2.2.1 Television Briefly

The Federal Communications Commission (FCC) adopted standards for commercial television broadcasting in the United States in 1941.¹ It is an analog system, in which the picture is transmitted via vestigial-sideband modulation of the visual carrier and the sound is frequency modulated on a separate carrier. In 1954, the NTSC “compatibly” extended the system to include color information by increasing the utilization of the 6-MHz spectrum occupied by the television channel while allowing continued use of essentially all older receivers with little performance degradation. In 1976, the FCC reserved line 21 for the closed captioning system, and in 1983 a teletext service was authorized. In 1984, the NTSC system was again compatibly extended to include stereo sound. Most recently, in June 1996, the FCC authorized additional ancillary data services in analog television. These technologies will be considered in this chapter.

The NTSC television system is termed analog because the signals representing the picture and sound information can take on any value between the minimum and maximum limits. In particular, the strength of the visual signal is inversely related to the picture brightness, with black portions of the picture having the most transmitted power. This negative modulation of the video information was found to reduce the impact of Gaussian white noise on the displayed image. Periodic pulses are included at powers greater than those used to represent black areas in the picture. These pulses provide the timing information required to synchronize the transmitter and the receiver, allowing the picture to be painted correctly on the screen.

Understanding television technology is aided by some appreciation of its history.

2.2.2 Scanning

A picture csxan be regarded as the composite of many points of light called picture elements, or pixels. The information for each of these points of light must be derived from the object or scene to be televised, organized in a logical manner, conveyed to the receive site, and then displayed as an image. Since the number of points of light needed to make up a picture is huge, it is not practical to convey to each of them via its own transmission channel. Scanning is a simplifying process that allows a single transmission channel.²

The process of scanning was first used on a large scale to transmit still images for use in newspapers (see Figure 2.1). A photograph is placed on a rotating drum and a light-sensitive device is focused on it. As the drum rotates, the light-sensitive device follows a path across the image. With each revolution of the drum, the light-sensitive device is advanced down the length of the rotating drum. This creates a helical scan that follows the path of an imaginary coil spring across the photograph on the scanning drum. The electrical output of the light-sensitive device is proportional to the reflectance of the image, bright areas yielding a larger electrical signal than dark areas. At the receive location, the process is reversed. A light-sensitive film is attached to a rotating drum, and a light source scans the same helical path across the film. High electrical input results in a bright light and a full exposure, or blackening of the film. Conversely, low electrical input yields a lower exposure, or a more transparent area on the film. In this way, a negative image is produced. Standard photographic techniques yield a comprehensible image. It is important that the rotating drums at the transmit site and the receive site be synchronized. They must rotate at the same speed, and the start of scanning must occur simultaneously in both locations. Likewise, the rate of advance of the helical path down the drum must be the same in both locations. If this synchronization does not occur, then geometric distortions are introduced in the reproduced image. Modest distortions involve stretching or compressing of the reproduced image as if it were printed on a rubber sheet and then the rubber sheet distorted. More destructive distortions are possible if the speeds are very different in the two locations.

Figure 2.1 Scanning of photograph on rotating drum for transmission to receiver with light source and photosensitive film.

Although scanning works well when the receive site includes a photosensitive film that can accept and retain the image, the problem becomes a bit more complex when the receiver is the human eye. The human eye exhibits a phenomenon termed persistence of vision that allows it to continue “seeing” a light for a brief time after the light is removed. The application of the scanning principle to direct human vision requires that the scanning process be executed repetitively and quickly enough for the eye to fuse all the scan lines into an image.

In 1883, the German engineer Paul Nipkow invented a scanning method consisting of a rotating mechanical disk with appropriately located holes arranged in a spiral (see Figure 2.2). As the disk rotates, only one hole at a time is over the image to be transmitted (see Figure 2.3). As one of the holes completes a left-to-right scan, another hole commences scanning from the left slightly below the path of the previous hole. When the bottom hole completes its scan, the scanning disk will have made a full revolution, and the top of the image is again scanned. Nipkow was granted a German patent but did not actually build the system. In 1923, John Logie Baird of the United Kingdom built a crude system based on the Nipkow scanning disk. It was clumsy and produced poor pictures, but it confirmed the principle of scanning.

Figure 2.2 Nipkow disk operation. (a) Nipkow disk with a spiral of apertures. (b) Nipkow disk’s scanning trajectory through the image plane.

Figure 2.3 Only one hole at a time is exposed to the image on the sending side and the light source on the receiving side.

Television did not become practical until electronic display tubes and pickup tubes were introduced. In 1923, Vladimir Zworykin, a Russian immigrant who became a U.S. citizen, applied for a patent for a completely electronic television system. The interest in television was intense at that time, and many inventors were active in the field. Space does not allow a discussion of the numerous contributions and the interesting patent fights among inventors. This fascinating story is described elsewhere in detail and is recommended reading.³

In 1929, Zworykin applied for a patent that made the monochrome (black-and-white) picture tube practical (see Figure 2.4). An electron gun launches electrons past a control grid that regulates the strength of the electron beam. The beam then passes through focusing elements that concentrate the electron beam into a small spot on a glass faceplate that is covered with a fluorescent coating on its inside surface. When the electron beam strikes the fluorescent coating, light is emitted in proportion to the strength of the electron beam. Magnetic coils on the neck of the tube deflect the electron beam vertically and horizontally to implement scanning. The scanned rectangle of light on the face of the picture tube is called a raster. The picture tube is also called a kinescope because it was meant to display moving or kinetic images. Yet another name for the kinescope is the cathode ray tube (CRT), since the part of the electron gun that emits the electrons is called a cathode.

Figure 2.4 Electronic television is based on the kinescope, or cathode ray tube.

In October 1931, Zworykin applied for a patent on a practical electronic television camera tube that he called the iconoscope (see Figure 2.5). As with the kinescope, an electron gun forms an electron beam that is scanned across a surface by magnetic deflection coils. In the case of the iconoscope, the surface is coated with a light-sensitive material. Light focused on the sensitive material causes it to develop an electrical charge in proportion to the strength of the light. When the electron beam strikes the light-sensitive surface, it interacts with the surface in proportion to the accumulated charge. A signal plate on the back side of the light-sensitive surface produces an electrical signal proportional to the light that fell on the surface. Again, many other inventors were working on similar devices, and more patent struggles occurred.

Figure 2.5 Electronic television is also based on Zworykin’s iconoscope.

After a scan line is completed, it is necessary to move the electron beam back to the left side of the screen.⁴ Ideally, this retrace would be done instantaneously, since no picture information is transmitted during the backward motion. Both the scanning drum of Figure 2.1 and the scanning disk of Figure 2.2 did not require retrace. However, electronic deflection is accomplished with magnetic fields, and the magnetic fields store energy. Deflection circuits require time to change the magnitude and direction of the energy stored in the magnetic deflection fields. As a result, it is not possible to instantaneously move the electron beam from the right side of the screen to the left or from the bottom to the top. During the time the electron beam is in retrace, it must be turned off (a process called blanking). The period during which the beam is turned off in horizontal retrace is called the horizontal blanking interval (HBI), while the much-longer period required to move from bottom to top is the vertical blanking interval (VBI). The combination of the two blanking intervals constitutes around 25% of the total time and is the price paid for using cost-effective magnetic deflection. It is interesting to note that modern liquid crystal display (LCD) panels do not require retrace time.

Thomas Edison is credited with developing practical moving pictures. Moving pictures depend upon a rapid display of a succession of still pictures. In modern motion picture practice, 24 pictures are each displayed twice per second, but in early movie history the rate was reduced, leading to noticeable “flickers.” The human response to flicker is very complex and is related to the color, size, and brightness of the image. A small increase in the brightness, or a decrease in the repetition rate, results in a much higher degree of noticeable flicker. The minimum rate that is not objectionable is called the flicker fusion frequency. A flickering light source at frequencies above that value appears as continuous illumination. For example, fluorescent tubes operated at 60 Hz (and thus extinguished 120 times per second) appear to be continuously “on” because they are operated above the flicker fusion frequency. Monochrome television developers settled on 30 complete images per second in the United States and 25 images per second in Europe. Because of the interlace technique, described later, 60 halfimages (or 50 in Europe) are displayed per second. These choices were originally tied to the power line frequency causing any unfiltered power line “hum” that got into the picture to be stationary on the screen.(It was determined that slowly moving patterns of interference were much more objectionable than stationary patterns.) In the European systems, the brightness level at which television flicker is not objectionable is only about 30% of that in the United States. Many viewers accustomed to 60-pictures-per-second television find that the European systems flicker excessively at the brightness levels normally used.

Many of the specifications for the NTSC television system were based on the desire to minimize the cost of receivers using then-current technology and are no longer required. For example, it is now practical to keep the power line hum completely out of video circuits. Consequently, the need to relate image display rate to power line frequency is no longer significant. In Japan, for instance, 50-Hz power is used in the Osaka region and 60-Hz power in the Tokyo area, yet the same television system serves both markets.

The amount of information required to produce pictures of acceptable horizontal and vertical resolution at a 60-pictures-per-second repetition rate is huge. At current resolutions but with a repetition rate of 60 complete pictures per second (that is, without the interlace technique described later), twice the information would be required, and a baseband bandwidth of 8.4 MHz would be required for the video alone. As we shall see, the modulated bandwidth using simple amplitude modulation would be 16.8 MHz — almost three times the current broadcast bandwidth of 6 MHz. Clearly, ways of reducing the bandwidth had to be found that did not impair perceived picture quality.

One of the earliest signal compression techniques involved modifying the scanning system. Sixty pictures per second are required in television because picture tube technology could produce pictures bright enough to watch in daylight with the window drapes open without apparent flicker. Since experience with motion pictures confirmed that 24 complete images per second is an adequate information rate for motion rendition, a technique called interlaced scanning was developed whereby half of the scan lines are displayed in one “field” and the other half of the scan lines in another “field.” Complete two-field “frames” are transmitted at a rate of 30 per second, providing better motion rendition and flicker frequency than movies yet requiring just 4.2 MHz of baseband bandwidth. Interlaced scanning is illustrated in Figure 2.6. A 525-line raster includes two fields of 262.5 lines each. The odd fields start with a whole line and end with a half line, whereas the even fields start with a half line and end with a whole line.

Figure 2.6 Interlaced scanning to reduce flicker while conserving bandwidth.

Even though interlace scanning works well with most “natural” video (that is, camera images), interlace scanning can produce inferior results with “synthetic video,” which is electronically generated and does not benefit from some of the averaging phenomena of practical cameras. For example, the electron beam in a television camera has a distribution of charge of roughly a Gaussian shape. As a result, it accepts and averages light from adjacent scan lines. This averaging causes a smoothing of the image. Synthetic video, especially when producing text, sharply defines the images of each line. This results in a brightly illuminated edge that is presented at 30 images per second, well below the flicker fusion frequency. This is unacceptable in computer displays that are bright and utilized at very close viewing range. As a consequence, nearly all computer displays use a simple top-to-bottom “progressive scan” technique. Since computer displays are driven by locally generated video, the higher bandwidth requirement of a progressive scan raster is not a problem.

2.2.3 Synchronization

As previously described, there are two fields in each frame of video.⁵ Frames have 525 scan lines and repeat 30 times a second, resulting in 63.5 microseconds per horizontal scan line. The reciprocal of this yields the 15,750-Hz monochrome horizontal line frequency. Since fields occur at the rate of about 60 per second, a field lasts 16.67 milliseconds. (All of these values were slightly modified when color was added — see Section 2.4.1)

A horizontal synchronization pulse is required for every line and a vertical synchronization pulse is required for every field in order to display a stable image. Synchronization (sync) signals are distinguished from the composite signal on the basis of amplitude. The voltage range for visible brightness variations extends from 12.5% to 75% of the unmodulated carrier voltage, leaving 25% for the sync pulses.

Figure 2.7(a) displays an example of a signal that might come from a television camera as it scans one line. At 75% of the maximum voltage, a blanking pulse is inserted, timed to occur during the retrace time. This is shown in Figure 2.7(b). The blanking pulse is nominally 10.9 microseconds, or 16.5% of the horizontal period, while active video occupies 52.6 microseconds. The blanking pulse is at the video signal’s black level, keeping the picture tube from emitting light during retrace. The sync pulse is installed on top of the blanking pulse and is at a level referred to as “blacker than black” (see Figure 2.7(c)). The sync pulse is nominally 4.7 microseconds long. Figure 2.8 shows the relative levels of black, blanking, and sync pulses in the modulated envelope of the signal.

Figure 2.7 Construction of a video signal. (a) Video signal from television camera. (b) Blanking pulse added to avoid diagonal retrace lines. (c) Horizontal synchronization pulse added to facilitate left-to-right timing.

Figure 2.8 Composite monochrome television signal levels (not to scale).

Figure 2.9 shows the horizontal sync pulses and video of Figure 2.8 plus additional sync pulses for the vertical blanking interval. Each field’s VBI is 21 lines long, leaving 483 lines for active video. The first nine lines of the VBI are devoted to pulses associated with the vertical synchronization process. The fourth, fifth, and sixth lines in each field contain the vertical sync pulse. This pulse is serrated, creating pulse edges properly located for the horizontal synchronization to be maintained during the vertical retrace. The vertical sync pulses are called broad pulses, in comparison to the narrow horizontal sync pulses.

Figure 2.9 Vertical and horizontal synchronizing pulses of field one and field two. Note locations of half lines (not to scale).

Simple resistor-capacitor sync separator circuits can be used to separate the vertical pulses from the horizontal pulses.⁶ In the lower portion of Figure 2.10(a), capacitor C1 is charged through series resistor R1 and discharges through the input impedance of the following circuits. The narrow horizontal pulses will not build up a significant voltage because the charge bleeds off during the relatively long time between horizontal sync pulses. However, the broad pulses will build up significant charge. This is shown in more detail in Figure 2.10(b). Those who have studied such circuits will recognize this configuration as a simple low-pass filter, also called an integrator. A properly adjusted level-sensitive circuit following the resistor-capacitor network will provide a vertical timing pulse. This pulse is used to synchronize a vertical deflection oscillator, which in turn drives the vertical deflection magnetic coils on the neck of the picture tube. Six “equalizing pulses” are inserted before and after the broad pulses to equalize the charge buildup on the capacitor, preserving the precision of the vertical timing. This is necessary because one field ends with a half line and starts with a whole line, whereas the other field has the opposite configuration. Without this precaution, the timing between the two fields will be upset, and the rasters they produce will not be evenly interlaced. The resulting distortion is called pairing because adjacent lines from both fields pair up and blur each other’s detail while leaving dark lines between them. Although modern digital sync separator circuits do not need the equalizing pulses, the original receivers using vacuum tubes had to rely on tailoring the signal to allow simple circuits.

Figure 2.10 Receiver synchronization. (a) Sync separation and deflection ramp generation. (b) Vertical synchronization pulse generation.

In the upper portion of Figure 2.10(a), a series capacitor C2 with a resistor R2 to ground will convert the edges of the pulses into positive- and negative-going spikes that can be detected by appropriate level-sensing circuits. Note that the serrations in the broad pulses allow a continuous flow of uninterrupted pulses at the correct timing for horizontal synchronization. When these pulses occur halfway through a horizontal line, the horizontal circuits ignore them. They are effective only when they are near the correct time. Those who have studied such circuits will recognize this configuration as a simple high-pass filter, also called a differentiator.

By allocating significant signal resources (about 25% of the signal time and about 25% of the voltage range), rugged synchronization of the video signal was achieved.

2.2.4 Spectrum of Scanned Images

It is well understood that any repetitive signal can be decomposed into a sum of sine and cosine waves of appropriate amplitude, frequency, and phase.⁷ There are at least three ways we know this: (1) mathematical proof, (2) geometric construction, and (3) measurement with spectrum analyzers. The most rigorous method is the mathematical. >Equations (2.1), (2.2), and (2.3) display the familiar Fourier series equations.

(2.1)

Equation (2.1) is the trigonometric Fourier series representation of f(t) over an interval (t₀, t₀ + T).

The various constants a_n and b_n are given by

(2.2)

and

(2.3)

We could derive these equations and apply them to scanned images, but that would duplicate what is already available in many other places. In particular, a paper published by Mertz and Gray in the Bell System Technical Journal in July 1934, titled “Theory of Scanning and Its Relation to the Transmitted Signal in Telephotography and Television,” showed that periodic scanning of an image yields a spectrum with energy clustered around the horizontal scanning frequency, with large gaps in between. This pioneering work was done long before the availability of modern computers. Donald Fink did follow-up work and the results are included in the 1957 Television Engineering Handbook.⁸ His work clearly showed that the picture’s energy is clustered around harmonics of the horizontal scanning frequency, as shown in Figure 2.11.

Figure 2.11 Video spectral energy clusters around harmonics of horizontal scan frequency, f_h, all the way up to the upper band edge.

2.2.5 Television Audio

Television sound is limited to 15 kHz because of the presence of strong spurious signals related to the monochrome horizontal scanning frequency of 15,750 Hz. For many years, audio in television was underappreciated. Only in the last decade has it received the attention it deserves as a fundamental part of the viewing experience. Perception experiments with human subjects at the Media Laboratory of the Massachusetts Institute of Technology have demonstrated that subjects shown two identical video presentations with differing accompanying sound qualities believed that the video with the better sound also had better video quality.

2.3.1 A Phenomenological Approach to Amplitude Modulation

A television camera produces signals that are proportional to the strength of the light in the image and includes synchronization pulses. Figure 2.8 shows the waveform of such a signal. This is an information-bearing baseband signal because the frequencies of this signal extend from zero to the maximum produced by the television camera. The zero-frequency component represents the average brightness of the image. The highest frequency conveys the finest detail of the picture.

A natural method of conveying the television camera signals would be to modulate the strength of a carrier wave in proportion to the amplitude of the original baseband signal, as shown in Figure 2.12.⁹ The outline of the maximum excursions of the carrier, termed the envelope, is seen to be a replica of the original baseband information-bearing signal.

Figure 2.12 Television modulation. Carrier wave (shown at much lower frequency for convenience of illustration) modulated with video signal.

At the receive site, the goal is to display the signal from one television camera. This can be done first with a tuned circuit that passes only the desired modulated carrier (actual receivers implement a more sophisticated version of this method based on the superheterodyne receiver principle), followed by a simple diode that will pass only the positive-going portion of the modulated signal, followed by a low-pass filter to remove the carrier pulses, as shown in Figure 2.13.

Figure 2.13 Envelope detection with a diode and low-pass filter.

2.3.2 A Mathematical Approach to Amplitude Modulation

Amplitude modulation is the process by which a high-frequency carrier wave is multiplied by a low-frequency information-bearing signal to generate a new signal, centered on the carrier frequency, which includes sidebands that carry the information. Through use of multiple carriers, each modulated by a different information-bearing signal, it is possible for multiple information streams to share a common transmission medium.

More formally, let us define the two signals as follows:

(2.4)

where:

Carrier = carrier wave

Information = information-carrying signal (simplified to a simple sine wave to simplify the equations)

If we multiply these, we get

(2.5)

That is, we get the original carrier plus the effect of multiplying the original carrier times the information-carrying signal. When we multiply two sine waves, however, we get two signals whose frequencies are equal to the sum and difference of the two original frequencies and whose amplitudes (voltage, in this case) are in accordance with the standard trigonometric identity:

(2.6)

In other words, we get a sideband on each side of the carrier that is spaced from the carrier by the information-carrying frequency and whose amplitude will be 6 + 20 log (B) dB less than the amplitude of the unmodulated carrier.

In the case where the information-carrying signal is a scanned television picture, it will have frequency components extending from zero hertz to an upper frequency limit that is related to the possible horizontal resolution of the displayed image. In the NTSC system, the information-carrying signal is limited to about 4.2 MHz of bandwidth. Thus, with the simple double sideband modulation just described, the modulated signal bandwidth would be 8.4 MHz wide just for video information, with the audio still to be fitted into the spectrum.

It is apparent, however, that double sideband modulation is wasteful of spectrum, because each sideband carries exactly the same information. An alternative is single sideband (SSB) modulation, whereby only one of the sidebands and no carrier is transmitted (saving both bandwidth and transmitter power). However, the detection process is more complex for SSB signals because a local carrier must be generated to mix with the incoming sideband in order to recover the original information-carrying signal. While this is simple today, it wasn’t when the NTSC system was being developed and so was rejected.

The compromise reached was vestigial sideband (VSB) modulation, in which a normally modulated double sideband signal is passed through a filter that passes one sideband (the upper sideband of the transmitted signal), but only 0.75 MHz of the other sideband, as shown in Figure 2.14(a). As can be seen, the sound signal has been inserted 4.5 MHz above the visual carrier, leading to a total visual + audio signal that is only 6 MHz wide yet includes picture information extending up to between 4.0 and 4.2 MHz.

Figure 2.14 VSB reception. (a) Vestigial sideband spectrum normalized to 0.0 Hz as seen at RF (at IF, spectrum is reversed). (b) Output of a “flat response” television receiver as seen at RF (at IF, spectrum is reversed). (c) Idealized and typical TV receiver response curves as seen at RF (at IF, spectrum is reversed).

If this signal is demodulated using a standard amplitude detector, the result will be that modulating frequency components below 0.75 MHz will be recovered from both sidebands, while those falling above 1.25 MHz will be recovered from only a single sideband, thereby distorting the relative amplitudes of the detected information signal, as shown in Figure 2.14(b). In order to correct for this, the modulated signal must be passed though a compensating filter before detection. The shape of the required filter is shown in Figure 2.14(c).

2.3.3 Video Bandwidth

When the NTSC system was being designed, a tremendous number of human visual experiments were undertaken. Estimates were made on the likely course of technology. It is a tribute to the work done that the NTSC system has lasted for more than 50 years and is likely to be in use for at least another decade or two.

NTSC television is meant to be a “five times picture height” system. That is, the compromises made between video quality and economics were such that they would not be noticed if the picture were viewed from a distance of at least five times the picture height. Consideration of this and of the bandwidth available, as well as of the ways of occupying that bandwidth in a manner that would allow affordable receivers, led to the choice of 525 total (483 active) scan lines. If a synthetic video pattern is generated by electronically making alternate lines black and white, and if the whole raster is placed in the viewing area of the screen, then 483 picture elements can theoretically be resolved in the vertical direction. In the “real world,” a large number of factors limit vertical resolution, including that the scanning system is a sampling system and that raster lines do not fall exactly on object boundaries. Rather than a clean edge, some of the light falls into two or more adjacent lines, blurring the image. Additionally, there is raster instability. Minor movements between scans add further blurring. The scanning spot in the pickup device and the scanning spot in the display device are not precise points but have a bell-shaped (Gaussian curve) distribution of effectiveness. All these complex factors yield less actual vertical resolution than the number of scan lines would indicate. The ratio between the number of scan lines and the actual vertical resolution varied from 0.5 to 0.85 in various experiments and is commonly accepted as 0.7. R. D. Kell is credited with much of the work on this phenomenon, so this ratio is called the Kell factor. That is, 483 active scan lines yield an effective vertical resolution of about 338 lines.

