How does Television work?

[Mike Watterson]

All fascinating!

But what about the basic principles, e.g:

If one were to slow time right down and watch an old CRT display, one would see a scanning spot with trail, painting the picture in orderly fashion, pixel by pixel, line by line. Regardless of picture content.

But if one were to do the same with a modern digital display panel - or say a mobile phone - what would one see?

There are very many display technologies now. Most will use a block of RAM for a frame, with the data three or more bytes per pixel, organised in sequential lines.
There are:
DLP (mechanical nano sized mirrors) Either one chip and a spinning colour wheel (R G B & Clear) or three chips. LED, laser or incandescent lamp
Plasma
CRT
LCD: Mono with backlight and coloured filter. There are three main filter layouts: vertical stripes, 2 x 2 Bayer like some cameras (R G and G B on alternatel lines) and R G B Y.
LCD projection: Mono with LED or incandescent backlight and coloured filter. Rarely three panels.
Mechanical Projection. Like a laser printer. No focus needed. Polygon disc mirror for lines, polygon drum for frame.
OLED: Not real LEDs, but diode like Electroluminescent dots with phosphors. Blue  wears faster!
Real LEDs. Both giant walls and regular displays. Very expensive. Sony calls it Crystal LED. longest life.
eInk (milky liquid bistable cells with black and white particles). Usually mono (white, black and 14 greys) but a colour filter on top can give dim pastel colour or very dim colour (4096 variations only). Rubbish at video and ghosts unless entire page is refreshed.

Lower resolution may multiplex by rows (lines) then columns (pixels) for the entire panel. Two edges minimum. Or the panel may  be in two parts separately refreshed, or four parts. A panel may have on board amorphous / thin film transistors at each pixel (LCD, OLED or real LED) and then there are less connections to the panel.
The display RAM can be two port or alternately accessed by the panel HW and the video output. Some panels thus need not be organised by TV line (row) and pixels per line (column).

DLP using a colour wheel in theory has to use four times the real frame rate. RCA before shadow mask had a mono CRT inside a drum with coloured panels.

The three panel (LCD or DLP) projectors use Dichroic prism block like in three tube or three chip cameras. The light source can be monochrome LEDs or incandescent or Xenon discharge on very bright projectors.

Single tube and then single chip cameras used vertical strip filters. Cameras now use
R G
G B
pattern  https://en.wikipedia.org/wiki/Bayer_filter
The stripe system requires 3x horizontal resolution on display or sensor
The Bayer pattern needs 2x horizontal and vertical resolution, which is better for fabrication. It's also brighter on display and more sensitive on the camera.
The camera image, no matter if stripe, Bayer or three chip, is processed to give most resolution to the Luminance and less to hue/saturation, at some stage even if the data is RGB, which will be serialised as RGB per horizontal pixel, then rows of pixels then frames. Digital compression may need the columns and rows to be exactly divisible by 16, even though compress may use  8 x 8 blocks.

Common resolution schemes are 4:2:2 and 4:2:1.

Basically no matter how the camera sensor or display  really does images, at some stage at origination and reception/display  there is a frame of video organised by lines, pixels and what ever number of bits or bytes per pixel there is. Traditionally full colour used 8 bits each for R, G and B, thus 256 levels which is about 16.7 million possibilities. But there is now video with more levels and also images can have a transparency channel (though not usually for video).

A Screen Capture of a video frame (rare) or static image on many systems with a 16 shade mono eink panel will be in full 24 bit RGB colour. This is the case for ALL display systems and cameras; somewhere between the decompression and the panel one or more frames of video as horizontal lines of RGB pixels exist. There might be a separate RAM frame for the R G & B, or the bytes might be sequential in the "pixels" for the line.

[Mike Watterson]

Note the real native internal refresh frame rate of a panel may be different to the video frame rate and it may be updating multiple areas (half, interlaced or quadrants) alternately or at the same time.

I have one 1600 x1200 laptop from 2002 which uses interlace for the panel, even though everything feeding it is progressive. You can see flicker with an image having alternate white and black lines one pixel high.

A modern HD naked LCD panel may have less connections than an old 640 x480 mono LCD due to ICs on the internal glass substrate and the TFTs driving the pixels.

[Panrock]

The depth of knowledge of some of the members here is quite incredible. This is where the 'heavy hitters' live. I'm so glad I chose this forum to pose the question. I shall now disappear for a few hours just to digest the latest answers!

Steve

[Mike Watterson]

Apart from having worked with Video and TV for the day job, on and off from 1973 to 2008 and the interest in Vintage R&TV (from 1966) I researched the concepts for an SF series with Aliens I started in 1989 and seriously was working on from 2012 till now.

