NBTV interlace

[AncientBrit]

In the NBTVA Newsletter there has often been mention of the use of interlace to improve the resolution of the 32 line image.

Over the last few years I have carried out several experiments using interlace and Andrew Davie has kindly agreed to allow me (to attempt!) to upload a Visual Basic6 compiled program that allows various interlaced NBTV clips to be displayed on the PC.

(BTW Andrew the 1MB guess was a little on the low side)

Interlace is only really practical using some form of image storage in the replay chain.
Without this the 6.25Hz flicker is really intolerable.

The principle is fairly simple. Both scanning and replay use lines half the normal thickness.
On even frames the scanned and display image are stepped sideways by this thickness to infill with the second frame. The net effect (on still images) is to improve the horizontal resolution by a factor of 2, giving the appearance of a 64 line image.

"There is no gain without pain".
This cosy theory breaks down when the image moves vertically. Edges from odd and even frames no longer align horizontally and the corrugations from fast movement are quite noticeable.

The application contains some descriptive info on NBTV file formats that are used.
Each NBTV frame occupies 2048bytes, consisting of 64 pixels in 32 lines.
Note this also applies to the interlaced format as well.
No additional storage is used in interlaced format.

Don't attempt to use interlace on non-interlaced pixs, the picture will look very pixilated.

The files are zipped and consist of several folders.
The application and the two folders of NBTV file clip material should be loaded into one new folder on your PC.

And this is where it gets problematic .......
If all goes well you should be able to launch the NBTVplayer.exe with no trouble if your PC has the necessary Visual Basic6 runtime files.

However I have also included runtime files for Win98 (on which this program was developed) and Win2000P on which it has also run successfully.

If you have one of these platforms then copy the runtime file(s) for the appropriate platform to the same folder.

If not then Google to obtain the necessary file(s).
The program will normally prompt for any missing files and you obviously know what platform you are running (don't you!!)
You may gather that a certain lack of confidence is creeping in here.......

You can see why I now opt to do all current development on Microcontrollers (PIC18F) and burn code into embedded chips with no hassle over PC Operating Systems.

I electronically convert my NBTV pics to 640 x 480 x 60Hz VGA format and display directly onto a VDU with no PC involved at all.
I did consider 625/50 but the VGA format is more universal throughout the world.

(Recording long clips without using a PC remains to be resolved)

Right. Let's see if I can attach some ZIP files to this text.......and good luck


Kind regards,


AncientBrit

[Andrew Davie]

In the NBTVA Newsletter there has often been mention of the use of interlace to improve the resolution of the 32 line image.

Over the last few years I have carried out several experiments using interlace and Andrew Davie has kindly agreed to allow me (to attempt!) to upload a Visual Basic6 compiled program that allows various interlaced NBTV clips to be displayed on the PC.

Very nice.

For Windows XP users, this appears to work 'out of the box' if you download the ZIP_NBTVapplication.zip and the materials files, unzip and then run the .exe. In other words, the runtimes appear to already be installed under XP.

I've had a bit of a play, and surely the interlaced checkbox makes a huge difference in apparent resolution. I would love to see how this looks on actual hardware, but my first impressions are very favourable indeed. I would imagine that flicker would be a real killer, though...?

As to use of this forum, and permissions to 'do stuff' -- this forum is for the NBTV community to use as it wishes. Asking me is not necessary. Don't worry too much about space; I monitor things every day, and will keep things running smoothly.

[AncientBrit]

Hi Andrew,

Glad it worked. I'm pretty amazed it did!

The flicker on a CRT makes the 6.25Hz image unviewable.

The eye perceives the image as more of a hopping patch rather than a discrete image

Regards,

GL

[Steve Anderson]

Gents,

I too did some experimentation with interlacing the NBTV signal. I drew the same conclusion as yourselves that it just doesn't work at our low scan rates. I tried standard scanning, (Fig.1), 2:1 (Fig.2) and 4:1 (Fig.3) interlacing as derived from the circuit diagram and waveforms attached.

The waveforms shown were used to drive the frame (horizontal) scan on a CRT display, the line (vertical) was simply a 400Hz ramp yielding a simple all white raster on the screen.

As you have observed it is not a workable arrangement without some form of storage (usually digital), in which case interlacing is not required anyway.

However, it seemed a good idea at the time!

Steve A.

[Stephen]

Unfortunately the "magic number" for imparting relatively smooth motion of moving objects is somewhere between 9 and 12 fps, depending on the individual. Much animation, such as one may find on the web, relies on a frame rate of 12 fps.

Flicker is an entirely different matter. Apparent flicker depends on display brightness and the relative size of the displayed images. A bright display requires a much higher frame rate than a dim one. Likewise "large area" flicker is much more apparent than "small area" flicker, and therefore large area objects, or coarse detail, in a display would require a higher frame rate to reduce apparent flicker than small area objects, or fine detail. Typically, relatively flicker-free displays at low or medium brightness require frame rates of 24 to 48 fps.

