To de-interlace or not?! (DV source to mpeg2)

[Cyberdave]

Hi,

I have been researching this and think I got the general idea but just wanted to see if anyone who knows all this stuff which I am trying to figure out can confirm what I have conlcuded?!

I am capturing streams (tv shows) from my DISH dvr via cables to my mini-DV camera which has a digital pass through via firewire into my pc. I am using WinDV app to capture it. Then I want to use TMPGenc 2.5 to convert it to mpeg-2 and then author to dvd.

I noticed the interlaced phenomenon when looking at the captured video on the computer monitor. At first I thought I had to filter it out during the conversion to mpeg2. However since I dont plan on watching these shows on the computer but standard definition tv's (although there might be the occassional flat screen tv in a hotel room that I hook the portable dvd player up to), then once the dvd is authored and played on the dvd player on a sdtv, then it should just look "normal". Is that correct? Or do I need to deinterlace / inverse telecine it?

Thanks for any advice/info!

[buzzqw]

if your final media is a DVD video don't deinterlace, you will loose only quality by deinterlacing.

hen once the dvd is authored and played on the dvd player on a sdtv, then it should just look "normal". Is that correct? Or do I need to deinterlace / inverse telecine it
you don't need to touch it, leave as is, just encode as interlaced mpeg2 (using hcenc for example) and then author on dvd

BHH

[manono]

you don't need to touch it, leave as is, just encode as interlaced mpeg2 (using hcenc for example) and then author on dvd
If it's film based (as are most major NTSC TV shows these days, as well as all movies), I'd definitely prefer to IVTC it. If it's really been shot on video (true interlace), then I agree that I'd keep it interlaced.

[buzzqw]

you are right manono!

since i live in PAL land , my DV files are only interlaced, not to telecine

BHH

[Cyberdave]

If it's film based (as are most major NTSC TV shows these days, as well as all movies), I'd definitely prefer to IVTC it. If it's really been shot on video (true interlace), then I agree that I'd keep it interlaced.

Thanks folks and that's very interesting. I will do a test disc with one as is and the other IVTC and see what it looks like and let you know!

[manono]

Better would be to figure out what you have before doing the encoding.

[manolito]

if your final media is a DVD video don't deinterlace, you will loose only quality by deinterlacing.
Yeah, this is the common advice. But is it still true these days when CRT TV sets are quickly entering the endangered species list?

Today it is most likely that your DVD will be shown on an LCD TV or a projector, and these devices can only display full frames. Which means that your interlaced sources must be deinterlaced somewhere in the chain (either the display device itself or the DVD player does the deinterlacing).

AFAIK the quality of these built-in hardware deinterlacers varies wildly. Not every device sports a Faroudja deinterlacer. Plus the hardware deinterlacer has to work in real time which also compromises quality.

Wouldn't it be better to do the deinterlacing in AviSynth then? It will slow down your encode, but you have full control over the deinterlacing process. If you use one of the very slow edge based deinterlacers (nnedi) then I suppose that the resulting visual quality you get on an LCD will surpass the quality you get when feeding interlaced material to this LCD.


Cheers
manolito

[Mc Onyx]

Yeah, this is the common advice. But is it still true these days when CRT TV sets are quickly entering the endangered species list?

Today it is most likely that your DVD will be shown on an LCD TV or a projector, and these devices can only display full frames. Which means that your interlaced sources must be deinterlaced somewhere in the chain (either the display device itself or the DVD player does the deinterlacing).

AFAIK the quality of these built-in hardware deinterlacers varies wildly. Not every device sports a Faroudja deinterlacer. Plus the hardware deinterlacer has to work in real time which also compromises quality.

Wouldn't it be better to do the deinterlacing in AviSynth then? It will slow down your encode, but you have full control over the deinterlacing process. If you use one of the very slow edge based deinterlacers (nnedi) then I suppose that the resulting visual quality you get on an LCD will surpass the quality you get when feeding interlaced material to this LCD.


Cheers
manolito


Exactly what I was thinking! Am I wrong?

[Bogalvator]

I would imagine most of these newer TVs would de-interlace a 29.97i stream in to a 59.94 full frame stream (and 25i in to 50 etc) during playback.

But you can't encode a 59.94 fps MPEG-2 stream (not one one that will play on a DVD player at least), so you'd have to deinterlace to a regular 29.97 stream, and the picture would not be as smooth (really noticeable in fast moving footage likes sports).

I hope the above makes some sense, I'm finding it hard to explain for some reason!

So basically, the best solution is to keep it interlaced and encode it as such.

[SomeJoe]

I agree with Bogalvator. If the final source is destined to be DVD, then you have to work within the confines of that format. DVD cannot hold 59.94 fps progressive video. So if you bob-deinterlace 59.94i material (giving you 59.94p material), you would then have to decimate that to 29.97p material before it can go on DVD. This halves the temporal resolution.

