[opendtv] Re: 1080P Question
- From: Kilroy Hughes <Kilroy.Hughes@xxxxxxxxxxxxx>
- To: "opendtv@xxxxxxxxxxxxx" <opendtv@xxxxxxxxxxxxx>
- Date: Tue, 22 Sep 2009 20:50:51 +0000
Sounds like you are talking about inverse 3:2 FIELDS, which is mostly
invertible on 60 field systems (aside from lag needed for cadence detection you
mention). Deinterlacers in modern displays will use all sorts of tricks to
generate 60 or 120 FRAMES per second for display, including automatic 3:2
detection; but even when they get top and bottom fields from different sample
times (normal for interlaced capture), they'll generate a couple decent frames
using lots of tricks including motion estimation (to "reposition" odd and even
lines from time separated fields to synthesize a progressive frame), so there
isn't a different update rate, and the reduction in resolution isn't dramatic,
like switching from "weave" to "bob" with dumb deinterlacing.
3:2 FRAME repeat is a different kettle of fish, and particularly stinky.
There's no "mush frame" made from fields of two different time samples like 3:2
FIELD pull down, so you have one progressive frame cleanly exposed for 84ms and
the next for 126ms ... that creates a very accurate and annoying judder; not
obscured by overlapping fields, phosphor persistence, twitter filtering,
Gaussian splatter, etc.
So STBs that generate 60P output (mostly Blu-ray and HD DVD players) usually
use "deinterlacers" that either work on the 60 field/s 3:2 field signal they
all generate for legacy video connections, or do some more creative conversion
of 24 frames to 60 frames. Either way, the results aren't invertible by
throwing away repeat fields ... partly because there aren't any, but mostly
because the frames sent over the wire are synthesized/blended from the 24
originals, and no "cadence" remains to identify original frames even if they do
remain untouched. After "deinterlacing" you can't put 24P Humpty Dumpty back
together again in the display.
It is better for an STB playing 24P content to output at 30i with 3:2 pulldown
if the display or interface doesn't support a 24P signal. Most displays will
recognize the 3:2 FIELD pattern, chuck the repeat fields, and use the undamaged
24P in the display (in a variety of ways) to generate progressive images.
24P over the wire is best, but 30i with 3:2 fields is a close second. 120Hz
displays can jump between 4 or 5 repeats without much trouble if they get the
messy mix of sample rates typical of a broadcast stream (telecine intercut with
commercials, news, sports, etc.), but can lock on reliably at 5X 24P or 4X 30P
or 2X 60P for most disc and Internet content, which tends to be designed for
progressive content and progressive displays.
Kilroy Hughes
-----Original Message-----
From: opendtv-bounce@xxxxxxxxxxxxx [mailto:opendtv-bounce@xxxxxxxxxxxxx] On
Behalf Of Tom Barry
Sent: Tuesday, September 22, 2009 12:24 PM
To: opendtv@xxxxxxxxxxxxx
Subject: [opendtv] Re: 1080P Question
Kilroy Hughes wrote:
-- quote --
PS. IMO, Devices that "deinterlace" 24P to output 60P should be recycled ASAP.
Forcing the display to guess whether the real sample rate was 42ms or 16 ms and
interpolate whatever mixed frames that came from the STB deinterlacer for 120Hz
update (vs. just starting with a clean 24P signal) is very bad system design,
and one reason display interpolation often looks bad.
---
I have written commercial code to do this, not counting the in-between
frames. On broadcast TV it may have to deal with source material
switching back and forth from interlaced to telecined fairly rapidly,
not even counting during commercial breaks. It can still be done more
or less reliably except for a couple frames at the transitions.
I wonder how fast an LCD multisync TV could change sync rates? It
might make for some interesting visual effects. The various code I've
worked on was still outputting frames (or progressive duplicates) at 60
Hz. I've never really considered the problem of delaying the
presentation time to 24 Hz. multiples after identifying telecined
material. Doing it accurately at 120 Hz still has many of the same
problems if you really want judder free smooth motion. But do-able.
Though I'm still in favor of giving all TV's the processing power to fix
all the legacy conveniences and sins of the broadcast industry.
Processing power is getting pretty cheap these days.
- Tom
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