Line crawl is a serious problem with interlace systems, particularly those with a low line count, as is the case with NBTV. John Logie Baird proposed a progressive scan system in his British Patent 391,924, wherein he describes a system that transmits coarse detail at a frame rate to eliminate flicker whilst transmitting fine detail at a lower frame rate that is still high enough to impart smooth motion.

The latest newletter shows a first part of an article that I have written about this approach. You may read the entire article at http://www.taswegian.com/NBTV/images/SDS.pdf . I propose optical means added on to ordinary equipment for implementing this. One example is a 48 × 48 pixel system that has an effective frame rate of 50 fps and a bandwidth of 19,800 Hz. Colour is also possible with no increased bandwidth.

[AncientBrit]

Steve,

I did find that interlace yielded improved horizontal definition.

Text that was not legible on 32 line camera/display became readable when re-scanned with interlace for camera and display.
Although I have not included "before" and "after" clips.

I gather that you may not be able to run the clips as you are Unix based, but the horizontal resolution is apparent in some of the clips I uploaded.

BTW I found that I had to electrically defocus the vidicon tube to get a larger scanning spot. Lag was a problem on the tube I used, see the "comet tail" on the Teddy clip

Signal current rose as the scanning spot covered more of the target area and "waterfalling" was eliminated, apart from edge of scanned area.

Strictly correction should be made for a "stepped scan" image when displayed on a sawtooth driven skewed raster, about 3%.

Not really a problem with 405/525/625 where the % is much less.


Regards,

GL

[Stephen]

JLB shows some very interesting interlacing combinations in his early British Patent 289,307 in the Patent and Articles section of the forum. As well as describing ordinary interlacing (page 2, lines 6-15 wherein he refers to interlacing as intercalated sets of parallel bands), he illustrates some very ingenious variants. See Figures 1, 2 and 3 in particular.

Interlacing with the line count and frame rate of NBTV results in poor image resolution with any sort of motion. However, an early patent that I cannot identify at the moment described subject matter that is a cross between progressive and interlace scan.

In this proposal, the first frame comprises an ordinary progressive scan of the image, such as a progressive 32 line scan. The second frame also comprises a 32 line scan of the image, but with the scanning lines displaced by half a line so that the centreline of each line in the second frame overlaps a corresponding pair of lines in the first frame to add detail. That is, the second frame scans from a slightly different perspective from the first frame to fill in information between lines in the first frame. This would tend to smooth out the pixelation that ordinary progressive scanning would produce without the artifacts of true interlace scanning, increasing perceived resolution by perhaps one half: about what one would expect with interlacing in practice.

This scheme would require a 64 line optical scanning element for both the camera and display. The first 32 lines would represent the first frame and the second 32 lines would represent the half line shifted second frame. There would still be 12.5 complete frames per second, so each scanning element would rotate at half speed compared to a 32 line scanning element.

I am still searching for the original patent.

[Stephen]

I found it! It is an American patent to Pierre Mertz who worked for AT&T, US1806638, filed 24 July 1928. I have uploaded a copy of this patent into the Patent and Articles section of the forum.

Notice the two sets of spirals for the scanning disc in Figure 2. These would produce two successive progressive scans, with apertures 11-15 overlapping 6 through 10 in the second progressive scan.

[Stephen]

I just realised today that Mr Baird's British Patent 321,389, filed 5 June 1928, describes overlapping progressive scans for better resolution, and thereby it anticipates Mr Mertz's patent by almost two months. Mr Baird's '389 patent describes his field sequential colour system and he recommends slightly overlapping each of the three progressive scan colour fields by one-third to achieve better effective resolution. He also mentions that the method is usable without colours on page 1, column 2, lines 76-77.

I have uploaded Mr Baird's '389 patent to the Patent and Articles section of the forum for reference. It seems to me that such overlapping sequential progressive colour fields are a natural choice for field sequential colour.

[Stephen]

As a final note on the method of scanning slightly overlapping scanning lines in sequential progressive scan frames as independently proposed by John Logie Baird and Pierre Mertz, the result would be to reduce or eliminate the so-called Kell factor. The Kell factor represents a reduction in effective resolution due to pixel-size detail straddling or overlapping adjacent scanning lines. It generally has a value of about 0.6 to 0.7. That is, with a 32 line scanning system the effective resolution would only be 19 to 22 lines due to the Kell factor.

Overlapping scanning lines in sequential frames by one-half line, as most clearly shown in the Mertz patent, eliminates the possibility of pixel-size detail straddling or overlapping adjacent scanning lines. That is, any detail so straddling scanning lines in one frame would line up perfectly with the scanning lines in the subsequent frame. This may improve the Kell factor for such a scanning system to a value close to 1.0. In other words, it might improve the resolution of a 32 line system to the equivalent of 32/0.7=45 lines or even 32/0.6=53 lines with no increase in bandwidth. This system may be worth trying, although of course it requires discs or drums with two sets of 32 apertures, lenses or mirrors.