[opendtv] Re: ClearLCD adaptive dual-pulse

  • From: Jeroen Stessen <jeroen.stessen@xxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Wed, 20 Dec 2006 09:43:18 +0100


> *"Adam M. Costello" <opendtv.amc+0+@xxxxxxxxxxxx>*
> Sorry if it reappears and you get both.]

I got both, but I can not reply from home now. I am on leave because
we have a newborn baby girl. I'll also use my vacation to prepare
my presentation on wide color gamut for ***The Tech Retreat***.

> Mr. Stessen, thank you for your very informative posts to the opendtv
> list.  I am intrigued by ClearLCD, and especially by a tantalizing
> remark you made a while back that we can "have our cake and eat it too",
> referring to the tradeoff between 60 Hz flicker and 120 Hz judder.

Okay, but 120 Hz (frame repetition) is not perceived as judder, it
is perceived as a double image. And a double backlight pulse is
equivalent to frame repetition, so it is not the desired mode of
operation for moving images.

> I'm not in the business, just an interested consumer who is hesitant
> to "upgrade" my CRT TV to a display technology with superior spatial
> resolution but inferior motion portrayal. 

Inferior to CRT TV, but still far superior to the movie theater...
That is, if at least the frame rate is increased to 50 or 60 Hz.
Then the further improvement (to movies) by flashing the image,
or by further increase of the frame rate, is modest. It is very
noticeable on certain selected images, but not quite so visible
on movies that were shot with a long shutter time.

> ClearLCD is the only non-CRT TV technology I've heard of that 
 > even attempts to retain the impulse characteristic of CRTs.

Well, some plasmas come reasonably close, but not quite, and
OLED could do it, and SED certainly promises to do it too.
And then there are many variants of motion-improved LCDs:
scanning backlight, blinking backlight, black field insertion,
grey field insertion, and increased frame rate (to 120 Hz).
The latter development is happening very fast, because it is
the only solution that does not create a flicker problem.

ClearLCD for Europe is a combination of frame rate conversion
to 75 Hz and a scanning backlight, but it can also be implemented
by 75 Hz and black field insertion. 75 Hz is high enough to
suppress flicker, and low enough for existing signal processing.
For USA 75 Hz is not a good choice, and 120 Hz is still out of
reach, so that is where the adaptive backlight comes in as an
easy solution. It does not require any change to the signal
processing or the display panel, but it does require a different
backlight technology (for higher peak power) and is thus a big
inconvenience to the rigid supply chain.

> (Is anyone developing scanning LED backlights?
> I haven't found anything on the web beyond speculation.)

It is not a preferred choice for LEDs, because the required
peak power (i.e. average power divided by the duty cycle)
makes the LED backlight too expensive. With HCFL the peak
power reserve came naturally, so it was an easy choice.
This reserve can also be exploited for backlight boosting,
as I have explained in our paper 26.4 on the SID 2006.

> I guess ClearLCD's solution to the flicker/judder problem is related to
> something called "adaptive dual-pulse" backlighting, but I have been
> unable to find any details about it.  Can you (yet) tell us anything
> about that?

Have you read paper 54.4 from the same SID 2006 ?! It was
presented by Pierre de Greef, who is now with NXP. Pierre
has also published an on-line paper in Journal of the SID.

> From the name, it sounds like the backlight switches between 60 Hz
> scanning (one pulse per refresh) and 120 Hz scanning (two pulses per
> refresh, 180 degrees out of phase, so that two horizontal stripes are
> illuminated at a time, 1/2 screen height apart), depending on whether
> the video currently contains any moving objects that the eye might
> track. 

Not "switching". It is a continuous transition. And in
principle it can be done for each of the lamps individually.
You'll find my name near the bottom of the list on a recent
patent that goes into the details. Wait until it comes out...

> When switching from 60 Hz to 120 Hz, does the duty cycle remain
> the same (the pulse width is cut in half), or does the pulse width
> remain the same (the duty cycle doubles, and the lamp is dimmed to
> 1/2 intensity)? 

The former, because in general analog dimming of fluorecent
tubes is not so popular. Even more so for CCFL lamps, analog
dimming leads to non-uniformity over the length of the tube.

> What is the pulse width (in milliseconds)? 

At 100% nominal output, the pulse width is approx. 35% of the
frame period of 17 ms, so it's approx. 6 ms. Less when dimmed.

> When the backlight is dimmed for dark scenes, is that accomplished 
 > by reducing the duty cycle or the intensity?

I am so far removed from the end product that I can not make
any statements with certainty. But based on the above, you can

> If the pulse width is 4ms (I'm guessing based on various imprecise
> statements I've found), that's a lot better than a full frame time of
> 17ms, but still a far cry from a CRT, where the phosphors have largely
> decayed after 0.1 ms and almost completely decayed after 1ms (right?).

Reasonably correct, yes, but you are assuming that less is always
better. The same assumption is generally made for the response
time (2 ms is better than 4 ms, etc.), but due to the sample-and-hold
effect that this discussion is about, a shorter response time is
irrelevant until you introduce a shorter on-time first.
Then, making rhe on-time shorter is only relevant to the point where
the motion smear due to integration on the retina of the tracking
eye becomes insignificant compared to the motion blur due to
integration of moving objects on the light sensitive surface of
the camera. Because the typical camera uses a long shutter time
(I think typically between 5 and 16 ms), the images themselves are
already blurred. That blur does not go away with better displays.
So there is a point where the law of diminishing returns applies.

> If something moves the full width of the screen in two seconds (which I
> wouldn't call especially fast), that's one pixel per ms, so a 4ms pulse
> would blur away about 3/4 of the horizontal resolution of the moving
> object (versus 16/17 for an always-on backlight, versus almost no blur
> for a CRT).  Is work underway to further reduce the duty cycle?

See above. Further reduction is only relevant for moving images
that are sharper than what the typical movie producer delivers.
But then you'll also get strobing artefacts, so I doubt that you
can convince Mr. Spielberg to go further than in Private Ryan.
Cinematographers use motion blur and camera tracking, and depth of
focus, as a tool for focussing the viewers' attention on foreground
objects. We don't want to deny them the use of their own tools !
So, for the typical moving images, you would have to judge for
yourself how large the gap is between the best LCD and a CRT,
and whether it is worth to further reduce this gap. I think that
with an on-time of 8 ms, the improvement is already the
proverbial 80% (of the 80-20 rule). This is probably why we are
seeing an explosion of 120 Hz panels (being offered by the panel
makers to the setmakers), with static backlights. The signs were
all there on the SID, and will probably also be at the next CES.

> By the way, why is 60 Hz flicker a problem?  I've been watching NTSC TV
> all my life, and never noticed a flicker (and still don't).  But I can
> see the flicker on a CRT computer monitor set to 60 Hz refresh.  What's
> the difference?

Probably viewing angle. Your peripheral vision is much more
sensitive to flicker than your central vision, and you are
sitting very close to a monitor. Possibly also (peak) brightness,
though monitors are usually dimmer than TVs. And content...
Good content can make you forget about all the artefacts.
Especially moving content.

-- Jeroen

| From:     Jeroen H. Stessen   | E-mail:  Jeroen.Stessen@xxxxxxxxxxx |
| Building: SFJ-5.22 Eindhoven  | Deptmt.: Philips Applied Technologies |
| Phone:    ++31.40.2732739     | Visiting & mail address: Glaslaan 2 |
| Mobile:   ++31.6.44680021     | NL 5616 LW Eindhoven, the Netherlands |
| Skype:  callto:jeroen.stessen | Website: http://www.apptech.philips.com/ 

Other related posts: