Hole shape in Nipkow disks

[Andrew Davie]

It seems that most Nipkow disks (?) are currently made using circular holes. Since display intensity is a major issue -- such a small amount of the total light 'gets through', it would seem desirable to have as large a hole as possible.

A quick calculation shows that a square hole would increase light intensity over a round hole of the same diameter by rougly 27%. This is a pretty major increase, IMHO.

I'm aware that some have experimented with oval and diamond holes -- mostly to (I think) alleviate the visible scanline gaps -- but can anyone report on their thoughts regarding the use of square holes?

[Jim W.]

The seminal work on diamond-shaped holes seems to have been done by one A.O. Hopkins, who was an active experimenter in the mechanical-TV age, and even became a member of the NBTVA shortly prior to his passing, in the '90s, if memory serves.

Although I am not a maths-whiz, his argument for this shape was quite convincing to me, enough so that I created a diamond-hole disc of about 12 inches in diameter (304.8mm to the rest of you!). At the time (1980s) circuit boards and print publications were still created with paste-ups and huge cameras. I was in a position to make the pattern of the disc at four times normal size, then have it shot-down to the proper diameter onto high contrast litho film. The negative was stuck-down to a vinyl phonograph record, which had been pre-drilled with oversize holes in approximate positions.

Even using every means at my disposal to position the holes properly, and with the advantage of the reduction to lessen layout slop, I was so disappointed with the precision of the scanning lines that I quit the project in disgust, never even trying to create a viewable image.

Since then I have become enamored of "bead disc" scanners, which are poor cousins to the lens disc. A local plastics emporium has quite perfect quarter-inch styrene beads available on the cheap; these can be glued into quarter-inch holes drilled or punched into a plastic or other disc. A single-LED light source placed well beyond the focal length of the plastic sphere shows up at a greatly reduced size, but every bit as bright as the (shrunken) original. What's more, with simple masking the shape of the scanning dot can be made anything you wish... square, diamond, etc. A bit of trial-and-error even corrects for the barrel distortion that the spherical lens imparts to anything but a round aperture. The main problem, of course, is drilling quarter-inch holes with sufficient precision, but a drilling jig should help here. 32 plastic beads cost only a few US dollars, and can be mailed in a JiffyBag® easily if anyone is interested in experimenting with this.

Furthermore, I do have somewhere an offprint of A.O. Hopkins' articles on diamond-scanning, published in a TV journal from the electronic period. I believe that I could locate this, scan it, and make it available for posting, if someone has a suitable site.

Many thanks to Andrew for setting up this forum! It's a great place to kick around ideas, and gives more instant gratification than the NBTVA Journal with its limited annual issues. Having said that, I do feel that it is very important to chronicle any new developments for publication in the Journal; many members depend on the printed edition.

[old_tv_nut]

You can see a photo fo the General Electric Octagon scanner at the Henry Ford Museum, showing that it had square holes, on my web site at
http://www.bretl.com/mech%20tv/mechanical%20tv.htm

[Stephen]

The ideal shape for the apertures of a a scanning element that is to provide equal lateral and vertical resolution is square. This is the only shape that may provide equal light distribution over the entire area of each pixel at any instant. A circle will provide a generally gaussian light distribution that is detrimental to resolution. In fact, the well known "Kell factor" for a cathode ray display is generally on the order of about .7, meaning that the effective resolution in the direction transverse to scanning is reduced by about 30 percent, due to the circular scanning spot and its resulting uneven gaussian light distribution. In contrast, the Kell factor of displays with discrete pixels, such as LCD, LDP and so forth all have square pixels and their Kell factor is 1.0, meaning that they have no loss in resolution due to aperture light distribution.

[Klaas Robers]

Oh Stephen,

but then you forget that Nipkow discs are not illuminated by flashing light, but by continuously burning changing light. In the mean time the hole, circular or square, moves along the line. If the light behind the disc is modulated with a symetrical square wave with a frequency corresponding to the highest frequency (10 kHz for the 32 line club standard), you will not get a square wave like dot pattern, but a triangular-wave like pattern. This is not the case on LCD screens.

The best form of apertures is, according to the late A.O. Hopkins, a cats eye shape. This is in fact the sin(x)/x form from -pi/2 to +pi/2 (-90 to +90 degrees). There should be some overlap with the adjacent lines. Then the line structure, which is very well visible in square hole discs, should be hardly visible. I have no possibility to punch my own discs, I got a square hole disc, so I have to do it with that.

[Stephen]

I would very much like to see Mr Hopkins analysis of scanning apertures. My first impression is that a 10 kHz square wave signal is going to have a 10 kHz fundamental sinusoidal component along with odd sinusoidal harmonics, mainly the third and fifth, so that the 10 kHz square wave signal fed into a data channel and modulated light source that is bandwidth limited to about 10 kHz is going to produce a generally sinusoidal (or triangular?) dot pattern along each scanning line, but it seems to me that this would be generally independent of aperture shape because it is due to bandwidth limiting.

Without such a bandwidth limitation, I can see that a slot-like (or cat's eye-like) aperture would have some advantage, but this would only be because the combination of greater data channel bandwidth and a narrower aperture in the direction of scanning would provide inherently better resolution, albeit non-symmetrical.

Having not seen Mr Hopkins work, my impression still is that for a scanning system with symmetrical resolution and a bandwidth limited data channel that a square aperture sized to the limit of resolution of the data channel will provide as good a result as the bandwidth permits. However, I look forward to seeing Mr Hopkins analysis and being proved wrong on this point.

[Stephen]

Of course, it would be ideal to reproduce perfect pixels right up to the limit of the bandwidth of a data channel. In the case of the 32 line system, it would mean being able to reproduce 48 square pixels with a 9600 Hz bandwidth. Of course, this is impossible with any aperture configuration or display, even LCD, since due to the limited bandwidth only a sinusoidal 9600 Hz fundamental will reach the display. Anything resembling square pixels would require a data channel of at least three times this, or 29 kHz, to be able to reproduce the third harmonic so as to approximate a 9600 Hz square wave.

There is a way around this. If the camera output is sampled in time slots of approximately 1/60000 seconds duration the scanning element may capture the necessary instantanous data to do this. The time slots would be spaced every 1/20000 second so that there would be two 1/60000 second duration samplings every 1/10000 second. These samples would then be integrated with a 10 kHz bandpass filter and sent to the display.

The display would sample the signal in a similar fashion, synchronised with the camera, displaying 1/60000 duration time samples every 1/20000 second. This is a way to reproduce a row of 48 almost perfectly square pixels for each of the 32 lines with no increase in bandwidth.

This time sampling technique was originally developed for radar during the second world war. RCA later used the technique for its compatible colour system, sampling each colour in rapid succession.

Electronic switching circuitry could easily perform this function. Interestingly, a chopper wheel of the sort that Mr. Baird used in his first television cameras to overcome the sluggish response of selenium sensors could conceivably be used in the camera and display, with these chopper wheels operating in synchronisation.

[Klaas Robers]

Yes Stephen,

this could be done if you have light enough. The basic problem of mechanical television is that the Nipkow disc gives an attenuation to the light of 99,9%. Only 0,1% is translucent. So each extra attenuation is too much. This counts for the display side. But may be we can flash the LEDs and over drive them with large currents. I have never done that.

For the camera there is nowadays no problem at all, as this is done by sampling a CCIR video signal. In some cases we have as much as 256 samples for an NBTV-line, so the resolution is plenty. Before making a CD the video is bandwidth limited to 10 kHz, but always to 20 kHz, the bandwidth limitation of the CD. There is little difference in picture quality between 10 kHz and 20 kHz when seen on a mechanical Nipkow monitor.

The aim of mr. A.O. Hopkins was to remove the very well visible line structure of the square holes. Even on a perfect disc, or may be especially on a perfect disc, you see the razor sharp transients from one line to the next one. By giving the lines some overlap in a way that a white field is almost equally lit, this sharp transients are eliminated. A cats eye horizontally oriented does this quite well. Hopkins showed photographs with no visible line structure and reasonable resolution. The pictures in the NBTV Newsletter in that time were very bad, but I will have a try to add them.

[moderator]

[Stephen]

Klaas, you have a good idea about flashing the LED(s) in an overdrive condition for time sampling of the received signal. In fact, the best way may be to convert the analogue signal samples to synchronised width modulated pulses wherein the maximum duration of each pulse corresponds to no more than 1/60000 sec.

With respect to available modulated light for the display, it is possible to direct most of the light from a modulated light source such as a high-power LED through each optical scanning element aperture by way of optical fibres. A central "optical commutator" that comprises inner ends of a group of optical fibres sequentially transmits modulated light through each fibre to the outer end that protrudes in its respective aperture.

I have posted this scheme on the forum. The PDF version is at http://www.taswegian.com/NBTV/images/OFSS.pdf .

[Stephen]

I take back my comment about square holes being ideal for a scanning disc. Upon re-reading John Logie Baird's British Patent GB329,664 I noticed that he says the following:

"It has hitherto been proposed to make the spirally-arranged holes in the disc either round or square in shape, but holes thus shaped tend to produce the appearance of lines across the image. To obviate this disadvantage, the present invention contemplates giving to the holes a sector-like shape, that is to say the shape of that part of a sector of a disc that is constituted by two concentric arcs of the disc joined by parts of two radii of the disc."

This makes a lot of sense. The scanning lines for a scanning disc are curvilinear, not linear, so each hole should have the shape of a section of a curved scanning line, not a straight one. I imagine that the best way to achieve this would be to place little masks with appropriate sector-shaped holes on oversize holes in the disc. The masks could comprise a set of transparent sectors on a black background made with a drawing application, such as Visio, printed out on a transparency sheet and cut up into individual masks to cover the oversize holes.