There is no single derivation of the bandwidth needed to carry video. The best that can be done is to give a plausibility argument for why something like 4.2 MHz is a reasonable number. Assuming a desire to have approximately the same horizontal resolution as vertical resolution and noting that the aspect ratio (the ratio of vertical height to horizontal width) of an NTSC picture is 3 by 4, we must multiply 338 by 4/3, yielding 450 horizontal lines of resolution. Dividing the aspect-ratio-adjusted number of horizontal picture elements by 2 yields the maximum number of cycles of video in the active horizontal video time. This yields 225 cycles. The 225 cycles must take place in 52.6 microseconds, yielding a frequency of 4.28 MHz.

The maximum number of picture elements possible in a video frame constructed electronically is 483 scan lines times 450 horizontal lines of resolution, or 217,350 pixels. European television systems that use the 50-pictures-per-second rate have more time per picture, allowing for 625 scan lines. To retain a balance between vertical and horizontal resolution, these systems have allocated more bandwidth per channel.

2.3.4 Frequency Modulation

The previous discussion of modulation considered only modifying the strength of a carrier signal to impress an information-bearing baseband signal upon it.¹⁰ A cosine wave, however, has three parameters that can be varied to convey information. Equation (2.7) shows them:

(2.7)

Here A is the amplitude, f is the frequency, and θ is the phase angle of the cosine function. The symbol s(t) is the signal as a function of time. As previously discussed, visual information is added to the carrier by varying its amplitude.

Sound information is modulated on a separate carrier by varying its frequency, in which case the signal can be expressed by

(2.8)

where the information-bearing signal is g(t). Without loss of generality, the constant phase angle 0 can be dropped, yielding

(2.9)

If the baseband information-bearing signal is a simple cosine wave of frequency f₂, the resulting expression is

(2.10)

The mathematics of cosines within cosines rapidly becomes very complex, involving advanced functions and concepts. There are excellent treatments of these details for anyone who is interested. The principle result for our purposes, however, is that the bandwidth of a frequency modulated signal can exceed that of an amplitude modulated signal given the same information-carrying signal. The signal, however, is more robust and can survive noise and other distortions better than amplitude modulation. Frequency modulation (FM) belongs to a class of signals called spread spectrum because it spreads the information signal over more bandwidth than it occupies in its baseband version. This technique trades off spectrum inefficiency for signal robustness.

The phenomenological concept motivating frequency modulation is the notion that noise, particularly static, was believed to be only an amplitude effect. If the amplitude of the signal did not carry information, the impact of noise added to the signal would be negligible. This simplistic model is mostly correct, and the main benefits of FM include much less susceptibility to noise.

The creators of NTSC considered it to be a video enhancement to radio. The intention was that, at the outer limits of reception, the audio signal would survive even if video failed,¹¹ and thus it was made extremely robust by using frequency modulation. In cable practice, we are able to take advantage of this by reducing the relative strength of the audio signal and thus reducing the overall signal loading on the cable system. In NTSC television, the sound signal is frequency modulated onto a carrier that is 4.5 MHz above the visual signal. Over-air broadcasters use separate transmitters for visual and aural signals and combine the signals before feeding a shared antenna.

2.4.1 Color Television

Color is the most significant addition to NTSC.^12,13 A large number of trade-offs were necessary in the original NTSC system to make it practical and affordable, but even more were required to add color. The color system carefully matches its design to the human visual system to minimize the incremental information rate required.

The most important discovery is that the human visual system responds to color in two distinct ways: (1) as a response to the frequency (or wavelength) of the light, and (2) as a response to combinations of primary colors. If we pass sunlight through a glass prism, we decompose the light into individual wavelengths that we see as many separate colors. If this were the only way in which the eye perceives color, a color television system, or even color printing and photography, would be all but impossible. Fortunately, the eye, in response to various ratios of red, green, and blue,¹⁴ perceives nearly all the possible colors.

The science of colorimetry is fascinating and complex and well beyond the scope of this brief chapter. Only a few basics can be covered here. In the color receiver, the picture tube is actually three picture tubes in one.¹⁵ There are three electron guns and three sets of colored phosphor elements on the screen. A shadow mask in the tube is carefully aligned to allow only the electrons from a particular gun to strike the phosphors of its assigned color. The shadow mask is also called an aperture mask because of its array of holes, or apertures (see Figure 2.15). The three electron guns are tilted at an angle with respect to the axis of the picture tube. The three electron beams pass through the same holes in the shadow mask, which is located very close to the screen. Because they aim from different angles in the neck of the tube, the beams land on adjacent but different spots. The spots have phosphors that emit red, green, and blue light. The spots are small and close together, and the eye fuses them into a continuous color from normal viewing distances. The reader is invited to take a magnifying glass and look closely at the screen of a color receiver. Clusters of dots or clusters of small bars of the three different colors will be visible when the screen is illuminated.

Figure 2.15 Color picture tube structure with colored phosphor dots, aperture (or shadow) mask, and cluster of three electron guns.

If a combination of three complete images in red, green, and blue were required to make a color picture, the required transmission bandwidth would be tripled. Even the improvement color brings might not be worth that drastic a loss of channel capacity. Fortunately, the eye is quite pleased with a “colored” television system and does not require a true “color” system. If we consider the Sunday color comics, we realize that we have a detailed black-and-white image with some color thrown on it. The eye and the imagination do the rest. In colored television, the same approach is taken. The detail is in the luminance (brightness) signal. The color signal contains much lower resolution and only approximately fills in the monochrome outlines. That this is the case can be seen when colored text is in the picture. For example, a yellow letter T on a blue background will have a yellow top bar because the image is large enough horizontally to be represented by the frequencies involved. However, the vertical component of the T will usually not be yellow. To represent it in yellow would require frequency components beyond what is available.

To make the color signal compatible with older monochrome receivers, it was necessary to combine the signals in a manner that yielded the luminance signal plus two narrower bandwidth “color difference” signals. Since the eye is more sensitive to detail in certain colors than in others, it is possible to reduce the bandwidth of some of the colors more aggressively. The color television system manipulates the color signals using a set of linear matrix equations. This separates the signal into a component that contains flesh tones and near flesh tones and another component that has quite different colors. The principal color signal, termed the I signal, has a transmitted bandwidth of around 1.5 MHz. The secondary color signal, termed the Q signal, has a transmitted bandwidth of around 0.5 MHz. In a color receiver, linear networks can then separate the signals into the three primary colors, red, green, and blue. Simultaneously, a monochrome receiver will respond only to the luminance signal and ignore the color signals.

Although the NTSC system uses bandwidths of 1.5 MHz in the I channel and 0.5 MHz in the Q channel, practical color receivers actually implement a bandwidth of only about 0.5 MHz (or less) in each channel. This compares with several megahertz for the luminance signal. The main reason for the reduced bandwidth is the problem of “chroma noise.” When the higher-frequency chroma signals are synchronously demodulated (a frequency conversion process), the noise added at those frequencies is also brought down in frequency. This causes the interference resulting from the noise to increase in size on the screen. Reducing the chroma bandwidth minimizes this unpleasant effect, but at a reduction in color resolution.

Even with only 1.5 MHz of I signal and 0.5 of Q signal, we still need to find a place to put 2.0 MHz of additional information. The principle of compatibility precludes simply adding more bandwidth. This is where the work of Mertz on the spectrum of scanned images provides an answer. Figure 2.11 showed that the signal energy is clustered around multiples of the horizontal scan frequency, with significant spectral gaps. Since this is a consequence of scanning and the color information likewise is a product of scanning, it too has spectral gaps. The trick then is to arrange the two spectra such that the energy clusters interleave. Figure 2.16 shows that if the color signal is modulated onto a carrier that is an odd harmonic of half of the line frequency, the color signal spectrum will interleave with the monochrome signal spectrum. The color signal occupies frequencies that are nearly empty in a monochrome signal.

Figure 2.16 Luminance spectral energy clusters around harmonics of horizontal scan frequency f_h all the way up to the upper band edge. Chroma spectra energy components cluster around odd multiples of half horizontal scan frequency and are interleaved with luminance.

This technique for hiding the chroma signal in the luma can be visualized in the time domain as well (See Figure 2.17). A cosine wave that is an “odd harmonic of half of the line frequency” will start with the opposite phase on adjacent lines. In the basic case, the first harmonic of half the line frequency will make only half of a cosine wave in the time of one horizontal scan. It will begin the next line with the opposite phase. So where the signal is brightest on one line, it will be dimmest on the adjacent line, and vice versa. These opposite brightnesses on adjacent lines will tend to average out and reduce visibility of the color carrier. The same is true of higher odd harmonics (even harmonics would cause the signal to have the same phase on adjacent lines, and the bright and dark spots would reinforce each other rather than tend to cancel). At the color subcarrier frequency, the bright and dark spots on the screen are very small and are positioned on adjacent lines relative to each other so as to average out the brightness they cause.

Figure 2.17 Time domain view of the frequency interleaving principle used to hide color information.

The two color signals, I and Q, are modulated in quadrature to each other onto the color subcarrier.¹⁶ Quadrature modulation means that two carriers of the same frequency are used, one phase shifted by 90° from the other. The color signal can be expressed as

(2.11)

The sine function is 90° phase shifted from the cosine function. The required detection process is called quadrature detection. In the receiver, quadrature detection cleanly separates the two signals. This separation is based on the following simple trigonometric identities:

(2.12)

(2.13)

(2.14)

Let f₁ = f₂ = f_c, which is the carrier frequency for the color information. Then, from Equation (2.12),

(2.15)

And from Equation (2.13),

(2.16)

And from Equation (2.14),

(2.17)

If we take the color information signal, c(t), of Equation (2.11) and multiply it by cos(2πf_ct), a cosine at the color information carrier frequency, we get

(2.18)

Applying Equation (2.15) to the I(t) term and Equation (2.17) to the Q(t) term, we get

(2.19)

(2.20)

The first term on the right-hand side of Equation (2.20) is the I(t) signal at baseband frequencies. The bracketed terms are modulated onto carriers at twice the chroma carrier frequency, f_c, and can easily be removed with a low-pass filter. Thus, the I(t) signal has been completely separated from the Q(t) signal through quadrature detection.

If we take the color information signal, c(t), of Equation (2.11) and multiply it by sin(2πf_ct), a sine at the color information carrier frequency, we get

(2.21)

Applying Equation (2.27) to the I(t) term and Equation (2.26) to the Q(t) term, we get

(2.22)

(2.23)

As with the I(t) signal, the Q(t) signal has been completely separated through quadrature detection.

The I signal is the in-phase color signal and is amplitude modulated onto a color subcarrier that is located at 3.579545 MHz, the 455th odd multiple of half the line-scanning frequency. The Q signal is the quadrature color signal. It is amplitude modulated on a carrier that is derived by a 90° phase shifting of the color subcarrier (see Figure 2.18). Note that the Q signal is a double sideband signal, whereas the I signal has a 0.5-MHz double sideband component and an additional 1.0 MHz of single sideband modulation. However, as already mentioned, practical consumer receivers limit bandwidth in both channels to around 0.5 MHz.

Figure 2.18 Location and bandwidths of the luminance and chroma signals.

It will be recalled from the study of geometry that there are two common forms of coordinate systems, rectangular and polar. In the rectangular system, a location is defined by the distance in the horizontal direction and the distance in the vertical direction. In the polar system, a location is defined by a distance along an angle. This is like giving driving directions of 3 miles east and 4 miles north. If the traveler has an airplane, he can simply be told to fly 5 miles on the heading of 30°. The diagram of Figure 2.19 applies these principles to the quadrature color Equation (2.11). The color signal c(t) can be given either as the sum of the I(t) signal and the Q(t) signal or as the resulting magnitude c(t), which is the square root of the sum of the squares of I(t) and Q(t) at an angle given by the inverse tangent of the ratio of Q(t)to I(t). In this “polar” way of interpreting the color signal, the color, or hue, is defined by the angle, and the intensity of the color (termed its saturation) is described by the length c(t). Figure 2.20 shows the location of various colors on a polar graph.

Figure 2.19 Rectangular versus polar representation.

Figure 2.20 Polar representation of television colors.

The changes made to accommodate color required the horizontal and vertical scan rates to be slightly modified. This is because the difference between the color subcarrier frequency and the aural subcarrier frequency is a frequency well within the video bandwidth. Any nonlinearity in that signal path will produce a beat product that is in the video passband and can produce an annoying pattern on the screen. This difficulty has been termed the 920-beat problem because the difference between the frequencies selected for the chroma subcarrier and the sound subcarrier is 920 kHz. Two techniques are used to minimize this problem. The existing 4.5-MHz trap strongly attenuates the aural carrier frequency before the video path. However, it was thought that this measure may not be sufficient in existing monochrome receivers that were designed and produced before the color standard was established. The remaining 920-kHz beat was made less visible by causing it to be frequency interleaved (just like the chroma signals) with the luminance signal. This required that the 920-kHz beat be made an odd multiple of half the scan frequency. To accomplish this (since the chroma subcarrier is already an odd multiple of half scan line frequency), the sound carrier must be any multiple of the horizontal scan frequency (that is, an even multiple of half scan line frequency). The 286th multiple of the monochrome horizontal scan frequency (15.75 kHz) is 4.5045 MHz. Using these numbers would require changing the aural subcarrier by 4.5 kHz. Some of the then-existing monochrome television receiver’s audio circuits might have failed to perform with this much of a shift in frequency, jeopardizing the compatibility mandate. Also, the effectiveness of the aural frequency trap would be reduced. The alternative was to change the horizontal frequency so that it was the 286th submultiple of the original 4.5-MHz aural subcarrier. This yields a horizontal frequency of 15,734.26 Hz. Since there are 525 scan lines in a frame and two fields in a frame, the field rate will be the horizontal rate divided by 525 and multiplied by 2, yielding 59.94 Hz. The chroma subcarrier is the 455th multiple of half of this line rate, or 3.579545 MHz.

Synchronous detectors in receivers are used to completely separate the two chroma signals. Then, using circuits that perform the functions listed in Equations (2.24) through (2.26), the luminance signal E_Y, the in-phase color signal E_I, and the quadrature signal E_Q are matrixed to yield the blue signal E_B, the red signal E_R, and the green signal E_G that drive the tricolor picture tube.

(2.24)

(2.25)

(2.26)

Synchronization of the color oscillator used in the synchronous detection process is facilitated by the addition of a burst of unmodulated color subcarrier on each horizontal blanking pulse after the horizontal sync pulse (see Figure 2.21).

Figure 2.21 Color burst relationship to sync pulse.

These techniques allow color television to be compatible with monochrome television. This compatibility is not complete because monochrome receivers built before the introduction of color had video bandwidths of up to 4.2 MHz, providing very sharp black-and-white pictures. When color signals were introduced, these receivers suffered from “dot crawl.” The color signal was not adequately rejected by the older receivers and appeared as a moving pattern of annoying dots. This was overcome in later monochrome receivers by introducing a notch in the frequency response that eliminated much of the color signal but reduced sharpness. Alternatively, the video bandwidth of monochrome receivers was rolled off, with much the same effect.

Adequate compatibility was critical to the rational introduction of color television into a market already populated with monochrome television receivers. Consumers who purchased color receivers derived more benefits from the same signals that consumers with monochrome receivers continued to enjoy. Those who could not afford a color receiver could buy a new monochrome receiver and still have access to television. No one was disenfranchised by the technological advance to color.

2.4.2 Stereo Sound

The next improvement to NTSC came in the area of sound.¹⁷ Television sound is frequency modulated on a separate carrier that is a fixed 4.5 MHz above the visual carrier. When stereo sound was added to the television system, the requirement of compatibility was again enforced. This avoided the chaos that would have resulted from making obsolete the existing sound system. Just as with compatible color, there were compromises to monaural receivers when stereo sound was added. But the net benefit to consumers was considered positive. The marketplace gave its approval to both compatible color and compatible stereo sound.

Stereo sound is implemented by first creating a spectrum that includes the sum of the left and right sound channels at baseband. As in FM radio broadcasting and monophonic television practice, the sum signal is preemphasized using a 75-microsecond preemphasis curve.(See the discussion in Chapter 8 under modulators.) The difference between the left and right channels is double sideband suppressed carrier modulated onto a carrier at twice the horizontal scan frequency (2 × 15,734 = 31,468 Hz). In order to improve the signal-to-noise ratio, this difference signal is companded; that is, the signal is compressed in amplitude at the transmission point and expanded at the receiver. The compression is frequency selective and is based on an algorithm originally proposed by the dBx Corporation, whose name it bears. This is in contrast to the FM radio stereo system, which is not companded but employs the same deviation of the sound carrier, by both the sum and difference carriers. Both changes to the television system were brought about by the need for improved signal-to-noise ratio. A limited-bandwidth (10-kHz) monaural second audio program (SAP) channel is frequency modulated onto a carrier at five times the horizontal scan frequency. The SAP channel is intended for a second language or other such purposes. A very narrow-bandwidth (3.4-kHz) “professional channel” is frequency modulated onto a carrier at six and one-half times the horizontal scan frequency. It is used for television plant intercommunications. This entire complex spectrum is then frequency modulated onto the 4.5-MHz carrier (see Figure 2.22). The relationship between the video and aural carriers is tightly controlled, since nearly all television receivers depend on this relationship. The visual carrier is used as the local oscillator to bring the aural spectrum down to baseband. This technique is called the intercarrier-sound method of TV receiver design. Since the final aural modulation process is that of frequency modulation, the TV receiver uses a “limiter” circuit to strip off any amplitude modulation that may have become impressed upon the aural signal. Typical sources of this amplitude modulation are noise, particularly impulse noise, and cross modulation in nonlinear elements in the signal path. The TV receiver’s aural system then becomes relatively insensitive to amplitude modulation on the aural carrier.

Figure 2.22 Stereo audio baseband spectrum.

A variation of the previously described aural recovery technique is the direct-sound method. It has also been called the split-sound method. In this approach, the aural component is extracted from the composite IF signal prior to video detection. This signal is subjected to limiting to remove incidental amplitude modulation components and presented to an FM detector.

When direct-sound techniques are applied, the aural signal is not dependent upon any properties of the visual carrier, as it is in the intercarrier-sound approach, but is now affected by the accumulated instability of processing oscillators (heterodyning) along the signal chain. These sources of instability include the transmitter, cable heterodyne processors, and local oscillators of intervening frequency translation devices, such as cable subscriber terminal devices and the tuner of the TV itself.

There are trade-offs in making a choice between intercarrier-sound and direct-sound designs. Some visual modulators suffer from nontrivial amounts of incidental phase modulation of the visual carrier. The intercarrier principle causes these visual carrier instabilities to be transferred to the aural carrier, and this can introduce sound distortions. A direct-sound design would not be subject to this problem. Essentially all consumer TVs and VCRs use the intercarrier principle.

2.4.3 Ancillary Signals

It was determined that more uses can be made of the electron blanking time. One important application is to use the VBI to carry analog test signals. The analog test signals can be used to measure the transmission characteristics from the signal source to intermediate points along its path to the final point of use. Another use for the VBI is to carry analog signals that represent digital data. The data signals can be of two or more levels that are resolved into data bits by appropriate circuits. Since the “digital” signals are of just a few discrete levels, the data detection circuits can discriminate against significant amounts of noise, distortion, and interference. This makes these data signals more robust than the analog video signal itself for most forms of interference.

The first U.S. attempt to use the VBI for ancillary data purposes was in 1970, when the National Bureau of Standards (NBS — now the National Institute of Standards and Technology, NIST) proposed to use it for the distribution of precise time information nationwide. The American Broadcasting Company (ABC) television network was a partner in that effort. Even though this initiative did not result in service, ABC recommended a captioning service for the hearing impaired.

Eventually the FCC reserved line 21 of field one of the television signal for the transmission of closed captioning (CC) in the United States in 1976. In 1990, Congress passed the Television Decoder Circuitry Act, mandating that after July 1, 1993, new television receivers of 13-inch diagonal display measure or greater must include caption decoding circuits. Approximately 20 million television receivers per year are covered by this requirement. In 1992, NCI worked with the FCC and the Electronic Industries Association (EIA) to develop captioning technical standards. The 1996 Telecommunications Act required the FCC to institute rules requiring closed captioning on video programming but allowing exemptions for programming that would suffer an “undue burden.”

The CC system supplies data to appropriate digital and analog circuits that place carefully timed text on the television screen, making it possible for the hearing impaired to read a description of the conversation taking place and have indications of other relevant sounds. Similarly, those who cannot understand the spoken words may have text in their language, allowing them to follow the program. The CC system uses very low-speed data in order to minimize the impact of transmission path problems such as reflections and interfering signals. The data rate for the CC system is 503.5 kb/s of binary (two-level) data. This allows only two 8-bit characters to be transmitted per VBI line. If only field one is used, there will be 30 of these lines per second. This yields 480 bps, or 3,600 characters per minute. If the average word is 5 characters long with a space, then 600 words can be conveyed per minute. The rest of the line is occupied with both a burst of seven sine wave cycles of 503.5-kHz clock run-in and a unique “start bits” pattern placed at the beginning of the line. These signals synchronize the detector circuitry. Since only line 21 is protected for captioning by FCC rule, the rate of transmission is slow, but it is perfectly adequate for the purpose. The on-screen display consists of a maximum of 15 rows of 32 characters each. The captions usually appear only on rows 1 through 4 and rows 12 through 15. The middle rows are usually transparent to show the action. A text mode provides scrolling text. Further details can be found as part of EIA standard number EIA-608.

The closed captioning signal carries four components. There are two captioning channels and two text channels. The first captioning channel is synchronized to the video programming, causing the words to carefully match the video. The second captioning channel is not synchronized.

The FCC expanded the captioning standard EIA-608 to allow use of line 21 of field two. This adds two more captioning channels and two more text channels. A fifth channel has been added to carry extended data services (EDS). EDS can carry a wide variety of additional information, including precise time information to set clocks in consumer products. Finally, the original intent of the NBS will be realized.

The channel’s name and call letters are included along with current program information, such as title, length, rating, elapsed time, types of audio services and captioning services, and intended aspect ratio. Also included is the data for the “V-chip” (violent programming advisory), which is intended to control children’s access to programming that parents deem objectionable. Public service announcements such as weather and emergency advisories will be transmitted. Cable system channel layout information will be provided, causing the channel number indicator to use the more familiar channel identification number rather than the number associated with the frequency utilized. This facility can bring the same “channel mapping” benefits subscribers have enjoyed in their cable set-top terminals to consumer electronics products.