It's much better to encode the 59.94i material to begin with, and allow the playback device or display device to deinterlace if that's necessary for their display technology (LCD, etc.)

Obviously, this applies only to true 59.94i material. If original material was 23.976p and telecined to 59.94i, then applying IVTC and encoding at 23.976p with pulldown flags is the correct choice.

By the way, on a side note, please do not refer to "29.97i" material. There is no such thing. The reason is that "29.97i" implies that there are 29.97 of something per second and that these things are interlaced together. Interlaced NTSC video does not have anything that repeats at 29.97 times per second, because an interlaced NTSC video stream does not have frames. There is no such thing as a "frame" in interlaced video. Interlaced video is a repeating sequence of fields that are not grouped together in any way. If fields A, B, and C arrive at the display device, one every 1/59.94th of a second, fields A and B don't "go together to form a frame" any more than fields B and C go together. The video is simply a sequence of fields that alternate in spatial location and have a 1/59.94th of a second difference in temporal sampling. So, with 59.94 fields per second, that are interlaced together, the proper notation is 59.94i.

Now, having said that, most codecs and containers are designed around the basis of frames. So for convenience, those codecs/containers that can handle interlaced video take two fields and group them together into a "frame" for compatibility with the same methods used to store progressive video in the codec/container. But that's just for storage. At some point when that stream is played back to an interlaced display device, the stream loses that grouping and again becomes a sequence of alternating fields.

[manono]

There is no such thing as a "frame" in interlaced video. Interlaced video is a repeating sequence of fields that are not grouped together in any way. If fields A, B, and C arrive at the display device, one every 1/59.94th of a second, fields A and B don't "go together to form a frame" any more than fields B and C go together.
Nonsense. 30i material (or 29.97i if you prefer), may be outputted at 59.94 fields per second, but it's almost always stored on a DVD at 29.97 frames per second. There is such a thing as a frame in interlaced video. They are grouped together as frames when stored. And referring to it as 30i (or 29.97i) is correct.

[SomeJoe]

Nonsense. 30i material (or 29.97i if you prefer), may be outputted at 59.94 fields per second, but it's almost always stored on a DVD at 29.97 frames per second. There is such a thing as a frame in interlaced video. They are grouped together as frames when stored. And referring to it as 30i (or 29.97i) is correct.

It's not nonsense, it's 100% factual.

Interlaced NTSC material will always be output at 59.94 fields per second, not "may". If it's not 59.94 fields per second, then it is not interlaced NTSC video, by definition.

I have an entire paragraph in my explanation that addresses how video is stored. Did you miss this when glossing over my post?

And again, you are incorrect with you say there is a "frame" in interlaced video. There is not. If you look at an interlaced NTSC (or PAL, for that matter) video signal on an oscilloscope, there is no "frame" marker. There is no signal anywhere that tells you what two fields "go together" to form a frame. The fields are completely independent from each other. That is why "29.97i" is completely incorrect. There is no 29.97 Hz frequency associated with the signal.

People have let the introduction of digital video, codecs, and containers cloud their thinking. None of that matters when you talk about the signal. By definition, an NTSC video signal is analog, and all parameters and characteristics of the signal are derived from that. The pictures that are captured by an NTSC camera and the associated analog signal it creates have no "frames", period, end of story. Only fields.

[manono]

Interlaced NTSC material will always be output at 59.94 fields per second, not "may". If it's not 59.94 fields per second, then it is not interlaced NTSC video, by definition.
By using your definition all distinctions between progressive sources and interlaced sources on NTSC DVD are erased. It's all output at 59.94 fields per second. Progressively encoded 23.976 with pulldown (soft telecine) is output at 59.94 fields per second. Already telecined 23.976fps film (hard telecine) is output as 59.94 fields per second. Progressive sourced 29.97fps material, whether encoded as interlaced or as progressive, is output at 59.94 fields per second. And truly interlaced 29.97fps video is output as 59.94 fields per second. By your definition it's all 59.94i. And that's simply not correct. DVD is an interlaced medium. All NTSC DVD players (at least initially) output 59.94 fields per second. But that says nothing about the source material - the material being stored on the disc. 29.97p or i and 23.976p, on the other hand, do describe the source material (or describe it better).

The encoder (almost always) encodes interlaced material at 29.97fps (and not 59.94 fields per second). It's stored on the DVD at 29.97fps (or at 23.976fps with pulldown in the case of film). The picture structure is (almost always) by frames. The stored frames are the important consideration.
If you look at an interlaced NTSC (or PAL, for that matter) video signal on an oscilloscope, there is no "frame" marker.
But there is in a DVD, which is the topic of discussion here. Of course if it's examining the output it's seeing fields. But I've already discussed that.
There is no signal anywhere that tells you what two fields "go together" to form a frame.
But there is for the DVD fields.
None of that matters when you talk about the signal..We're talking MPEG-2 encoding and DVDs.
By definition, an NTSC video signal is analog..Again, we're talking DVDs. My path from player to HDMI cable to TV is all digital. It doesn't drop to analog anywhere along the line. You can talk oscilloscopes and signals all you want, but none of it has any relevance here.

[Guest]

There is no signal anywhere that tells you what two fields "go together" to form a frame. Sure there is:

http://lurkertech.com/lg/dominance_sgi.html

'For "F1 Dominant," a "frame" is an F1 field followed by an F2 field. This is the protocol recommended by all of the industry video specifications.'

[SomeJoe]

Sure there is:

http://lurkertech.com/lg/dominance_sgi.html

'For "F1 Dominant," a "frame" is an F1 field followed by an F2 field. This is the protocol recommended by all of the industry video specifications.'

Yes, but that is only when the video signal is on a storage medium where the field dominance now matters. These are storage mediums where the fields must be dealt with in pairs. In this case they were talking about an editing VTR that can only accurately edit with pairs of fields. Once an edit is made, now you have to know what field (F1 or F2) the edit started with, and all other edits have to be done the same way.

See the related link on that site Where Does Field Dominance Originate? (http://lurkertech.com/lg/dominanceorigin_sgi.html) for their graphical explanation.

But in the absence of a storage medium that forces pairing of fields, there is no field dominance. Also on that page is the quote: "Video material is born when video fields come out of a video camera, a video signal generator, or a field-based computer graphics algorithm. At this point, there is no need to impose any frame boundaries on the material: it has no field dominance."

Which was my point previously.

I believe we're all on the same page as far as the analog signal goes. Now let's turn our attention to the DVD storage and why I believe that certain notations shouldn't be used.

By using your definition all distinctions between progressive sources and interlaced sources on NTSC DVD are erased. It's all output at 59.94 fields per second. Progressively encoded 23.976 with pulldown (soft telecine) is output at 59.94 fields per second. Already telecined 23.976fps film (hard telecine) is output as 59.94 fields per second. Progressive sourced 29.97fps material, whether encoded as interlaced or as progressive, is output at 59.94 fields per second. And truly interlaced 29.97fps video is output as 59.94 fields per second. By your definition it's all 59.94i. And that's simply not correct. DVD is an interlaced medium. All NTSC DVD players (at least initially) output 59.94 fields per second. But that says nothing about the source material - the material being stored on the disc. 29.97p or i and 23.976p, on the other hand, do describe the source material (or describe it better).

Ah yes, but this is an allusion to my point. To me, a number which represents a field/frame/update rate, and a letter which represents interlaced/progressive is too ambiguous to describe the MPEG stream on the DVD.

These notations came from how we describe display devices. And for display devices, they're very clear as to what they mean. If I talk about a 59.94i display, it's very clear that I'm talking about something like a standard CRT TV, which is scanning interlaced fields at 59.94 Hz. If I talk about a 23.976p display, it's very clear that the display is doing progressive scan (or progressive display, in the case of LCD/DLP/LCoS), and updating the screen at 23.976 Hz.

But for the MPEG stream on a DVD, these notations by themselves don't fully describe the stream. Let's take film material shot at 24.000 fps and then slowed down for digitizing to 23.976 fps. As you know and I know, this can be stored legally on DVD two ways:

1. MPEG-2 encoded as frame structure, progressive frames, at 23.976 fps, with TFF and RFF flags applied to pulldown the stream to 59.94i upon playback. (Better known as soft telecine).

2. MPEG-2 encoded as frame (or field) structure, interlaced, at 59.94i fields per second, with the 3:2 pulldown done before encoding. (Better known as hard telecine).

How would you notate these two streams using only a number and letter? "23.976p" is ambiguous. That describes the original film material, but not the encoding. As you've pointed out, "59.94i" may be technically accurate for both streams, but is even more ambiguous because it can describe any DVD stream. "29.97i/p" is not very descriptive either.

To go even further, what if the storage medium is Blu-Ray and not DVD? Now we actually have a 3rd possibility to store this stream:

3. MPEG-2 encoded as frame structure, progressive frames, at 23.976 fps, with no flags.

On Blu-Ray, this is legal because Blu-Ray players can output 23.976p directly. Now how do we describe this stream using only a number and letter? "23.976p" now has 3 possible meanings.

The encoder (almost always) encodes interlaced material at 29.97fps (and not 59.94 fields per second). It's stored on the DVD at 29.97fps (or at 23.976fps with pulldown in the case of film). The picture structure is (almost always) by frames. The stored frames are the important consideration.