1) Our colour TV only really works for humans and not perfectly for all humans. Even on Earth different creatures have different colour perception in our visible spectrum. Some creatures really only have monochrome, some like dogs have a very different response basically two "primaries". Some birds have IR vision. Some birds and insects can see UV (some women may have this tetrachromacy but the UV is mostly blocked from the retina). The pigeon may be pentachromatic. See below about Rods & Cones.
2) Frame rate. Guinea pigs need to spot movement so it might be why their perception of motion is much higher than ours which kicks in about 12 fps but some of us can see 50Hz flicker, but only at the edge of vision. Regular TV might be simply separate frames to guinea pigs.
3) Organisation. Leaving aside mechanical scanning and film frames, or colour, even a purely electronic system needs to serialise image transfer either on cable, RF or inside the electronics. You either do rows and then the pixels along the rows (R to L or vices versa), or columns and then the pixels on a column (Up to Down or vice versa). Obviously if you start with scanning film frames as an electronic source the same applies, except a movie projector inherently wants vertical spools (Stills can use either and has done), so you have rows first and the order depends on the film transport direction, which seems to have been an arbitrary choice.
4) Cameras vs display. A camera is a lot trickier than a display. Your display can have a brighter source if mechanical. Mechanical cameras and Philo Farnsworth's camera only sense light at one pixel element at a time as it's scanned, so as the resolution is increased it's quickly uselessly insensitive. A suitable electronic camera needs to accumulate charge from the light during the entire frame. It's true that the sensor sensitivity does drop as the resolution is increased, but you can keep the pixel element the same size and use a larger image area and bigger lenses. Not possible with the mechanical cameras and Philo Farnsworth's camera. So you can imagine an Alien civilisation that uses film cameras, doesn't have Baird's clever processing and wet scanning in the camera, but has electronic displays.

Rods and Cones. Generally rods are thought to be more or less insensitive to hue (but are not be the same range of spectrum in different creatures here) and Cones do colour, somehow. Colour perception is always overlapping curves. So a set of R G and B LEDs matched to peaks may give a good simulation. More colours or wider band sources may be better, but LEDs will produce some colours more saturated than filters on LCD or phosphors. OLED may use phosphors and filters. LCDs might use R G B LED backlights, CCFL tube backlights (mercury discharge + phosphor) or "white" LEDs (really blue to violet or UV with phosphors) or in a projector: LEDs, incandescent lamp or xenon discharge.

There are three theories as to how the Cones actually work . Unlike a single tube/chip camera they may not use pigments (a filter).

So Aliens and us can use the same monochrome TV system if the frame rate is high enough, but colour TV would be rubbish unless the aliens had multiple sentient creatures or have a culture encompassing multiple worlds. It's possible to have a wideband camera and display system with very many more colour channels than three. Of no value to the majority of people here. I'll not bore with details of one possible implementation.

Maxwell may have been the first to realise a simple "trick" three colour system would work for humans and Newton decided a rainbow has seven colours because he liked the mystical connotations of seven. A rainbow has maybe thousands of colours to our perception and we can "see" some colours not actually in a rainbow or white light or in any spectrum. If a signal has red and blue components of the spectrum) but no components in the green area, then we see colours that don't exist in any monochromatic light. The Cyan Yellow Magenta (or C Y M K) models used for colour photography and printing in subtractive or overlay mode use this trick to produce images that look realistic to life to use. The earliest colour film used R G B random starch based filters on one layer so was very insensitive. Using C Y M layers for a negative was a massive improvement. The "holy grail" for a bistable colour display alternative to monochrome eink (which is opaque) is to have bistable layers that are clear or C Y or M over a bottom monochrome eink. This would also work for brighter projectors with LCD. The coloured layers could be 1/3rd resolution or less of the monochrome.

[Panrock]

I have to admit I am finding some of this rather indigestible. Television has certainly 'moved on'.


Panrock: But if one were to do the same with a modern digital display panel - or say a mobile phone - what would one see?

This seems to be a question impossible to answer, since it depends so much on the display technology!


Frank.C: The vertical blanking lines referred to above I would assume are to accommodate CRT TV's. In a digital system it would appear to be wasteful to transmit them.

Just as in mechanical systems, like my 120-line mirror screw. At the end of one line (or frame), it's merrily straight on with the next, with no gap whatsoever.


Mike Watterson: There are very many display technologies now. Most will use a block of RAM for a frame...

So (in general) there are still frames then? I presume these will be rendered progressively. How many per second? Quite a lot, I guess, because with the many display technologies, you can't rely on 'phosphor lag'. 

Though scanned frames as such may no longer exist. Rather, the 'frame rate' may simply signify how often this-block-or-that is reviewed for a change in picture content.  I have an awful feeling that even asking a question like this may show I've lost the plot.


Steve

[ppppenguin]

While some (many?) displays do all sorts of clever things with refresh the fact remains that the input to the display system is a sequence of pixels and lines, starting at the top left. We haven't yet adopted any other model for visual information sent by electronic means. In the route from sender to display we may process the data in many and complex ways but it's still an array of pixels. We could choose to send a frame's worth of pixels in any order we like but by convention we still start at the top left corner. En route we may put pixels in blocks,  describe them in the frequency domain (cosine  and fourier transforms) or by other transforms (fractal, walsh, hadamard) all in the service of sending as few bits as possible while keeping good pictures.

