A best, one cycle of an analog video frequency can provide information to two pixels.

A conventional NTSC image has 525 lines scanned at 29.97 Hz with a horizontal resolution of 427 pixels. This gives 3.35 MHz (assuming 2 pixels per video cycle) as a minimum bandwidth to carry the video information without compression.

If one decides to move to an HDTV image that is 1050 lines by 600 pixels (keeping the same frame rate), then this means a bandwidth of 18 MHz. Clearly we have a problem here -- as the current terrestrial channel allocations are limited to 6 MHz!

(The word "terrestrial" is TV people jargon for conventional wireless TV transmission. This is to differentiate it from satellite or cable.)

The options for terrestrial broadcast (assuming a 20 MHz bandwidth) are roughly as follows:

1. Change the channel allocation system from 6 MHz to 20 MHz.
2. Compress the signal to fit inside the 6 MHz existing bandwidths
3. Allocate multiple channels (2 with compression or three without) for the HDTV signal

Options 1 and 2 are virtually incompatible with current NTSC service. About the only possibility for maintaining compatibility is simultaneous broadcast of NTSC information over certain channels and HDTV information over other channels.

Option 3 does allow compatibility -- as the first 6 MHz of the signal could keep to the standard NTSC broadcasting and the remaining be additional augmentation signal for HDTV. Typically, in this type of augmentation system, an existing VHF channel would be tied to one (or two) UHF channels. The VHF channel would carry information similar to the current NTSC signal and the UHF channel (or channels would carry augmented high resolution information).

Distribution -- terrestrial? satellite? cable?

There is a lot to consider when you wish to compare satellite tv services with each other, and with other methods of programming distribution.

Advocates for HDTV systems fall into two major categories. There are those that feel that these systems will ultimately be successful outside the conventional channels of terrestrial broadcasting. Equally vehemently, are those that think HDTV can and must use existing terrestrial broadcast channels.

NTSC terrestrial broadcast channels are essentially 6 MHz wide. Service in a given area (roughly a 50 mile circle around the broadcast station) is typically offered on every other channel in order to avoid interference effects. A relatively small range of channels are available (channels 2-69, 55-88, 174-216, 470-806 MHz).

In 1987, the FCC issued a ruling indicating that the HDTV standards to be issued would be compatible with existing NTSC service, and would be confined to the existing VHF and UHF frequency bands.

In 1990, the FCC announced that HDTV would be simultaneously broadcast (rather than augmented) and that its preference would be for a full HDTV standard (rather than the reduced resolution EDTV).

These two decisions are very interesting, as they are almost contradictory. The 1987 decision is clearly leaning toward a augmentation type format -- where the NTSC service continues intact and new channels provide HDTV augmentation to the existing. The 1990 decision is a radical and non-conservative approach -- one which basically removes the requirement for compatibility by allowing different HDTV and NTSC standards to exist simultaneously for a period of years. Then the NTSC is gradually faded out as the HDTV takes over.

Now, the FCC does not have jurisdiction over channel allocation in cable networks. Thus, there is the rather interesting question of what the cable TV companies will do. They have a number of interesting options. They can continue to broadcast conventional NTSC, they can install 20 MHz MUSE-type HDTV systems (or other types of HDTV systems), or they can go with the digital Grand Alliance systems. This presents the interesting possibility of two different HDTV standards, one for terrestrial broadcast and one for cable broadcast.

Interlaced versus non-interlaced.

The maximum vertical resolution promised by a particular TV system is greater than the actual observed resolution. The reduction in resolution is due to the possibility of a picture element (pixel) falling "in-between" the scanning lines. Measurement gives a effective resolution of about 70% of the maximum resolution (the Kell factor) for progressively scanned (i.e. not interlaced) systems. If the image is interlaced, then the 70% factor only applies if the image is completely stationary. For non-stationary interlaced images this resolution falls to about 50%.

Interlacing also produces serrated edges to moving objects, as well as flicker along horizontal edges (glitter) and misaligned frames. As a consequence of the many problems associated with interlacing, a number of the HDTV proposals are for progressively scanned (not interlaced) service. Notice that these apply both to new ideas for HDTV, and to upgrades of the existing NTSC, PAL and SECAM systems as well. (Although initiating progressive scanning on conventional service does create compatibility problems, some of these techniques offer improved performance to NTSC/PAL/SECAM without the associated problems of moving to "true" HDTV.)

Compression

Even if extra channel space is available -- it is usually not enough for the very wide bandwidths of HDTV. As an example, the current NHK satellite broadcast system in Japan (the only "in-service" HDTV system) requires 20 MHz, but only has 8.15 MHz available per channel from direct satellite broadcast.

Thus, some type of compression is typically required. Interestingly enough, although these compression schemes result in analog signals -- they are digitally implemented.

Thus, the line between "digital" and "analog" HDTV begins to blur.

1. Signal compression in the MUSE system

The MUSE currently used for satellite HDTV service in Japan is a modification of the NHK HDTV standard for direct broadcast satellite service. The wide bandwidth of the NHK HDTV system is too large for the 8.15 DBS service. As a consequence, the signal must be compressed.

The NHK HDTV signal is initially sampled at 48.6 Ms/s. This signal controls two filters, one responsive to stationary parts of the image -- one responsive to moving parts. The outputs of the two filters are combined and then sampled at the sub-Nyquist frequency of 16.2 MHz. The resulting pulse train is then converted by to analog with a base frequency of 8.1 MHz.

What is happening here is that the subsampling results in successive transmission of signals representing every third picture element. Thus, three adjacent picture elements in the receiver actually represent three successive scans of the same line. Stationary objects are not bothered by this, and appear at their full resolution. However, moving objects do not reoccur in their proper positions and create a smearing effect. This is not a real problem with moving objects in the scene (as the human eye is not very sensitive to this either). However, it does present a problem during camera panning, where the overall image suffers about a 50% drop in resolution -- while the human eye does not.

2. Signal compression in the MAC system

The MAC system was originally proposed as the analog compression standard for European HDTV. Under the original plans, HDTV broadcasts using MAC would be standard in 1995. However, for a variety of reasons, MAC did not make it in Europe. In fact, MAC has died so hard that Europe may simply wait until the US develops an all digital HDTV standard and then use a 50 Hz modified version of it. (As an aside, an interesting situation occurs with European HDTV systems. The peripheral vision is much more sensitive to contrast and movement than foveal vision. As a consequence, the 50 Hz field rate (25 Hz frame rate) has been found to be too slow. The edges of a 50 Hz HDTV image will flicker. Thus, most European HDTV systems advocate 100 Hz.)

However, in spite of the political death of MAC, the technological aspects of the compression are very interesting and worth knowing about. Basically the MAC (multiplexed analog components) compression system fits the luminance and chrominance information into the horizontal line scan in a sequential way. In other words, the R-Y information is sent on one scan, and the B-Y on the next scan.

The color difference and luminance information is sent in a time multiplexed fashion. Looking at the signal in time, the first part of the signal is audio information, followed by chrominance (R-Y or B-Y), followed by luminance.

In order to get the signal into this form require some serious digital processing. Initially, the luminance, R-Y and B-Y signals are sampled and stored digitally. The luminance is sampled at 13.5 MHz and the color difference signals at 6.75 MHz. Then a 3/2 compression on the luminance and a 3/1 compression on the chrominance is performed.

Now, the three signals are read out to produce pulse trains, and then back converted into analog form. The time compression resulting from this operation allows them to be time domain multiplexed in order to fit within the 64 uS horizontal scan time.