Colour decoder fundamentals

[Mike Phelan]

I thought this might be useful for our members who have just started with colour sets, especially hybrid ones. I will archive the thread after it has run for a while - any boo-boos and suggestions welcome.

I have deliberately simplified this – there is scant reference to Simple PAL and very few mathematics! We will also forget U and V and refer to the R-Y and B-Y colour difference signals.

Transmission
To understand how the decoder works, it is necessary to know a little of how the colour transmissions are sent. A requirement was always that monochrome signals could be received by a colour set, and that the converse was true – colour transmissions had to be received by a monochrome set.
So a normal monochrome signal, complete with sync pulses is sent (the Luminance or Y signal) and the colour information is additional. Colour can be defined in terms of three primary colours – Red, Green and Blue. If we send the Y signal as well, only two other colours are needed.
I had terrible problems trying to get my head around the fact that we only sent the red and blue, and somehow the green mysteriously appeared from nowhere!
Then the light dawned – it was not red and blue that were sent, it was R-Y and B-Y. As Y is a mix of R,G and B in various proportions to make white, R-Y and B-Y contain the green (actually G-Y) information, as can be proved mathematically, but we wont bother with that stuff here.
Look at it as a 2-dimensional map of every possible colour – we only need two co-ordinates to find any point.
R-Y if you look at it on the screen, goes from green to red, B-Y from yellow to blue.
Oh, all right, then. All we need is:
Y = 0.3 Red + 0.59 Green + 0.11 Blue
R-Y = 0.7 Red – (0.59Green + 0.11 Blue)
B-Y = 0.89 Blue –( 0.59 Green + 0.3 Red)
OK – At the transmitter, the colour is modulated on to a 4.43MHz subcarrier, out of the way of the video sidebands and the 6MHz sound. It is AM, but to modulate both the R-Y and B-Y each bit is put on at a particular point in the phase of the carrier. R-Y is shifted 90 degrees from B-Y and reversed on alternate lines. The carrier is suppressed, but so we can get it back at our end, as 10 or so cycles of burst are transmitted during the back porch of the line sync period.
Finally, the set needs to know which way round the R-Y signal is on each line; this is achieved by swinging the burst either side of the carrier by 45 degrees on each line. Enough of that – on to the decoder.

The decoder.
Many of us have never met a discrete chroma section in the flesh, and never needed to know how any of it worked; it is all in a chip nowadays. Even the G8 had a couple of these in it.
Treating the decoder as a “black box” we have a chroma signal (modulated subcarrier) going in, together with line pulses for various purposes, and R, G and B coming out.
There are really three main things to achieve; the chroma signal needs to be amplified and separated into its R-Y and B-Y components; there needs to be a subcarrier oscillator to replace the one suppressed at the transmitter and this oscillator needs to be phase-locked to the incoming burst; the R-Y and B-Y need to be matrixed to give G-Y and these three colour difference signals need to have the Y component added back in to give R. G and B. Many hybrid sets, including the G6, do this by using the CRT itself.

The chroma amplifiers
These are two or three stages in number, and have tuned circuits like IF transformers. There is provision for chroma AGC (ACC) and a means of turning the channel off altogether (colour killer) when a monochrome transmission is received – not often nowadays!
Often it is necessary to override the killer when fault finding.

Originally the idea was to use Simple PAL so that any phase errors on the chroma signal had the R-Y component reversed on alternate lines, and the eye compensated up to a point; what should have beenm say, a red, would have had alternate orange and purple lines for a severe error.
This did not give very good results, so we introduced PAL-D with the delay line - this averages the alternate lines electronically and also as a bonus, separates the two signals.
To separate the R-Y and B-Y components, we use a delay line that delays the signal by one complete line period. Nothing like its luminance counterpart, it is a block of glass with a pair of piezo transducers – one for input has the signal reflected across the glass and picked up by the output transducer.
Later delay lines are much smaller and use multiple reflections.
So – we can get the direct and delayed chroma simultaneously. If we add these together, because the R-Y is inverted on alternate lines, it cancels, leaving 2 x B-Y. Inverting either the delayed or direct chroma and adding them together cancels the B-Y leaving 2 x R-Y. We have separated the two colour difference signals, but need to put the subcarrier back in before we can use them.

The subcarrier oscillator
A crystal-controlled oscillator is used here, as the phase can be changed sufficiently to lock it to the burst.
The oscillator demodulates the R-Y and, with a 90 degree phase shift, the B-Y. These two signals are fed to a pair of synchronous demodulators, either a diode bridge or a par of diodes and a centre-tapped inductor. Whatever, the idea is to switch on the demodulator to pass the signal though at the correct time or phase. Unless we are using Simple PAL the signals are already separated by the delay line so the output from each demodulator outputs are really just giving us the amplitudes and whether +ve or -ve.
There is still something to do – the R-Y is inverted on every line, so either the subcarrier or R-Y signal fed to the demodulator needs to be switched on each line. A bistable fed by line pulses does this. It’s termed a PAL switch.

The burst channel
The chroma signal is taken from the amplifiers and passed through a burst gate, preceded or followed by a stage of amplification or two. The burst gate is normally closed; a delayed line pulse timed with the burst on the back porch opens it. The amplified burst stripped of its chroma goes to a phase detector – normally two diodes and two capacitors; this gives a DC output that varies with phase and goes to either a varicap diode or a “reactance valve” to control the subcarrier oscillator phase.
A free-running oscillator because of a fault, will show coloured horizontal stripes running up or down across the screen, usually it is necessary to override the killer (q.v.)
The burst signal can be rectified for ACC as well, as it is a constant part of the chroma signal.

Bistable, ident and colour killer
Wait a minute, though – the burst is swinging (it was in the sixties, remember?) through 45 degrees on each line. Doesn’t matter as far as the phase detector is concerned because of time constants.
We mentioned the bistable earlier – this operates the PAL switch. It will happily start in either phase; you will get reversed R-Y – green faces and pink grass – when it feels like it!
Because of the swinging burst, a half-line-frequency square wave (7.8KHz) termed the ident signal, appears at the burst phase detector. This is amplified, usually by an LC circuit, and used to “steer” the bistable.
If we rectify it and iron out the bumps with an RC circuit it can be used to turn off the colour killer as well - see above

Colour difference
The last piece of our jigsaw. As the R-Y and B-Y signals emerge from the demodulators, a soupcon of mixing proportions of them together gives us G-Y.

On sets using colour difference output, the three colour difference signals are amplified and fed to the CRT grids. As the luminance goes to the three CRT cathodes, the tube itself forms the addition of the Y component back into the colour difference signals, leaving us with R, G and B. Later sets to this by adding the Y component to the CD signals and then just supply RGB to the CRT cathodes; the grid voltage is not fed with any signals, so can be used for beam limiting and the like.
Back to the colour difference – as we have completely lost the DC component in all the processing, each CD output needs to have some sort of black-level clamping. Typically, this is done by a triode for each CD channel, it is normally biased off but connects the grids to a line pulse, turning the anode current on.

[Ray Cooper]

...The system could have just sent R and B or any combination of the colour information such that each component is not redundant...

It's a monochrome-compatibility and bandwidth-saving issue. If you transmit primary colour signals rather than colour-difference signals, you should really be doing so using the full picture-bandwidth of 5MHz+ for each component. This would extend the bandwidth required in the subcarrier part of the signal, and cause excessive cross-patterning into the luminance (black-and-white) part of the signal. One of the triumphs of the former NTSC system was in its realising that colour-difference signal bandwidth could be heavily restricted without making much of an impact on the final viewed picture. Also, using difference signals aids compatibility with monochrome signals enormously: when you transmit a black-and-white picture, the colour difference signals fall to zero, and you are left with a luminance-only signal which will have no cross-patterning effects (unlike SECAM). If you transmitted primary colour signals instead of difference signals, then their amplitude would be very high on a monochrome signal and severe patterning would result.

...by averaging the colour information between the lines the alternating R-Y encoding causes the hideous colour effects of small phase errors to cancel out. Presumably the original idea was that it would cancel in the eye - but using the delay line works better and maybe simplifies the circuit although it does create a quite strange change in the vertical colour resolution especially when you bear in mind that the frames are interleaved.

PAL, as originally proposed, did not use a delay line ('Simple PAL') and relied on the eye to average out chrominance phase errors, as you say. It does this quite well for small phase errors, but as their magnitude increases, disturbing patterns emerge on the picture in the form of a fine horizontal line patterning: 'Hanover bars' or 'Hanover blinds' (cf Venetian Blinds...). With large phase errors this becomes awful. Use of a delay-line removes this 'blind' effect, but at the cost of reducing the saturation of the coloured areas as the phase errors increase (PAL = 'pale and lurid'...)
This is subjectively much less distressing than the 'blind' effect. 'Delay-line PAL' was proposed and accepted even before the European colour services started in 1967. I don't think that any 'Simple PAL' sets were ever marketed, though they would have been quite feasible: it's just that the delay-line version gives you that much more protection against phase errors.

[Station X]

After reading Mike's original post, which I thought contained just enough information, I thought I understood how a colour decoder works. In the light of subsequent posts I'm not so sure.

I never had any problems in understanding simple PAL. The human eye does the work. However I never understood the more complex version where lines are delayed. Could someone explain how this works please. To a simple soul like me it seems that delaying (alternate?) lines would result in a line being displayed on top of another which was undelayed.

[ppppenguin]

Originally Posted by Station X

To a simple soul like me it seems that delaying (alternate?) lines would result in a line being displayed on top of another which was undelayed.

That's near enough correct. The delay line system is averaging adjacent pairs of lines. This effectively halves the vertical resolution of the colour. This doesn't matter because the eye can't see it. Same argument as the reduced horizontal bandwidth. Rather more subtly, the averaging moves the colour information down the screen by half a line. Again you can't see it but it matters for some processing.

More sophisticated decoders use more complex sets of delay lines to keep the Y and C coincident. They also use yet more delay lines and other clever bits to perform accurate separation of Y and C. These are known as comb filter decoders. This avoids the horrible cross colour effects that plague fine Y detail.

[Station X]

Bearing in mind that the picture is interlaced does this mean that odd numbered lines are averaged with the next odd numbered line and even numbered lines with the next even numbered line?

[ppppenguin]

Absolutely right. When I talked about a half line vertical shift I meant within the same field. If you look at both fields, an averaged pair of lines in the odd field will be the same vertical position as the luminance line on the even field. It will be 20ms apart in time since it is on a different field.

If we take this much further we'll start getting into the realms of spatial frequency. It can get very mathematical and if your maths is up to it it's much easier to explain in those terms. I don't think I'll be doing a Mike Phelan and writing a tutorial piece on it

I'll just say that some of the problems of TV are to do with getting some bits of a picture misinterpreted. The best known example is cross colour where fine luminance information, specifically fine luminance at particular diagonal angles, looks just like colour as far as the decoder is concerned. We have all seen the lurid colour patterns.

Another example, now much more relevant due to LCD displays, is interlace. Because of interlace, certain fine vertical detail is almost indistinguishable from motion at certain speeds. This makes de-interlacing, essential to LCD displays, a tricky business. If you get it wrong you canget all sorts of horrible effects, commonly a combed edge effect on verticals.

This is really getting a bit too far from decoders. Although I'm not really prepared to do a lot of writing on the subject if anyone wants to start a new thread with questions like this I'll do my best to give sensible answers. There's at least one other forum member (JJL) who understands this sort of thing so I hope he'll help too.