ABC and ESPN HD only 720P - Why not 1080I? - Page 4

Spectrum folding (not sure where foldback came from) was discussed a lot in papers I read extensively to do with digital standards conversion a good few years ago.

One paper I particularly remember reading - which may be of interest is here : http://www.bbc.co.uk/rd/publications..._1984_20.shtml

It doesn't specifically use the phrase spectrum folding - though others I read (can't find them online) did. (They used it in relation to non-ideal interlaced systems - such as non-anti aliased cap gens causing issues. Actually - here is one discussing "spectral overlap" http://downloads.bbc.co.uk/rd/pubs/reports/1984-14.pdf Fig 8 (iii) shows the issue quite nicely )

The quoted paper pre-dates CCD broadcast cameras (and probably is mainly concentrated on non-CCD telecines as well) - and is very much based on 70s and 80s "real world" sources rather than what we have today.

However the discussion about the diagonal vertical vs temporal graph (Fig 29) that is required for interlaced content (rather than the square variant that would be allowed for progressive - then called sequential) shows the issue. If you are outside the triangle section of that ideal filter - you are in the realms of ambiguity in an interlaced signal (you simply can't absolutely know whether the information is vertical temporal or vertical spatial information)

It is a good 20 years since I put my Fourier and Laplace stuff to good work - and that paper is 25+ years old. However the fundamentals are sound.


The basic fundamental spectrum folding is fine detail flickering when content isn't vertically filtered (usually where the source is film or progressive video). That is where vertical spatial information has been folded into the temporal domain - and thus appears to flash/flicker (i.e. is altering in the temporal domain) rather than appearing static.

Similarly - the converse is true. Content moving vertically at certain speeds can appear to be 'distorted' in vertical resolution terms if not properly filtered. If you imagine a picture made up of fine black and white alternating lines. If this moves up the screen at certain speeds at a certain resolution, you might only ever sample the black lines or the white lines. The temporal information has distorted the spatial. Or if a similar image was static, you might sample all the black lines in one field, and all the white lines in the next. You'd therefore end up with an image that flashed black and white - the worst case scenario where the vertical spectrum has entirely folded into the temporal. If you pre-filtered by line-pairing with a line offset, you'd end up with a continuous grey field - so would have lost vertical resolution, but gracefully, and without folding into the temporal spectrum. Obviously worst case - but it illustrates the point?

Whether there is flicker on the de-interlaced image displayed progressively will depend on the de-interlacing approach - and assumptions as to the source - but these are assumptions. (AFAIK there is no 25pSF/30pSF vs 50/60i source signalling in use these days - PALPlus included it to improve chroma decoding - though MPEG2/H264 encoders may make assumptions in their encoding streams they aren't explicitly signalled in the HD-SDI sources AFAIK - and would break-down where pSF and i sources are on-screen at once)

^^^Appreciate the added details/sources. Also discovered Googling couldn't find anything (video "spectrum foldback") but Altavista.com coughed up that academic paper (about anti-aliasing) cited earlier when I substituted 'fold back'. The paper has lots of images, such as a brick wall. And that rings a bell with current HD or SD images showing distant citiscapes or walls and sometimes (right focal plane)--perhaps with vertical motion-- when distant closely spaced horizontal lines appear to flash; that's temporal, clearly, but hard to grasp without mentioning flashing. Recall a slow upward pan over a jagged cliff face in a feature movie, years back, when the entire cliff was flashing strongly. Not sure, but such flashing seems rarer with my current pro-model progressive-imaging plasma compared to my former interlaced CRT RPTV. -- John

Yep - the chances are that the de-interlacer in your plasma is using a different algorithm to that used by the interlaced CRT + eye/brain combination!

You concern me.

Wouldn't the 60 updates per second of 1080i be better for rapid motion video (like sports)?
Complete and utter BS. The standards were picked by the US FCC which determined a 6 megahertz TV channel could transmit either a progressive or high resolution, but not both at the same time.

So you've spun this elaborate story about evil suppliers, when it was actually the Government and the limits of the technology which mandated either 720p or 1080i. The high res progressive 1080 simply wasn't possible at the time.

1080i is 60 half frames a second which de-interlace to 30 full frames a seccond wheras 720p is 60 full frames a second which is why sports channels like ABC and ESPN use 720p. See the following link:

http://www.hdtvprimer.com/ISSUES/what_is_ATSC.html

I really wish people would stop saying that 1080i deinterlaces to 30 frames per second. Any normal display or set-top box will deinterlace to 60Hz, with techniques ranging from a dumb bob (each field is scaled to become a frame) to intelligent method that pass through static areas and use motion adaptive techniques and adjacent fields to reconstruct the missing data on moving areas.

Oh. I thought 720p was 720p/30. My mistake. So then 1080i really is more suited to still images, just like Japan's old 1088i MUSE TVs did better with still or slow-moving video versus rapid-paced video (like sports).

(shrug) I watch my HDTV on an NTSC so it all looks Greek.... I mean, 480i to me.

In theory, yes. In the real world, maybe, maybe not.

On still images with an adaptive deinterlacer, 1080i is superior. On 24p content carried in a 60Hz format through pulldown or frame duplication when used with a display that can reverse the pulldown, 1080i is superior. On high motion 60Hz content, 720p held a slight theoretical edge, but this has mostly disappeared as the quality of deinterlacers has increased. At this point, picture quality depends a whole lot more on the infrastructure than the choice of picture format. CBS would still look great at 720p, and ESPN would still be a crapshoot at 1080i.

Alright, you guys seem to know your stuff, so I'll ask here.
This is my tv http://www.sonystyle.com/webapp/wcs/...specifications

Max resolution is 1920x1080. If I wanted to set my HD device such as my TiVo or Moxi to one resolution, which one should I choose? 720p or 1080i ? Which is the best resolution for my specific TV?

1080i should look best on your set. Try each and judge which you like for yourself. Why not just feed it 720p and 1080i and let the TV do the scaling?

I would but when, for example i switch from channel to channel the screen blanks out for a second while it switches to the other feed. Its annoying. lol

I mean if I could use 1080p full time I would, that would be best of both worlds, but the TiVo doesn't upscale to that, unless the program source is natively that. And the Moxi DVR is having some software issues with its chipset which doesn't correctly display 1080p on alot of tv's via HDMI.

Assuming the scaler in your TV is better than the scaler in your Tivo, you would get the best possible video quality by sending both 720p and 1080i natively to your TV and letting the Sony do the upconversion. As you have noticed, switching resolutions takes a half second to get signal lock and is kind of annoying.

In practice, having the Tivo output everything as 1080i works pretty darned well.

*hugs* As I found, my exact issue was detailed in a thread about my model series on this forum, I'll link to it now so it'd help others in the same scenario.
http://www.avsforum.com/avs-vb/showt...227927&page=97

Also I found a "full pixel" mode on my display that only shows up when using 1080i, and its improved the picture a good 30%, which was already 90% awesome, so I think I'm overwhelmed with the wow factor.

Is that the same as a 1-to-1 mapping of the Video Pixels to the LCD pixels? (i.e. no scaling)

I'm assuming so.
For example under normal mode, the fox bug is touching the bottom of the screen, but when I put it to full pixel mode, I have more screen real estate.
The user manual says this under Display Area

Full Pixel: Displays pictures in their original size when parts of the picture are cut off.

+1: Displays pictures in their original size.

Normal: Displays pictures in their recommended size.

-1: Enlarges the picture so that the edges are outside the visible display area.

As you say, it also disappeared as bit rates dropped. If you step through the frames of something in motion of any 15 mbps 720p station, you won't see the crystal clear motion that proscan people promised. You'll see macroblocking and mosquito noise from overcompression.

This wasn't the case when our ABC affiliate throw their whole bandwidth into their 720p HD stream.

Technically that's how pictures are supposed to be. Just like with analog NTSC, the 5% edge of the picture is not meant to be seen. It's the overscan area.

Actually it means that area of the picture may or may not be seen, so the important stuff should be kept within the safe area. Stuff that is not meant to be seen should not be anywhere in the frame.

but it can't hurt, right?

It is not that the content of the overscan is not meant to be seen. It is that none of the movies contain or video contain any meaningfull content or text in the outer 10% of the video content since it is assumed that the TV will overscan the imput.

Sorry - that is NOT the case...

1080i30 aka 1080/60i native cameras capture 60 images per second, convert them into 60 half-vertical resolution fields with a frame-line offset (as discussed earlier in the thread). These are then either displayed in sequence on an interlaced CRT (and a few native interlaced plasmas) - as 60 half-resolution fields for the eye/brain to merge - or they are de-interlaced to a 60 frames per second progressive image (for most LCD, DLP and Plasmas and some progressive CRTs) to display.

No decent display will de-interlace to 30 progressive frames per second - it would look lousy and kill the motion / temporal resolution of the source.

On a progressive display 1080/60i and 720/60p are both running at 60Hz refresh rates and will be displaying 60 progressive frames per second. These 60 progressive frames will still be different from each other and carry proper motion information at 60Hz. The interlaced source will need to be de-interlaced to create these 60 progressive frames, the progressive source will not need de-interlacing. As de-interlacing is an imperfect science there is scope for a 1080/60i signal to de-interlace in a way that doesn't perfectly reflect the original scene.

(Some very cheap and nasty solutions, particularly PCTV cards and some PC DVD player apps - MAY do a de-interlace from a 60Hz native interlaced source to 30 frames per second progressive - but they are not proper solutions...)

I completly agree with your post
I should never have infered that all TVs de-interlace 1080i content to 30 fps content.

For folks with tiny <60" televisions 720p might be acceptable but larger screens really emphasis the shortcomings of 720. ABS and FOX pictures look sad on a big screen home theater. Like looking through dirty contact lenses.