LCD displays are inherently zero order hold. In other words  a pixel stays at its last setting until it is refreshed. Think infinite persistance in a phosphor. This means no large area flicker, regardless of frame rate. This can also give blurry motion. Other displays (plasma, OLED) don't have zero order hold. Hence the portrayal of motion will differ on different types of display. And that's before any interframe processing for data compression.

H and V blanking are now historic artefacts. CRTs and camera tubes needed them, modern displays don't. In a studio context the space isn't wasted, it carries all sorts of data such as timecode and embedded audio. For transmission it's a waste of data so blanking intervals are omitted.

The attached paper shows how 2 BBC research engineers saw the early development of digital TV. Well worth reading. In 1974/5 I spent a gap year (the term hadn't been invented) between school an uni at BBC Research (Kingswood Warren). I met the authors and many of the people mentioned in the article.

[Mike Watterson]

While some (many?) displays do all sorts of clever things with refresh the fact remains that the input to the display system is a sequence of pixels and lines, starting at the top left.

Exactly. And there are not a lot other ways to do it. All will be frames, lines and pixels per line. Though the lines could have been vertical and they could have started bottom, or right.

I'm next going to write about common consumer methods of video to a display:
Composite
S-Video
R G B (SCART/Peritel, RCA, CGA, EGA, VGA)
Component Video
USB methods (but not USB-C video)
Firewire
DVI, HDMI, Display Port and USB-C video are all related.
Analogue RF (still used)
Analogue Video Senders, sometimes called Digisenders
Common Digital tuners built in (Satellite, DVB-C, DVB-T, ATSC)
Video over Ethernet,WiFi & BT in screen casting.

My Toshiba HDTV has: Composite, RGB Scart, RCA Component, S-Video, VGA, Multiple HDMI, PAL Analogue RF, DVB-T
My newer 4K HDR has no RCA Component, S-Video or VGA. It does have 4K capable DVB-S2 as well as DVB-C and DVB-T2. Also DVB-S and DVB-T). It also supports Bluetooth, WiFi and Ethernet for streaming, screencasting and Apps. But you'd be mad to enable those.

Perhaps later Jeffery can explain the various professional video connections including SDI.

Summary:
1) All video, analogue or digital, somewhere at source and destination, is rasterised frames (lines of pixels).
2) Colour is a trick. It's not perfect even for all humans that are not "colour blind". Normal human colour blindness is a difference of perception commonly in red-green and rarer blue green. Mostly men. Actual monochrome vision is very rare in humans. Cattle see monochrome, dogs have reduced colour perception.
3) There are several ways cameras acquire colour and many methods of displaying it.
4) Only some raw R G B data from 3 chip or 3 tube cameras has the same resolution for colour and luminance. RGB at the consumer end has the same resolution for colour content as any other consumer signal of the same luminance resolution. Consumer colour resolution is always less than Luminance.

[Mike Watterson]

BTW, I did read a BBC article on Digital TV (calling it that), in 1970. At the 1971 Young Scientist Expo in Dublin RDS (Jan or Feb?) I outlined (with charts) how TV would be digitised and then could be transferred over laser links. I didn't know about fibre optics till 2 years later. One judge was from RTE and said while interesting we'd never have Digital TV or communication on lasers. Well, they only started TV late at night on the 31st December 1961 (FM Radio a similar date) and were using portable 78 rpm record cutters in late 1950s instead of tape-recorders to gather rural music & folk tales.

Oh, a project on Dragonflies won that year, but I did get highly commended. I was maybe the only one from the UK and one of the few with zero school support. They only left Royal out of the name of the School because otherwise it wouldn't have fitted on the cert!

[Mike Watterson]

Also some computer image formats started as Fax. Which is like really slow video printed on paper, so naturally it does horizontal lines. Invented in 1851. Because TIFF format was used to store faxes, it's one of the few multiframe file formats that's not animation or video.

[Refugee]

30 years ago I worked for a company that made specialist CCTV cameras.
There first solid state camera used a double height CCD chip with half of it masked physically with a blackened label.
The pixels in the exposed bit absorbed light and charged to an analogue level. During the blanking period the whole lot was transferred parallel port style to the masked bit where it was read of serially and buffered for onward transmission between the sync pulses of the 75 ohm output. Yes that is right early CCD cameras really were analogue.
The whole camera was about the same size as a hand held colour one of the period. I got to see colour CCD cameras after I moved on to another company.
Modern cameras use more layers in the head chip with a layer of RAM to store the analogue pixel outputs after transfer as above where they will just be digitised as they are read off serial port style.
There will be some kind of frame marker to set the resolution of the display software.
The processing software dismantles the signal and converts it to a single frame plus updates.
Those sweeping pan shots of the analogue days will disappear due to the amount of bandwidth they use up.