Why I haven't taken the plunge -yet. My take on all this. (1 Viewer)

[Cees Alons]

David,

On standard TV sets, it's virtually impossible to mix up the odd and even fields, because of the frame sync pulse. In fact, the image would be really terrible when odd and even frames were not "recognized" as such, because the TV set physically draws on odd and even lines, intermittently (which is achieved by having an odd number of lines, so half a frame starts at half a line).

So, in my perception, the 2:2 pull-down is rather straight-forward (both fields belonging to one and the same film frame) and it's the 3:2 pull down that presents a real problem - if the (full-)frame sync isn't supported properly. But, honestly, I never observed a DVD frame with odd and even lines transposed.


Cees

 

[ChristopherDAC]

Cees, it's perfectly possible, depending on the operation of the film-to-video equipment, for even the PAL 1 film frame-2 video fields system to create two different "cadences". There is no guarantee that the film frame will be changed on a "top" field as your description requires : if it is changed on a "bottom" field, every "reconstructed film frame" will actually consist of lines from two different frames, if you use the video frame sequence as your reference.

"1080i" is actually an 1125-line format using odd-line interlace, unlike the EUREKA 1250-line system, so as far as this goes there is no difference from standard TV.

 

[DaViD Boulet]

It's not that the TV has a problem telling the odd/even fields apart...that works just fine as you say and so you can watch your interlaced show on your interlaced TV no problem.

The difficulty is when have a *progressive scan* monitor and your display takes on the added task of converting the interlaces signal to progressive on-the-fly as you watch it.

The cheapest/easiest way to convert to progressive is to take each field one at a time and repeat each line twice to "fill in the gap" left there by the missing odd (or even) lines.

That creates a progressive image with no aliasing...but it also looks pretty soft.

The next best way is to digitally create a "new" line between the existing lines by averaging the points of information to interpolate...sort of like the oversampling filter in a d/a converter for audio. That looks sharper than just repeating each line twice to fill in the gap, but it's not as sharp as a "true" native progressive image.

The next best way is to compare the fields before and after the field you're looking at, and figure out what moves and what stays stationary in the image. Whatever moves...then for that portion of the image do your "averaging" thing like was just mentioned. However, for what stays stationary...you'll get a more "true" progressive resolution if you zip in the missing odd (or even) lines from the precending or following field. This called "motion adaptive" deinterlacing and it's also something that Faroudja pioneered.

Ok...that all works great for video that was recorded natively in interlaced. But film offers a new twist. In video each field happens in a different moment in time... so when we deinterlace really the best we can ever do is guess at what was supposed to be where those empty lines exist.

But with film since every two pairs of fields (ingorning 3-2 for a moment) represent a single instance, or frame of time, if a computer algorithm can analyze the sequence of fields and figure out which ones go back together to form "frames"...you can restore the original frame as if it had never even been split into fields.

That's what film-recognition is about...and it's the same for both PAL-speed up and NTSC encoded film sources. NTSC regions add in the extra complication of 3-2 pulldown.

 

[DaViD Boulet]

Cees,

I just re-read your post and now I realize what you are saying. You're suggesting that the "field number one" of a frame would always be, for instance, an odd field and the "field number two" would always be the even. That way when a processor grabed ahold of a odd field (which it could easily recognize) it could just assume that the next even field to come along would be the other half of the "frame" and zip them back together.

I don't know enough about how interlacing from film-source material works in the mastering houses to know for sure...but I don't think it's always a given which side of the odd/even field split represents the first and second field.

But in any case... if you are correct about that, then yes, that would certainly make 2:2 pulldown for PAL users a much more simplified task!

My assumption as to why this might not be a given is because that would also have made it very easy to recognize 3-2 cadence and reverse for it (so it would seem to me) but I'll leave that as a question for anyone who has more knowledge on the subject.

Thanks for bringing up that point... very interesting to consider.

 

[DaViD Boulet]

Good for you! I think that you've got both of perhaps the handful of consumer options for 3-2 pulldown for film-based 1080i content.

This point needs to be emphasized... consumers thinking that they'll get "real 1080P" from their 1080P displays when fed a 1080I output from a device like the Toshiba HD DVD player are mistaken unless their set does proper film-dedection and frame reconstruction.

HD enthusiasts...accept nothing less!!! If you have a 1080P display...you DESERVE to see the original 1080P film-source image!

 

[Mattias_ka]

I will also wait until 2nd generation player with 1080P output! EVEN when I have a Sony Ruby, I think the main reason to wait a little is:

1. Software. 15 titles so far, with maybe 2-3 I would like to have.

2. People can scream blue murder about the Toshiba player not filtering the quality when outputting 1080i, but I rather wait and see. Maybe they are right, maybe not. I think that either way, next generation player WILL have better image quality.

3. Too many problems with the player for my taste.

I really WANT HD-DVD/Blue-ray, as I have always complained about DVD and all the digital artifacts that are present in DVD's. On a big screen, DVD can really suck and I prefer LD sometimes, with it's softer image, over bad MPEG compression from DVD's. Now we have a format with the potential for no problem with MPEG compression artifacts, and I'm really glad!

So when are there coming out 1080P players?

/Mattias

 

[DaViD Boulet]

So far all announced BD players will do 1080P.

I think at the moment the Samsung is slated for mid-June.

 

[Mattias_ka]

Will, I think that late summer will be a perfect time to get some serious 1080P contant and players.

 

[Lew Crippen]

I probably did not write clearly: what I meant was that although it is not possible to view Blu-Ray in the States today, it may well be possible elsewhere.

I did not mean that the initial Blu-Ray players here (or elsewhere) would not be 1080P.

Since I have been reading your posts in the HD sections, I believed your DTS support to be indicative of the HD formats. As I now understand that not to be the case, I’ll wait patiently (along with everyone else) to listen to both of the new formats—and I look forward to reading your views.

In the prediction business, I expect that there won’t be much difference (and probably no detectable difference to the human ear on normal, moderately high-end equipment). Of course lossless formats will probably make the discussion moot.

 

[Cees Alons]

David,

But that's how an ordinary TV set works. If we call the first vertical line #1 (and not #0), then it always starts a frame with line #1, so the odd lines, then exactly half-way an overscan line (because the no. of vertical lines is odd, so halfway a frame is in the midst of a vertical line), a frame interrupt occurs and the beam is reset (drawing a few invisible vertical lines in the upward direction) to the zero position + 1/2 double line distance (positioned therefore below our #1, half-way to #3) and so it starts the even field. (Please note that the naming "odd" and "even" is arbitrary here, but the physical meaning isn't).

It simply can't work in a different way, it's hardwired to work like that, so to speak.
And as I said before: I never noticed a TV image (on a non-defective TV set) that had the odd and even lines the wrong way. Nor have you, I guess. The image would have been ridiculous.

A full frame always starts with line #1 (numbering according to the above definition) in the first interlaced field.


Cees

 

[Dave Moritz]

Its going to be difficult to say which format will end up wining this war. I was suprised when Beta did not win during the VHS vs Beta war. And I was hoping there would have been a winner in the DVD-A vs SACD war but they both now coexsist and there is still a small number of discs available for both formats.

Even though I believe that Blu-ray is the best choice we got. I can only hope that what ever format wins this war. I only hope that the format will be around for at least 10 years. Because I am not really to crazy about having to buy another format in the near future. I hope that ether way this format war is over within 2 years max.

 

[Nils Luehrmann]

See, now I bought into Betamax with the strong suspicion that it would never win. After all, VHS provided all the most important advantages that appeal to the mass market. VHS hardware and software was less expensive, offered far more players and titles to chose from, and the longer recording times were more convenient.

However for me, I appreciated having 50% more lines of resolution, and as I had almost no interest in owning pre-recorded tapes, I never felt like I was going to be investing thousands of dollars in media. Besides, by the time Sony finally gave up on Betamax and tapes became hard to come by, I already had more than enough titles and tapes and the only reason I stopped using it was that I had replaced it with an even better performing S-VHS recorder.

Interestingly enough though, if you want to use the Betamax VHS war as a road map, currently with the exception of cost, Blu-ray looks much more like the "VHS" party this time around (more industry support & capacity), while at the same time enjoying specific performance advantages. One big difference though this time around is that neither format offers anything truly substantial to the mass market due to the limited infiltration of HD displays, and the public's general lukewarm reception to products that in their mind are not a major step forward from what they have available to them right now at substantially lower costs.

 

[DaViD Boulet]

Now I see where we're coming from. You're describing a "frame" to be the definition of a video frame...which is a pair of fields, odd lines first, then even (since you start with line 1). That's a video definition...and it's correct. But it doesn't actually tell you anything about the picture information contained inside those lines if the content was mastered from a film-source.

An ordinary TV doesn't distinguish or care which fields have which half of the original film frames...they just show all the fields one after the other.

Whether or not an odd and even field sequence contain fields split from the same frame, or whether they each have a field from a different frame of film makes no difference.

But it *does* make a difference to a deinterlacer.

The point is that even if we define a "frame" of interlaced video to be an odd/even pair (since it starts with line number 1 as you say), that does not guarantee that the odd/even pair contain picture information relating to a frame of film... the mastering house could have off-set something in the editing room.

That happens all the time actually...a single "field" gets dropped and throws the whole sequence off by one (bad edits are something that deinterlacers need to be able to handle gracefully). Also, think about 3-2 pulldown...since every third *field* is repeated twice...it's constantly flip-flopping the "start" of the *film* frame information from a odd field to an even field and back and and forth each time that 3rd field gets repeated. So clearly that breaks the rule that a frame of picture information is represented by the odd/even pair...odd lines first.

 

[Cees Alons]

Exactly David. But it would only be a problem with 3:2 pull down (because it's an odd number of frames), and not influence 2:2 pull down.

The monitor/TV set/interlacer only "sees" consecutive fields without knowing from what source. E.g. an NTSC player will output F1odd, F1even, F1odd, F2even, F2odd, F3even, F3odd, F3even, F4odd, F4even, etc.

So a single full-frame might be reconstructed wrongly (e.g. one frame between full-F1 and full-F2 might contain F1odd and F2even), which is an improper image, although it's hardly wrong to the eye (in its time-sequence). However, a purist might want to have it reconstructed correctly, taking the proper 3:2 sequence into account.

Note that this never happens with 2:2 pull down: F1odd, F1even, F2odd, F2even, etc.!

But all this has nothing to do with the inherent quality (resolution) of a 1080i or a 1080p image. It's a conversion problem and, in particular, should never be confused with "scaling", as some people seem to do (not you! ).


Cees


PS I admit that I do not always use the word frame (meaning a film-frame, a TV full frame, or a TV half-frame) univocally, but I trusted the context would always clarify the respective meaning. My apologies if it didn't.
C.

 

[DaViD Boulet]

Cees,

what about a bad edit (dropped field) in a 2:2 PAL scenario? Couldn't that throw off the "pairs" in the odd/even sets as well?



Very true. When talking about film content being deinterlaced...it's not a matter of scaling at all. That happens *after* proper frame-reconstruction has been acheived...and only when the output resolution needs to differ from the input signal.

 

[Cees Alons]

David,

I believe that that's when the sync comes in. A sync at the start of a vertical line means a new frame, half-way a vertical line signals that the even field is coming. TV broadcasting has never been very robust in not losing frames or part of the information.

And, frankly, I knew that we basically agreed all the time.


Cees