F YOU'VE arrived here from Test Cards, Tuning Signals, Idents and Clocks zip up your anorak and prepare to get your teeth into some real technical stuff. Each of the elements in a well-designed test card has a reason for being there and this page, which includes extracts from BBC and IBA information sheets and articles derived therefrom, attempts to explain exactly what their purpose is.

A rather puzzling aspect (!) of the changeover from 5:4 to 4:3 is that the extant photographs of the test cards and tuning signals used before 1950 show them to have been draughted in the 4:3 aspect ratio. Here are three possible ways in which the 4:3 Test Card C (above) could have been radiated in a 5:4 raster prior to 3 April 1950:	1: Stretched
	2: Letterboxed
	3: Cropped

HE ASPECT ratio of a picture is its width compared to its height. A television picture is made up of a number of parallel lines created when a spot of light moves from side to side and down the screen to form a 'raster'. The waveforms that create this movement can be varied independently in amplitude - that is, the height and width of the picture can be varied. As long as the relationship between height and width of the raster in the receiver is the same as in the camera, the picture will be the correct shape and circles will be displayed as circles.

The book 'Television Up-To-Date', by R.W.Hutchinson, published in 1935, second edition 1937, states that the aspect ratio of the EMI 405-line standard was set at 5:4, the same as half-plate photographs, while the Baird 240-line standard used the slightly wider 4:3 (the so-called Academy Ratio used by the film industry). According to Edward Pawley, the BBC's technical historian, in 'BBC Engineering 1922-1972' the 5:4 aspect ratio of the 405-line standard was changed on 3 April 1950 to 4:3.

It is not clear how this change was effected. The original specification for the 405-line standard had a blanking period per field of 'at least ten lines', but later it was standardised at 14 lines. If this change was made at the same time as the aspect ratio change it doesn't account for it entirely. Reducing the number of active lines per frame from 385 to 377 gives a change in aspect ratio of 1.25:1 (5:4) to 1.28:1 rather than 1.33:1 (4:3). The number of active lines on that basis would have to be 361, so it has to be assumed that the camera scan amplitudes were adjusted as well.

Theoretically all the domestic receivers would also have needed adjustment, but in practice the stability of the circuitry of the day meant that the height and width would vary by more than that in an evening's viewing. However, television display tubes (cathode ray tubes, or 'crts') continued to be made to the old 5:4 aspect ratio until the nineteen seventies, when the 'flatter squarer tubes' started to make an appearance. That meant that the 'sixties set had to be adjusted either to give the correct width and reduced height, or correct height and excessive width. In order to avoid black bars appearing at the top and bottom, the latter approach was used.

Test Card F has castellations with arrows to enable the dealer to adjust the scan amplitudes and thus ensure that the picture just fills the screen whilst circles remain the correct shape.

	On the left is the display of a correctly adjusted receiver having a 5:4 aspect ratio crt from the sixties...
	..and on the right a 4:3 tube from the seventies. The slightly wider screen allows the castellations down the sides to be displayed as well as those along the top and bottom.

I was surprised to find that none of the BBC test cards that I have unearthed dating from before 1950 show signs of having been designed for the 5:4 aspect ratio. Indeed two, dating from 1936 and 1949 complete with circles and castellations, are most definitely 4:3 in aspect ratio and they cannot be cropped or reshaped to 5:4 without producing an elliptical circle or missing castellations. I have also found adverts for tv sets and display tubes from the period 1936-48 that quote picture sizes as 10"x7.5", etc; in other words 4:3, despite the tubes themselves being circular or perhaps 5:4 in shape.

At the start of the twenty-first century there are currently two aspect ratios (and therefore shapes of display screen) in common use - 4:3 (which can also be expressed as 12:9 or 16:12) and the so-called widescreen 16:9. That is the ratio that defines the shape of the raster (that is the grid of lines scanned out by the electron beam), and on many modern sets the raster can be displayed in either shape, resulting in deep black bars along the top and bottom of a 'square' set or down each side of a 'widescreen' set when the 'wrong' shaped raster is displayed. Since black bars are now in fashion, the broadcasters use them to blank out parts of the 12:9 or 16:9 coded frame in order to present more (or in some cases less) of the original picture on both traditional and widescreen sets.

There is a shorthand way of describing the aspect ratio of a television picture in the form of a letter between two numbers. The first number defines the shape of the picture (4:3, 14:9 or 16:9) and is the numerator of a fraction whose denominator is 9. The letter indicates whether the picture fills the screen (F), is letterboxed with black bars top and bottom (L) or is pillarboxed with black bars left and right (P). The second number indicates the shape of the coded frame and is 12 for the traditional 4:3 (12:9) shape and 16 for the new widescreen 16:9 'anamorphic' shape. Shapes currently used in the UK are 12F12, 12P16, 14L12, 14P16, 16L12 and 16F16.

Overscanning is something that his been with us throughout the history of television, but it has started to confuse the issue with multiple aspect ratios. A professional-grade television monitor (and indeed most computer visual display units, now also called monitors) are designed to project a picture (called a raster) that is wholly displayed within the face of the display tube, whatever the shape of either may be, so that the whole of the broadcast picture is visible. Domestic receivers however, almost invariably project a raster that is larger than the display screen and a significant border of the transmitted picture may be cut off by the edge of the tube or a mask surrounding the front of the tube. The reasons for this are manifold: early cathode ray tubes (crts) were small and rounded (or indeed circular) and so to obtain the largest possible picture consistent with displaying most of what was broadcast a degree of overscan was employed. Also, circuits were not very stable to begin with, and height, width, centring and linearity were all liable to vary over a few hours. Overscanning helps to avoid the danger of black bars appearing at the edge of the picture - which would almost certainly involve a call-out to put right.

This was all very well when the transmitted picture was roughly the same shape as the screen, but today we have two shapes of screen and many more shapes of picture. The raster is still overscanned, but because the picture transmitted sometimes doesn't reach the edges of the raster - the broadcaster adds an area of black as padding - one pair of edges gets cropped while the other pair doesn't, which results in a displayed picture that has a different aspect ratio from the transmitted one (even if the geometry is correct - that is circles appear circular). This means that in the case of 12P16 pictures the digital widescreen viewer sees the full width of the original picture but perhaps not the full height, whereas with 16L12 pictures the analogue viewer sees the full height of the original picture but probably not the full width. Of course in the case of 14L12/14P16 the original 16:9/4:3 pictures are cropped for width/height by the broadcaster.

