1.1: Hotlinks:
The following WWW hotlinks lead to commercial info on PALplus TVs:
1.2: Reviews:
Reviews of PALplus equipment
will be included as and when members of Europe-LD (or anyone else
who reads this) get a chance to see or even obtain such stuff.
1.3: Rumours:
The following information is provided in good faith, but no
warranty is given (nor should be implied) that any of it is accurate.
If you plan to buy equipment, it is
your
responsibility to check it offers the facilities that you want.
Duff info will be removed or ammended if readers e-mail me
with suggestions.
Sony's 1995 technical brochure appears to mention a "new" TV, type number KV-32W3(S). It does:
Nokia have had a trio of PALplus TVs available since spring 1995 (see hotlinks section above). However, several sources seem to think that they don't offer the Motion Adaptive ColourPlus "Clean PAL" colour decoding facility for Camera Mode signals. It is unclear as to whether they offer even the simple ColourPlus decoding in Film Mode. Their WWW page states that there is a "comb filter" for colour decoding which is a bit ominous. Also (being 50Hz interlaced sets) they don't use the Film Mode helper signal (there's a good reason for this, given below in section 5.1.1).
The Nokia decoder plugs into a TV's "tuner out" SCART connector. Most probably all such decoders will operate this way, taking the composite signal from the TV and feeding the decoded RGB back into the TV via the same SCART. If the TV is a 4:3 model, the decoder will force it into 16:9 squeeze mode (via pin 8 of the SCART connector) when appropriate. If the TV is a 16:9 model, it would force it into 4:3-with-side-panels mode when appropriate. The advantage of this is that the TV's own tuner supplies the signal to the decoder and thus the TV's remote-control still works normally once you've set everything up correctly. Also aspect ratio switching is automatic unless overridden by the user.
It seems Sony plan two PALplus-aware VCRs. One will have a partial decoder inside which will (by the look of it) just recover the helper. The other will have a more complete decoder (with MACP?). It seems that both will yield a 16:9 anamorphic normal PAL signal (probably with an SVHS style Y/C feed), and will (maybe?) rewrite the WSS signal on replay to reflect that fact (or otherwise use the SCART aspect ratio pin to force the TV into 16:9 squeeze mode).
However, before talking about PALplus, we'd best mention how a PALplus TV knows that the incoming signal is indeed PALplus. This is done by an new European Broadcasting Union (EBU) signalling standard intended for 625-line TV of any format in Europe.
Basically, scan line 23 of all future broadcasts (including most or all non-PALplus broadcasts - even SECAM broadcasts) will contain a data burst which details the signal format in use. This line will be known as the Wide Screen Signalling (WSS) line (though it carries more data than just that).
PALplus broadcasts must carry this WSS line, which will tell the TV not only if PALplus processing is to be done, but also whether the signal is "Film Mode" or "Camera Mode". A film mode broadcast has both fields of a frame coming from the same image - as would typically be the case with a movie scanned on a telecine. The fields of a frame in camera mode are assumed to be independent of each other - as would be the case when an electronic camera scans a scene in motion.
A Wireless World "Update" news article [FOX93] states that it is hoped that the GCR system will alleviate problems caused to about 1.1 million TV viewers in London due to echoes from the Canary Wharf tower in Docklands. (Why they don't just site a repeater on top of the offending building beats me. What are the owners of "normal" TVs in the troubled areas supposed to do? Pay for new Ghost Cancelling TVs out of their own pockets?).
PALplus makes use of filtering in both the horizontal and vertical axes. Knowledge of horizontal filtering is assumed since it has been a part of TV technology since the start. Vertical filtering is however a more recent introduction since it requires delay-lines. Basically the idea is that you consider any vertical column of pixels in the original image to be the "signal" to be filtered, and the result of the filtering operation appears as that same column of pixels in the processed image. (The filtering algorithm is repeated on all the vertical columns of pixels to yield a complete image).
The difficulty is that because of the scanning procedure, the information pertaining to a column of pixels gets split up. In the case of sequential scan, two pixels that are vertically adjacent appear one line-period apart in the transmitted signal.
Interlace makes the situation worse. If there are three vertically adjacent pixels A, B and C then pixel 'B' appears 312 lines after 'A', and pixel 'C' appears only one line after 'A'!! Delay lines (or read/write framestores) are used to get the data back into a suitable sequence for filtering - see later.
The real spec of System I PAL TV as used in the UK is given in [DTI84].
The real spec of the PALplus process is currently not published. The info however does get out in various different write-ups, which is where the following details emerged. References are given where possible.
WSS may (for instance) flag a Clean PAL encoded 14:9 picture, in which case the PALplus decoder will be expected to use its Clean PAL decoding circuits, but not to try to process the helper signal (which won't be there).
The encoding of WSS is defined in [WSS93], though Fig 8 of [ELLI94] is a diagram showing the basic structure of line 23, with a data packet looking like Teletext [MOTH90] occupying the first 27.4uS of the active line. The signal levels are indeed designed to look like Teletext (logic '0' is black level, logic '1' is black level+500mV) so that existing data-slicer chips can operate on both, and because 20 years of experience with Teletext shows it to be robust. The raw data rate of WSS is lower though, based upon 5MHz divided by 6 (i.e. 833KHz bit rate). This is apparently so that the signal will survive recording and playback on domestic VCRs.
Starting from a rising edge datum which is 11.0uS after the start of the line sync pulse, the run-in code looks like this (each '-' symbol being 0.2uS long):
----- --- --- --- ---
--- --- --- ---
10 6 6 6 6 6 6 6 6 <-timings in 0.1uS units
This is then followed by a start-code looking like this:
---- ---- ----- --- --- ----- 6 8 6 8 10 10 <-timings in 0.1uS units
Obviously the run-in code sets the phase of the 5/6MHz (833KHz) symbol timing clock, but the exact reasoning behind the timing of the start-code pulses escapes me!
The rest of the line following the start-code is the WSS data, consisting
of 14 symbols, each 1.2uS long, numbered B0 to B13 in that order.
Each '0' symbol looks like:
--- --- 6 6 <-timings in 0.1uS units
..and each '1' symbol looks like:
--- --- 6 6 <-timings in 0.1uS units
The data decodes into four groups as follows [WSS93], [CONN94], [BUCH94]:
Group 1: Aspect Ratio (3 bits + odd parity): B0 B1 B2 B3
Full format 4:3 0 0 0 1
Letterbox 14:9 Centre 1 0 0 0
Letterbox 14:9 Top 0 1 0 0
Letterbox 16:9 Centre 1 1 0 1
Letterbox 16:9 Top 0 0 1 0
Letterbox > 16:9 1 0 1 1
Full format (shot-and-protected for 14:9 centre) 0 1 1 1
Full format 16:9 anamorphic 1 1 1 0
Group 2: Enhanced services (4 bits): B4 B5 B6 B7
Mode:
Camera 0
Film 1
Colour Encoding:
Normal PAL 0
"ColourPlus" Clean PAL: 1
Helper status:
Not present 0
Present 1
Unallocated: 0
( The two PALplus-specific flags (Helper Status and Colour Encoding)
are not shown in [WSS93], [CONN94] or [BUCH94]. But the current PALplus
draft allocates them as shown above. )
Group 3: Subtitles (3 bits): B8 B9 B10
Teletext subtitles:
No 0
Yes 1
Open subtitles:
Not there 0 0
In active picture 0 1
Out of active picture 1 0
Reserved 1 1
Group 4: (Unallocated (3 bits): B11 B12 B13
Surround Sound (Unofficial):
No 0
Yes 1
Unallocated 0 0
The various unallocated WSS bits are supposed to be set to zeroes, though
Hans-Juergen Reggel spotted a couple of cases where ARD and ZDF transmitted
signals with B12 and B13 set to '1'.
PALplus is defined as being Letterbox 16:9 Centre, Camera Mode or Film Mode, Helper present, "ColourPlus" Clean PAL used. A widescreen movie could be transmitted further letterboxed within the 16:9 PALplus image, but the WSS bits will still claim Letterbox 16:9 Centre. Teletext subtitles may or may not be present, Open subtitles may be present, but only in active picture. In addition, during a PALplus transmission, the last 21uS (appx) of line 23 is used to carry a reference signal consisting of 10uS of black level followed by a burst of 48 cycles of -U phase subcarrier (used to fine-trim the phase and gain of the helper-line demodulator).
Also, only during PALplus transmissions, the first half of line 623 carries a reference signal, this time consisting of about 9.5uS of black level followed by 10uS of peak white, used for setting the levels in the letterbox processing circuits.
To keep the arithmetic simple, the original anamorphic picture is assumed to consist of 576 active scan lines, which is rescanned into a 432 line letterbox and 144 lines of helper. However, line 23 and line 623 are "stolen" for transmission of reference signals (see later). Only the second half of line 23 is active anyway due to field blanking, and only the first half of line 623 is active due to field blanking (and it lacks a colour burst due to burst blanking).
The upper 72 helper lines are therefore moved down the picture by a line, and the bottom 72 helper lines are moved up. This means that only 430 lines of actual letterbox image is transmitted. The various minor reconstruction errors are hidden by the overscan on the TV.
Before transmission, the anamorphic 16:9 image is squeezed into only three- quarters of its height. If this is to be done without horrific aliasing problems, the vertical resolution must be reduced (by vertical low-pass filtering) first. In the case of the chrominance signals, this is accomplished by simple lowpass filters.
In the case of luminance, where the 'lost' resolution must be regained later, the filtering is done by use of a Quadrature Mirror Filter (QMF) highpass and lowpass pair. The characteristics of the QMF process permit the highpass and lowpass outputs to be resampled, transmitted and later recombined without loss (see [VAID87]). In this case the lower three quarters of the luminance vertical detail components are split from the top quarter.
Both the lowpass and highpass outputs of the QMF pair still initially occupy a full picture. If you looked at them on a screen, the lowpass output would look just like the original but with a bit of blurring on horizontal edges. The highpass output would be blank except for a bit of signal corresponding with each of those blurred horizontal edges. Initially the output of the chrominance vertical filters also occupies a full picture too, but with a lot more blurring on horizontal edges as the roll-off frequency is well below that of the luminance.
The QMF lowpass luminance output and the chroma lowpass outputs are now re-scanned into three-quarters of their original heights which is a process which loses no information because there are no vertical components in those signals with sufficent vertical detail frequency to cause aliasing. After Clean PAL encoding, they become the letterbox signal that non PALplus TVs will display.
The luminance highpass output of the QMF pair becomes the "helper" signal. It contains only the top quarter of the vertical frequency components of the luminance part of the original signal and can be transmitted without loss using the spare lines above and below the letterbox picture. Contrary to some things printed in the press, the helper lines are transmitted as an analog signal, not digital. ( The only aspect of PALplus which can really be considered digital apart from the WSS line, is the optional NICAM sound system. )
To reduce visibility, the amplitude of the helper signal is companded (curve unknown), then modulated on a U phase colour subcarrier with a maximum amplitude no greater than the colour burst of conventional PAL. The transmitted luminance component of these lines is zero (i.e black level). The modulation is effectively single lower sideband, suppressed carrier with the signal rolled off at 3MHz.
The vertical resolution of the luminance image has to be reduced to 216c/aph before it can be squeezed into the 432 active lines which it allegedly occupies on transmission. As mentioned above, a QMF highpass and lowpass filter pair is used, splitting the luminance vertical detail into DC to 216c/aph and 216c/aph to 288c/aph.
The chrominance signals are lowpass filtered to start rolling off at about 80c/aph. This has an extra advantage of making a normal PAL-D decoder generate fewer artefacts on horizontal chroma edges.
These lowpass outputs are now re-scanned into 432 lines - losing no information because there are no vertical components in any signal with greater frequency than 216c/aph. They become the letterboxed image.
The luminance filter highpass output (216c/aph to 288c/aph) is "decimated" by a factor of 4 (meaning that only 1 in 4 lines is transmitted). These 144 lines will become the "helper" signal. It is a feature of the QMF bandsplitting that this decimation process loses no information.
The 144 helper lines are split 72/72 above and below the central 430 lines of the letterbox picture for transmission. We're now back to a 574 active line picture, but with the original anamorphic image reduced to its correct A/R in letterbox form, sandwiched between the helper signal. Now all that remains is to make the helper lines look "black" by companding and modulating on the U phase chroma subcarrier as described above.
These 574 scan lines are now transmitted in interlaced form, 287 active lines per field - simply because that's what the 625 line transmission spec requires. The TV will collect all the data from a pair of fields before reversing the processes described above.
The vertical resolution of the luminance image has to be reduced to 108c/aph before it can be squeezed into the 216 active lines which constitute the letterbox image. As mentioned above, a QMF highpass and lowpass filter pair is used giving two signals, one containing vertical detail from DC to 108c/aph, the other containing 108c/aph to 144c/aph. Chrominance is lowpass filtered by simple filters with a -3dB point around 40c/aph (not sure about the exact value).
These lowpass outputs are now re-scanned into 216 lines - again, no information is lost because there are no vertical components in any signal with greater frequency than 108c/aph. These become the letterbox signal.
The luminance highpass output (108c/aph to 144c/aph) is decimated by a factor of 4, giving 72 lines which will become the "helper" signal.
The 72 helper lines are split 36/36 above and below the central 215 lines of the 216 original letterbox picture. This gives a 287 active line picture, but with the original anamorphic image reduced to about its correct A/R in letterbox form, sandwiched between the helper signal. Now all that remains is to make the helper lines look "black" by companding and modulating on the U phase chroma subcarrier.
These 287 scan lines are now transmitted as they are, in interlaced form. The TV will process the fields as they arrive, reversing the processes described above.
Only the letterbox part of the PALplus signal is encoded with Clean PAL. The helper lines are not strictly PAL at all, they merely look enough like PAL to get through the transmission chain and remain fairly invisible on normal TVs.
Common to both variants of ColourPlus is the use of processing across a delay of a field period. Various BBC research reports have shown that the best comb filters for PAL luminance/chrominance separation need delays of 312 line-periods (see [DREW75], fig 38 for instance). They rely on the facts that:
Filters based on field delays are not used on normal PAL TV decoders because they introduce motion-blur, but with the additional knowledge that there is no motion between field pairs, that objection can easily be removed.
Fixed ColourPlus arranges to transmit the same chroma and HF luminance (i.e luminance above 3MHz) on lines N and N+312 during the letterbox portion of a frame. (These lines are spatially next to each other and so this processing is almost invisible to the viewer). At the decoder the lines are bandsplit into an LF portion and an HF portion by simple filters. If the HF portions of lines N and N+312 are added, the chroma cancels out, if they are subtracted, the luminance cancels. This allows complete separation of luma and chroma with no crosstalk - and the full luminance bandwidth of 5.0 or 5.5MHz (the transmission system limits of systems B/G and I respectively) is preserved.
Luminance below 3MHz is not processed in this way. It is clear of interference from the sidebands of the chroma signal and thus can contain different information on lines N and N+312 - allowing the full vertical resolution of 287c/aph to be re-constructed with the aid of the helper signal (which is also limited to 3MHz bandwidth).
Luminance above 3MHz is thus limited vertically to 108c/aph (because of the fact that it occupies pairs of lines within the letterbox, with no contribution from the helper). The PALplus system plays on the fact that the eye is not so sensitive to such fine diagonal frequencies.
The decoder's problem is how to tell which bits of the picture were encoded MACP, and which bits CP. To do this, a "normal" PAL-D decoder is first used on the incoming signal, and its output is reserved for later (*). The PAL-D decoded U and V outputs for lines N and N+1 are delayed and are averaged with the PAL-D decoded U and V outputs one field later (from lines N+312 and N+313 therefore). They are then further low-pass filtered (filter characteristic not known). This, if you like, yields a "reference" chroma signal for a frame.
Individual chroma straight from the PAL-D decoder (marked with a (*) above) is compared with the "reference chroma" for every point in the image. If it matches, then evidently "Fixed ColourPlus" must have been used at that point in the encoder, so adding and subtracting across the frame is now used to produce clean luma and chroma for the display.
If there is a mismatch, then evidently "Motion Adaptive ColourPlus" must have been used at that point, in which case the original PAL-D decoded signal (marked with a (*) above) is used for the displayed chroma instead. It will already be "clean" because the encoder didn't send any HF luminance at that point.
Switching between these two decoder modes must be able to be done many times within a single scan line. Response speed isn't known yet.
Some details of the current proposals are given in a paper by Gardiner [GARD94] indicating that line 318 (unused by teletext in any part of Europe) is to be used to carry a radar-like swept frequency chirp starting at DC and ending at about 5.0MHz in the first 24uS of the line, the rest being left blank. In order that the GCR signal does not intrude below black level, it is stood on a DC pedestal of (currently) 0.35V so it might swing between 0V (black) and 0.7V (peak white). Pulsed Klystron transmitters (common in Europe and especially common in the UK) don't like blacker-than-black signals. Mind you, the helper signal in PALplus gets away with it....
The phase of this GCR signal is inverted on each successive frame so that the time-invariant components of the signal (i.e. the sync pulse and DC pedestal of the GCR line itself plus any ghosts of the time-invariant lines to either side) can be subtracted out over two frame periods.
For countries using PAL, there is no colour burst on line 318 (it is within the PAL burst blanking period) which simplifies things a bit. In addition, the burst on line 319 is only present every other frame (due to the 4-field PAL burst blanking interval) and each time a burst does appear on line 319 it is in antiphase with the previous occasion. This again helps, and over a 4-frame period the GCR can be recovered cleanly.
I am not sure about SECAM's equivalent, but the system is designed to be workable regardless of colour system in use.
Gardiner also states that this proposed European GCR is of the same type as the GCR which has recently been standardised for the NTSC system [ATSC92], thus permitting commonality of research, and allowing any future Ghost Cancellation hardware to be made dual standard with the minimum of fuss.
Some recent work in Canada [FIAL94] on the NTSC version of the GCR idea shows that there is possible mileage in adding a luminance staircase signal in the unused latter half of the line, but this does not seem to be under consideration in Europe, nor is it part of the current NTSC standard.
In order to work, the GCR must be preceded and followed by a period of non time-varying signal. In Europe, the use of Teletext (virtually unknown in NTSC-land) means that the only space availiable is line 318. This line is preceded by the lines containing frame-sync equalisation pulses, and followed by a line currently not used for Teletext. As it stands then, the GCR can cope with pre-echoes delayed by several line-periods, and a post-echo delayed by up to 104uS (one line period plus the unused 40uS of the GCR line itself).
If, in the future, line 319 gets bagged for teletext then post-echoes of up to 40uS would still be immediately correctable (and that is believed to be sufficient anyway). European tests seem to show that most ghosts are delayed by no more than 10uS anyway.
Actually, even in the presence of Teletext on adjacent lines, the GCR can still be used to cancel long delay echoes if more frame-by-frame averaging is applied. In the US GCR standard, line 19 is used. The following line is used for insertion test signals (non time-varying) and in theory the preceding line might be used for Teletext, which would complicate the GCR processing for NTSC pre-echoes (the reverse of the potential European problem). Teletext is sufficiently rare in NTSC-land that US GCR processors appear to make a test for Teletext on line 18 and do the processing with one of two algorithms depending on whether or not it is present.
