cameras have a global shutter, I guess a bit like Philips LDKs?)

The shallow DoF is also popular for pop concerts - though it isn't popular with camera operators working live - particularly if you are shooting 4K with HD lenses (which are a little soft to begin with)

I know Sony's 4K system camera concept is based around the F55 - a single sensor camera with a clever fibre back that will talk to HDC-series CCUs via a converter (so you use the HD CCU for talkback, reverse, prompt, racking data etc., but the video is split out upstream with a fibre pre-processor). Probably cost effective for operators who already have a pile of HD CCUs in their trucks too.

The Arri Alexxa/Ikegami hybrid system camera (not 4K but being pushed for high-end concert production, which already use Alexxas on Copperhead-type fibre back) is also single sensor.

I don't know whether the GVG LDX and Hitachi 4K system cameras are also single sensor or 3 sensor?

The NHK 8K cameras were originally 3 sensor (with a half-pixel offset on the green sensor to increase resolution as the sensors weren't 8K), but I think the new handheld is single sensor?

The only Philips LDK series I worked on were the 90/91's with the modified domestic camera sensors and rotating mechanical shutter. Sounds horrendous on paper but I have to say they worked very, very well. We had 5 of each type and in all the years they were in use we only had one failure, and only on one camera, and that was a classic dry solder joint on a pcb.

Thinking about it, I think we also had some LDK 200's, but they never went wrong - a few talkback level issues at times but no faults - that's why I have trouble remembering anything about them - because they were so reliable. I was a great admirer of Philips camera products.