Origins of PAL: 1956 radio engeenering airticle from UK mag -- phase alternations (and effects) considered...

In article , Doug McDonald wrote:

The colour gamut is determined by the characteristics of the dichroic
filter system in the camera and the primary colours emitted by the CRT.
(The encode/decode system is completely transparent as far as colour
analysis is concerned and simply results in RGB to RGB with no signal
level changes). When the same cameras with the same filters are used to
produce both NTSC and PAL, and the results are viewed on the same
monitors, how can they produce different results?

They don't. What you don't understand is the the 1953 NTSC spec
for the color of the phosphors is DIFFERENT from PAL.

see http://www.aim-dtp.net/aim/technolog...yz/cie_xyz.htm

I can undwerstand primary colours being specified differently, but if you're
going to insist that "true PAL" and "true NTSC" can only exist where both
camera and display have kept exactly to this part of the book specification,
then it makes a nonsense of what happens normally in the real world. In
fact, by this definition, true PAL and true NTSC can only be theoretical
concepts that don't exst outside laboratories.

Many items of television equipment, including cameras, telecine machines,
and displays, are multistandard, which in my experience means that the parts
of the system that are switchable (or offered in different versions at the
time of purchase) are the parts that handle the encoding from and to RGB
signals that have already undergone the usual vision control processing
(such as linear matrixing and gamma correction) at source.

Programmes are routinely exchanged between countries that use either system,
and I have never seen, nor heard of, any transcoding system that decodes the
input standard down to RGB, undoes gamma correction, re-matrixes the linear
signals to simulate origination from the other set of primary colours, then
reprocesses and re-encodes. It would be impossible to do this properly
anyway, because you would also have to standardise the black level settings
of the linear signals and remove any operational effects such as "black
stretch" and "knee", which being operationally adjustable and not part of
the spec, are unknowable quantities. And all for the sake of an academically
small difference in colorimetry that is insignificant beside normal
operational variations and differences between display devices. What value
of gamma would you use anyway? Nobody ever lines up a camera with a gamma of
2.8, which is the strictly correct value in the PAL spec; the adjustments in
the cameras don't have the range, and nobody makes the test charts. It just
doesn't happen. In fact, even gamma correction is often a front-panel
operational control these days, operated by people who have no clear idea
what it means, so the likelihood of a video signal being anywhere near the
theoretical spec for either system is remote.

I think that most people who talk about the relative merits of PAL and NTSC
are only talking about the line standards and encoding and decoding
processes, which in practice these days are the only parts of the systrm
that are different. Assuming that everyone takes it to include camera and
display colorimetry maky be technically correct, but will result in talking
at cross-purposes and misunderstanding.

The green is MUCH greener ... look at the places on the CIE
chart. Note that the 1953 NTSC spec is very similar to the
1998 Abode spec.

Thus, Dyson's (an my) statements apply only to TV sets
that actually obey the 1953 specs. As I have said, modern
RP sets with mercury lamps are much nearer the NTSC 1952 specs
than most CRTs, and result in very gorgeous greens.

The greens might be gorgeous, but are they "correct", in the sense of
accurately reproducing what was in front of the camera? Unless the entire
system, including the camera or telecine machine, has been held rigidly to
the same colorimetric specification, you cannot be sure of this.

Rod.

"Roderick Stewart" wrote in message
om...
In article , Doug McDonald wrote:

The colour gamut is determined by the characteristics of the dichroic
filter system in the camera and the primary colours emitted by the
CRT.
(The encode/decode system is completely transparent as far as colour
analysis is concerned and simply results in RGB to RGB with no signal
level changes). When the same cameras with the same filters are used
to
produce both NTSC and PAL, and the results are viewed on the same
monitors, how can they produce different results?

They don't. What you don't understand is the the 1953 NTSC spec
for the color of the phosphors is DIFFERENT from PAL.

see http://www.aim-dtp.net/aim/technolog...yz/cie_xyz.htm

I can undwerstand primary colours being specified differently, but if
you're
going to insist that "true PAL" and "true NTSC" can only exist where both
camera and display have kept exactly to this part of the book
specification,
then it makes a nonsense of what happens normally in the real world.

They don't have to match the book spec perfectly, but with NTSC, I have
seen significantly better greens because of better NTSC phosphors and
better filters on other kinds of displays. With PAL, it is just impossible.

So, it is a degenerate argument to claim that everything has to exactly meet
the spec (because it doesn't.) However, there is significant opportunity
(and
is realizable in reality) to display a better green on NTSC.


Programmes are routinely exchanged between countries that use either
system,
and I have never seen, nor heard of, any transcoding system that decodes
the
input standard down to RGB, undoes gamma correction, re-matrixes the
linear
signals to simulate origination from the other set of primary colours,
then
reprocesses and re-encodes.

I sure hope that an NTSC/PAL conversion does a rematrix. In the old days,
they were sloppy about it, and NTSC display of PAL broadcasts (after
conversion)
would sometimes look 'brownish' with incorrect colors.

FACT: whether or not you have accurate NTSC phosphors -- the NTSC colors
are different from PAL.



The greens might be gorgeous, but are they "correct", in the sense of
accurately reproducing what was in front of the camera? Unless the entire
system, including the camera or telecine machine, has been held rigidly to
the same colorimetric specification, you cannot be sure of this.

Not quite: if your phosphor reaches deeper into NTSC green, then you can
get a better green. I have seen it, and it works. Perfection isn't
necessary to
simply get better rendition. Think about it this way (in analogy), 1080i
HDTV mostly looks MUCH MUCH better than PAL, even though our 1080i
broadcasts might only show 3/4 of the max horizontal resolution. 1080i
would
probably still look much better than PAL 720Hx576V, even if the 1080i
only had effective resolution of 720Hx1080V!!!

The display device with the better phosphor (like a pro monitor) does
probably
know that the CRT has a wider gamut, and the matrix will likely be
implemented
to support that better gamut (depending upon the spec of the source format.)
Given
PAL, you just cannot legally represent that deeper green.

This thing isn't all or nothing. However, when your spec hits a hard limit,
then it
is difficult to surpass it. :-).

John

"ivan" wrote in message
...

Five minutes' drive away from where he was staying is a shopping complex
called Cribbs Causeway, if he had paid a visit to John Lewis in the
Shopping
Mall, or the Comet store, he would have seen literally dozens of large W\S
plasmas LCD's and CRT's (hardly any 4:3 receivers on display) but even so
there is such a very large disparity between the picture quality of
different manufacturers, that I find it difficult to see how he can assess
the overall quality from a few different makes of receivers, and I can
assure him that IMHO Lowe are certainly not the very best receivers I've
ever seen.


Whilst in many TV shops there is very poor distribution (leading to
appalling pictures), even on paper the Loewe TV's are some way from being
the best.

BTW IMHO the new large LCD screens give the best picture (e.g. Philips 42PF
9986), but they are dear.

In message
John S. Dyson wrote:

PAL TV sets just cannot give an accurate green (period.) Claiming that
NTSC is perfect would be wrong, but it is amazing when you do an a-b
with a closer-to-NTSC phosphor gamut.

Thank you. I'm sorry that when someone says 'PAL', I think of a system
for encoding colour difference signals on a sub-carrier. When the
actual topic of the conversation is chromaticity of display phosphors
then the arguments being put forward make more sense.

Can you enlighten me as to which specific aspect of PAL encoding is
responsible for providing 'inaccurate color' ?

Actually, the fact that PAL isn't as completely decoded as NTSC
kind-of makes those color flashes more likely

Aside from differing I/Q and U/V weightings applied to the colour
difference signals, the fundamental difference in PAL is the V-Axis
switch which inverts the V axis subcarrier line-by-line.

I must admit that I find it difficult to explain technically why this
would lead to a situation such as "incomplete decoding", and would
welcome further clarification. Is this a momentary red/green
transposition by any chance ?

(which even appeared on some apparently digital broadcasts due to
BBC not using their cool 3D comb and/or not using component instead
of composite in critical parts of the infrastructure.)

All 'critical parts' of the infrastructure run in component digital.
With very few exceptions, mostly involving some Outside Broadcasts, a
digitally sourced picture should make it all the way to the terrestrial
or satellite viewer without seeing a single PAL coder or decoder. Of
course, what happens at the viewers' end is beyond control.

Maybe I ought to point out that on installing my own Sony DVB set at
home, the first thing I felt compelled to do was switch the internal
presentation of the digital terrestrial baseband from the factory set
PAL to RGB.

That kind of color flashing artifact provides an essentially TOTAL
FAILURE on the section of the scene.

Can you describe more on the "color flashing artifact" please ?

Thanks,

Gareth.

--
http://www.rat.org.uk gareth at lightfox dot plus dot com

In message
I wrote:

a digitally sourced picture should make it all the way to the terrestrial
or satellite viewer without seeing a single PAL coder or decoder.

I should, of course, point out that I omitted to qualify this as

"the digital terrestrial or satellite viewer in the UK"

Apologies,

Gareth.

--
http://www.rat.org.uk gareth at lightfox dot plus dot com

In article , John Dyson wrote:
I sure hope that an NTSC/PAL conversion does a rematrix. In the old days,
they were sloppy about it, and NTSC display of PAL broadcasts (after
conversion)
would sometimes look 'brownish' with incorrect colors.

Because of the complications that would be involved, as described in my
previous posting, I doubt that this has ever been done, or ever could be done
properly in the real broadcasting world. Re-matrixing to different primaries
would have to be done on linear signals representing what came from the pickup
device (tube or chip) in the camera, *before* any other processing. This is
complicated enough even if you just have to cope with gamma correction (even if
you know what value has been used, and the reality is that the book value is
*not* what is used in practice), but when some of the processing before gamma
correction in the camera will have been subjectively-judged operational tweaks
whose values it is not possible to know, then you should realise the whole
exercise would be pointless.

Many cameras in use by real programme makers don't have their linear matrixes
lined up properly because most people have no idea what this piece of circuitry
does, and don't have access to the necessary test gear. Most service manuals
make no reference to the matrix (or "masking" as the Japanese ones usually call
it) except to tell you to switch it off to line up the other circuits, and then
they don't always remember to tell you to switch it on again afterwards. The
chances of finding a real camera that is anywhere near any theoretical
specification are practically nill.

FACT: whether or not you have accurate NTSC phosphors -- the NTSC colors
are different from PAL.

This may be a fact on paper (and I have seen the specifications and fully
accept that it is a fact), but it is not a fact that has any practical
relevance in the real television programme-making and programme-exchanging
world.


The greens might be gorgeous, but are they "correct", in the sense of
accurately reproducing what was in front of the camera? Unless the entire
system, including the camera or telecine machine, has been held rigidly to
the same colorimetric specification, you cannot be sure of this.

Not quite: if your phosphor reaches deeper into NTSC green, then you can
get a better green. I have seen it, and it works.

Define "better" in the context of a display that is supposed to show a
reproduction of an original scene. If you use a display with the official NTSC
primaries to look at material that was photographed with a camera that used
different ones, then how are the brighter greens "better" if they don't
represent what the camera saw?

Rod.

"Roderick Stewart" wrote in message
om...
The greens might be gorgeous, but are they "correct", in the sense of
accurately reproducing what was in front of the camera? Unless the entire
system, including the camera or telecine machine, has been held rigidly to
the same colorimetric specification, you cannot be sure of this.

I'm not going to win any awards for this one, but what the hell ... ( and
for the
record, I'm not particularly addressing Roderick Stewart's post.)

The accuracy of the colors is fairly pointless as long as the picture is
pleasing. On
TV, the vast majority of programming presents unlikely characters depicted
in improbable
situations which are all neatly resolved in about 44 highly artificial
minutes, replete with
assorted processing that alters whatever was in front of the camera.

.... yet, for the sake of reality, some people are allowing themselves to
worry about the
color accuracy of the actors' makeup. Heck, it isn't even their skin you're
seeing,
or their clothes, or their walls, or their furniture or their food -- on and
on and on.

I like having good quality TV pictures. If I didn't, I would never have
given this
newsgroup a second look. On the other hand, I'd rather not have my reality
calibrated in nanometers.

There! I've vented. Now, chop me into dog food -- I'm sure I have it
coming.

In article [email protected],
"Sal M. Onella" writes:

"Roderick Stewart" wrote in message
om...
The greens might be gorgeous, but are they "correct", in the sense of
accurately reproducing what was in front of the camera? Unless the entire
system, including the camera or telecine machine, has been held rigidly to
the same colorimetric specification, you cannot be sure of this.

I'm not going to win any awards for this one, but what the hell ... ( and
for the
record, I'm not particularly addressing Roderick Stewart's post.)

The accuracy of the colors is fairly pointless as long as the picture is
pleasing. On
TV, the vast majority of programming presents unlikely characters depicted
in improbable
situations which are all neatly resolved in about 44 highly artificial
minutes, replete with
assorted processing that alters whatever was in front of the camera.

Note that one of my observations in the UK was that I saw MUCH MORE
of the 'green face' syndrome than I have recently seen in the US
while watching NTSC. The major conclusion that results from this
observation is that the color 'robustness' of PAL is relatively
less important than the studio setup. The old, claimed inferiority
of NTSC is mostly manifest on old tube equipment.

In a way, the results of the observation got slightly distracted,
(probably by me), where some of the so-called advantages of PAL are
of nil value nowadays. Even in the case where everything is perfect
in PAL and NTSC, then NTSC can provide better color.

I can agree that color accuracy isn't very important in non-critical
viewing, but there is a certain visual impact when the picture is
excellent. It probably doesn't make much of a difference in quality
of life, but alot of technical issues are alot like 'splitting hairs'.

John

On Mon, 18 Oct 2004 00:14:28 +0000 (UTC), (John S.
Dyson) wrote:

If a 'very good green' isn't in your gamut, you cannot display it. The
'Green' in the NTSC gamut is certainly better.

Looking at the spectral sensitivity curves for the three types of
cones in the eye, there is a considerable overlap between the various
sensitivity curves. For an acquisition system (such as video camera or
film), the spectral characteristic should be the same as in the human
vision system for each channel, i.e. peaks at the same wavelengths and
the same spectral width (slopes). For green, the peak is somewhere in
the 550-555 nm area (yellow-green).

Ideally, the channels should be connected directly to the
corresponding cones. However, even if the display only produced
monochromatic light at the three peak wavelengths, the other cones
will also be excited, so there is going to be some colour purity
problems.

Apparently the NTSC 1953 phosphors tried to avoid this by moving the
green phosphor to a shorter wavelength to minimise the red and blue
cone excitation. From the 2 degree CIE 1931 photopic standard observer
curve, a good place would be around 520 nm, with only 0.1 response for
red and blue. Since the green cone sensitivity has dropped to 0.8 at
this wavelength, the amplitude of the monochromatic phosphor 520 nm
radiation would have to be multiplied by 1.25.

Thus, provided that the original recording was done using the eye
spectral sensitivity curves (with green peak around 550 nm) and
displayed with three monochromatic lights (with green at 520 nm) the
colour purity could be improved.

However, I do not see anything that would make this NTSC specific, it
should work as well in PAL, SECAM, ITU-R Rec. BT.601 or the various
MPEG2 and HDTV formats, since it would require that the green
amplitude has to be increased at the _display_end_only, if nonstandard
displays with 520 nm green is used.

Paul

In message
John S. Dyson wrote:

The major conclusion that results from this observation is that the
color 'robustness' of PAL is relatively less important than the
studio setup.

The "color robustness" of PAL is an electrical property derived from
alternating the phase of the V axis subcarrier line by line.

What you perceive as the "better color" of NTSC is an optical
property of the colourimetry of the NTSC display phosphors.

Cheers,

Gareth.

--
http://www.rat.org.uk gareth at lightfox dot plus dot com