Widescreen Test Card Development

Snell & Wilcox's first attempt at a widescreen version of Test Card F

With the forthcoming digital television services using a 16 x 9 aspect ratio, a widescreen version of the test card was required.  Although it may not be transmitted very much (if at all) we would certainly need it internally to check the alignment of Aspect Ratio Converters (ARCs) and the alignment of widescreen monitors.

In 1997, when PMS were installing the SDV (Serial Digital Video) matrix in CAR, they also installed a Snell & Wilcox rack with five Frame Pattern Generator cards.  These can generate a large number of line-repetitive test signals, as well as one complete frame of any picture.  The obvious candidate for the picture was the Test Card, although ideally we needed two versions - the normal 4 x 3 Test Card F, and also a widescreen version.  The 4 x 3 test card has been generated digitally since 1984, and so Research Department were able to provide S&W with the data file for storing in the FPG's EPROMs.  S&W then quickly hacked that data to provide a 16 x 9 variant.  As some of you may have seen, they did this by taking the two edge squares from the grid, including the corner diagonals, and inserted these on opposite sides.  When re-sampled this gave the correct aspect ratio, but aesthetically was not particularly pleasing, as can be seen from the above picture. As each generator card can only store one or the other, some of the cards have the 4x3 version, and other the 16x9 version.

The original test card data had a number of minor errors in it which had been barely noticeable in the analogue version, but which were now much more apparent without the PAL footprint all over them:

• The top and bottom lines (23 and 623) were blank, instead of having half a line of video.  (And in the digital domain, there are no half lines - they are all full, resulting in a 576 line picture, not 574 and two halves!)
  
  
• The data was pre-blanked to nominal (analogue) active line width.  Digital lines have nine pixels outside of the nominal analogue blanking position (about two/thirds of a microsecond).  Even if it was only required in the analogue domain, assuming the DAC re-blanked the signal (in exactly the same place!) double blanking a signal increases the overall blanking width by about 120ns due to the rise-time of the blanking waveform.
  
  
• The second step down on the greyscale actually comprised two levels 10mV apart.
  
  
• There was a grey spot on the top line of the colour bars in the magenta section.
  
  
• The arrow heads at the sides drooped down slightly.
  
  
• The anti-aliasing of the black/white/black vertical lines under the centre circle left something to be desired.
  
  
• The centre picture had originally been grabbed on a Quantel which did not co-time the chrominance samples, but this had not been compensated for in the SDV data, resulting in noticeable chrome/lum delay.
  
  
• The centre picture was generally 'orrible, being soft, noisy, ringy, and sat up.
  
  
• S&W's widescreen card still had the ident letter F under the centre circle.
  
  

Richard Russell in Research Department (Kingswood Warren, near Tadworth, in Surrey) created a rather more aesthetically pleasing widescreen test card, correcting most of these errors:

Richard Russell's interim widescreen test card

Original picture obtained and re-scanned

George Hersee, the SCPD ("Studio Capital Projects Department") engineer involved in the creation of the original Test Card F, and whose daughter, Carole, it is in the centre picture, died of a heart attack on Wednesday the 11th April 2001, aged 76.  Back in 1998 I contacted him, and luckily he still had the original 6cm square transparency from which the centre picture was taken.  He kindly lent me the slide, and I arranged for an external professional scanning firm (Masterlith, of Mitcham in Surrey) to provide a high quality scan.  They were able to provide an image just over 5000 pixels square, with a colour depth of 48 bits (16 each from Red, Green, and Blue).  Arranging that was a saga in itself.  They have one of the few Agfa Selectscan 2000 scanners in the country, but had to get special software over from the manufacturers in America to provide the 48-bit data!  The results justified the effort involved, however.  On the left below is a reduction of the scan showing the whole area of the slide, while on the right is the detail obtained from around her right eye.

And so a new Test Card is born

With this vastly improved source picture, to which Richard Russell has applied appropriate colour- and gamma-correction, a clean new test card has been born.  It has been created with the full 10-bit accuracy available on serial digital video, and a number of new features added.

