Aris Silzars Display Technology Consulting


More On Rows and Columns...October 2002

In the June 2002 issue the subject of this column was the fundamental difference between CRTs and all flat-panel displays -- the difference being that all flat-panel displays that exist at this time require row-and-column addressing, while CRTs do not. In describing the various ways of modulating the electron beam that writes the information onto a CRT phosphor screen, I posed the questions "Why do we write the image from left to right? Why not scan back and forth with a triangle waveform?" Well, being the knowledgeable group of readers that you are, I received quite a number of interesting responses. As we discussed this topic via e-mail, it occurred to me that your letters were beginning to form an interesting story all in themselves. Therefore, I have decided to dedicate this month's column space to your stories. My modest contribution was to arrange the letters in the sequence that seemed to have a logical flow to it. So here we go -- with seven of the most interesting letters about how to write images onto a CRT screen.

Mr. Silzars,
Your question about the reason for using sawtooth raster scanning rather than some other method makes me think that there probably are a multitude of very interesting engineering stories that could be told about the development of what later turned into standards and conventions. In regard to sawtooth raster scanning, it is easy to show in hindsight why some other methods such as triangle wave or sine wave scanning would not work. The question is, who and with what process developed the technique that is now standard? Were there a multitude of experiments or was the technique developed mostly from theory? In addition to sawtooth scanning, how was the decision to use 2:1 interlace made? Why not 3:1 or 4:1? If any of your readers have the real scoop, I hope you can mention it in a future column.

Roger Weise

Hello Aris,
You asked, in Information Display 6/02, about scanning in both directions. I recall several papers in SID and IEEE publications over the last 20 years describing efforts to implement scanning in both directions. I don't have the papers in hand, but I recall that aligning adjacent scan lines was a problem. Maybe others will write with more details. And maybe you are really asking why left to right scan was incorporated into the NTSC standard.

Last year you may remember me asking why horizontal scan is used in television instead of vertical scan. I received many answers. However, the Director of the David Sarnoff Library, Dr. Alex Magoun, directed me to the answer in the literature. Donald Fink writes, in Television Standards and Practice (1943) on page 206, "…The Panel 1 outline also recommends the horizontal scan for lines, because horizontal motion is much more common than vertical motion and is more clearly depicted when lines lie in this direction. Vertical motion across the line structure often produces an effect known as optical pairing of the lines, which arises from the persistence of vision in the eye and reduces the apparent detail in that direction..."

Roger Casanova Alig, PhD

Dear Mr. Silzars,
The subject you discuss about why CRT raster is only left to right has captured my interest too. I was not in the display business when the standards were developed but I know of one good reason why they would have difficulty in making a commercial left-to-right and right-to-left scan. Magnetic deflection has hysteresis. That is, the position of the beam on the screen is not only a function of the instantaneous value of the deflection current but the recent history of the current as well. This would vary from yoke to yoke since there could be a different amount or type of ferromagnetic core that is used to produce the magnetic field. This would cause an alternate line misregistration.

My interest in this topic is that I am developing electrostatic deflection CRTs for various applications. One advantage is that electrostatic deflection has no hysteresis so reversing the sweep direction does not cause misregistration errors.

Michael Retsky

Dear Mr. Silzars,
In my limited knowledge of TV, I would assume that the present scanning method dates back to the `Nipkow' scanning disks and before Farnsworth, etc. I recall back in '41 when my deceased ex-boss brought Baird British theatre TV rear-projection equipment from New York to Rauland in Chicago, which used an 80-kV polished metal back phosphor screen goose-neck tube. The horizontal deflection system used a pair of simple pancake coils shaped over and under the tube neck. Because the tube neck was on an angle to the phosphor screen the vertical deflection coil was on the bottom of a U-shaped laminated transformer iron yoke between the horizontal coils and the gun end of the tube neck. The arms and two short extensions of the arms could be manipulated in and out to correct for the trapezoidal pattern. Large optical projection lenses were perpendicular to the flat polished-face-tube phosphor screen. The cabinet was lead lined. After Zenith took over, the unit was scrapped. (Curses!) Two other British console TVs had the same deflection system. One of the sets is in the office lobby of Rauland's Skokie, Illinois manufacturing building.


Dear Mr. Silzars,
To my understanding, TV was quite mechanical at the very beginning. Initial "cameras" had mechanical scanner units, e.g. the Nipkow wheel, with a spiral order of holes as a means for sequentially scanning an image. This determined scanning in parallel for all lines and did not allow back and forth (clockwise or counterclockwise would have been equivalent). The display also consisted of such a wheel to control an illumination light path to a screen. This was relatively soon replaced by the CRT, which then had to follow the former mechanical scanning sequence for obvious reasons of compatibility. This also holds again for the later introduction of the electronic camera.

All revisions of TV systems mainly changed resolution, aspect ratio and so on, but never the basis scheme. I was able to find some further details in an old book by Eduard Rhein (published in the 1950s), unfortunately without precise references. P. Nipkow filed a patent on a wheel or disk with a spiral pattern of holes in 1883. No practical use or demonstration was claimed at that time. P. Weiller invented a drum with tilted mirrors in 1928. This was similar to today’s laser bar-code scanners. The earliest proposal for the use of a CRT, as far as I found, is a patent by M. Dieckmann and G. Glage (both coworkers of Ferdinand Braun) in 1906. A CRT was also proposed and used as a scanning light source. The number of lines for mechanical scanning remained in the order of 100. All these means were mainly applied to scan movie film with only an occasional demonstration of that could be considered as "live TV".

There is an illustrating episode. At the 1936 Olympics in Berlin, a movie camera was used to take the scene. The film was immediately processed and was then scanned for TV broadcasting with an approximately one-minute delay. The problem was to synchronize the speaker’s comments with the image. I hope you will find this to be of interest, even if the precision of the historical data is weak, not to mention the completeness.

Werner Fertig

Dear Aris,
A colleague, who has been in the industry 25 years longer than I (and so is about to retire), has provided the reasons for CRT scanning using fly-back between lines. There are two principal reasons why a boustrophedonic scanning system (the word derives from the method of ploughing a field with oxen `bos') was not practical.

One concerns the end-to-end chain from camera to display, and relates to timing. Any timing discrepancy between picture and syncs will of course create a zig-zag horizontal offset between successive lines. In a conventional scanning system, timing errors only cause a horizontal shift of the whole picture, or at worst, a slight oblique slant.

The second concerns the vertical drive waveform (in both a tube camera and a CRT display). In the simplest conventional system the beam descends at a constant rate, such that during the time taken to traverse a single line, plus the fly-back period, it has descended by the required line pitch. Thus, the lines actually slope an unnoticeably small amount. On the contrary, in the boustrophedonic system, the beam would have to remain horizontal during the line-scan period, and then rapidly descend to the next line in preparation for the following horizontal transit. To obtain a decent picture, the vertical drive waveform would thus need almost infinite bandwidth, and much more complex circuitry.

To move on to a pint made later in your article, your suggestion that large displays of more that 700 vertical lines will be confined to the cinema (where in any case, release prints hardly even reach that resolution), found a resonance in our own discussions amongst broadcasters in Europe. However, I wonder how this will change over the medium term. I would be very interested to hear from people working across the display industry as to their views of whether, in ten years time, they expect the best domestically affordable 40- to 60-inch progressive displays will be in the 720/768-line or 1080-line class.

Richard Salmon, Senior Engineer, BBC R&D Department

To answer your question (Why not scan back and forth with a triangle waveform?):
Because the return beam would have to exactly match the forward beam's velocity nonlinearity. For example, if the forward beam moves fast, then slow, the return beam would have to move slow, then fast, to maintain proper registration. This can be done digitally, but is practically impossible with an analog sweep.

Sid Deutsch, author of Theory and Design of Television Receivers, McGraw-Hill Book Co. New York, 1951.

Well, there you have it. I'm sure that some of these letters will generate yet other interesting discussions. For example, Richard Salmon's request for your thoughts on what will happen with television resolution over the next several years would for sure be suitable for another column of reader responses. Let me know your predictions and we may do this again.

One further comment on row and column addressing was made to me via telephone by Allan Kmetz, our current President. He pointed out that there is an error in the way I have stated that, "For every doubling of linear resolution, the number of drivers increases by a factor of four." That is only true for active matrix displays if the drivers, to which I refer, are the active elements at each pixel. However, since most of the discussion can be assumed to apply to passive-matrix displays and all other flat panels with external row and column drivers, that statement is not correct for those situations. For displays such as plasma panels, the number of row and column drivers will increase only in direct proportion to the increase in linear resolution.

Finally, in my closing paragraph of the June column I posed the question, "Is there a large-screen display in your immediate future?" The most enthusiastic response came from Trisha Murray ( Among her many reasons for buying a 20.1" flat panel were, "... weight savings that the FP monitor offers." and "Big monitors showing planes flying around are very attractive eye candy at trade shows." A useful resolution of 1200x1600 was a further motivator. The $1250 price proved not to be a deterrent to her enthusiasm. Is she the harbinger of a new trend?

Starting with next month's issue, these View From the Hilltop columns will appear only every other month. I have been asked to relinquish some of this space so that others may have an opportunity for contributing their ideas and opinions to the international display community. My contact coordinates, however, remain the same, and your responses will, as always, be most welcomed and appreciated. You may contact me by e-mail at Email, by telephone at 425-557-8850, by fax at 425-557-8983, or by mail at 22513 S.E. 47th Place, Sammamish, WA 98075.

See you in December!