Standards Conversion


Section Headings


The Basic Approach

The most fundamental problem when attempting to convert from PAL/625 to NTSC/525 and vice versa is the difference in frame rate. With NTSC/525 fitting 30 frames into a second, and PAL/625 only having 25 frames in the same time period, the frames will only start in the same place in the signal on (at best) five occasions within each second.

The only way to effectively perform such a conversion is to be able to freeze a frame and hold it until it's the right time for it to start in the other TV system. The only viable way to hold such a large lump of signal electronically is to digitise it and store into a large block of fast computer memory. This is exactly what a digital standards converter does, with the picture quality of the converter governed by the amount of memory used and the resolution of the digitisation process.


Coping With Movement

The biggest single problem to contend with when doing standards conversion is that of making the motion in the finished image look as smooth as possible. This is an extremely difficult thing to do and there is still no really effective solution to this problem, leading to the purist view that material should always be watched in the TV format it was originally made in. Usually this is only possible for people with multi-standard high end equipment such as multi-standard LaserDisc players and who are prepared to import large amounts of material from other contries.

The diagram above illustrates some of the solutions to the problem. The simplest solution is shown in Method A, and is the method used by the majority of domestic TV standards converters such as the Panasonic W1, Awia HX-M1, etc. The next solution (Method B), namely to mix two adjacent images together to create the in-between frame where needed, is used by many industrial converters and some of the high end domestic standards converters such as the Sharp AN-300SC. The final solution (Method C) with motion tracking involves storing a sequence of frames for processing and extremely fast micro-processors to try and analyze the motion. This is hugely expensive by comparison with the other techniques with standards convertors featuring this technology costing a mimimum of around US$25,000.

Even the most advanced standards converters have their problems. While they can usually find any pronounced "edges" in the moving image and predict where it would be half way between the two images; the more indistinct the motion the more likely it will fail to predict the position correctly. With extreme motion such as sporting events, these converters can even sometimes fail to track the object at all. For example, a gymnast's legs can disappear during motion because there is no way of tracking the vectors of the motion between images taken 1/25th of a second appart.

Scanline artifacts

As ever, the problems don't end there. As well as the problem of dealing with motion, there is also the quest for fine detail in the image. This in itself presents a problem since the differing line structure of NTSC/525 and PAL/625 means that each line of the image is out of proportion in reference to it's neighbour in the other system.

The simpler standards converters merely drop or repeat every fifth line or so in the output picture, resulting in some slightly strange effects particularly on text, but retaining much of the definition. The median approach is to mix the appropriate lines together, often on a proportional basis, as illustrated in the diagram. The extreme top end approach is to try and detect edges of objects within each scan line by areas of rapid transistion and to predict where the exact position would be based upon the position of the edge in the scan line above and the scan line below.

This area in particular is made much worse by the two field to a frame interlace used by all the world's TV systems. The idea is that to keep the flicker of the screen down, each TV frame is drawn onto the screen in two halfs. The first trip down all the odd-numbered lines of the TV picture are drawn, the second time down the even-numbered lines are drawn. This means that a low end converter with only a field store will only be able to do comparisions between adjacent lines from the field, rather than adjacent lines from the full frame. This means that the lines being compared will in fact have a missing (even if the pair are odd) line that should have gone between them.


Choose: Digital Television Building Bridges... Main Index


Bevis King <B.King@ee.surrey.ac.uk>
14th June 1996