Analog and digital television
Transmitting moving pictures by radio meant solving a formidable problem: turning something two-dimensional that changes over time into a single signal that travels along a cable or through the air. The solution —scanning the image line by line— is still in use eighty years later.
01Scanning and interlacing
| Parameter | 625-line system | 525-line system |
|---|---|---|
| Lines per frame | 625 | 525 |
| Frames per second | 25 | 30 |
| Fields per second | 50 | 60 |
| Line frequency | 15,625 Hz | 15,750 Hz |
| Line duration | 64 µs | 63.5 µs |
| Used in | PAL, Europe and much of South America | NTSC, North America and Japan |
At 25 frames per second motion looks smooth, but the screen flickers in a very annoying way. Doubling the frames to 50 would cure the flicker and double the bandwidth required, which at the time was not feasible.
Interlacing achieves both: each frame is divided into two fields, one with the odd lines and one with the even lines. The screen is refreshed 50 times per second —the flicker disappears— but only 25 complete images travel through the channel. It is one of the most elegant engineering tricks in the history of communications.
625 lines per frame × 25 frames per second = 15,625 lines per second. That is exactly the high-pitched whine given off by the flyback (line output) transformer of CRT TV sets, at the edge of audibility: many people heard it and never knew where it came from.
And of the 625 lines, about 576 carry picture: the rest are used up in the vertical retrace, time that was later put to use for teletext and other data.
02The composite video signal
| Component | What it carries |
|---|---|
| Luminance (Y) | The brightness. It is the signal that black-and-white TV sets used to see, which is why color had to be compatible. |
| Chrominance (C) | The color, modulated onto a 4.43 MHz subcarrier in PAL. Amplitude gives the saturation and phase gives the hue. |
| Horizontal sync | A pulse at the start of each line that tells the receiver where it begins. |
| Vertical sync | A special sequence that marks the start of each field. |
| Color burst | A few cycles of the subcarrier on the back porch: it is the phase reference without which the colors come out shifted. |
| Black and white level | They define the scale: the sync pulses sit below black, in the “blacker than black” region. |
When color arrived, there were already millions of black-and-white TV sets. The solution was to fit the chrominance inside the same channel, on a subcarrier chosen to interleave with the luminance spectrum and interfere as little as possible.
A black-and-white receiver sees that subcarrier as a fine pattern and ignores it; a color receiver demodulates it. The whole PAL system is a consequence of that constraint.
In NTSC, a phase error in transmission directly changes the hue: faces look green or purple, which is why TV sets came with a tint control.
PAL inverts the phase of one of the components on alternate lines, so that the error on one line is canceled by the error on the next. The result is a slight loss of saturation instead of a color shift, and no tint control.
03Digital television
| Stage | What it does |
|---|---|
| 1 | Digitization: the image is converted into pixels and the signal into numbers. Uncompressed, a standard-definition picture requires on the order of 270 Mbit/s. |
| 2 | Compression: MPEG-2 or H.264 reduce that rate about a hundredfold by taking advantage of the fact that almost nothing changes between consecutive frames. |
| 3 | Multiplexing: several video, audio and data programs are combined into a single transport stream. |
| 4 | Channel coding: error correction is added to withstand multipath and noise. |
| 5 | COFDM modulation: thousands of narrow carriers, each with its own QAM, spread across the same 6 MHz channel. |
In analog television, a signal that arrived bouncing off a building produced a displaced “ghost” image. In COFDM, the thousands of carriers are so narrow that each symbol lasts much longer than the delay of the reflection, and in addition a guard interval is added between symbols.
The result is that the echo, instead of causing trouble, adds energy. This makes possible something impossible in analog: single-frequency networks, in which all the repeaters in a region transmit on the same channel.
Analog television degraded gradually: first a little snow, then more, until it became unwatchable. Digital works perfectly up to a point and then cuts out abruptly, with a frozen or blocky picture.
