The end of the mass produced CRT

In article , Jukka Aho
writes
Kennedy McEwen wrote:

With 1366x768 completely independent pixels, any blockiness produced
in the image comes from the source, not the display

Or it might come from the completely different gamma characteristics
of the LCD display. Or it might come from the penny-pinching design
where the individual colour channels are (possibly) internally
processed only with a 6-bit depth - as is the case with many LCD
panels - leading to banding and shades of a colour collapsing into a
single shade.

No, it wouldn't come from those, since these affect levels, not
pixels or blocks.

The above-mentioned technicalities affect levels, which in turn affect
how visible the edges of the DCT blocks will be to the viewer - that
is, how big a jump there is from one shade to the other. If these jumps
are finer on a CRT and coarser on certain types of LCD panels, of
course the DCT blocks will stick out (more) like the proverbial sore
finger on an LCD screen while the viewer with a CRT-based set will
notice less of that when watching the same signal.

If the number of levels is so limited that the edges between blocks
become more visible, then the detail within the blocks would also appear
more posterised. It doesn't, so your argument is baseless.

The fact that the limited bitrate in the MPEG stream results in
blocks with fewer levels is actually working in favour of a low bit
depth technology.

Since MPEG quantization is done in the frequency domain, the quilting
caused by low bitrates is not exactly the same thing as "blocks with
fewer levels". The blocks may have a fine continuous range of levels
within, but given a sufficiently low bitrate, the edge pixels and
shades don't necessarily align too well with the adjacent blocks. If
the display device, for some reason, reduces the decoded colour channel
values to a lesser bit-depth (after those values have been converted to
the RGB space for display), it is very likely that this will make these
kind of edge differences more visible.

For that to happen, the display would also show serious posterisation
within each block since, as you correctly note, the blocks "have a fine
range of levels within". In that case, the discontinuities between
blocks would be less of a problem than the limited number of levels
within them. That isn't the case.

Also, since LCDs are inherently much more linear than CRTs and the
gamma encoding must be implemented before display, 6-bits would result
in the equivalent of only 3 or less bits on a natural gamma CRT. Since
the displays are clearly much better than this your argument that they
only use 6-bits is complete crap.

See http://google.com/search?q=6-bit+lcd for lots of articles about
6-bit LCDs.

Most of which appear to be either completely naive and misleading or
referring to the input resolution, not the internal precision, which
requires to be much higher due to the gamma simulation.

It's actually easy to test this with greyscale and colour ramp test
images. If you see banding that is not visible on a CRT screen, surely
there is some sort of bit-depth reduction going on in the panel.

And if you don't (which you don't on the better LCDs) then by your
assessment there isn't.

However it isn't that simple - LCDs are close to linear display devices,
not perfectly so, but close enough for this discussion. CRTs are very
non-linear, with an intrinsic gamma of around 2.5 depending on the
phosphors and drive circuits. That is, the intensity of light they
produce is of the form I = (V + o)^g, where g is the gamma and o is an
offset pedestal level. By sheer coincidence, because it was never
planned that way (CRTs were just the only realistic display technology
at the time), the natural gamma of the CRT is a close match to the
inverse of the eye's logarithmic response to light in most of the region
of interest for normal display. Hence evenly spaced quantisation of the
video signal (which is gamma compensated) results in evenly spaced
brightness steps being perceived on the display, even though the
intensity difference between steps in the dark regions are much less
than those in the light regions. For example, in 6-bit video the
intensity difference between the highlight steps on an ideal (perfect
eye matching) CRT is roughly 1-(62/63)^2.5 = 4% of the peak white level.
For the extreme shadows it is roughly (1/63)^2.5 = 0.003%, about 1000
times less, yet the steps appear pretty close to evenly spaced due
purely to the eye response.

LCDs don't have any significant gamma, but they still have to simulate
it, not only because they are provided with a gamma compensated input in
the form of the video signal, but because they must also approximate
inverse response of the eye to stimulation with light. To achieve that
0.003% spacing of the black levels on a linear display technology
requires something close to 15-bits of internal processing precision. If
it were only 6-bits, as you and the articles your search suggested, then
black quantisation steps would only be around 1.5% and not only easily
visible, but highly objectionable. So 15-bits of internal accuracy are
required merely to reproduce the gamma curve of the ideal CRT
sufficiently well to avoid posterisation in the blacks with just a 6-bit
input to a linear display technology such as LCDs, and a lot more for
8-bit video.

There is no "bit-depth" reduction going on, quite the opposite: extra
precision is required to compensate for the eye response, which the CRT
approximates reasonable well.

However, even the CRT response departs significantly from the inverse of
the eye response in the blacks, and this is addressed in the video
encoding scheme by the application of a toe level, below which the gamma
is fixed at 1. In the digital video standards this toe corresponds to
8.1% of the peak video level, or around level 5 if 6-bit video
quantisation was used. The toe reduces the number of bits required by
linear displays, since they only need to simulate the gamma down to the
toe level and their own natural linearity for lower levels. I'll leave
you to work out that the result requires 10-bit precision of the LCD
brightness to reproduce even a 6-bit video signal accurately.

The better displays do use more than 10-bits, but your 6-bit internal
argument is complete rubbish. A lot do use 6-bit *input* signals, as do
many high quality digital CRT circuits, which is perfectly adequate and
certainly wouldn't give rise to noticeable posterisation, let alone the
blockiness that you are objecting to - that comes from having adequate
resolution (well beyond the video signal bandwidth) to clearly see the
block edges caused by inadequate bit-rate coding in the first place.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's ****ed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

"charles" wrote in message
...
In article , Ivan
wrote:

"Graham" wrote in message
...



Only this week I heard an item on Radio 5 about the recent
(worldwide) growth of organised criminal gangs, who are stealing
millions of pounds worth of copper wire and sometimes disabling whole
railway systems

I heard that item on Radio 4.

Radio 4, isn't that the wireless station that posh people listen to?...
the sort of people who think that 'sex' is something that coal is
delivered in)

Only if they come from Mairningsade (Morningside)

Shurley you mean Mornington Cresent?
--

Graham

%Profound_observation%

Thanks for the ideas, folks.
Turps

In article ,
Graham wrote:

"charles" wrote in message
...
In article , Ivan
wrote:

"Graham" wrote in message
...



Only this week I heard an item on Radio 5 about the recent
(worldwide) growth of organised criminal gangs, who are stealing
millions of pounds worth of copper wire and sometimes disabling whole
railway systems

I heard that item on Radio 4.

Radio 4, isn't that the wireless station that posh people listen to?...
the sort of people who think that 'sex' is something that coal is
delivered in)

Only if they come from Mairningsade (Morningside)

Shurley you mean Mornington Cresent?

Nope, Morningside, Edinburgh

--
From KT24 - in "Leafy Surrey"

Using a RISC OS computer running v5.11

"charles" wrote in message
...
In article ,
Graham wrote:

"charles" wrote in message
...
In article , Ivan
wrote:

"Graham" wrote in message
...



Only this week I heard an item on Radio 5 about the recent
(worldwide) growth of organised criminal gangs, who are stealing
millions of pounds worth of copper wire and sometimes disabling
whole
railway systems

I heard that item on Radio 4.

Radio 4, isn't that the wireless station that posh people listen
to?...
the sort of people who think that 'sex' is something that coal is
delivered in)

Only if they come from Mairningsade (Morningside)

Shurley you mean Mornington Cresent?

Nope, Morningside, Edinburgh

Ay, You've had your tea?
--

Graham.
%Profound_observation%

Graham wrote:
"charles" wrote in message

Nope, Morningside, Edinburgh

Ay, You've had your tea?

The version as I have it (from a totally unbiased source in Glasgow)
is:

"Come away in. You'll have had your tea ...".

André Coutanche

"tony sayer" wrote in message
...
In article , Andy Wade
writes
Laurence Taylor wrote:

Hands up those who remember "Its the tube that makes the colour"

One hand here! (Only just). Be nice to see it again, in fact.

And who remembers "clean and prime, coat, wrap and lay" and "switch-off,
isolate, dump, earth" ...?


Old whatsisname coming back form the dentist ..humm..better see if
wossisname has reversed the coil properly, where's that earth stick
before I touch it !!BANG!!!

Whoops got the cap reversed instead..

Yes like it was yesterday

And, much more recently, the BBC windmills film in the early days of DTT.


There was also a long sequence of material sourced from VT, TC and stills
that was played out during the early NICAM tests, to determine what
artifacts were present.
--

Graham.
%Profound_observation%

In article ,
Graham wrote:

Only if they come from Mairningsade (Morningside)

Shurley you mean Mornington Cresent?

Nope, Morningside, Edinburgh

Ay, You've had your tea?

I understand the current question is (as the bottle goes round) "But,
you'll be driving?"

--
From KT24 - in "Leafy Surrey"

Using a RISC OS computer running v5.11

In article ,
André Coutanche wrote:
Graham wrote:
"charles" wrote in message

Nope, Morningside, Edinburgh

Ay, You've had your tea?

The version as I have it (from a totally unbiased source in Glasgow)
is:

"Come away in. You'll have had your tea ...".

and (as told by Stanley Baxter): the hospitality is Aberdeen is quite
different. They show you into the front room where the table is groaning
with food, beautifully cooked - and very reasonably priced.

André Coutanche

--
From KT24 - in "Leafy Surrey"

Using a RISC OS computer running v5.11

On Fri, 8 Dec 2006 23:03:08 -0000, "Graham" wrote:

And, much more recently, the BBC windmills film in the early days of DTT.


There was also a long sequence of material sourced from VT, TC and stills
that was played out during the early NICAM tests, to determine what
artifacts were present.

None that I can recall. We didn't really know what artefacts were
until the start of digital transmission, ironically the system that
was supposed to eliminate them.

I think it's probably the result of trying to be too clever and
inventing a system wherein the bandwidth saving (i.e. bit-rate
reduction) is picture-related. The inevitable consequence is that the
problems are picture-related too, so whatever any measurements with
instruments might say, they are far more subjectively annoying than
most analogue artefacts which are unrelated to picture content and can
be mentally tuned out.

Rod.