In Film Mode on an interlaced display, the maximum attainable vertical resolution is appx 0.7 times the above figures due to the onset of interlace twitter. (This is the so-called "Kell Factor", already taken into account when the bandlimits of "ordinary" 625 line TV were set, way back in the 1950's.) This puts the usable resolution limit at "only" 216c/aph, but the luminance bandwidth would still have to be 7.38MHz to achieve parity. But bear in mind that the Kell factor is subjective, and should only be taken as a guide.
Though camera mode signals only really attain 144c/aph vertically per field, the interlace "fills in" the gaps on non-moving objects making the above observations for film mode applicable in those cases. For moving objects in camera mode, the 144c/aph limit is more applicable and the system only requires 256c/apw horizontally to attain parity. However, for coloured moving objects in camera mode, Motion Adaptive ColourPlus limits the horizontal frequency to 3MHz (156c/apw), and so horizontal resolution falls short again!
If I understand it correctly, a PALplus TV will remove these artefacts as a by-product of QMF reconstruction with the helper signal. But the viewers of standard TVs will still see them.
ATSC92: "ATSC Standard - Ghost Cancelling Reference Signal for NTSC", Doc A/49, United States Advanced Television Standards Committee 1992. BUCH94: Buchner P: "Line 23 signalling", (c) IEE 1994. CONN94: Connelly M.C, Hopper R.D: "Wide Screen Signalling in the Studio and Production Environment", International Broadcasting Convention 16- 20 Sept 1994, Conference Pub No. 397, (c) IEE 1994, pp20-25. DREW75: Drewery J.O: "The Filtering of Luminance and Chrominance signals to avoid Cross-Colour in a PAL Colour System", BBC Research Dept. Report, 1975/36. DTI84: "Specification of Television Standards of 625-line System I Trans- missions in the United Kingdom", Dept. Trade and Industry, London 1984. ELLI94: Ellis R.J.G: "The PALplus Project: Conception to Introduction", International Broadcasting Convention 16-20 Sept 1994, Conference Pub No. 397, (c) IEE 1994, pp8-19. FIAL94: Fiallos E, Jarmusz J, Caron B, Ledoux B: "An Enhanced Ghost Cancellation Reference Signal", IEEE Trans. Consumer Electronics, Vol 40, No. 3, Aug 94, pp640-644. FOX93: Fox B: "Ghostbusters move into Canary Wharf", Wireless World, Jul 93, page 533. GARD94: Gardiner, P.N, Brockhurst D.M, Curtis J.S: "Ghost Cancellation of 625- line Systems", Proc IBC94, IEE Conference Pub. No. 397, pp673-678. GREE93: Greenberg C.B: "Ghost Cancellation System for the US Standard GCR", IEEE Trans. Consumer Electronics, Vol 39, No. 4, Nov 93, pp928-933. HUAN93: Huang J: "A Ghost Cancellation System for the NTSC Television", IEEE Trans. Consumer Electronics, Vol 39, No. 4, Nov 93, pp896-904. MOTH90: Mothersole P.L, White N.W: "Broadcast Data Systems: Teletext and RDS", Butterworths 1990. (Classmark TK5105.M84, ISBN 0-408-04815-8). READ83: Read D.C: "Improving Colour Television Decoding", Wireless World, Dec 83, Jan 84, Feb 84, Mar 84, May 84, Jun 84, Jul 84, Feb 85. SIMS69: Sims H.V: "Principles of PAL Colour Television and Related Systems", Newnes-Butterworths 1969 (reprinted 1970,1974). (ISBN 0-592-05970-7). VAID87: Vaidyanathan P.P: "Quadrature Mirror Filter Banks, M-Band Extensions and Perfect Reconstruction Techniques", IEEE ASSP Magazine, Jul 87 (Vol 4, No. 3), pp4-20. (ISSN 0740-7467). WSS93: Draft prETS 300 294:1993 "625-line Television Wide Screen Signalling (WSS)". Presumably available from the EBU.