Digital widescreen viewers may notice that the pictures in 12P16 (eg sports from overseas) and 14P16 (eg news from overseas) appear wider than their 12F12 and 14L12 analogue counterparts and assume that the broadcasters are somehow cutting out something at each end of the lines in the latter cases, but this is not so. Setmakers have become so fond of overscanning that they have employed a trick to use it when a widescreen set is displaying a reduced-width raster, such as would be used for 12F12 or 14L12. They electrically chop a margin off the left and right of the picture (called 'blanking') so that less than the whole width is displayed. This gives a vertical straight edge to the picture with no danger of 'foldover' which might be visible if blanking were not used.

The overscan on a domestic receiver can be in the order of 5%-10%. For this reason broadcasters specify a 'graphics safe' area within which the text of captions should appear. The 'forbidden zone' was originally made rather generous to allow for the excessive overscan (and curved display tubes) found in early television sets but it has been relaxed in recent years. In the graphic on the left, the magenta inner curved line shows the Safe Title Area specified in 1966, and the outer cyan rectangle is the 1998 Graphics Safe Area for 4:3 pictures.

The following graphics represent how some currently transmitted formats in the UK might appear on widescreen and 4:3 sets of equivalent size (32" 16:9 and 25" 4:3 models for example). I have added a template to indicate the visible area with 5% and 10% overscanning. 12F12, 14L12 and 16L12 pictures displayed on a widescreen set will have the same cut-off as shown for the 4:3 set. The 14:9 pictures are instructive - with 10% overscan, no black bars are visible so that such 4:3 viewers rarely notice that most shows are now 14L12, and many 16:9 viewers think that 14P16 shows are true widescreen.

Since the cards I've used to demonstrate the aspect ratios below have frequency gratings that have been scaled, I've shown the actual frequencies in MHz that would be found in the pictures shown. Bear in mind that although the theoretical maximum frequency transmitted in the digital domain is 6.75MHz, in analogue System I the highest luminance frequency carried (for example by the PAL encoded composite baseband or uhf output of a set-top box) would be 5.5MHz. The figures therefore give some idea of the amount of original detail lost in the Aspect Ratio Conversion (ARCing).

[2.00] [3.33] [4.67] [5.33] [6.00] [7.00] 12P16 (AFD=0, 1 or 7) A 4:3 picture coded into a 16:9 frame (also as a 4:3 12F12 transmission should be correctly displayed on a widescreen set). 4:3 action contributions to BBC 16:9 sports shows (eg Grandstand) often appear like this, with 16F16 half-time expert analysis. If the show is ARCed to anything other than 12F12 for analogue broadcast, the result on a 4:3 set is a border all around as shown in "Floaters" below.	[2.00] [3.33] [4.00] [4.67] [5.33] [6.00] 16L12 (AFD=2) A 16:9 picture coded into a 4:3 frame (as widescreen feature films are sometimes transmitted on analogue).
[1.71] [2.88] [4.00] [4.57] [5.14] [6.00] 14P16 (AFD=3) A 4:3 picture zoomed to 14:9 and coded into a 16:9 frame (as a 4:3 source on CBBC, CBeebies, News 24, and inserts into entertainment shows, documentaries national and local news programmes etc, would be displayed on a widescreen set).	[1.75] [2.91] [3.50] [4.08] [4.66] [5.25] 14L12 (AFD=6) A 16:9 picture zoomed to 14:9 and coded into a 4:3 frame (as most 16:9 shows, excluding most sports and some feature films, are transmitted on analogue).
[2.00] [3.33] [4.00] [4.67] [5.33] [6.00] 16F16 (AFD=0, 2, 6 or 7) A full-height anamorphic 16:9 picture as displayed on a widescreen set.	[1.50] [2.50] [3.00] [3.50] [4.00] [4.50] 12F12 (AFD=7) A 16:9 picture zoomed to 4:3 and displayed on a 4:3 set (as most sports shows are transmitted on analogue).
[1.50] [2.50] [3.50] [4.00] [4.50] [5.25] "Smart" or "Just" mode A 4:3 picture displayed on a widescreen set with special non-linear horizontal scan designed to eliminate 'black bars'.	Most widescreen television receivers have a special display mode that attempts to squeeze a 4:3 picture into the full 16:9 screen. This is achieved by cropping the top and bottom off the picture and progressively increasing the slope of the horizontal timebase so that the edges af the picture are stretched out to reach the sides of the display tube, whilst keeping the centre of the picture roughly the correct shape. It is most effective in landscape shots - the widening of the edges is similar to the effect of using a very wide-angle lens - which compensates for the fact that the 4:3 picture is actually 'tighter' than a similar 16:9 shot. Across the centre of the picture I have superimposed a sliver from Test Card J zoomed linearly to the same height in order to show how the 'smart' function fattens the picture towards the edges.

Floaters This is what happens when producers of 16:9 shows reduce the size of a 16:9 or 4:3 picture in order that all viewers shall see the entire frame.

[2.33] [3.89] [4.67 [5.45] [6.22] [7.00] In this case a 16F16 picture has been reduced in size so that when the picture is ARCed to 14L12 for analogue transmission the full width of the card is seen. On a 4:3 set the picture looks identical to 16L12 above, but on a widescreen set displaying a 16F16 anamorphic picture there is a border all around. With some decoders it's possible to select a 14L12 output, in which case the picture can be zoomed by the set to fill the screen. Advertisements are occasionally broadcast like this - advertisers like to put their obligatory text as near to the edge of the screen as allowed by the ITC, but in some cases this would result in it being cropped off when ARCed to 14L12 unless the picture size is reduced. The effect may also be seen on "The World", a news programme broadcast simultaneously in 16F16 on BBC4 and 14L12 on BBC World. As is now standard on BBC News, perfectly good 4:3 pictures are butchered to produce 14:9 versions as a compromise that means that both 16:9 and 4:3 viewers lose the top and bottom of the picture. Although the studio part of this show is made in 16F16 with the sides sliced off for the the 14L12 version, when it comes to 16:9 contributions they shrink the picture so that the whole frame can be seen in the 14L12 version of the show, while 16:9 viewers see the postage stamp version depicted above.

[1.71] [2.88] [4.00] [4.57] [5.14] [6.00] Here a 4:3 picture has been incorporated into a 16:9 production which has then been ARCed to 14L12 (above) and 16L12 (below) for analogue transmission. On a 16:9 receiver the results are the same as 12P16 above, but on a 4:3 set the result is a border all around. There is nothing the analogue viewer can do, but a digital decoder could be switched to 12F12 to give a full-screen picture. Producers often put coloured 'curtains' each side of the 4:3 image when using this technique. [2.00] [3.33] [4.67] [5.33] [6.00] [7.00] This effect will also be seen on 4:3 transmissions from the BBC and five when using a terrestrial decoder that does not understand AFDs (see below) and is set to display letterbox pictures on a 4:3 set. That is because those broadcasters do not send full-width 4:3 pictures on terrestrial digital, but 'pillarbox' them in the middle of a 16:9 coded frame.

Wot? no AFDs? Active format descriptors (see below) enable a 4:3 telly to give its best shot at displaying a compromise widescreen picture of maximum size, with minimum black bars, and including as much of the intended picture as possible, automatically for each show. However, some terrestrial receivers, and all Sky digiboxes, ignore AFDs and give the 4:3 viewer a fixed choice of centre cut-out or letterbox only. Such viewers, who have 'upgraded' their 4:3 sets via this route are stymied when it comes to watching ITV1 regional and national news, for those in charge of digital transmissions in most regions have chosen to slice off the top and bottom of the 4:3 pictures and transmit them in 14P16, in order to con 16:9 viewers into thinking that they are widescreen. In order to avoid losing even more of the picture, or having a small 'floating' picture, the only option is to tune back to analogue, where the original 4:3 pictures are broadcast (at least there is this choice - the BBC never offers it).
ITV1 News letterbox on digital	ITV1 News on analogue	ITV1 News centre cut-out on digital
This sequence highlights another problem with widescreen. Subtitles, which are generated by the receiver at a fixed position within the raster, often do not appear where the subtitler intended when the underlying picture format is changed (in most cases they are designed to fit correctly in the 14:9 analogue version, as here).

W R O N G !
A 4:3 picture stretched to fill a widescreen receiver (as often set up in the home by the 'installer', or seen in shop showrooms). There are special non-linear screen modes available on most widescreen sets that attempt to squash a 4:3 picture into the 16:9 frame (see above). For those who dislike black bars, that is an arguably more elegant solution than putting up with uniformly fat actors.	A 16:9 picture as received on a digital set-top box or DVD player set to '16:9' or 'widescreen' and displayed on a 4:3 set without the capability of reducing the raster height to 16:9. This is also the shape in which digital 4:3 viewers see the reduced-size pictures on BBC Parliament and the BBCi newsloops, since they cannot be rescaled to 4:3 by the receiver.

How does the receiver know what shape the transmitted picture is? In the days of analogue transmissions and 4:3 receivers when the transmitted picture was always 4:3, the broadcaster simply blanked off part of the picture in order to produce an active field of the required shape. The receiver didn't know or care about this - it just continued to churn out a 4:3 raster. Viewers cared though - the black bars looked like a fault in their telly and so transmitted pictures that completely filled the screen were the order of the day for many years.

The first time anything needed to be done was when PAL Plus was introduced, giving pictures that were intended to be displayed on a tube with a 16:9 aspect ratio. Such a receiver needed to know when a PAL Plus transmission was being received so that it could switch in its decoding circuitry and present the picture as a full-height-anamorphic 16F16 raster. A system called WSS (Wide Screen Signalling) was devised, comprising a teletext-like signal occupying the first half of scanning line 23 (this is officially the last part of the field blanking period in 625/50 analogue transmissions). The signal comprises four cycles of clock run-in (833kHz) followed by a framing code and then fourteen bits of data which carry information about the picture.

The first group, comprising four bits, B0-B3, carries information about the picture shape, while the remaining groups are concerned with other characteristics of the PAL Plus transmission.

Here are some full-size 'square pixel' versions of WSS signals. If you download the files and paste them into the top left-hand corner of a 768x576 pixel picture they should produce the appropriate switching action when fed via a video board to a standard widescreen receiver that can decode WSS. The files are one pixel high, but I have asked your browser to fatten them for display here to make them more visible.

 WSS = 0 (12F12)

 WSS = 1 (14L12 Centre)

 WSS = 2 (14L12 Top)

 WSS = 3 (16L12 Centre)

 WSS = 4 (16L12 Top)

 WSS = 5 (Deeper than 16L12)

 WSS = 6 (16F16/14:9-safe)

 WSS = 7 (16F16)

With digital transmissions, part of the data stream is devoted to information relating to the shape of the picture. There are two methods of signalling, the first simply indicating whether the coded frame is 4:3 or 16:9 ("anamorphic") in shape. A digital receiver equipped with scart connections sets the voltage on pin 8 of the scart connected to the tv set to either +6V for a 16:9 picture or +12V for a 4:3 picture. (0V tells the tv to display its own off-air picture.)

This is fine as far as it goes, but in the case of 4:3 coded frames there are several available letterboxed formats, and many widescreen receivers are capable of displaying them full-height rather than with a small picture floating in a sea of black.

A second digital signalling system called AFD (Active Format Descriptor) is used to tell the receiver (or indeed, any other relevant equipment in the broadcasting chain) what the current picture shape is, and what the required display shape is to be. This is important because unlike an analogue receiver, the digital one is able to change the shape of the picture under the control of the broadcaster and/or the viewer to produce an appropriate video signal for the type of display device that is connected.

For example the BBC currently distribute all their digital pictures as 16F16 or 12P16 - that is, in 16:9 coded frames only. An AFD sent with the pictures instructs the ARC (Aspect Ratio Convertor) in the satellite chain to convert 12P16 pictures to 12F12 before transmission. On the terrestrial network the pictures are broadcast 12P16 and the home receiver is expected to interpret the AFDs and generate the 12F12 picture itself. Any receiver that failed to do this would diplay 4:3 pictures as a postage-stamp picture with broad black borders all around on a 4:3 receiver that is set up to display 16:9 pictures in deep letterbox.

However, a terrestrial receiver equipped with comprehensive AFD decoding facilities can not only produce correct 4:3 pictures, but can render 16F16 pictures in an appropriate shape as dictated by both the broadcaster (who knows the optimum display shape for the programme material) and the viewer (who knows what he likes). Thus a 16:9 show could be displayed automatically as 16L12, 14L12 or 12F12 as dictated by the broadcaster, or the viewer could opt to change to another of those formats.

It is interesting to note that while both BBC and ITV Sport generally choose to transmit 12F12 pictures on analogue, their 16F16 digital transmissions carry AFD7 (BBC, 4:3-safe) and AFD6 (ITV, 14:9-safe).

While the majority of digital receivers continue to communicate the required aspect ratio to the display device by means of the scart switching signal, a number also use WSS signals inserted on line 23 so that widescreen receivers can display a full-height picture. This is relevant when for example a film is broadcast as 16L12 or 14L12, as occasionally happens with old telecine transfers. It also means that recordings (including VHS) played through a suitable 16:9 display will assume the correct picture shape automatically.

How pictures with different AFDs are output from the decoder

		Intended output when decoder is set to...
	(CCO=Centre cut-out, ie top/bottom or L/R of transmitted picture discarded)	4:3		16:9
		..and transmitted coded frame is...
AFD number	AFD description	4:3	16:9	4:3	16:9
0	Same as coded frame	12F12	16L12 [1]	12F12	16F16
1	4:3 only	12F12	12F12 CCO	12F12	12F12 CCO
2	16:9 only	16L12	16L12 [1]	16L12 [2]	16F16
3	14:9 only	14L12	14L12 [1]	14L12 [2]	14P16
4	Reserved: decoders should behave as if AFD=0 were being transmitted.
5	4:3 (12F12) coded image framed to be "14:9-safe"	12F12	-	14L12 CCO [2] [3]	-
6	16:9 (16F16) coded image framed to be "14:9-safe"	-	14L12 CC0 [1]	-	16F16
7	16:9 (16F16) coded image framed to be "4:3-safe"	-	12F12 CC0 [1] [4]	-	16F16

Notes:

^[1] In these instances the decoder may often be set to output a different shape of picture, from full height to deep letterbox, under the control of the user.

^[2] Widescreen displays are often capable of 'zooming' 14L12 and 16L12 pictures to the full screen height, either under the control of the user or a WSS signal from the decoder.

^[3] In this instance the decoder might not add extra blanking the to 12F12 picture, leaving the widescreen display to mask the picture by zooming it to '14P16'.

^[4] The DTG in its document DIGITAL RECEIVER IMPLEMENTATION GUIDELINES and Recommended Receiver Reaction to Aspect Ratio Signalling in Digital Video Broadcasting (160KB file) recommends that decoders output a 16L12 picture when 16:9/AFD7 is transmitted, presumably because those viewers whose decoders have the option of only letterbox/full-height in 4:3 would have no other way of choosing to display a 16L12 picture.

Feature films shown on UK television are generally now cropped to 12F12 or 16F16/16L12 (using a technique known as 'pan and scan' in order to keep as much of the action as possible within the narrower screen), with a few films with original aspect ratios near to 4:3 or 16:9 being shown with thin black bars either at the top/bottom or sides. The problem with these odd aspect ratios is that it is sometimes impossible to avoid a black border all around the picture on one or other of the tv set formats, and the broadcasters rely on overscan in the receiver to mask these. This is also true of the 14:9 analogue television format, but because that is so prevalent in the UK, widescreen setmakers now include a 14:9 zoom mode so that analogue 14L12 pictures may be displayed at their full height. DVD film releases often strive to preserve the original cinema aspect ratio, and present the film with black bars in either a 4:3 or 16:9 coded frame. Here are some common examples:

1.66:1 (15:9)
15P16	15L12
15:9 is roughly the aspect ratio of a super-16mm negative. The actual image size is 11.65 x 7.00mm, so the precise aspect ratio is 1.66:1, or 14.98:9. Much television material has been shot on this stock for many years, but was generally framed and cropped for full-screen 4:3 showing. For widescreen television it is now framed 14:9-safe and cropped for full-screen 16:9 showing (14L12 on analogue). Occasionally the full frame is shown, generally letterboxed into a 4:3 frame for television, perhaps where the negative has been blown up to 35mm cinema prints. Television films are frequently shown in this 15L12 format in France.

1.85:1 (17:9)
1.85:1 is the cinema format nearest to television's 16:9 (1.78:1). In fact, given the tolerances in film production and telecine framing, they can be considered equivalent. The thin black bars top and bottom of the 16:9 television frame (11 scanning lines each - half the height of a teletext row), even when they are transmitted, are very likely to be hidden by the crt overscan.

2.35:1 (21:9)
21L16	21L12
2.35:1 is an ultra-widescreen cinema format, almost always cropped to 16:9 on UK television, but often shown deep-letterboxed in 4:3 frames in mainland Europe, even though half the tv lines then contain no picture. 2.35:1 DVD releases are generally in 16:9 coded frames, though it's almost impossible to tell from the sleeve labelling information whether or not this is so.

Because the 625-line and 525-line standards share approximately the same line repetition frequency (15.625kHz and 15.734264kHz respectively) a decision was made to use a common sampling frequency of 13.5MHz when digitising either standard. Using this sampling frequency, a whole number of samples per scanning line is produced in each standard - 858 samples in the 525-line 63.556µs line period, and 864 in the 625-line 64.000µs line period. Only the active portion of each line need be digitised and the EBU proposal of taking 720 samples per active line has been adopted so that the whole of the active line, even in the presence of timing errors, is sampled, these errors being corrected either in the digital domain or on conversion back to the analogue domain.

However, the number of sampling periods contained in the actual active line is smaller than 720 in each case. For the 52.90µs active line period of the 525-line standard this figure is 714.15 sampling periods, while for the 51.95µs period of the 625-line standard it is 701.325. In each case the figure is often rounded up or down to 704, a multiple of 16, which is desirable for MPEG encoding purposes. Indeed, 704 samples represents an active line period of 52.15µs, which is just within tolerance for both standards. However, this has led to some confusion over the actual active length of the line that contributes to the aspect ratio. The 625-line 4:3 (or 16:9) picture area comprises 702 x 576 samples, while the 525-line 4:3 (or 16:9) picture area comprises 715 x 480 samples.

esolution, or definition, is a measure of the amount of fine detail contained in a picture. On a photographic film it represents the finest grid of lines that can be distinguished on the final product before it merges with the background grain. In a digital medium such as your computer it's limited by the number of discrete picture elements (pixels) that go to make up the display. An analogue television picture is a combination of both. Comprising discrete scanning lines, in the vertical direction the picture cannot show a grid finer than those lines, but in the horizontal direction the resolution depends on the maximum vision frequency handled by the transmission system and this has a smoother roll-off.

Resolution is expressed as 'lines per something', the lines being alternately black and white. In photography it would be lines per unit length. In television a convenient measurement is 'lines per screen height' - that is the maximum number of vertical lines that can be distinguished in a square the same height as the picture. One would expect the horizontal resolution to match the vertical - in computer terms the pixels would be square - but that's not the case.

Because of the discrete scanning lines, and the fact that they are not displayed progressively, but interlaced - that is lines from one field fit in the gaps between the lines from the previous one - the vertical resolution is somewhat less than the number of displayed scanning lines would suggest. If a camera were to be pointed at a card printed with a grid of horizontal lines that exactly concided with the tv scanning lines the picture would flash alternately white and black at field rate, or if the lines on the card fell between the scanning lines the result would be an even grey. In between there would be a mixture of the two. Similarly if the camera zoomed in slightly the grid would not be reproduced correctly - there would be interference between the grid and the raster producing strong moiré patterns, called aliasing. The apparent vertical resolution is given by multiplying the number of active scanning lines by a 'Kell factor' and it's this modified figure that the horizontal resolution should be designed to match.

The Kell factor is entirely psychological and subjective and so cannot be determined by calculation or measurement, but a figure has emerged of 0.7 for still pictures, rising to 0.9 for pictures with fast movement and falling to 0.5 to take into account 2:1 interlace. You pays your money and takes your choice... Certainly a Kell factor of 0.7 was specifically allowed for in the NTSC 525/60 system at the time of its design in 1940, whereas the 405-line system was over-endowed with bandwidth by comparison.

In the analogue domain the horizontal resolution is a function of the scanning frequencies and the video bandwidth of the system. The faster the spot traverses the screen, the higher the bandwidth needed to maintain the same resolution. For example, in the 405-line standard the active line period is 80µs and the maximum modulation frequency is 3MHz. 240 cycles of that frequency can be displayed across the screen, and since each cycle comprises a positive-going (light) and negative-going (dark) half, that means 480 lines may be accommodated, equating to a resolution (with a 4:3 aspect ratio) of 360 lines per picture height, comparable to the 377 active display lines.

In a digital television picture the horizontal lines are sampled at discrete intervals instead of being continuous, and it's this sampling frequency that defines the maximum horizontal resolution. In both the European 625/50 and the US 525/60 systems the sampling frequency is 13.5MHz as that is a convenient value that coincides with requirements of both systems. The 525/60 system has 715 samples per active line while the 625/50 system has 702, though the number of samples to be coded per line is 720 for each system. That's equivalent to a maximum vision frequency (given by Nyquist) of 6.75MHz - half the sampling frequency.

Since the scanning line is divided into discrete samples, some sort of Kell factor should be applied when assessing the horizontal resolution of digital pictures, though in the horizontal direction there's no interlace to consider and on a crt at least, the line is produced as a continuous stream rather than discrete steps, so there are no possible gaps or overlaps. The value of an appropriate Kell factor could be anywhere between between 0.7 and 1.0 depending on whom you ask.

Below is a list of the horizontal resolutions, expressed as lines per picture height, and the Kell factors of various television systems past and present. A Kell factor of unity represents 'square pixels', while less than 1 means 'fat pixels' and greater than 1 means 'tall pixels'.

It would be instructional first of all to apply these calculations to the Baird 30-line standard, in which a raster of aspect ratio 3:7 was produced by thirty vertical scanning lines with a picture repetition frequency of 12.5 per second. This gives figures of f_H = 12.5Hz and f_V = 375Hz. As the raster is produced by a spiral of holes punched at the rim of a rotating disc, there is no line or field flyback period (nor are there any synchronising pulses!). For a 100% vertical resolution of 30 lines per picture width (70 lines over the full picture height) a vision bandwidth of 13.125kHz would be needed. With a Kell factor of 0.7 applied, that requirement falls to 9.1815kHz, within the nominal 10kHz audio bandwidth of the radio channels that were used to transmit Baird's pictures.

The system letters in the table below are those adopted by CCIR (Comité Consultatif International des Radio Communications - the International Radio Consultative Committee). Note that with digital systems, the number of active lines/samples is usually given, rather than the total lines per frame period as with analogue.

Notes:

^[1] For a time, the French Band I/III 819-line monochrome programmes of TF1 were duplicated on a few low-powered UHF transmitters, also on 819-lines, but fitted into System L 625-line channels with a horizontal resolution similar to that of the 405-Line Standard A.

It should be noted that when a PAL, NTSC or SECAM colour signal is carried by any of the above analogue systems, the horizontal resolution is reduced - by the visible presence of the colour subcarrier when viewed on a monochrome receiver and by filtering of the luminance channel on a colour receiver. The chrominance resolution also has to be considered separately - in PAL, SECAM and DVB the vertical resolution of the chrominance is half that of the luminance. In digital systems the horizontal resolution is also half that of the luminance, while in analogue systems it depends on the bandwidth allocated to the double-sideband modulated coded chrominance signal - in System I that's 1.3MHz per sideband, so the horizontal resolution of the recovered chrominance signal is a quarter of that of the (unfiltered) 5.5MHz-wide luminance signal.