This completely changes the installation work: it is not enough to “see that it looks good.” You have to measure the margin —signal-to-noise ratio, MER and error rate— to know how far you are from the cutoff. An installation that looks perfect may be one decibel from the cliff, and fail with the first rain.
| Display | How it works | Characteristics |
|---|---|---|
| Cathode-ray tube | An electron beam sweeps the phosphor screen, deflected by coils. | Historical. Bulky, heavy and with high voltage inside. |
| LCD | Liquid crystals that rotate the polarization of light from a backlight. | The most widespread. Black is never completely black. |
| Plasma | Gas cells that emit light when ionized. | Very good contrast and viewing angle. Discontinued because of power consumption. |
| OLED | Each pixel emits light by itself, with no backlight. | Absolute black and a very thin panel. Lifespan and cost are its limits. |
04Closed-circuit television
- Resolution and field of view: the figure that matters is how many pixels fall on the object. To identify a person you need on the order of 250 pixels per meter; to merely detect presence, much fewer.
- Lighting: almost all image problems come from light, not from the camera. Backlighting, reflections and darkness are solved by placing the camera well, not by raising the resolution.
- Storage: bit rate per camera × number of cameras × retention days. It is what sets the size of the disk and is often underestimated.
- Power: over the same network cable (PoE) it simplifies the installation enormously, but you have to check the total power of the equipment that supplies it.
Video surveillance is regulated. You must post signs that the area is being filmed, limit the view to your own property —not the sidewalk or the neighbor’s window—, set a retention period and restrict who has access to the recordings.
It is part of the technical job: an installation that points at other people’s spaces is a legal problem for the installer and the owner, as well as a lack of consideration.
05In the lab
With a bar generator and an oscilloscope, observe a complete line of video: identify the sync pulse, the porch, the color burst and the levels of each bar. Measure the duration of each part and compare it with the standard.
With a field strength meter, survey the level, signal-to-noise ratio, MER and error rate of the digital channels at different points in the school. Determine how much margin there is to the cliff at each outlet, and document it.
Set up a distribution network with taps and splitters, calculate the losses and check that the level at the worst outlet stays within the receiver’s range. Check what happens when an output is left unterminated.
Mount a camera and calculate the number of pixels per meter at different distances. Verify experimentally from what distance a person can be identified and beyond what distance they can only be detected.
06Common mistakes
| Symptom | Usual cause |
|---|---|
| Looks perfect and sometimes breaks up into blocks | Insufficient margin: digital gives no warning before cutting out. You have to measure MER and error rate. |
| Wrong colors in analog | Problem with the color burst or a phase error, characteristic of NTSC. |
| Ghost image in analog | Reflected signal. In digital, COFDM turns it into useful energy. |
| High level and poor quality | Interference from another channel or saturation of the installation’s amplifier. |
| Only the high channels fail | Cable loss, which grows with frequency, or poorly chosen taps. |
| Unterminated distribution outputs | Mismatch and reflections that degrade all the outlets. |
| Camera facing the window | Backlighting: you see a black silhouette. It is a placement problem, not a camera problem. |
| Disk full within a week | The bit rate per camera and the retention days were underestimated. |
07Self-assessment
What is interlacing for?
To refresh the screen 50 times per second —and eliminate flicker— while sending only 25 complete images, without doubling the bandwidth.
Where does the 15,625 Hz line frequency come from?
From 625 lines × 25 frames per second. It is the high-pitched whine given off by CRT TV sets.
What are luminance and chrominance?
Luminance carries the brightness —it was the complete signal in black and white— and chrominance carries the color, modulated onto a subcarrier within the same channel.
What is the color burst for?
It is the phase reference of the subcarrier. Without it the receiver cannot demodulate the color correctly and the hues come out shifted.
What advantage does PAL have over NTSC?
It inverts the phase on alternate lines, so that a phase error is canceled between lines: instead of changing the hue, it produces a slight loss of saturation.
What does COFDM do with reflected signals?
Thanks to narrow carriers and the guard interval, the echo arrives within the same symbol and adds energy instead of producing ghosts.
What are single-frequency networks?
Networks in which all the repeaters transmit on the same channel. This is possible thanks to COFDM, because the signals from several transmitters add up instead of interfering.
What is the digital cliff and what does it mean for the installation?
Reception is perfect up to a point and then cuts out abruptly. It means that it is not enough to look at the picture: you have to measure the margin with MER and error rate.
Why is the digital television signal compressed?
Because uncompressed it would take on the order of 270 Mbit/s for standard definition, and the channel is only 6 MHz wide. Compression takes advantage of the fact that almost nothing changes between consecutive frames.
What determines whether a camera allows a person to be identified?
The number of pixels on target: on the order of 250 per meter to identify. It depends on the resolution, the lens and the distance, and also on the lighting.