by chris_vk3aml » Mon Jun 02, 2008 9:50 pm
Just a few figures and a warning about some (perhaps) unexpected problems:
You propose 72 lines = 5 angular degrees per Nipkow disc sector.
With 3:4 aspect ratio = 72 x 96 pixels = 6912 pixels x 6.667 = 46082.3 pixels per second. Top frequency required = 46082 / 2 = 23041 Hz.
This will probably be OK with CD recording, even better at 48 KHz CD-ROM sampling BUT problems are likely to occur with low frequency phase shifts as the frame repetition rate is so incredibly low. To get a system PHASE FLAT to 6 Hz requires a frequency response extending well BELOW even that frequency. We are already stretching this to the limit of most audio A/D cards at the 12.5 Hz rate.
There is also NO WAY that you can legally put a 23 KHz wide uncompressed video signal on hf ham radio below 29 MHz. Here in telecommunication region 3 (Australia) the legal bandwidth limit for the whole signal is 8 KHz on those hf bands. In your part of the world I believe the bandwidth limits are even more stringent. Even on 10 metres (above 29 MHz) you are limited to 16 KHz, which means 8 KHz of legal modulation on a.m. or about 15 KHz on vestigial sideband. NBFM will be a washout, owing to its higher order sidebands demanding even greater bandwidth, and due to most commercial NBFM transceivers having pre-emph/de-emph curves causing wild video phase shifts.
Are you aware of the difficulty of watching a 6.7 Hz frame repetition picture with no storage or interlace? Interlace, perhaps more than 2:1 interlace might be worthwhile, but the electronics to generate this may be a bit fearsome. The Sanabria concept of 3:1 interlace at 45 lines was a good one.
There are also good reasons for NOT using horizontal scanning OR the 3:4 ratio on a picture of less than 100 lines where head-and-shoulders are the most suitable subject - as Baird empirically determined all those years ago. Facial images simply suit vertical scanning and narrow vertical aspect ratios better, by disposing the limited number of pixels in a more economical way around the face. Indeed, another good reason for using a tall narrow picture was to decrease the number of lines while retaining a fair number of image pixels. Why should we want to do this? Well, the sector angle for a 30 line scanning disc is 12 degrees - and that gives a fair leeway for angular errors in laying out the scanning disc. The usage of a narrow picture also avoids the bogey of low frequency phase shift, as a tight close-up of a face covering the full width of the vertically scanned picture will have minimal transverse shading - shading that would be resolved into low frequency video components down near the frame rate. It is actually my opinion that our 2:3 ratio is not narrow enough, and that Baird hit the desirable target somewhat better at 3:7. However in setting a standard I bowed to Doug Pitt's and Stan Kujawinski's wishes in the early 1970s.
Let's say you increase the number of lines to 72, and decrease the sector angle from Baird's 12 degrees (with 30 line) to your 5 degrees (with 72). The accuracy you're going to need in positioning the holes in that 72-hole disc to maintain 1/4 pixel positioning accuracy will be, as we discussed in another posting, incredibly stringent. The angles decrease by a factor of 11.25/5 = 2.25, the picture area goes down by a factor of about 6, the hole positioning accuracy required goes up by a factor of something like 12.
To cut the disc to anything like this degree of accuracy you will definitely need:
(1) An accurate and unworn turntable dividing head from a metal-worker's lathe, suitable for high-accuracy gear-cutting.
(2) A rigid and accurately centered boss to hold the disc axially in the dividing head while you punch the holes.
(3) A suitable and rigidly mounted jig for the hole punching to one side of the dividing head, with micrometer adjustment for the radial measure.
Unconvinced? Consider the scanning disc:
With a 32 line disc, one pixel = 1/48 of 11.25 degree = 0.23 degree = 14 arc minutes.
One quarter pixel positioning accuracy (to avoid image streaks and jagged edges on reproduced lines) = approx 3.5 minutes of arc.
With your 72 line disc, one pixel = 1/96 of 5 degrees = 0.05 degrees = 0.3 minutes of arc.
One quarter pixel positioning accuracy (to avoid image streaks and jagged edges on reproduced lines) = 0.75 minutes of arc.
Angular layout of the 72-line scanning disc MUST be 3.5/0.3 greater accuracy = about TWELVE (12) TIMES MORE ACCURATE!
I wish you luck but you'll need VERY accurate tools - and even if you find those tools and manage to make a disc, who else will be able to make one? There are very good reasons for a group keeping to 32 lines as a general standard where mechanical scanners are concerned.
When one gets up to those line numbers, something like the Mihaly-Traub stationary mirror drum - adjustable while the scanner is in operation - becomes a much more attractive constructional proposition - or maybe even keeping the system wholly electronic by using CRT's.
I would LOVE to see such a scanner, but I just felt that I should emphasize the problems and perhaps indicate a few potential solutions. I'd be seeing a firm of engineers with an ACCURATE metalworker's lathe as a first priority. Then I'd be building a custom vestigial sideband transmitter for use on the broadband section of the two metre amateur band before I proceeded.
Or - look at the alternative - use digital systems, a webcam and software compress to .mpeg in real time, and transmit those compressed images in by DRM system on hf, which I suspect would be more practical - and may even be EASIER to achieve.
I honestly wish you luck...
And if you find this type of engineering facility PLEASE let me know, please. I wouldn't mind getting some discs of that accuracy myself.
Best wishes,
Chris Long VK3AML.