Alternative test patterns for Richard Russell’s Test Card Generator

Richard Russell has produced a television test pattern generator whose default image can easily be changed - see his web page at http://www.rtrussell.co.uk/products/tccgen/tccgen.html.  A Windows’ program is supplied with the generator that will encode a suitable image file and upload it to the generator, creating signals in all of the common TV standards.  My offerings as alternatives to the supplied test patterns are linked below.  They can be used as is, or you can add your own caption before letting the encoding program do its thing.

Richard’s encoding program is very versatile in its requirements for a source image file, accepting BMP, JPEG, and GIF file formats.  These are all assumed to have “square pixels” and to have RGB ranges of 0 - 255 for black to white, as typically produced by computer image processing programs.  The generator uses a 12 Mc/s sampling clock, so the native width of a 625 picture is 624 pixels ( = 52µs * 12), but any (reasonably) sized image can be used, and the program will resize it as appropriate.  This width of 624 pixels results in a height of only 468 lines (= 624 * 3/4) for a standard 4:3 aspect ratio picture, not the 576 actually needed for a 625-line signal (or 486 for a 525 signal), so the program will expand the height appropriately.  Note that this vertical “oversampling” will result in ringing on the edges if they are not correctly anti-aliased.  Note also that the image is expected to be in the 4:3 aspect ratio: if a widescreen pattern is required, the user must produce an anamorphic image by squashing the width to 75% of its square-pixel size.

Should the user want more control over the final image, a 720 x 576 REC601 YUV image file can also be encoded.  In this case, the encoding program sub-samples the 720 pixels to the required 624, rather than over-sampling the number of lines, and so it is again up to the user to ensure the supplied image is properly vertically anti-aliased.

A little while ago, Richard released an updated version (3.2a) of the programming software which added the PAL-M, PAL-N, and NTSC 4.43 colour standards, and also provided a means to upload animated images - see below.  As a very useful bonus, there is also a new program included - WaveView - which gives an oscilloscope type display of the lines of a prepared ROM image, as Y, R-Y, B-Y, or composite!  The current version of the program (3.2c) was released on the 8th of July, includes a couple of minor bug fixes, and allows the maximum length of image path/filename increased to be up to 255 characters

The Test Pattern Collection

The list above shows all the test patterns available here, and they link to brief descriptions of the pattern.  For the purposes of testing and demonstration, most of the images are available in all the acceptable formats.  However, as BMP and YUV images are inherently uncompressed, they would both be over 800 Kbytes per image.  These file types have therefore been ZIP’d to make download times more reasonable (not to mention reducing the loading on my server!).  With the exception of the plain test signals, the GIF format should be avoided if possible, and JPEGs should always be avoided, due to the impairments introduced by their respective compression schemes.

Barney Wol’s 4:3 test card

As it is not allowed to use the famous Test Card F (or its newer replacement, J) for copyright reasons, Richard had to design his own test card that would be programmed by default.  I have created my own, as I felt it would be useful to have circular patterns in the corners to aid checking scan linearity - particularly as the newer “reduced height” televisions often have a nasty display mode where only the sides of a 4:3 picture are stretched to avoid the “ugly black borders”!  The plain grey background has a 16 by 12 white grating on it, so the squares should thus appear truly square, and the same size all over the screen.  The centre circle should of course be perfectly circular, essentially, the same height as its width.  The line of coloured blocks are full saturation (100%), so each of colours is in turn fully on or fully off.  The greyscale (or “staircase”) just below centre should have no colour, being a plain, neutral grey.  Incidentally, the steps are a linear progression (100mV each), and are not the same level as the luminance component of the colour-bars above.  The frequency gratings are the same as on test Card F, namely 1.5Mc/s (only one square of this), then two squares each of 2.5Mc/s, 3.5Mc/s, 4.0Mc/s, and 4.5Mc/s; and finally one square of 5.25Mc/s.

At the top is a “streak-box” designed to show up ghosting etc.  On the BBC test cards, this has traditionally been black, but I believe a dark grey would be better.  Low level ghosts would be unlikely to be seen on a black streak-box, as even if the display brightness wasn’t turned down too far, the logarithmic response curve of the display would render the ghost invisible.  To demonstrate my point, see the image below.  A ghost image has been artificially added 5µs late, and 30dB down - it is readily apparent on the dark grey streak box, but not on the black one - yet I promise you it is there just the same!

The image is available in each of the supported formats, and can be downloaded from the links below.  If you want to add your own caption, I would recommend the ZIP’d BMP file: despite having been saved with minimum compression, the JPEG has some noticeable differences from the original.  The GIF, despite having only 256 colours, is actually better!  If you don’t wish to edit the image at all, then use the ZIP’d YUV file.

Barney Wol’s 16:9 test card

With thanks to Dave Young for spotting that the corner circular patterns were actually oval, all the 16:9 test card patterns were corrected on the 7th of August 2005.

For those with a new, reduced height television, here is a 16:9 test card.  The frequency gratings and most of the features are the same as on the 4:3 test card above.  If viewed on a 16:9 television, the smaller central circle on this 16:9 test card should appear the same size as that on the 4:3 test card, if the latter is displayed full height with black borders.  However, if this is viewed as a 16:9 letterbox on a 4:3 television, the large circle should appear the same size as that on the 4:3 test card.  If this is displayed full-height on a 4:3 television, and the width expanded to just show the central 4:3 section of the image, the two outermost squares at the left and right will be lost, as indicated by the inner arrow-heads with the 4:3 notation.  Broadcasters usually convert their widescreen digital networks to a 14:9 letterbox before broadcasting on analogue transmitters.  This results in narrower black borders top and bottom, and only one square is lost at each side as shown by the outer arrow-heads. (see below)

14:9 letterboxed 16:9 test card

As mentioned above, broadcasters usually convert their widescreen digital networks to a 14:9 letterbox before broadcasting on analogue transmitters.  This image indicates the resulting shape and size.  The active picture height is reduced to 493.7 lines (576 x 12/14), and 6.25% of the width has been trimmed off at each side.  (How can you have 0.7 of a line?  Very easily, by altering the position of the anti-aliased edge.  That seems like a cue for an article on sampling, but it’ll have to wait for now!

16:9 letterboxed 16:9 test card

Occasionally the broadcasters will convert a widescreen programme to a 16:9 letterbox for transmission on the analogue network.  It is also the other option (besides 4:3 centre cut-out) provided on digital set-top boxes.  This image indicates the even smaller shape and size that results.  The active picture height is reduced to 432 lines (576 x 12/16), but none of the sides of the image are lost.

4:3 Safe Title Area outline

This is the standard Safe Title Area outline for 4:3 pictures.  The said safe area is a very old, conservative limit, the borders each being 10%, and the corners rounded with a radius of 21% of picture height.  This standard is still adhered to, even though modern televisions show much more of the picture - typically overscanning by less than 5%.  More information about safe areas (particularly for widescreen where allowance has to be made for the images being shown on a 4:3 display) can be found on the Commissioning section of the BBC’s web site at http://www.bbc.co.uk/commissioning/tvbranding/picturesize.shtml

Grille

A luminance grille is the standard test signal used when adjusting the convergence of a colour television or monitor, that is, adjusting it such that the red, green, and blue images correctly overlay each other, producing a white display with minimum coloured borders.  The cross in the middle is dead-centre, and so used for the static controls.  On a 4:3 display, the squares should be truly square, and evenly spaced across the whole screen.  There is no JPEG version of this, as GIF can save a perfect copy of a monochrome image, and on simple test patterns like this, its LZW compression is very efficient.

PLUGE

PLUGE - standing for Picture Line-up Generating Equipment - is another monitor setup pattern.  The two feint vertical stripes on the left are just above and just below black level.  The monitor's brightness control should be adjusted so that the super-black stripe is lost, but the other can just be seen.  In a studio gallery with lots of monitors, the contrast control will then be adjusted with the aid of a spot photometer to ensure they are all the same.  The different grey blocks can also be used as a quick test to ensure there is no colour balance error at any level.  Richard includes a simple two-block version in his software, with the feint near-black stripes in the centre of the picture, as advocated by the S.M.P.T.E. in their recommendation RP 167-1995 - "Alignment of NTSC Color Picture Monitors".  They suggest the near-black stripes should be within the middle 25% of the picture, to avoid errors caused by luminance shading at the sides.  However, most commercial generators produce the four-block signal like this.  The near-black stripes are well outside of the central 25%, so I have reversed the normal order to place the super-black stripe nearer the middle.  (Note that the brightness of this thumbnail image has been deliberately raised so that the super-black stripe is visible.)

“Tartan” colour-bars

This apparent mish-mash of colours is useful in setting the saturation on a television/monitor.  Every colour has a border with every other colour, and so if the display can be switched to show, for example, just blue, all blocks should be the same brightness.  Note: these bars (including the white) are only 75% amplitude to prevent overload problems that might occur with full level bars.

Luminance Frequency Sweeps

A simple sine-wave with a linearly increasing frequency will show up any high frequency response limitations, or anomalies like notches at the colour subcarrier frequency, for example.  Separate patterns are provided with calibrated markers for 625, 441, and 405 line systems.  Extending the frequency sweep right up to half the sampling rate on the 625 pattern is a bit cheeky, but at least it demonstrates the equipment limits.  And hey, Richard has done the same on his test card!

S.M.P.T.E. color-bars

Specifically intended for use as a 525 test signal are these SMPTE (Society of Motion Picture and Television Engineers) color-bars, drawn from the SMPTE specification: “EG 1-1990 - Alignment Color Bar Test Signal for Television Picture Monitors”.  The narrow band of reversed order bars makes it possible to set the saturation on a monitor by eye if just the blue gun can be switched on.  Unless you really want to add your own personal caption, it is strongly recommended that the YUV file be used in preference to either of the RGB versions, as this signal has “illegal colours” in the bottom section that can’t be accurately recreated in the RGB files.  (Note: Richard’s program expects all source files to be the standard size for 625 use, and then scales them accordingly for the other line standards, hence the YUV file is 576 lines high, not 486 as would be expected for a native 525 file.  Further, Richard’s program creates a 525 image that is the full 485 lines high, and not 480 as is typically produced by 525 DVD players.)

Simple Test Signals

Here are some simple, standard test signals:

	Plain EBU colour-bars
	ebubarsbmp.zip	640 x 468 RGB	ZIP’d BMP file - 5 KB
	ebubars.jpg	640 x 468 RGB	Low compression JPEG - 18 KB
	ebubars.gif	640 x 468 RGB	64-colour GIF - 13 KB
	ebubarsyuv.zip	720 x 576 YUV	ZIP’d YUV file - 6 KB
	Luminance Sawtooth (or “ramp”)
	sawtoothbmp.zip	640 x 468 RGB	ZIP’d BMP file - 4 KB
	sawtooth.gif	640 x 468 RGB	256-greys GIF - 95 KB
	sawtoothyuv.zip	720 x 576 YUV	ZIP’d YUV file - 6 KB
	10-step Luminance Staircase
	staircasebmp.zip	640 x 468 RGB	ZIP’d BMP file - 5 KB
	staircase.gif	640 x 468 RGB	64-greys GIF - 16 KB
	staircaseyuv.zip	720 x 576 YUV	ZIP’d YUV file - 7 KB
	Multi-burst (0.5, 1, 2, 4, 4.8, & 5.5 Mc/s)
	multiburstbmp.zip	640 x 468 RGB	ZIP’d BMP file - 4 KB
	multiburst.gif	640 x 468 RGB	64-greys GIF - 34 KB
	multiburstyuv.zip	720 x 576 YUV	ZIP’d YUV file - 8 KB
	Pulse & Bar
	pulseandbarbmp.zip	640 x 468 RGB	ZIP’d BMP file - 5 KB
	pulseandbar.jpg	640 x 468 RGB	Low compression JPEG - 17 KB
	pulseandbar.gif	640 x 468 RGB	128-colour GIF - 11 KB
	pulseandbaryuv.zip	720 x 576 YUV	ZIP’d YUV file - 5 KB

Note: A standard 625 multi-burst pattern would have 5.8 Mc/s as the top frequency, but reproduction of that was rather poor due to the sampling rate of this generator, and so has been replaced by 5.5 Mc/s.

50 Hertz

A deceptively simple, but positively evil test signal!  When thinking about video bandwidths, we usually worry about the high frequency end, completely forgetting that it also needs to go down to D.C.!  For faithful reproduction of this 50 cycle square-wave, the video path needs either an excellent low-frequency response to avoid sag on the flat topped signal, or alternatively an excellent clamp.  Traditionally this signal would be full amplitude black and white, however, for the same reasons as discussed above for the test card streak/ghost block, I have made this white and 25% grey, so any sag should be visually apparent on-screen by a non-uniform grey below the white top-half.  Extreme distortion may even interfere with the sync-separator operation, resulting in line-tearing after the black-to-white transition at the top of the frame, and/or the white to grey transition in the middle of the frame.

Red and Black Fields

Yes, plain coloured fields!  I’m slightly embarrassed to include such simple “patterns”, but they have their uses.  The red is useful when “de-gaussing” a CRT display, ensuring the electron beam from the red gun only hits the red phosphor (the other colours will fall in line when red is right, and errors are most easily seen on red).  The red is just 75% amplitude so that it doesn’t overload and defocus the beam.  Plain black can be used for a true “screensaver”; as a background for simple captions generated by a video titler maybe; and in video editing, to tidy up the top and tail, etc.  They are trivial to create, of course, but are here to save you the trouble.

Specials: 1000c/s Audio Tone

Finally, some pre-encoded ROM images.  As the generator has an eight-field store, simple repetitive animation sequences can be produced, and the current versions of Richard’s software (since 3.2a) include an option to upload pre-encoded ROM-image files.  In normal operation, when the program is in the process of programming the generator, it creates some temp files which are saved in the C:\windows\documents and settings\<user>\temp\ directory (Windows-XP) (or C:\windows\temp\ in Windows-95 etc.).  This presents the opportunity to encode several static images, and then edit the ROM.TMP files produced into a moving sequence; or to create a totally different type of file.  We recently had a requirement at work for a feed of Line Drive pulses at 300mV.  When I started work umpteen years ago, Line Drive was one of seven signals produced by a pulse generator, and used by all signal sources (cameras, telecine machines, etc.).  These days, however, everything uses the single “black-and-burst” signal, recreating internally whatever pulses it needs, and so our modern pulse generators aren’t even configured to produce Line Drive any more :-(   So, to the rescue comes RTR’s Test Card Generator!  Creating the ROM data for the simple line-repetitive pulse required a simple external program, and then Richard’s program uploaded the data to the generator.

I imagine there will be very few uses for a 300mV 625 Line Drive signal, but the ZIP’d file is available here anyway, as much as anything to demonstrate the principle.  To use, expand the ZIP file, and choose "Program from file" in the Options menu.

The is also the possibility of using this generator to produce repetitive audio signals.  Whilst the 8-bit sampling might normally be considered woefully inadequate for audio, the tremendous oversampling of the 12Mc/s clock allows dithering to be added that gives excellent results.  By way of example, the 1000c/s file available here gives noise and harmonics measured at 65dB below the signal - not HiFi maybe, but pretty good for what is intended as a video generator, and perfectly usable.  Note, even when ZIP’d, this file is 503 Kbytes, due to the random dithering applied making it difficult to compress.

Back to my front page

This page last updated 28th August 2006