[opendtv] Re: RGB mania ('translated' for Prin)

  • From: jeroen.stessen@xxxxxxxxxxx
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Mon, 17 Jan 2005 09:39:33 +0100


Bert Manfredi wrote: 
> This is a fascinating thread. If I understand
> what you're saying, a normal TV transmission
> assigns Y values of, say, 235, both to large
> areas of normal white and to tiny highlights
> which should be super-white. 

This is not entirely correct. The target value for a white 
highlight may well be 235, but because of the ringing in 
the anti-aliasing filter on such a small transient the peak 
value will rise above 235, up to the highest value allowed 
(i.e. 254 in ITU-R.601, because 255 is reserved for sync). 

Thinking digital is okay, but keep in mind the analog world... 

> And because CRTs react to Y values in strange ways, 
> they will naturally blast a very powerful beam on a
> tiny area identified as Y 235, to render the specular 
> highlight expected there, 

This is called "punch", and it is thought to be a positive 
quality of CRTs, and in fact of most other emissive displays. 
It is specifically not a natural quality of light valve 
displays. In CRT this occurs because for large bright areas 
the average power limiter takes precedence over the peak 
drive limiter. So images that contain large bright areas are 
limited, to perhaps 20-25% peak brightness, whereas images 
that contain only specular highlights on a dark background 
are not limited. The average power limiter is added to 
prevent overload of the EHT power supply and to safeguard 
the life time of the picture tube and limit the temperature. 
As a side effect it limits the blinding effect of an all-
white scene, limiting it to e.g. 250 cd/m2. But in fact we 
appreciate this feature for its opposite effect, that the 
highlights in a dark scene can reach 1000 cd/m2 or so ! 

> even without having been told specifically that
> this tiny area was actually brighter than
> the large areas marked Y 235.

One big IF.... the behaviour as described above ONLY 
happens if the user contrast has been turned up high 
enough for the limiters to activate. Joe Kane advises you 
to turn down the contrast so far that the peak brightness 
is between 100-200 cd/m2, and then neither the peak drive 
limiter nor the average power limiter will EVER be active 
anymore. Then the peak brightness will be proportional to 
the video signal, as it was probably intended !! 

Activation of the peak drive limiter (which acts like an 
automatic gain circuit) causes the nasty effect of 
"pumping": the appearance of a subtitle can cause the 
brightness of the entire scene to be reduced. This is 
most annoying because a local cause has a global effect. 
(Our patents on "contrast reserve" aim to avoid that.) 
Activation of the average power limiter is less annoying, 
because a high average power level is by definition a 
global cause, and it has a global effect. This is natural. 
It is even appreciated that large bright scenes are 
reduced below the level of being blinding. 

> So I'm trying to see how an LCD would react to this, 

A snow scene on a typical LCD, with 400-800 cd/m2 peak 
brightness, is blinding ! Not so on a plasma display, 
because this is severely average power limited. 

> to see if I can understand why it was so much harder 
> to adjust my LCD than it is to adjust a CRT TV.

This could have several other causes as well, e.g. the 
black level of many LCDs is not very good due to light 
leakage. The black rarely goes below 0.5 cd/m2, unless 
the backlight is dimmed appropriately. It is not always 
easy to see if and how you can dim the backlight. 

> An LCD would paint the specular highlights
> with the same intensity as the larger normal
> white regions. 

Except for that effect of ringing of the AA filter. 

> So perhaps if one sets the screen to look right for 
> specular highlights, the effect is that all the 
> white or brighter regions of the image look too pushed. 

It would be similar to a CRT set to below the limiting 
levels, as per advice of Joe Kane and the ISF, except 
that it all happens at much higher brightness levels. 

> It's especially noticeable on b&w images.

That's a different aspect: in order to get maximum 
efficiency out of a light valve it is tempting to drive it 
up to the clipping level. The whites will look crushed. 
The CRT has no real hard clipping mechanism, and even if 
it had then the two limiters will usually keep it well 
below the clipping limit. 

> Which makes one want to back off on the
> "contrast" control (i.e. white level), which
> in turn causes the darker parts of the
> scene to fade to black too fast.

With LCDs... the familiar "contrast" and "brightness" 
controls are sometimes connected to different circuits. 
Do not always expect the familiar behaviour ! 

> If this reasoning is about right, it would
> describe what I was seeing. My reaction was
> to back off on the "contrast" and turn up
> the "brightness" (i.e. black level), to
> bring up the darker parts of the image.

That is a correct reaction. Your contrast ratio will be 
even worse, but you will see more details in the blacks. 
It has to do with the differential gain of the display 
(defined by its gamma curve) and the differential gain of 
your eyes (defined by a similar inverse gamma curve). 

Best regards, 
-- Jeroen 

| From:     Jeroen H. Stessen   | E-mail:  Jeroen.Stessen@xxxxxxxxxxx    |
| Building: SFJ-5.22 Eindhoven  | PHILIPS Applied Technologies           |
| Phone:    ++31.40.2732739     | Visiting & mail address: Glaslaan 2    |
| Mobile:   ++31.6.44680021     | NL 5616 LW Eindhoven, the Netherlands  |
| Pager:    ++31.6.65133818     | Website: http://www.pdsl.philips.com/  |

You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at 

- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: