DBS Digital Video 101: The Flame Filled Rant ;-)

This page last modified 1-9-2002.

Table of Contents

January 2002 Update
Introduction
Lesson 1. "TV Lines" vs. a film photographers "Lines of Resolution".
Nyquist and TV Lines, what is the resolution of a digitized TV image.
Analog waveforms and TV lines: What is the resolution of an analog TV signal.
"Enhanced" NTSC, getting more horizontal resolution.
Digitization of the NTSC signal for transmission
The Dish vs. VHS Myth, the ultimate FUD
Discussion about MPEG-2 encoding.
Opinion section: The MPEG variability factor.
References and Links
About the Author

January 2002 Update

This page was originally written in 1998 and went through a fairly extensive revision in 1999.

Since then, a lot of things have happened in the world of DBS.

For one thing, both providers have upped their compression considerably over the years.

On January 1 of this year, a government regulation called "must carry" kicked in. This regulation states that the providers must carry all the local TV stations in any market where they carry any of the locals.

Both providers use a technology called "spot beam satellites" to accomplish this. Spot beam satellites re-use a few transponder frequencies over and over, beaming them down in narrow "spots" that only cover two or three metropolitan areas each. The spots carry the locals for the metropolitan areas they illuminate.

Trouble is, DirecTV's spot beam satellites were delivered by January 1, 2002. Dish Network's were not. (Both of the providers ordered their spot beam satellites at about the same time, both were delayed beyond their promised delivery dates. Dish's satellites were delayed a few months more than DirecTV's.)

As a result, at the moment I write this, Dish Networks picture quality is truly awful as a result them being forced to carry way more channels than their non spot beam satellites have the capacity for.

The "worse quality than VHS tape" mantra, that I go through so much trouble to rebuke in the text that follows, is unfortunately a fairly accurate description at the moment.

However, when Dish Network's spot beams come on line later this year, their picture quality should improve a great deal.

On an additional note. I have learned that Dish Network's Pay-Per-View channels are in fact, digitized to 720 X 480 pixels, whereas the "basic" channels are digitized to 480 X 480 as discussed below. Sort of cool news!


Introduction

I was inspired to create this document as a result of the plethora of awful misinformation about the technology behind the transmission of DBS television signals that has been proliferating through various newsgroups of late. Much of this misinformation consisted of junk science and made up technical information, designed to support certain opinions about which DBS provider was "better" in the opinion of the poster.

As a person who's professional specialty is digital video processing, it was causing me great pain to read these misleading posts, and I decided to write this tutorial, in order to clear up some of these misunderstandings regarding the technology.


Lesson 1. "TV Lines" vs. a film photographers "Lines of Resolution".

There is a fundamental difference between how resolution is measured in the context of television and in the world of film photography.

When a film photographer speaks of "lines of resolution", he is referring to the maximum number of straight parallel black and white alternating lines that a piece of film can resolve over a given area. However, the photographer only counts the black lines. That is, the photographers target looks like alternating black and white pin-stripes. When the photographer refers to 100 "lines". He is counting only the black ones. In other words, each pair of one black and one white bar is being called a "line".

In the world of video, each black and each white line are each counted as a line. That is black bar is line one, subsequent white bar is line two, black bar is line three, etc.

What this means is, a picture that from the photographers definition would have 200 lines of resolution would have 400 lines of resolution by the conventional video definition.

The video definition has a further complication. When a video resolution figure is expressed, it is given as "resolvable vertical lines per frame height". What this means is, if a picture is said to have 480 "TV lines" of resolution, that means in a square the same width as the screen is high, 480 lines could be displayed. Since an NTSC standard TV screen has an aspect ratio of 4/3, to calculate the total number of vertical lines that can be displayed across the entire width of the screen, you multiply by 4/3. This means that an image rated at 480 "TV lines" can display 640 lines across the entire screen width.


Nyquist and TV Lines, what is the resolution of a digitized TV image.

When an analog NTSC TV signal is digitized, it is converted into discrete blocks of information called "pixels". The images is mapped to an array of these pixels of a certain width an height. Each pixel is a "sample" of the continuous analog information.

The Nyquist sampling theorem tells us that the maximum frequency that can be accurately represented by digital sampling, is one half the sampling rate. For example, in audio, to obtain a frequency response of 20,000 HZ, your sampling frequency, at minimum, must be 40,000 hz.

NOTE: Email flaming me for not discussing how waveforms right at the Nyquist limit have no phase information stored will go straight to the bit bucket. I understand this but the discussion of this aspect of Nyquist sampling is beyond the scope of this document, and is an UTTER red herring with respect to the point I'm making.

One scan-line of the "vertical line" test pattern used for determining "TV Line" resolution results in a square wave, with each white bar being a maxima on the wave, and each black bar being a minima. Each "cycle" of this wave contains 2 "lines", one black and one white.

So,
Each wave cycle requires a minimum of two "samples" (pixels) to represent it.

Each cycle contains two "lines".

Ergo:
Each pixel in the horizontal direction can resolve one line.

Shown graphically:

---     ---     ---     ---     ---      <--White lines
   |   |   |   |   |   |   |   |   |   |  
    ---     ---     ---     ---     ---  <--Black lines

--- --- --- --- --- --- --- --- --- ---  <--Pixels.

Each pixel can resolve one line.
So, a video image of 640 (H) by 480 (V) pixels, can resolve 640 lines all the way across. To get the "TV line" rating (remember, vertical lines per picture height) we must multiply by 3/4. We get a picture that would be rated at 480 "TV lines". (Well, what do you know, square pixels!)

Analog waveforms and TV lines: What is the resolution of an analog TV signal.

So, now that we have analyzed how resolution of a digital TV signal is determined, you may ask how do you determine the resolution of an analog TV signal.

In a straight baseband signal, essentially infinite horizontal resolutions would be theoretically possible (a bit more on this later), however when signals are broadcast via radio waves, practicality requires that their bandwidth be restricted so that a reasonable number of signals can be fit in a reasonable amount of spectrum space.

The FCC allows a total bandwidth of 6 Mhz per TV channel for UHF and VHF. The following figure, shows a breakdown on how that 6Mhz is utilized:

Spectrum of a TV channel in North America

The broad chunk of spectrum on the graph is the "lumanance" signal, the black and white portion of the picture. The lumanance signal is derived from the red (R), green (G) and blue (B) components of the color picture according to the following formula:

Y (lumanance) = 0.3R + 0.59G + 0.11B

The I and Q segments of the graph represent phase encoded representations of the signals known in the digital video world as "U" and "V", where U = R - Y and V = B - Y.

>From these three signals, the receiver can derive, in analog or digital (analog most commonly, of course, but this is rapidly changing) the correct R, G, and B values.

The total bandwidth of the lumanance signal is 5.33 Mhz, lets break down what this bandwidth is utilized for. Lets take a look at the waveform of one scan line in the video image:

A TV Line
The low square dips on the graph represent the horizontal sync pulses. These pulses trigger the horizontal sweep in the TV to reset back to the left to begin another scan line.

The flat areas on either side of the sync pulse are known as the "horizontal blanking interval". This gives the electron beam time to reset back to the left of the screen to begin the next scan line.

Superimposed on the horizontal blanking interval, is the "colorburst". This is the actual physical manifestation of the Q and I signals upon the demodulated signal. A closeup of what the colorburst looks like is shown here:

colorburst signal

So, how do we determine just how many lines can be displayed upon one of these scan lines?

Well, remembering that "TV lines" refers to a row of black and white bars where all the black ones and all the white ones are counted, consider the following:

Imagine if the video information on the scanlines above consisted of a high frequency sine wave extending all the way from the 2 volt white level to the 0.5 volt black level, and suppose that all the scanlines in the image consisted of the identical sine wave, perfectly aligned with one another. What would the image look like?

I hear you say, "a corrugated metal roof."

Yes, yes, yes, but... remember that TV resolution was defined by video engineers who didn't want their format to sound too awful. Imagine, since this a very steep wave, that it looks like a row of black and white lines! Each positive lobe of the wave is a white line and each negative lobe of the wave is a black line. There are TWO lines per cycle of the wave.

So, what is the most of these black and white lines that we could display. Let's do a little math:

We have 5.33 Mhz available, and there are 15,750 scan lines per second (30 X 525 = 15750)

So 5,330,000 cycles / 15750 lines = 338.4 cycles/line.

Note, however that we don't have the entire line available for displaying video information. We only have 42 out of 63.5 microseconds of each line available. So we have to multiply our 338.4 cycles by 42/63.5, the result is 223.82 cycles.

Since there are two lines being displayed per cycle, we have 447.6 lines being displayed across the entire width of the screen.

Now, since horizontal video resolution is expressed as "Vertical lines per picture height", we must multiply this by the aspect ratio of the screen, we get:

447.6 * 3/4 = 335.7.

Expressing this as an easy to remember round number we get:

330 lines.

This slight downrating of the figure takes into account the rolloff at either end of the bandwidth, portions of the picture masked at the end of the screen, etc.

Even 330 represents a bit of specmanship however. If the frequency spectrum encroaches too far into the chroma spectrum (even this calculation is based upon it encroaching into it completely), the croma circuitry of the TV is likely to interpret it as croma information. The result is "rainbows" on the screen. In fact, lines of a density of greater than 180 TV lines or so create the risk of producing "rainbows". When that doofus sportscaster on the local news wears his loud sportcoat with the pinstripes and it makes you go blind with the rainbows... This is what's happening.

Actually, the color information is phase encoded as well as frequency encoded, so it takes fine details coupled with some coincidences regarding their phase information for this effect to be really bad.


"Enhanced" NTSC, getting more horizontal resolution.

So, how can some formats like Hi-8, SVHS, LaserDisk, DVD, and DBS claim to have a resolution Higher than 330 TV lines given the limitations above?

Remember, the 6 MHZ bandwidth is imposed by the FCC to make for reasonably frugal use of the finite RF spectrum. Any RF modulated signal, including OTA signals, Cable TV signals, and the signal coming out of the RF outs on your DBS receiver are going to be limited by these restrictions.

However, as long as we don't have to RF modulate the signal, we can overcome these restrictions.

The composite output on your LaserDisk player, your DVD, or your DBS receiver, could theoretically have an infinite bandwidth. However, the chroma signal is still embedded in this signal, just like it is shown in the graph above. So, this composite signal has to be band limited to, at least almost the same 330 line level, or your picture could be rainbow city.

The solution: the Svideo jack.

The Svideo jack carries the chroma signal (Q and I) and the lumanance signal on separate physical wires. This results in a bandwidth limited only by other limitations of the source or the receiver. This is what gives DVD's and DBS a potential to display nearly 500 TV lines.


Digitization of the NTSC signal for transmission

In simple terms, the DBS provider is transmitting his information in the digital domain. The source material, and your TV set, however, are analog NTSC.

The first step is to digitize the signal into an uncompressed digital representation of the signal. This uncompressed stream of data is then compressed, using the compression rules specified under MPEG-2, (or something similar but proprietary in the case of Primestar), so as to squeeze the most possible channels into the available bandwidth, then at your receiver, it is expanded back into an uncompressed digital representation, and finally, converted back to an analog NTSC signal (or "extended" NTSC, in the case of the signal coming from the Svideo output jack).

Graphically:

Analog-->Uncompressed video-->compressed video
-->Transmitted to you via satellite-->uncompressed video-->Analog.
The MPEG-2 compression standard provides for an arbitrary frame size in terms of vertical pixels and horizontal pixels.

The number of pixels in the vertical direction is a function of the number of active scan lines, so is fixed. For NTSC, its 480 pixels.

The horizontal frame buffer size, which can also be expressed in terms of the sampling rate on each scan line, can be varied by the DBS provider.

