[opendtv] Re: Half Truths - Was More 1080p@60
- From: Jeroen Stessen <jeroen.stessen@xxxxxxxxxxx>
- To: opendtv@xxxxxxxxxxxxx
- Date: Fri, 7 Dec 2007 09:24:06 +0100
Hello,
Tom Barry:
> For that matter, over maybe 60-70 hz I'm not even sure the display
> refresh rate even has to be an even multiple of the source rate.
Yes it does, at least until you get in the hundreds of frames per
second. Which is a typical (sub-)frame rate for a color-sequential
(single panel) projector, to keep the color breakup acceptable.
So you may see those running asynchronously. Our late LCoS projector
ran at 540 Hz, IIRC. Which is very very fast for an LCD.
> At some speed judder probably becomes invisible anyway on a
non-flickering
> fixed pixel progressive display. The error in presentation time is less
> than half the display refresh time, just by rounding to the nearest time
> stamp.
It's more than that. If I understand you correctly, you are saying
that frame repetition (nearest neighbour up-sampling) is sufficient
once the frame rate gets over 60-70 fps, right ? But we already know
that this is not true, because of the artefacts from 3:2 pull-down
when we repeat frames from 24 to 60 fps. Because of the unequal
repetition factors, the base frequency of the motion judder becomes
12 Hz. Which is very low, therefore very visible. Also the width of
the repeated edges as they are projected on the retina of the
tracking eye is twice as much as what it could be, if for example
we had used 3:3 pull-down and converted to 72 fps. So it is not so
much the absolute frame rate that matters, but the ratio of the
output frame rate to the input frame rate. If that is an integer
then the frequency of the motion judder equals the input frame rate,
which is the highest that you can get from frame repetition.
Of course, true motion-compensated up-conversion is far superior to
frame repetition. And if it is done perfectly then indeed the output
frame rate does not matter anymore. (But see below.)
> Though I don't know of any studies to see where smooth slow motion
> actually appears continuous at increasing (non-multiple) refresh rates.
When using only frame repetition it will always judder like hell,
because the judder frequency will never be higher than the input
frame rate. Which is presumably a lot lower than 60 Hz in case of
slow motion. It's more a sequence of still images then.
Cliff Benham:
> Not making the new display refresh rate an even multiple of the
> source rate means lots more processing is necessary which can make for
> expensive hardware.
Any multiple would do to make it simple, e.g. x3 is fine too.
But there is one problem with non-integer up-conversion rates...
With an integer ratio, say xN, we can display 1 original frame and
N-1 interpolated frames. For N=2 this means that 50% of the output
frames still have the original quality. The other frames will suffer
from various processing errors, like halos around occlusion areas,
and de-interlacing and scaling artefacts for the sub-pixel phase
shifting. With a non-integer ratio, say xN/M, meaning that M input
frames are converted into N output frames, only 1 in N output
frames equals an original (1 in M) input frame. Like, when we still
made 75 Hz displays (the first PixelPlus CRTs, and ClearLCD with a
scanning backlight), and we converted 50 fps video to 75 fps output,
then only 25 frames per second are originals, i.e. one in three.
The other 50 must be interpolated. It is hard to do it perfectly.
So that is why, whenever possible, we stick with integer up-
conversion factors, and why we are so happy with LCDs capable of
100 and 120 fps. These are now suddenly readily available. :-)
Of course that does mean that we have to spit out processed pixels
at a rate of 250-300 MHz, over quad LVDS links. The new PNX5100 is
doing a mighty fine job of that, and all in TriMedia software too !
This will be (in 2008) a must-have TV, at any price. I promise !
Greetings,
-- Jeroen
+-------------------------------------------+------------------------------------+
| From: Jeroen H. Stessen | Phone: ++31.40.27.99650
|
| Deptmt.: Philips Consumer Electronics | Mobile1: ++31.6.1389.2492
|
| Innovation Lab Eindhoven | Mobile2: ++31.6.4468.0021
|
| Address: High Tech Campus 37 - room 8.042 | Skype: Jeroen.Stessen |
| 5656 AE Eindhoven - Nederland | VoIP:
Jeroen.Stessen.at.Philips |
| Website: http://www.consumer.philips.com/ | E-mail:
Jeroen.Stessen@xxxxxxxxxxx |
+-------------------------------------------+------------------------------------+
Other related posts:
