Flipped RGB colours in a TV

[DaveC426913]

TL;DR Summary

This TV appears to be inverting RGB colours but not black-white.

So this is the TV in our room in Cuba. It's an RCA - not ancient, since it's a flatscreen. It gets maybe two dozen channels, all of which look fine.

But a couple of times a day a half dozen contiguous channels would turn wonky-coloured. The effect would survive channel-changing, but would spontaneously correct itself after an hour or so.

Pink faces appear a cyan tint; solid red logos appear solid cyan:

Here's the correct CNN logo:

Blue logos appear red:

Green footie fields appear magenta:

and dark red peppers appear dark blue-green:

I know that, for VHS tape signals, the black/white signal is distinct from the colour signals, but this is not VHS. And it corrects itself after an hour.
Since it occurs across multiple channels (but not all channels), presumably this means it's happennig at the broadcasting end?

The TV normally operates just fine.
Boot up:

Discuss!

 

[DaveE]

What do your neighbors see? I'd first focus on inside your house vs. outside.

PS: WAY OFF TOPIC. This sounds like a great excuse to connect with them, maybe even after you know the answer. You can start with TVs, but move on to children, hobbies, etc.

 

[DaveC426913]

What do your neighbors see? I'd first focus on inside your house vs. outside.

We were at a resort. Not about to go bangin' on the doors of other hotel guests and asking in broken Spanish if I can watch their TV...

It would be kind of wierd, don't you think, for only a select half-dozen contiguous channels to be affected, and stay this way for a half hour - while all other channels remain fine, no?

 

[Baluncore]

Looks like you were watching NTSC, not digital. Maybe the TV in your room was tuned to an old local analogue transmitter by a previous tenant. Did you do a channel scan?

 

[berkeman]

So this is the TV in our room in Cuba.

We were at a resort.

Sigh.

It's a defective TV. Request a different room when you go back to that resort.

 

[DaveC426913]

Sorry I didn't make this clear: what I'm really interested in is the technical aspect. This is a science observation; not a thing to be fixed.

I would not have been surprised if the entire signal was flipped, so that a white values are black and green hue is magenta. i.e. this makes sense to me:

But to have the black/white value be normal, while the colour spectrum gets flipped suggests they are somehow processed as separate signals and combined for display. I think.

i.e. this does not make sense:

The phosphors or LEDs or whatever are RGB, right? For the sake of argument, let's say they are analogous to a computer monitor: 256 increments of RGB.

So, dark red [255,0,0] has been converted to dark cyan [0,255,255].
For simplicity's sake, lets say skin is light magenta: [255,192,255]. It has been converted to light green: [192,255,192]. I would have thought light values would get flipped to dark values.


Can anyone suggest a simple, plausible way the algorithm that's converting RGB signals to screen colours could get corrupted (temporarily) to flip colour hues but not black/white values?

 

[Tom.G]

flip colour hues but not black/white values?

Black is all OFF, White is all ON.
And the color decoder is very likely completely bypassed without the color subcarrier. (at least here in North America)

Cheers,
Tom

 

[DaveC426913]

Black is all OFF, White is all ON.

Yes. The b/w values are unaffected; only the hues are reversed.

And the color decoder is very likely completely bypassed without the color subcarrier. (at least here in North America)

This is what I"m trying to get at, yes.

 

[pbuk]

Yes. The b/w values are unaffected; only the hues are reversed.

Do you understand how NTSC works?

 

[DaveC426913]

Do you understand how NTSC works?

No. Thats why I'm asking! (I mean, the suggestion that NTSC is the protocol to look at is new to me.)

I read the technical primer section on Wiki but OMG it just rambles.

It looks like the significant point may be seen on this diagram:

As I suspected, the luminance data is distinct from the chroma data.

 

[pbuk]

Thats why I'm asking!

Not sure what you mean.

If you do understand NTSC then what do you think it might look like if there was something wrong with the chrominance signal path but luminance was working fine?

If you don't understand NTSC then can I suggest Wikipedia?

 

[Baluncore]

Luminance; the brightness information as per original B&W TV.
Chrominance; two colour signals, IQ modulated on the sub-carrier = QUAM.

With NTSC (Never The Same Color), the I and Q signals are color differences:
In phase = I = red-cyan. Quadrature phase = Q = magenta-green.
https://en.wikipedia.org/wiki/YIQ

Phase lock failure of the color sub-carrier (on 3.58 MHz) results in a rotation of the YIQ color space being displayed.

 

[Vanadium 50]

You can still buy 3.579545 MHz crystals for cheap. Heaven knows why someone would want one.

 

[DaveC426913]

Not sure what you mean.

I mean I am asking about this because I do not understand how TV signals work.

If you don't understand NTSC then can I suggest Wikipedia?

As I said in the post that you responded to:

I read the technical primer section on Wiki but OMG it just rambles.

In other words, it was not at all illuminating for the layperson.

Well, except for the diagram I posted in that post, which seems to suggest I was right about white values being sent separate from colour.

 

[Baluncore]

Heaven knows why someone would want one.

They work well with low power microcontrollers, and they are cheap.
Here is the list.
https://en.wikipedia.org/wiki/Colorburst#Crystals

 

[Tom.G]

Here are some details of the NTSC color video signal. (it has been so long since I've bothered with it that I had to look it up to refresh my memory!)

a) The Color information is sent as phase changes of a signal on a subcarrier of around 3.5MHz. (Well actually the 3.5MHz carrier is suppressed so only the modulation sidebands are sent.)

b) Several cycles of the 3.5MHz subcarrier are sent on the Horizontal sync signal as a reference. This is used to phase lock the 3.5MHz crystal oscillator in the receiver.

c) The phase of the received color signal is compared the local 3.5MHz oscillator to determine the Color, and the color signal amplitude determines Brightness, or Intensity.

In weak signal areas, there may not be enough signal to phase lock the local 3.5MHz oscillator, causing a loss of color synchronization. Likely what you were seeing; or it may have been a defective phase lock circuit in the TV.

Here is a link to a more detailed explaination; the beginning assumes a fair amount of knowledge but the details are shortly filled in. You may want to watch it twice.

Oh, the first half covers NTSC analog broadcast color, the 2nd half is devoted to the PAL color system. I haven't watched the PAL part.



Cheers,
Tom

 

[jtbell]

Looks like you were watching NTSC, not digital.

The only NTSC signals I've seen (before over-the-air NTSC shut down in the US about 15 years ago) have had a 4:3 aspect ratio. The image in this case looks to have a 16:9 aspect ratio, not pillarboxed or cropped. Is there such a thing as 16:9 NTSC?

 

[Baluncore]

The only NTSC signals I've seen (before over-the-air NTSC shut down in the US about 15 years ago) have had a 4:3 aspect ratio.

Post #1, 6th image, identifies as "Cable 10 NTSC".
It seems there is an archaic cable TV system that distributes within the premises.

Is there any evidence that other TVs have the same problem?
Is the problem with the digital to NTSC conversion for the cable system?
Or is the problem only with this one TV set?

 

[sophiecentaur]

Looks like you were watching NTSC, not digital. Maybe the TV in your room was tuned to an old local analogue transmitter by a previous tenant. Did you do a channel scan?

Probably the cheapest way to migrate all the rooms to a digital source would have been to use a cheapo converter to NTSC so no need to change the sets in the rooms. The fact that the colour subcarrier phase was flipping would have been the nasty converter. I bet that could be the only explanation for the pictures - except if there was a local TV station still broadcasting on NTSC.

@DaveC426913 's diagram

Shows the transmitted spectrum. NTSC (or even moreso , the later PAL coding) was incredibly inspired. A narrow band compatible mono signal was available for reception by old B/W TVs. That signal is Vestigial Sideband AM. Double Sideband AM carries Video frequencies up to 1.25MHz and higher video frequencies up to ab out 4MHz. The HF components still get detected by a simple envelope detector. The carrier presence still allows the simple 'crystal set' detector to work (like a SSB recfeiver) with just a bit of equalisation. to give almost 4MHz of mono (Y signal) video bandwidth (brilliant!) The Colour Subcarrier is set to be be a strange multiple of TV line rate (cant remember the number) so that all the comb of spectral frequencies of the chrominance signal sit right in between the comb of the mono spectrum. So, for a stationary part of the poicture, they interleave and can be demodulated independently. (clever or what?). There is a burst of pure colour subcarrier at the start of each line to sync the chrominance demodulator. Plenty of opportunity to get a phase flip.

There was another smart thing. 'Intercarrier Sound'!) The sound carrier was demodulated by an SSB detector, using the vision carrier to which it had been locked.

The Chrominance 'subcarrier' uses two phase quadrature components with two 'colour difference' signals. Phase slope across the transmission band is a problem for basic NTSC but PAL reduces this by alternating the subcarrier phase to cancel out any overall tilt.

AND it was all done with valves and analogue circuits.

The so-called compatible mono signal was essential so that both existing and new TVs would all work. Some disgusting beat patterns used to be generated on mono by high saturated coloured areas so mono was not really compatible. Mono viewers soon suffered from colour 'improvements' but every night they watched black and white, the potential of a new colour TV was thrust in their faces. Good marketing. A great time and all water under the bridge.

 

[DaveC426913]

So it seems we know the where the colour flip can happen - does that tell us anything about how it could happen?

I still don't know the format of the actual colour data, and how it could get flipped, especially only temporarily (for less than a half hour at a time before fixing itself).

 

[sophiecentaur]

does that tell us anything about how it could happen?

Dodgy phase lock loop with minimal components, perhaps.

 

[sophiecentaur]

I still don't know the format of the actual colour data

This is a huge subject and I really doubt that you want to get too deep in there. I guess you could search for a source of NTSC coders to find how available they are. lol Or just trawl round for historical NTSC tutorials.

The 'data' is an analogue phase modulated signal which is demodulated into two analogue 'colour difference ' signals. They are matrixed with the Luminance signal to produce R G and B signals. The demodulation uses an oscillator that's locked to bursts of a reference 'colour burst' signal. Producing the two difference signals without crosstalk relies on a very stable oscillator and good group delay distortion across the whole channel. PAL (and SECAM) deal with channel group delay / phase tilt. Whatever coding unit is producing the distributed signals, it will need to be well lined up reliably.
A possible signal chain will start with a digital feed (satellite?) and a number of channels will be etracted and each one would go to an old fashioned (surplus from somewhere else) unit which will feed a number of NTSC coding racks. Each NTSC coder would need to be set up.

 

[Baluncore]

So it seems we know the where the colour flip can happen - does that tell us anything about how it could happen?

Do we know if it was in the cable TV generation, on the premises, or in your TV set alone? Did it happen also in the next room?

I still don't know the format of the actual colour data, and how it could get flipped, especially only temporarily (for less than a half hour at a time before fixing itself).

I gave the format in post #12.
It is not being flipped in RGB format, it is being rotated in the NTSC IQ sub-carrier space, which requires reliable phase locking to the short colour burst reference transmitted on the porch of each line.

Chrominance; two colour signals, IQ modulated on the sub-carrier = QUAM.
With NTSC (Never The Same Color), the I and Q signals are color differences:
In phase = I = red-cyan. Quadrature phase = Q = magenta-green.
https://en.wikipedia.org/wiki/YIQ

That link includes all the transformational matrices. Rotate the colour picture to see what colour might move where, when the IQ phase reference gets confused.

 

[DaveC426913]

Do we know if it was in the cable TV generation, on the premises, or in your TV set alone? Did it happen also in the next room?

In retrospet, I did have the opportunity, since my neice was in the room next door. I didn;t think of it at the time.

I gave the format in post #12.

I'll take your word for it. That post is dense with words I do not know. I did not realize it was this complicated.

It is not being flipped in RGB format,

No. that was just a hypothetical I was using in leiu of actually knowing what I was talking about.

it is being rotated in the NTSC IQ sub-carrier space, which requires reliable phase locking to the short colour burst reference transmitted on the porch of each line.

That link includes all the transformational matrices. Rotate the colour picture to see what colour might move where, when the IQ phase reference gets confused.

This is what I had been wondering. (Even though I didn't know the actual details) I was trying to see if it was flipped (i.e. Rd to Cy, Gr to Mg, Bl to Ye) or rotated (Re to Ye or Gr, Gr to Cy or Bl, Bl to Mg or Re) since a rotation seemed more logical. I concluded it was a flip. I still think so. I was pretty careful to note that
Red was showing up disinctly Cyan
Cyan was showing up disinxtly Red
Green was shoeing up distinctly Magenta.

I guess a rotation of 180 degrees on that table is the same as a flip (or is it??).

I think I am struggling with a rotation in a 2D color space (Mg-Gr, Re-Cy) versus a mirrring in a 3D volume (Re-Cy, Gr-Mg, Bl-Ye) space.

Also, I was unable to find any unequivocal blue/yellow flip.

 

[sophiecentaur]

Do we know if it was in the cable TV generation, on the premises, or in your TV set alone? Did it happen also in the next room?

I guess that one faulty TV is the most likely but pretty much the same mechanism would have the effect at either end; a bad RF or IF stage anywhere could mess up the phase response around the colour subcarrier. OR even a distribution amp / transmitter with poor frequency response.
Moreover, the dealership that provides the sets or the distribution system is unlikely to have a whizzo service person to chase and solve the problem. I wouldn't imagine that there are many 'old tv' enthusiasts in Cuba.

Also, I was unable to find any unequivocal blue/yellow flip.

There are many hits for a google search on NTSC testing and vectorscope. This one has a vectorscope display and you could get your answer from that, probably. But you want a particular phase switch. Would it necessarily require 180 degrees of the CSC reference? The CSC regeneration could have a number of phase errors. I wouldn't lose too much sleep about the details as we don;t know the actual situation.

 

[Baluncore]

I was pretty careful to note that
Red was showing up disinctly Cyan
Cyan was showing up disinctly Red
Green was showing up distinctly Magenta.

I guess a rotation of 180 degrees on that table is the same as a flip (or is it??).

It depends. Since the I and Q channels encode colour differences, locking with a 180° phase shift will invert both differences.
Diagnosis of the fault in the PLL or matrix computation by colour alone is possible, but not without a test pattern when the TV is in another country.

Moreover, the dealership that provides the sets or the distribution system is unlikely to have a whizzo service person to chase and solve the problem. I wouldn't imagine that there are many 'old tv' enthusiasts in Cuba.

Analogue NTSC TV transmitters, and NTSC cable TV systems, are still operating throughout Cuba.

 

[sophiecentaur]

Analogue NTSC TV transmitters, and NTSC cable TV systems, are still operating throughout Cuba.

That's interesting - I should have done my homework earlier!!. They are still proud to announce that Cuba was the second country to have a colour TV broadcast system. Searches mostly took me to 'past glories' of analogue TV but I did find this link which talks of a digital service, based on the chinese system. This has been around for ten years, apparently.

 

[DaveC426913]

I guess that one faulty TV is the most likely

I still think it's notable that the fault

1. happens across a small, yet contiguous range of channels, but leaves the all rest alone
2. survives channel flipping (i.e. I can channel surf all I want and the effect only affects the specific channels)
3. doesn't flicker in or out but stays steady for the half hour or so.

Point 1 and 2 I think strongly suggest the cause is independent of the TV itself.

 

[sophiecentaur]

Sounds logical.
NTSC is very fussy about phase and filters are much easier to tweak for amplitude.
Now, why some channels but not others? (Not just one channel?) Assume the broadcast transmitter / cable head takes digital feeds and has a bank of ntsc coders. Each ntsc signal (at an IF frequency) will be mixed up to a suitable RF channel and combined with the others. From the fact that it’s only in one group it’s probably happening in one particular frequency multiplexer.
If DTV has been available for ten years or more can we conclude that nothing has bee spent on this analog system for years and they’re waiting for a full upgrade to DTV? It’s just quietly dying . RIP

 

[Baluncore]

The problem I see with NTSC is multi-path propagation, where the colour burst is phase shifted by the addition of a delayed reflection.

That could also happen with reflections in a mismatched or faulty cable network.

 

[Algr]

They are still proud to announce that Cuba was the second country to have a colour TV broadcast system.

That is ironic! The Castro regime both shut down the color broadcasts, and drove out the people who had set them up.

---

Inverted colors is an easy flaw in NTSC, because the hue is determined by measuring the phase relationship between a wave off screen (in the sync pulse), and another on screen (hidden in the signal). If they are in phase, that is supposed to be red. 120° out of phase would be green, and 240° would be green.

But if you need to synchronize multiple NTSC broadcasts together, as you might in a cable TV arrangements, just a few feet of cable at the broadcast station can delay one NTSC signal so that it misaligns with the sync pulse generated elsewhere. (Perhaps in the hotel.) Sometimes at video studios, you'd see huge coils of cable laying on the floor that are just there to delay the signal slightly less than 360°, so that it would sync with something it was supposed to be mixed with! It was a pain, and lots of people didn't understand it, and did it wrong.

 

[sophiecentaur]

the hue is determined by measuring the phase relationship between a wave off screen (in the sync pulse), and another on screen (hidden in the signal

It's not really like that. The colour burst signal is the reference for the phase (and amplitude, btw) of the solour subcarrier in that line - it's all within the signal from each source. The timing equalisation you refer to is necessary because the colour burst on all incoming signals needs to be synced. Without this equalisation, each switch between sources would involve a period where the subcarrier generator drifted into a new phase relationship.

Sometimes at video studios, you'd see huge coils of cable laying on the floor that are just there to delay the signal slightly less than 360°, so that it would sync with something it was supposed to be mixed with! It was a pain, and lots of people didn't understand it, and did it wrong.

BBC Television Centre (London) was built in a circle with studios all placed at equal distances from the switching equipment at the hub so that they could between sources and still keep sync. (Shorter lengths of equalising cable needed for each camera).

In the old days of the UK independent TV service, there was no sync between the regions at all (just the mains 50Hz). Every commercial break had that annoying 'exploding white star' to mask the change in (coarse) timing.

It became relatively easy to re-code colour TV signals when they passed through a common point so switching between sources could be done seamlessly. But CATV systems have often been done as cheaply as possible, squeezing multiple channels into a very narrow space. Without tight regulations, picture quality suffers yet the consumer is happy to suffer that in order to see a 'coloured' picture.

 

[pbuk]

BBC Television Centre (London) was built in a circle with studios all placed at equal distances from the switching equipment at the hub so that they could between sources and still keep sync. (Shorter lengths of equalising cable needed for each camera).

I am not sure that is correct, do you have any source for that (other than the spurious information on Wikipedia, which I have removed)?

 

[sophiecentaur]

I am not sure that is correct, do you have any source for that

OH ye of little faith. I have been in the basement of TC (years ago) when I worked there. (Good enough for you??? ) Here is some evidence for the basement under the fountain.

The architects made a virtue of necessity and put a visible statue at the centre to make it all look nice.

PS I will accept a bouquet as an apology for your ever doubting me.

 

[pbuk]

PS I will accept a bouquet as an apology for your ever doubting me.

Oh there is no doubt about the basement under the fountain (later replaced by the statue), and the video recording suite being located there.

However the Central Apparatus Room was on the second floor of the 'central wedge' of the doughnut: are you saying that cables from studios 3 and 4 (adjacent to the central wedge) were routed out into the centre under the fountain and then back in via the video recording suite and up to the CAR (presumably also with a feed from the CAR back to the recording suite)? Or was the editing and switching actually done in the tele-recording suite?

This drawing may help jog your memory (although I believe this is as designed not as built):
http://www.r-type.org/timeline/time145d.jpg