DVD players digitize 665 pixels in the horizontal direction. DBS receivers, at least those that are DVB compliant, like the Echostar (DiSH Network) models, can handle this resolution and probably higher resolutions as well. Canada's ExpressVu (who uses Echostar receivers) has, in fact, announced plans to launch at least one "DVD Quality" PPV channel, which will be digitized to this resolution and carry the same bitrate as a DVD. (Put another way, it will be compressed as much, or as little, as a DVD.)

At present, however, none of the DBS providers in North America digitize to this resolution, and all of them run quite a low bitrate, less than 3 megabits-per-second, average per channel.

DirecTV stated upon their initial launch that their horizontal resolution was 544 pixels. It is quite possible that they have snuck it down since then. Do to the proprietary nature of their receivers, however, there is no way to measure this, and they, of course, keep it a closely guarded secret.

If a provider is using the DVB standard, however, their digitization rate can be measured with commercially available equipment. The author of The CoolStuff page on North American MPEG-2 reports that he used a Nokia Mediamaster receiver to pipe the transport stream from Dish Network into the SCSI bus of a PC and analyze it. He found that the Dish Network program he was measuring was digitized to 480 X 480. Canada's ExpressVu has been measured at the same value (which is not surprising since Echostar provides their receivers and set up their uplink center for them). NOTE: Added 1-8-2002 -- Dish Newtwork has published recently that their Pay Per View movies are transmitted at a digitization of 720 X 480, most other channels at the above mentioned 480 X 480.

So, lets calculate the "TV lines" resolution figure for some of the digitization rates discussed above:

For the DVD standard sampling of 665 pixels: 665 X 3/4 = 498.75 TV lines.

So: The theoretical maximum resolution you might ever see out of any DBS provider (e.g on ExpressVu's forthcoming "DVD quality" channels):

498.75 TV Lines for "DVD quality"

Or, 665 pixels of total horizontal resolution.

For a provider using 544 pixels, the resolution would be: 544 X 3/4 = 408.

408 TV lines (perhaps) on DirecTV.

For DiSH network and ExpressVu's "regular" channels: 480 * 3/4 =

360 TV Lines on most North American DVB channels.

One additional note:

DiSH (and I'm positive DirecTV also) is transmitting half the resolution in the horizontal and vertical directions on the red and blue signals as on the luminance (green is derived from subtracting the red and blue from the luminance). In broadcast television, the red and blue have about half the bandwidth of the luminance. This makes them have about half the resolution of the luminance in the horizontal direction. Psycho-visual experiments have tended to suggest that having these color signals symmetrically lower in resolution than the luminance is less disturbing to the eye/brain than an asymmetrical reduction (like in broadcast signals), and it saves bandwidth too! Of course, this is purely subjective and would always be a source of debate.


The Dish vs. VHS Myth, the ultimate FUD

Now we get to the usenet myth that inspired the writing of this page more than any other.

It was once circulating in USENET groups that DISH Network has a resolution of something less than VHS tape. VHS tape, in SP mode, has a resolution of about 240 TV lines.

One guy put up a post explaining how he "proved" that the resolution of DISH network was inferior to VHS tape. In his post, he explained how he had used a pixel meter (essentially a spectrum analyzer that converts the frequency spectrum of the signal to pixels) to analyze the spectrum of the luminance channel output from a DISH network receiver.

He explained how it showed that the picture contained effectively the equivalent of 550 pixels in the horizontal direction. He then went on to explain how the Nyquist theorem tells us that it requires two pixels to resolve one line. Therefore, the resolution of the DISH signal was about 205 TV lines, less than VHS. (550/2 * 3/4) = 206.25.

Did you catch what he did wrong?

Right! He used the photographers definition of "lines of resolution". It only takes one pixel to resolve one TV line. So what he measured was a resolution of 412 TV lines! (550 * 3/4)

Now: This value is above the 360 TV lines we calculated for DiSH above. It is even greater than the 408 we calculated for DirecTV. What's up with that?

A few possibilities:

However, assuming he didn't make the whole thing up:

In spite of his FUD spreading agenda, this guy actually proved that the image he was measuring contained actual bona fide details within the resolution range given above, and well above OTA signals!


Discussion about MPEG-2 Encoding.

This is where I will be putting a technical discussion of MPEG-2. How it works, combined with opinion about when it works well and when it fails. I haven't found the time to write it up yet.

Meanwhile, check out these excellent links about MPEG-2:
MPEG page from "CoolStuff."
Video compression for direct broadcast satellite (DBS) and broadcasting. From HEI.


Opinion section: The MPEG variability factor.

NOTE: Added 01-09-2002:

As has been mentioned at the top of this page, both providers have upped their compression considerably since this page was last significantly updated in 1999. In fact, at the present time, Dish Network is extremely compressed due to having to add must carry channels before their spot beam satellites were delivered. As of this date, Dish Network does look significantly more compressed than DirecTV. Around mid 2002, when Dish Network's spot beam satellites are operational, it is likely that the quality of the two providers will once again be comparable.

Now, back to the obsolete text written in 1999!

When I originally wrote this page, there was a huge FUD war going on in the DBS oriented newsgroups over which DBS carrier, primarily DISH Network or Direct TV has the best quality, with the advocates of one side making ludicrous, way way out there clams about the "poor" quality of the other. The "DISH has lower resolution than VHS tape" nonsense being a prime example.

However, the grain of truth behind this FUD is that these mediums employ a lossy compression scheme. This offers an element of extreme variability that, in theory, could result in a serious quality difference between DTV and DISH. But does it really?

Both systems use MPEG-2 rules for compression and multiplexing of their channels onto the transponders. Actually Direct TV uses a slight variation of MPEG-2 and DISH uses "pure" MPEG-2. The reason being that Direct TV went on line when only a preliminary draft of MPEG-2 existed, they based their system on this draft and filled in the holes in the draft with some ideas of their own. DISH was able to use the final draft. However, these differences have no impact whatsoever on the quality potential of the two providers.

DISH has 21 transponders available at their main 119 degree orbital location. DTV has 32 transponders available at their 101 degree main orbital location. However, DTV is running more channels from their main location, and both services are running 7 to 8 video channels per transponder, meaning that they each have identical bandwidth available per channel.

(Update 8/25/99: While they juggle their satellites around, DiSH is temporarily up to 10 channels per transponder. This temporary situation will end sometime in September of 1999, which will put them back on par with DirecTV. Ironically, this very real and drastic temporary reduction of DiSH's quality has not re-ignited the flames and the FUD of about a year ago when I first put this page up.

Now, it is possible to set your MPEG parameters badly and get a vastly inferior signal to someone else with the same bandwidth that has set them well. However, DISH network is a division of EchoStar Corporation, world renown for manufacturing the most well respected C-band and Ku-band downlink and uplink equipment. Their equipment is sold and widely used around the world. DTV is a division of Hughes Spacecraft Corporation. I find it hard to believe that, with their respective pedigrees, that one of these companies has an engineering staff composed of a bunch of Eienstiens and the other has one composed of a bunch of chimpanzees!

However, it is very possible that each of these providers may selectively filter some "throw away" channels so as to lower their resolution and increase their compressibility. (On DISH network, TV Land, Game Show Network, and especially Noggin do appear rather low res to me.) However, measured readings, common sense, and my own eyeballs tell me that on critical channels like movie and sports channels both providers keep the compression and filtering admirably mild.


References and Links

Some of these links were used as a reference for writing this page, others may be of interest along the same subject lines.

The DBS Online technical reference page.
And their front page.
Slides from a lecture on OTA transmission.
Lyngsat satelite charts.
A page on world TV standards and conversions.
Leopold's comparison of home video formats.
Leopold's "How Film is Transfered to Video" page.
Leopold's front page.
DBS DISH Satelite news and Information
The "How Television Works" chapter of "How Stuff Works"


About the Author

Here is my personal homepage.