A subsequent VBI data transmission system, called teletext,¹⁸ was invented to provide ancillary services to television users. The teletext system can display up to 24 rows of 40 characters on the television screen (but a specification of 20 rows was selected for use in the United States). Teletext quickly evolved into a transmission system for more complex data, including the downloading of software to computers. At the time teletext was introduced, electronic components were still relatively expensive, which severely limited the affordable capability of the teletext systems.

Teletext is a more aggressive form of data transmission than CC. It has been successful in Europe but has failed to enjoy commercialization in the United States. Teletext originated in Great Britain, with experimental transmission commencing in 1972. The British Broadcasting Corporation (BBC) branded their service Ceefax, and the Independent Broadcast Authority (IBA) called their service Oracle. France developed a packet-based teletext system called Antiope, based on a transmission system called Didon. Later, Canada developed another system, called Telidon, which featured higher-resolution graphics. The Japanese system, called Captain, featured “photographic coding” to accommodate Chinese Kanji characters and the Japanese Kana set.

There are a number of reasons for the difficulties in the United States. Principal among these was the failure to find a strategy that made money. Without this, the system could not be supported. Additional difficulties included the high cost of memory at the time of implementation. Even though a teletext page requires only about a kilobyte of storage, that small amount of memory was expensive then. Further problems centered on the quality of the graphics. The less expensive World System Teletext (WST), based on the British approach, had crude “Lego-style” graphics in its basic form. The other contender, the North America Presentation Layer Protocol System (NAPLPS), used a higher-resolution graphics system that painted itself on the screen painfully slowly, trying the patience of the user. Still another complication was the FCC’s 1983 decision to allow two standards and hope that the marketplace would resolve the confusion. One of the systems is WST, and the other is the NAPLPS evolution of Antiope, Telidon, and efforts by AT&T. A final problem was reliability of data reception. In a test in the San Francisco Bay Area, only about 25% of installations of the NAPLPS system were trouble free. The remainder suffered from various degrees of multipath impairment. The more robust WST system was not tested in that environment.

Both U.S. teletext systems have a data rate of 5.727272 Mbps, which is 364 times the horizontal rate and 8/5 of the color subcarrier. The data signal has a nonreturn to zero (NRZ) binary format. The WST data line consists of eight cycles of clock run-in (16 bits) followed by a unique 8-bit “framing code” followed by 16 bits of control codes and a payload of 32 8-bit display words. The displayed row has 40 characters, 8 more than can be conveyed in one VBI line of data. This is accommodated by transmitting 4 rows of 32 characters on 4 VBI lines and a supplementary VBI line with the 4 groups of 8 characters needed to finish the previous 4 display rows. This process is called gearing, because of the ratios between the display and the transmitted VBI data. Since the page format is 40 characters by 20 rows, with an additional “header row,” 21 field lines are required plus 5 field lines of supplementary 8-character groups, for a total of 26 field lines per full page of WST teletext. The payload of 296 bits per line allocated means that if one VBI line in each field is allocated, a data rate of 296 × 2 × 30 = 17,760 bps is obtained. Ten lines of VBI are possible (line 21 is reserved for captions, and the first nine lines form the vertical synchronization pulses), yielding a maximum of 177.6 kb/s for full VBI utilization.

The WST system maps the data location in the VBI line to memory locations and to screen locations, and always puts data in the same memory place. This allows for a very simple error-protection scheme. Since the instructions in the header are Hamming-code protected, a measure of the quality of the received signal is obtained. If the signal is of low quality, it is not loaded into memory. Only good-quality data is stored. As a result, good data can be accumulated from repetitions of the page until a good page of data is built up. It is also possible to use a “voting” approach to obtain very robust transmission.

The fundamental difference between the WST and the evolving set of Antiope, Telidon, and NAPLPS systems is that the latter systems all used a packet structure. They have been characterized as asynchronous because there is no mapping between the transmission scheme and memory and screen locations.

The Public Broadcasting System (PBS) has developed a packetized data delivery system based on teletext called the PBS National Datacast Network. The standard teletext data rate of 5.72 Mbps is used, yielding 9,600 baud per VBI line allocated per field. There are a wide variety of commercial applications for this signal. Currently, the StarSight Electronic Program Guide (EPG) signal is distributed via PBS.

A system by the WavePhore Corporation utilizes a teletext-like system in lines 10 through 20 in each field, for a data speed of up to 150 kb/s. WavePhore added substantial error-detection and -protection bits to its structure to protect against multipath and other transmission problems. Underutilized portions of the NTSC spectrum can be employed to “hide” data. In many cases, the process of hiding the data is incomplete, and results in artifacts under certain conditions. In other cases, the preparation of the NTSC signal to better hide the data reduces video quality. So the challenge is to both hide the data and not impair video quality while retaining signal robustness and the potential for an economic implementation.

The National Data Broadcasting Committee (NDBC) was formed in 1993 to establish a single standard for data transmission in video. On June 28, 1996, the FCC approved digital data transmission in the video portion of broadcast television transmission in its Report & Order (R&O) “Digital Data Transmission Within the Video Portion of Television Broadcast Station Transmissions,” in MM docket No. 95-42. It allows four formats: two of the formats, by Yes! Entertainment Corporation and A. C. Nielsen Co., place low-data-rate signals in the overscan region of the picture; the other two systems, Digideck and WavePhore, embed the digital signal into the video signal.

The WavePhore system begins by reducing video luminance and chrominance bandwidths. The “luminance” is reduced from its theoretical value of 4.2 MHz to a lower value of 3.9 MHz, and the upper sideband of the color signal is reduced by about 300 kHz. It is then possible to insert a data signal in this region at a carrier frequency of approximately 4.197 MHz above the video carrier and a strength approximately 20 dB above the noise floor of the video system. The data is synchronous with the video carrier and thus with the horizontal line frequency. As an odd multiple of one-quarter of the horizontal scan frequency, it interleaves between the luminance and chrominance bundles of spectral energy. Data is not sent during the vertical and horizontal blanking intervals. Thirty bits of data are sent per video line. There are 240 available lines per field (not counting the VBI, during which the signal is blanked). This yields a raw data rate of 431.6 kb/s. After error-correction coding, the raw data rate is reduced to approximately one-fourth of a telephone T1 rate, or 384 kb/s, and referred to by WavePhore as their system TVT1/4.

WavePhore shuffles the data before applying biphase modulation and filtering out the lower sideband. Shuffling the data reduces its visibility in the video. An adaptive equalizer is used in the receiver. A major advantage of the WavePhore approach is that once inserted into the video, the data can be conveyed through the video path without giving it further attention. The WavePhore VBI system and the WavePhore subvideo system can be combined to provide more than 500 kb/s.

There is some degradation of pictures using this system. However, the FCC is willing to let the broadcaster determine what its individual marketplace values are and to respond to that decision.

The Digideck system adds a differential quadrature phase shift keyed (DQPSK) signal carrying about 500 kb/s placed 1 MHz below the video carrier. In this regard, it is similar to the European NICAM system for adding digital audio to analog television broadcasts (see Chapter 7). This places the new carrier in the vestigial side band region of the signal. To accommodate this, the lower VSB slope is increased. Rather than starting at the traditional 750 kHz below picture carrier, in the Digideck system it starts at 500 kHz and drops more rapidly. The carrier is about 36 dB below peak power and has a raw capacity of 700 kb/s. Forward error correction and other overhead burdens reduce the data capacity to around 500 kb/s. Digideck calls the new carrier the D-Channel. The data signal is clocked synchronously to the television signal for ease of recovery and for better hiding in the video. The Digideck receiver also depends on an adaptive equalizer. A consequence of the D-Channel is that it must be inserted at the transmitter site and be brought there by an alternative path. Like the WavePhore system, Digideck introduces some artifacts. A marketplace approach will allow the broadcaster to determine acceptability.

The Yes! Entertainment Corporation’s system introduces a pulse in the video between 9.1 and 10.36 microseconds after the start of the horizontal synchronization pulse. The data rate is very low, about 14 kb/s. Its application is to deliver audio to a talking toy teddy bear! A. C. Nielsen uses line 22 of one field of the video for transmitting a program source identification. This ID is used to measure the viewing population for statistical purposes.

EnCamera Sciences Corporation has presented papers on a system that compatibly embeds up to 4.5 Mb/s in an analog 6-MHz NTSC television signal. Up to 3.0 Mb/s is modulated using multilevel signaling onto a carrier that is in quadrature with the visual carrier. This approach would normally result in visible artifacts in the analog video. However, techniques have been developed that avoid these problems. An additional 1.5 Mb/s is amplitude modulated onto the aural carrier using multilevel signaling and specialized filters.

2.5.1 CableLabs Updates Early Testing

Early testing was done by the Television Allocations Study Organization (TASO) in 1958.¹⁹ Further testing was done in 1983 by the network labs of the Columbia Broadcasting System (CBS), the National Aeronautics and Space Administration (NASA), and the Canadian Research Council (CRC).²⁰ Most recently, in 1991, work sponsored by CableLabs added to the base of knowledge concerning signal quality and customer expectations.²¹

The most significant conclusion is that customer expectations are rising. The TASO study proclaimed that in 1958 consumers found a weighted signal-to-noise ratio of about 30 dB to be “somewhat objectionable.” In 1983, subjects’ expectations were for a much better picture. About 40-dB signal-to-noise ratio was considered to be “somewhat objectionable.” The CableLabs study found 45–46 dB to be considered “slightly annoying.” This is a very substantial increase in demand for picture quality.

The CableLabs tests were conducted under carefully controlled circumstances and used both expert and nonexpert viewers. The expert viewers served primarily to set the endpoints of the scales for the different parameters. The selected visual material was presented on a custom-made analog video disc and on D2 tape.

Well-established procedures defined by the International Consultative Committee for Radio (CCIR) were employed.^22,23 The subjective scale ranged from “imperceptible,” to “perceptible, but not annoying,” “slightly annoying,” “annoying,” and “very annoying.” The double-stimulus pair-comparison method used allows subjects to compare pictures to reference images. This eliminates the need to remember what images looked like at other points in time.

Subject material was selected for sensitivity to the subjective impairments under study. Sources for the material included the Society of Motion Picture and Television Engineers (SMPTE), the European Broadcast Union (EBU), and Kodak Corporation. All these entities have long histories in the analysis of images.

Five impairments were studied in the CableLabs test: (1) random noise, (2) phase noise, (3) chroma/luma delay, (4) microreflections, and (5) composite triple beat (CTB). These impairments are briefly described here (see Table 2.1). Random noise is the result of thermal motion of electrical charges and yields “snow” in the picture. Phase noise is the visual manifestation of instability of oscillators that are used in heterodyning devices to shift frequencies. Several of these systems are in the signal path from point of origination to the subscriber’s screen. Phase noise usually appears as near-horizontal streaking of the picture. Chroma/luma delay results in the colored part of the image misregistering from the luminance part of the picture. The color is smeared and located in the wrong place, giving rise to a comparison with comic book drawings. Microreflections cause delayed versions of the signal to add destructively or constructively. In many cases, the picture is blurred. In a few circumstances, the subjective sharpness of the picture may appear to improve even though the objective measurements may lead to a different conclusion. Composite triple beat (CTB) is the creation of new, undesirable frequency components due to the signal’s passing through nonlinearities. The nonlinearities cause new frequencies to be created that interfere with the quality of the picture.

Table 2.1 Suppression of Impairments for Various Subjective Reactions

Table 2.1 summarizes the results. Quantitative data was not taken for microreflections and so is not included in the table. However, microreflections below 20 dB were not of consequence to most viewers. Below 10 dB, some pictures looked sharper, whereas others looked softer. The most important conclusion is that subscribers have become more critical and are likely to become even more so as they see more images delivered with digital television (DTV).

2.5.2 The W Curve and Direct Pickup (DPU)

In 1986, CBS Laboratories and the NAB tested subjective reaction of viewers to co-channel interference.²⁴ The desired signal was at broadcast channel 23 (visual carrier at 525.5 MHz) and consisted of three images: a gray flat field, a still image (of a boat), and moving video. The undesired signal was a series of sine waves at intervals of 500 kHz, starting at the frequency of the lower band edge of the channel. The ratio of the strength of the interfering sine wave to the strength of the visual carrier was adjusted until expert viewers agreed that the interference was “just perceptible.” To be “just perceptible” at the lower band edge, the strength of the interfering sine wave needed to be attenuated so that it was only some 15 dB below the visual carrier strength. As the frequency of the interfering carrier was increased, more attenuation was required to bring it down to levels that were “just perceptible.” Maximum attenuation, some 68 dB, below visual carrier strength was required at a frequency 250 kHz above the visual carrier. Less attenuation, about 55 dB, was required at the middle of the channel. Again more attenuation, in the low 60s, at the chroma carrier was required. And then, as the upper band edge was approached, less attenuation was required. These measurements yielded a W-shaped curve, as shown in Figure 2.23. The video content of the desired signal had some influence on the amount of attenuation required in the middle of the channel. Somewhat surprisingly, the flat field hid the interference better in the low to mid-frequencies, whereas the moving picture hid the interference better around the color subcarrier. Most surprising was the finding that the difference between the “just perceptible” level and unacceptable levels was just 2–3 dB.

Figure 2.23 W curve: Level of “just perceptible” interfering carriers for three images.

Although co-channel interference is not generally a problem for cable television, direct pickup (DPU) interference is a major headache. In the case of a strong broadcast channel that is carried on the same (or close) frequency on cable, the DPU may result in a ghost or smearing of the image, since the propagation time through the cable is longer than through the air. In the case of other users of the spectrum, second-level-order distortion products, local UHF broadcasters, or cable systems using other than off-air frequencies, interfering patterns may appear on the screen. One manifestation of this is in the cable channel 19 area, where pager transmitters are authorized to operate. Direct pickup can occur in the cable plant or in equipment connected to the cable drop. Those DPU signals can propagate between products connected to the same drop and cause picture degradation. If the tuner of a TV or VCR is not adequately shielded, the DPU will cause ghosts and other interference. An occasional problem arises when subscribers connect FM radios to the cable. These products generally have almost no shielding on their tuners and may conduct DPU signals back into the cable system to interfere with reception by other devices. The W curve gives an indication of sensitivity to DPU at various frequencies.

In 1993, the 1986 CBS tests were repeated, with emphasis on their application to DPU. Since the TASO observations were found to no longer be applicable, it was reasonable to wonder whether consumers had become more sensitive to co-channel and DPU problems. The tests were done on three cable channels (6, 12, and 78), at four frequencies relative to visual carrier (+0.25, +0.75, +1.75, and +2.55 MHz), at two desired signal strengths (0 and +10 dBmV), and at two carrier-to-noise ratios (43 and 50 dB). Only the gray flat field was used for the desired signal. The 1993 tests generally confirmed the results of the 1986 CBS experiments. The W shape of the curve remained. Across the three channels, four frequencies, two desired signal levels, and two carrier-to-noise ratios, the average value of the sine wave strength attenuation relative to the carrier strength to achieve “just perceptible” impairment levels was 56 dB. Since these tests involve human subjects and a wide range of variables, they are difficult to repeat and obtain closely similar results.