Although the sampling rate at the analogue-to-digital and digital-to-analogue stages at the start and end of the DVB system is 13.5MHz, or 720 samples per active line (702 for 625/50), in the actual transmission path the number of samples can be any one of 352, 480, 544, 702 or 720. 'Short' lines are resampled by the receiver to give a full-width picture of reduced horizontal resolution. Many, if not most, UK "Sky" digital transmissions are of reduced sampling rate. The BBC and five (the rebranded Channel 5) use 702 samples, while ITV and Channel Four use 544, giving three-quarters of the full resolution, and many others use the half-resolution of 352 samples. (ITV and C4 have recently crammed more channels into their satellite and terrestrial multiplexes and appear to have reduced the sampling rate to 480 - or possible 352 - in order to do so.) However, both the BBC and five distribute all the their pictures in 16:9 coded frames. 4:3 shows are padded with 88 black-level samples before and after the 526-sample active 4:3 line which is expanded to fill the full 702 samples just before satellite transmission (the complete 'pillarboxed' picture is sent on digital terrestrial, with the receiver instructed to expand the central portion to full-width). So in fact the resolution on all five 'terrestrial' stations is about the same for 4:3 shows but worse on ITV and C4 for 16:9 shows.

Using test patterns to assess resolution

A detail from a 35mm projector test slide showing the square-wave resolution gratings

A detail from Test Card W showing the sinusoidal resolution gratings and anti-aliased transitions

You might think that by defocussing or filtering a pattern of sharp black and white stripes you would end up with a nicely rounded black and white sinusoid, but the reality is a little more complicated, as the original designers of test cards discovered. A square wave, whether it's synthesised electronically, or produced by scanning black and white stripes on a card, comprises a sinusoid of the fundamental frequency, which gives the pattern its pitch, combined with a series of odd multiples of the fundamental which gives the waveform its shape. If the amplitude of the fundamental is 1, the amplitude of the resultant square wave is pi/4, or about 0.79.

If vertical black and white stripes are scanned by a camera with a sufficiently wide frequency response, they have a squarewave p-p amplitude of 0.7V, but when the harmonics are filtered out to leave the fundamental on its own, as happens when its frequency is higher than one third of the system bandwidth, the p-p amplitude rises to 0.9V (-0.1V to +0.8V), causing overmodulation of the transmitter. That is not a helpful characteristic of a test signal, so the amplitude (contrast ratio) of all the squarewave frequency gratings has to be reduced accordingly in order to provide a realistic indication of frequency response.

In the diagrams, the blue waveform represents the idealised squarewave (with an infinite number of harmonics), while the brown, orange, green and violet waveforms are the fundamental with successive odd harmonics added, up to the seventh in this instance. The relative amplitudes of the harmonics are: first (fundamental) 1, third 1/3, fifth 1/5, seventh 1/7, and so on, and they are all in phase at 0° of the fundamental.

The top diagram is symmetrical about 0V, as with an audio signal for example, but the bottom ones show a video waveform with the squarewave sitting between black and white levels. In the middle diagram, with peak white/black gratings, the reduced-bandwidth waveforms are clearly overshooting above white and below black levels, which would cause distortion in equipment further down the chain. The solution is to reduce the amplitude of the squarewave by making the gratings dark/light grey. The fundamental in the bottom diagram now has an amplitude of 100% of peak white and there is no overshoot. Note that even when sinusoidal gratings are used, their amplitude is often less than 100%.

This chapter has concentrated on the luminance, or black-and-white, resolution of television standards. In colour television the chrominance (hue and saturation) resolution is generally much lower than that of the luminance. This, and its implications on picture quality, are discussed in the chapter on Colour in the World TV Standards section of this web site.

525-line optical monochrome SMPTE Test Card

625-line optical monochrome SMPTE/BBC Test Card

LTHOUGH IT was designed primarily for 525-line use, the SMPTE (Society of Motion Picture and Television Engineers) test card was adapted for use on 625-lines by the BBC for tests prior to the introduction of a specially designed 625-line test card. However, when one was produced in the form of Test Card E, there were aesthetic problems with the frequency gratings and so a 625-line version of Test Card C was pressed into service instead. However, Test Card D, the 405-line twin of Test Card E, was used on both BBC1 and ITV between 1964 and 1969.

Despite having been designed for the NTSC system with its relatively low resolution, the frequency gratings on the SMPTE card extend to 500 lines (per picture height), well beyond the horizontal resolution of all the 625-line standards, though the BBC replaced the wedges with finer ones, reading up to a rather optimistic 600 lines of resolution. (The vertical resolution of the 625-line standard is 575 lines, or 400 lines using a Kell Factor of 0.7.) The points on the wedge scales corresponding to the Test Card F frequency gratings would be: 1.5MHz - 115, 2.5MHz - 195, 3.5MHz - 275, 4.5MHz - 350 and 5.25MHz - 410. The NTSC System M with 4.2MHz bandwidth would resolve the gratings at 345. This description of the BBC version of the SMPTE test card comes from a BBC press release of 1962.

Click on an area in the test card to jump to an explanation of its function.

405-line optical monochrome Test Card C

HE FOLLOWING account is taken from the Collins Radio Diary, 1968, probably quoting from a BBC information sheet of two decades earlier, judging by the Cholmondly-Warner style of prose. [I have added amendments and notes in square brackets.]

Test Card C was introduced in 1949 and continued in use on 405-lines until 1964 when it was replaced on both BBC and ITA transmitters by Test Card D.

625-line international version of optical monochrome Test Card C (Redrawn by Dantus and Alan Pemberton)

625-line BBC version of optical monochrome Test Card C

These two variants were modifications of Test Card C for 625-line operation. The one marked G was used in various 625-line countries, but not the UK. The one marked BBC2 was used instead of Test Card E (which had been the intended BBC2 test card) on the BBC2 monochrome network before Test Card F took over in 1967. There are minor design differences from Test Card C - the circle on G for example and the arrowheads in the castellations of the BBC2 card, but the chief difference was in the frequency gratings. On Test Card G they were 1, 2, 3, 4 and 5MHz, while on the BBC2 card they were 1.5, 2.5, 3.75, 4.5 and 5.25MHz.

405-line optical monochrome Test Card D

625-line optical monochrome Test Card E

EST CARDS D and E were updated versions of Test Card C designed jointly by the BBC, ITA and BREMA, the British Radio Equipment Manufacturers Association, and were more suited to the 'swinging sixties'. Whilst most of the tests are the same, the technical specifications were tightened up and the design made more aesthetically pleasing. Obvious changes are the use of arrowheads in the castellations, corner stripes at 45° instead of on the diagonal and a large space at the bottom for station identification. The background grid of grey squares and white lines was rationalised so that with the castellated borders - each half the width of one square - there is a 12x9 grid, the same as the aspect ratio. This feature has prevailed through the subsequent test cards F, J and W, with the widescreen version having a 16x9 grid.

The two cards D and E are similar apart from the frequency gratings - Test card D was designed for 405-line use while Test Card E was for 625 lines. Both cards demonstrate a departure from the norm by having sinusoidal frequency gratings of reduced amplitude (ie not reaching peak white or black level). These look much softer on the screen and were not entirely liked by the trade. Test Card E had a further innovation - instead of a peak white border around the gratings it had one the same brightness as the peak of the sine waves, and each row of gratings started and ended at this brightness level. The BBC decided that the visual effect was unacceptable, and so Test Card E was rarely transmitted by them and the 625-line version of Test Card C was used instead until the start of colour transmissions in July 1967. However, the Portuguese broadcaster RTP seemed pleased with the pair and regularly radiated both D and E interchangeably, despite the fact that Test Card D was only designed for the 405-line standards and Portugal has only ever used 625 lines.

During its life Test Card D was modified slightly, in order to appease the trade who demanded that the frequency gratings should look 'brighter' on their receivers, and to identify the new version a white dot was placed in the centre of each square to the side of the one containing the letter D.

Click on an area of either test card to jump to a description of its purpose.

625-line optical colour Test Card F

625-line electronic colour Test Card F

EST CARD F was introduced to BBC2 at the start of the experimental colour service in July 1967, having been designed jointly by the BBC, ITA, EEA and BREMA. This account is taken from BBC Information Sheet 4306(1).

The original Test Card F comprised two transparencies - one colour and one black and white - sandwiched into a 35mm slide mount. In 1984 a digital version was produced using computer technology to draught the card and scan in the original central photograph. There are several differences between the two and a pixel-by-pixel comparison is available on Peter Vince's web site.

Click on a feature on either card to jump to a description of its purpose.

625-line BBC 'Test card G' version of electronic colour pattern PM5544

625-line international version of electronic colour pattern PM5544

HE PHILIPS PM5544 pattern generator was designed as a test for PAL receivers and was soon taken up by broadcasters as a useful alternative to optical test generators. There have been many variations over the years, and when the BBC began using it for insertion at regional transmitting centres on the rare occasions that the main transmission feed from London was being used for other purposes, they called it Test Card G, and it contained many features that were based on Test Card F. This account has been compiled and paraphrased from various magazine articles.

"Colourless" subcarrier

When viewed on a PAL-S (simple PAL, not having a delay line) receiver, the 'colourless' subcarrier decoder test part of the pattern is not colourless at all. Alternate pairs of lines (one from each field) contain colours which are electrically complementary, and therefore produce grey in a PAL-D (delay line) decoder, but are not cancelled out in the human eye/brain combination. The diagrams show enlarged squares of the PM5544 pattern containing 'colourless' (B-Y) and (R-Y) subcarrier. The smaller diagrams show the squares actual size. See Colour Standards: PAL for a description of PAL-D and PAL-S decoding. Back to picture

Variations

625-line animated digital version of electronic colour pattern PM5544

The intervening years have spawned many variations on the PM5544 pattern, and on the satellite channels it is rare to find any two identical. This version is designed for digital use and presumably carries no colourless subcarrier information since it would be irrelevant to a digital transmission. However, as a test of the integrity of the transmission circuit, an animated feature is included. If the path of a digital circuit fails at any point the usual result is a still picture of the last available frame. Since a test card is a still picture anyway such an event is unlikey to be noticed.

On this pattern the white needle pulse in the lower black rectangle swings from side to side - in this time exposure you can see it on two consecutive fields. In addition, as the pulse passes the two outer of the three white pips from left to right the audio tone in the left-hand channel is interrupted, thus providing an indication of sound-vision synchronisation and an identification of the stereo sound channels.

PM5544 on 525 lines

525-line version of electronic colour pattern PM5544 used by NHK in Japan (Redrawn by William Brown)

The PM5544 test pattern has also been used in 525-line countries. The picture above, and the two below, were redrawn by William Brown of New York who has provided the following technical details:

Where the 625-line pattern has a grid of grey squares having a height of 40 scanning lines, the 525-line version has grey squares 32 lines high (separated in each case by two white lines, making each row 42 or 34 lines high, respectively). The height of the centre circle is 504 lines in 625 lines and 408 lines in 525 lines. The grey background is 48% of peak white and the colour bars are 75%.

In the Japanese version the frequency gratings are 0.5, 1.0, 2.0, 3.0 and 4.0MHz. The square wave above the colour bars lies between 0% and 75% white, with a frequency of 300kHz. At the top left, where BBC Test Card G has 'colourless' subcarrier, this version sometimes has extended 180° burst.

In the United States, the multiburst frequencies are the FCC standard of 0.5, 1.0, 2.0, 3.0, 3.58 and 4.2MHz. In Canada, 0.5, 1.25, 2.0, 3.0, 3.58 and 4.2 MHz have been used. The square wave frequency in North America is usually 307.5kHz.

SInce the 'colourless' subcarrier test of unswitched (R-Y) and switched (B-Y) is meaningless in NTSC, the two extreme left and right hand columns of grey squares sometimes carry instead unswitched (R-Y) on the left and unswitched (B-Y) on the right as a subcarrier phase check. The Canadian version omits the 'ears' carrying the subcarrier phase checks altogether and also includes 180° colour burst superimposed on the greyscale, which has eleven steps instead of five.

525-line version of electronic colour pattern PM5544 used by WBOY-TV in Clarksburg, West Virginia (Redrawn by William Brown)

525-line version of electronic colour pattern PM5544 used by SRC [Radio-Canada] in Montreal (Redrawn by William Brown)

---

|Top | Aspect Ratio | Resolution | SMPTE | C | D & E | F | PM5544 | ETP1 | J & W | Colour Bars | Bookmarks |

---

ETP1

Click on any feature of the pattern to jump to an explanation of its purpose.

625-line electronic colour pattern ETP-1

The IBA ETP1 electronic colour pattern in development, photographed in 1978

The IBA introduced its new ETP1 electronic test pattern in 1979 in preparation for the subsuming of the fourteen existing control centres around the country into four regional operations centres (ROCs). It was necessary to employ a test card generator that once installed required no adjustment at all. The pattern met with some resistance from the trade however because of its lack of a circle, picture, arrowheads and square grid.

On the left is a rather unusual photograph of the ETP1 pattern in development. It appears to comprise only the black, white, magenta and cyan components of the pattern. More fundamentally, only the five sets of frequency gratings of the PM5544 are present - later a sixth was added and the Test Card F frequencies adopted.

The prototype test pattern was transmitted from the Rowridge, Isle of Wight transmitter and its dependants for evaluation purposes in 1978. The rack-mounted equipment that generated ETP1 was notoriously unreliable. Boards would come loose, or edge connectors become noisy, resulting in components missing from the transmitted pattern, similar to the prototype pictured left.

The following description together with the colour picture above is taken from IBA Engineering Information Service information sheet number EIS 120 9/79. [I have added explanatory notes in square brackets.]

Features of the IBA test pattern include: Crosshatch pattern - for convergence check. This may best be seen with the colour control turned down to give a monochrome picture. So far as possible, the white grid should be free from colour fringeing. In practice, most receivers tend to give some slight fringeing, particularly near the edges of the picture. Back to picture. EBU colour bars (75% amplitude, 100% saturation). [See the Colour Bars section for an explanation of these terms.] Back to picture. Grey scale - 0%, 20%, 40%, 60%, 80%, 100% amplitude. The difference in luminance signal voltage between adjacent rectangles should be approximately constant. Back to picture. Multiburst - for bandwidth/resolution check. Six sets of sine-wave gratings corresponding to the following frequencies (MHz): 625-line UHF 1.5 2.5 3.5 4.0 4.5 5.25 405-line VHF 1.0 1.6 2.25 2.6 2.9 3.4* *[This last frequency (3.4MHz) falls outside the System A 3MHz passband and is not transmitted on 405-lines - the block appears grey.] It is normal for a colour receiver to exhibit a bluish-yellow pattern (known as 'cross-colour') on the 4.5 MHz and 5.25 MHz gratings. Because of a special filter incorporated in colour receivers, which prevents the colour subcarrier from appearing on the screen, the 4.5 MHz bars are likely to be indistinct. Also, they are likely to be insistinct on most 405-line receivers. [This is because a similar filter is applied at the transmitter prior to standards-converting the signals to 405 lines, in order to eliminate any visible dot crawl pattern from the colour cubcarrier which would be converted to 2.8791855MHz - within the 3MHz System A passband.] Back to picture. 150 kHz squarewaves - for transient response check. Just above the colour bars there is a train of 150 kHz squarewaves (0% and 75% amplitude). This is to facilitate a check on any ringing, overshoot or preshoot. Ideally, there should be sharp transitions between the black and white rectangles, without 'smudging'. The transmitted transitions are as fast as the UK 625-line standard permits. Back to picture. Black rectangle within white rectangle - for low frequency response check. Low frequency response can be assessed by the appearance of the black rectangle within the white rectangle near the top of the pattern. Poor low frequency response shows as streaking at the right-hand edges of the rectangles, and from the border castellations. Back to picture. White needle pulse - for reflection check. Any reflections of the television signal, from hills or large buildings, can result in displaced 'ghost' images. The effects of short-term reflections are revealed by secondary images of the white needle pulse within the black rectangle. Back to picture. Yellow-red-yellow rectangles - for chrominance/luminance delay check. The redness of the rectangle near the top of the pattern should fit snugly between the yellow rectangles. Back to picture. Line synchronisation castellations - the left, right and bottom borders are formed by a pattern of alternate rectangles in black and colours with high luminance value and with a white rectangle in each corner. On monochrome receivers these rectangles appear either as black or as various lighter tones ranging from grey to white. The right-hand side border serves as a test signal for checking the line synchronisation of receivers - faulty line synchronisation shows as horizontal displacement of those parts of the picture on the same lines as the lighter toned rectangles on this side. These castellations, being yellow and white, provide a check on sync separator performance in the presence and absence (in 625-line transmissions) of the colour sub-carrier. The spacing of the left-hand and right-hand castellations has been staggered to identify the side from which any disturbance arises. Back to picture. Colour receiver reference oscillator castellations. The coloured border castellations can be used in checking for correct decoding; top: cyan, bottom:green, left-hand side: red and blue, right-hand side: yellow. Back to picture. Picture centring castellations. The width of each border castellation along the sides of the picture is the same as that of one of the grey rectangles within the crosshatch grid. Similarly, the height of the castellations along the top and bottom is equivalent to the height of the grey rectangles within the crosshatch grid. The picture size on receivers would normally be set for some slight overscan at the edges, but castellations should be clearly visible along all four sides of the picture. Back to picture. The average picture voltage has been set (nominally) at 50% of the white level voltage.

---

|Top | Aspect Ratio | Resolution | SMPTE | C | D & E | F | PM5544 | ETP1 | J & W | Colour Bars | Bookmarks |

---

Test Cards J and W

625-line animated digital 4:3 electronic colour Test Card J

625-line animated digital 16:9 electronic colour Test Card W

EST CARDS J and W are the current patterns used by the BBC, for internal if not external consumption. They each feature most of the tests carried by Test Card F plus some others required for digital distribution and transmission systems. Both cards were designed by Richard Russell, and full technical descriptions appear on Peter Vince's web site.

A high-definition version of Test Card W, called Test Card X, has also been designed. The frequencies of the gratings for all three patterns is given below. Since Test Card W could be displayed or distributed in 12F12 centre cut-out or 14L12 as well as its native 16F16 format, all three sets of frequencies are given.

Frequencies of gratings in MHz

Test Card F, Test Card J	Test Card W			Test Card X
12F12	12F12 CCO	14L12	16F16	16F16
1.50	1.50	1.75	2.00	5.00
2.50	2.50	2.90	3.33	10.00
3.50	3.00	3.50	4.00	15.00
4.00	3.50	4.08	4.67	20.00
4.50	4.00	4.66	5.33	25.00
5.25	4.50	5.25	6.00	30.00

---

|Top | Aspect Ratio | Resolution | SMPTE | C | D & E | F | PM5544 | ETP1 | J & W | Colour Bars | Bookmarks |

---

Colour Bars and PLUGE

COLOUR television picture comprises three separate pictures of red, green and blue superimposed. If the red, green and blue signals are switched fully on and off in a certain pattern across each line (in fact square waves with the approximate frequencies f_H, 2f_H