1. The cross on the blackboard in the picture was always stated as being NEAR the centre of the picture.  It was indeed near, but not precise.  Having obtained the rescan, we have taken the opportunity of making it PRECISELY the centre of the picture.  This necessitated zooming out and panning right a little, which also resulted in a rather better composed picture than the original, giving Carole some headroom, showing her elbow, and the whole of the doll.  The colour of the doll's dress is noticeably duller, but is more realistic.  The original looked fluorescent, and George admits that the lab tinkered with it for Test Card F.  The new picture is also not so sat up or desaturated.
   
   
2. Probably the most obvious addition is the green block in the top black streak-box.  This is to show chrome/lum delay.  It was supposedly possible on the doll's yellow button, but that was notoriously difficult.  This green block will quickly show, by means of a grey edge, if there is any delay, either horizontally or vertically.  (The white arrow-head on the Magenta bar at the top of the screen is a rather more sensitive indicator to horizontal delay errors, but will often be out of sight on over-scanned domestic tele's.)
   
   
3. The coloured blocks at the edges of the old test card were to show sync-separation and burst-gating problems - both irrelevant to digitally transmitted signals.  The blocks on the right have been replaced by horizontal colour bars, with a fast, non-anti-aliased transition between each bar.  Any line-averaging chrominance processing will thus readily be seen.  The blocks on the left individually have only R-Y and B-Y components, decoding to minimum and maximum legal values of Red (top), and Blue (bottom).
   
   
4. The colour bars at the top of the picture (and those on the right) are 100% saturation.  The old test card used so-called 95% bars, where no subcarrier goes below black level.  This was only relevant to analogue transmission on negatively modulated UHF carriers, and again is meaningless to digital transmission.  100% bars have thus been substituted to test the full legal range.
   
   
5. Along the bottom are various ramp waveforms:
   
   
   • Bottom left is a slightly over-range luminance ramp which goes from 240.0 down to 12.0 (235.0 and 16.0 being nominal White and Black respectively.)
     
     
   • Just above that is a shallow luminance ramp, rising at one 10-bit level per pixel, and centred about the 3/8ths level.  It thus goes through both the 1/4 and 1/2 full range points, and will show any DAC non-monotonicity and/or quantisation depth truncation (8-bit processing, for example).
     
     
   • On the right are shallow R-Y and B-Y ramps, changing at two 10-bit steps per (chrominance) pixel in order to cover a reasonable range, while still indicating quantisation depth truncation.
     
     
6. Extra arrow-heads have been added on the centre-line at the sides to show where the edge of 4 x 3 and 14 x 9 centre cut-out pictures should be.  This is to aid in setting up aspect ratio converters and monitors.
   
   
7. The frequency gratings have different frequencies, having been increased in proportion to the aspect ratio so that they are spatially similar to those on the 4 x 3 test card.  The frequencies (when in the analogue domain) are, from the top down: 2 Mhz, 3.33 Mhz, 4.0 MHz, 4.67 MHz, 5.33 MHz, & 6.0 MHz.  That the last is outside the usual analogue pass-band of 5.5 MHz is irrelevant, as this signal will be distributed digitally where it is only limited by the Shannon/Nyquist rule of half sampling frequency (= 6.75 MHz).  (Where it is available internally as an analogue PAL signal, the bandwidth limit doesn't apply.)
   
   
8. Signals are increasingly being passed through synchronisers and other frame-stores, and we have seen in the analogue world how they can be shifted up or down by one line to effectively swap fields.  This results in the loss of the top or bottom line though, which is often not immediately apparent.  The top and bottom lines of the new test card are therefore distinctive, and not anti-aliased with the adjoining lines.  As such they appear to flicker, but this makes their presence more obvious, and hence also their absence.  The top line (23) has the Wide Screen Signalling data on the left, and continues with castellations to the end of the line at a similar frequency to the basic WSS data (833KHz).  The bottom line is castellated all the way across at this same frequency.
   
   
9. The dot in the bottom step of the greyscale was intended to assist setting the brightness of a monitor, ensuring it wasn't too sat down.  But it was quite a bright dot, so setting was inaccurate.  Also, there was no way of telling if the monitor was too sat up.  For many years we have used the special PLUGE signal (Picture monitor Line-Up Generator Equipment) which has super-black and slightly sat-up stripes at ±3%.  This principle has been introduced onto the test card, which now has two dots on the black step of the greyscale.  Unlike genuine PLUGE however, they are at ±7%, as 3% dots weren't easily visible with such a bright overall picture.
   
   The top step of the greyscale was previously at 91%, with the centre dot at 100%. This has been changed so that the top step is now 100%, and there are two dots, again at ±7%, thus showing if any premature clipping has occurred.
   
   The dots on the original Test Card F slide were circular, but they were made square (presumably for simplicity) on the electronic test card.  With the state of current computer technology, it was easy to make them circular again.
   
   
10. Finally, some form of movement was required.  This was difficult, as it had to be aesthetically acceptable.  A digital clock was considered, but the only one available (by Miles, as used in the DTA) was £12,000, the keyer only passed 8-bit data, and it didn't re-calculate the EDH!  However, the current Snell & Wilcox Frame Pattern Generator has the ability to switch cleanly at the rate of one a second, to and fro between any two patterns.  The (lossless) compression used in the storage of the signals in this generator meant that two test cards could be stored if there were only minor differences.  A second version of the test card was thus created without the dots in the greyscale, and the generator set to switch continuously between the two.  Because the dots are relatively faint, this is very unobtrusive, and so far no-one has noticed until it has been pointed out to them.  Once you know it is there, however, it is quite easy to see, and thus determine whether you have a "live" signal, or a frozen frame-store.  (SCAR also have a pair of Test Card generators, but theirs are manufactured by Crystal Vision, and regrettably don't have this "bounce" facility.)
    
    
    

Test Card 'M'

Snell & Wilcox are running the Test Card 'M' Development Group.  Seeing the need for a test signal to check MPEG decoders, they are developing a suite of test signals, carefully pre-encoded as MPEG data streams.  The main signal is their new test card, M (see above), which has lots of movement on it:

• The yellow arrows rotate within the centre circle.
  
  
• The coloured cube (directly above the centre circle in this picture) moves across the screen rotating and changing colour as it goes.
  
  
• A "clock-hand" rotates once per second on the small clock face at the bottom centre.
  
  
• The chrominance and luminance zone plates bottom left and bottom right respectively) are slowly moving.
  
  
• On the latest version (not shown here) there are pointers moving against graduated scales at both sides, and simultaneously there is a click on both left and right sound channels.  The intention being to check for lip-sync errors.
  
  
• The rainbow coloured column to the left of the centre circle is also slowly moving.  This is to check for accurate gamma and colour-balance correction, as the luminance is intended to be constant.  Any inaccuracies show as apparent coggles in the width of the column.
  
  

S&W are keen for the industry to standardise on this test signal.  It does, however, present a number of operational difficulties for broadcasters on a day-to-day basis:

1. The signal is only available as a pre-encoded MPEG data stream.  This would be difficult to transmit, as DTA works at 270Mb SDV.  It would need switching in the Coding & Multiplex room in the same fashion as the spare MPEG coder is switched to line.
   
   
2. A separate higher bit-rate version would be required for the sustaining feeds to the Regions (which would thus suffer decoding and recoding before transmission), and to be simultaneously switched when the "normal" signal was fed to the London transmitters.
   
   
3. The bit-rate is pre-set, so is unlikely to be easily operationally changeable (especially dynamically) to suit our Network requirements.
   
   
4. Sound is also pre-encoded, and is thus limited to what can be stored - a few seconds worth at most.  This is fine for internal lip-sync type tests, but the public are not likely to be too impressed by clicks, or tone.
   
   
5. Teletext is also pre-encoded, and again limited to perhaps one or two pages.  There could be no header clock showing the correct time.  ("PDC" implications?)
   
   

The above would seem to preclude the use of this test card for the usual public "trade test transmissions".  If we had a spare channel (???) or prior to the start of a service, then perhaps this could be used, but it doesn't seem very likely.

Further, it is no use internally for general widescreen use, as there isn't directly a 270Mb signal available, and an MPEG decoder would need to be tied up.

Being only available in the MPEG data domain, this test signal is thus limited to specific in-house applications, and production line testing for receiver manufacturers, but Joe Public ain't gonna get to see it.

Views and opinions expressed on this page are entirely those of the author, and not of the BBC or BBC Technology Ltd.