Yes, they are, but they're only one element of understanding exactly what you have on the DVD. To know exactly the entire processing chain, we have to know:

How the material was originally shot - frame rate and interlaced/progressive.
How the material is stored and encoded
How the material will be sent to the playback device
How the playback device will treat this material and display it


A simple number and letter is completely insufficient to describe this process.

You illustrated that "59.94i" can be used to describe anything on DVD and that it's an ambiguous notation. That is exactly my point.

When someone talks about a "29.97i" stream, I have no idea at all what they're talking about. To me, this could be referring to:

1. Actual 59.94i pure interlaced material that has been encoded in an MPEG-2 stream using frame picture structure.
2. 29.97p material (DV camcorder progressive mode) that will be played back on an interlaced display device (2:2 pulldown).
3. 29.97p material encoded in an MPEG-2 stream using field picture structure.

The term "29.97i" is just as ambiguous as referring to all DVD streams as "59.94i". There is simply more to it than what this notation can represent.

On top of all that, when dealing with camcorders, there's even more oddities. There's the Panasonic 24p Advanced mode, that delivers a 2:3:3:2 pulldown stream. There's Adobe Premiere Pro's alternate 24p mode that repeats every 4th frame. (I've actually seen this encoded on a commercial DVD directly. The stream required Decimate(cycle=5) to recover the 23.976p material, but did not require Telecide() ).

My point is just to be clear and accurate when talking about streams, and with all the oddities out there and multi-step processing/decoding/display chains, a simple ##.##i/p notation just isn't descriptive enough.

[manono]

You illustrated that "59.94i" can be used to describe anything on DVD and that it's an ambiguous notation. That is exactly my point.
No it's not, at least not initially. Here's your point:
So, with 59.94 fields per second, that are interlaced together, the proper notation is 59.94i.

My point is that 59.94i is meaningless since all NTSC DVD players output 59.94i, no matter what the source. 59.94i isn't proper notation for anything. Since you (apparently) agree, I consider the discussion closed. Feel free to have the last word.

[Guest]

SomeJoe, you're just wrong. The F1 and F2 fields are defined by the analog waveforms, and the video specs tell how to combine the F1 and F2 fields to make a full picture. That refutes your statement that there is nothing in the signal to specify how two fields are combined, which is the statement I responded to. Bringing up editors and dominance is now an equivocation on your part.

You're also equivocating with manono as he showed. So I'll consider this closed as well. :)

[smok3]

I would imagine most of these newer TVs would de-interlace a 29.97i stream in to a 59.94 full frame stream (and 25i in to 50 etc) during playback.

agreed, what i wonder now is...., is there a way to do a good bob/weave path before the encoder, that could actually improve later bob deinterlacers on such devices ? (forgive me, it was a long day ;))
-----
edit: about the XYz debate; imho XY stands for 'changing images or half-images per second' and z can be 'i or p', so for example:
25i = 25 half-images per second (there is no known format like that)..., that would be 12.5 full frames per second

[SomeJoe]

So, in other words, after a very carefully crafted explanation of my opinion in which I cited many examples that illustrate my point, the response is:

"Now you agree with us. The discussion is now closed."

I do not recall agreeing with anything. I have tried to illustrate that I do have the requisite knowledge to be qualified to participate in this debate, but instead of my points being addressed, we have no debate, and a "discussion closed".

OK. You are the moderators, I am just a guy with an opinion. From now on, I will keep my opinions to myself.

[Guest]

"Now you agree with us. The discussion is now closed." I didn't say you agree with me. I said you are wrong and that you avoid that by changing your tune.

Now you've added putting words in people's mouths to your bag of tricks. No matter, I'll continue discussing it with you.

I addressed your point and refuted it. Will you address that?

[SomeJoe]

I didn't say you agree with me. I said you are wrong and that since you avoid that by changing your tune, I'm no longer interested in discussing it with you.

Now you've added putting words in people's mouths to your bag of tricks.

I addressed your point and refuted it. Will you address that?

No, I will not go back and address my points any further. I apparently am not at all good at conveying my points or explaining myself. I have tried and failed to do so.

I am not putting any words in anyone's mouth. I still feel exactly the same way as I did when I started this conversation, I have not "changed my tune". If my explanation gives you that impression, then I have again failed.

What I do know is that I fully understand the topics and I fully understand the systems. I work with them every day, and everything I do is done correctly.

I do not accept your assertion that I am wrong in my understanding of the topic. What I am wrong or inadequate on is my ability to explain that in a manner that is convincing.

I do not know how to explain myself in any other way than what I have already posted. I have exhausted everything I know how to do to prove that I understand the topic at hand. You remain unconvinced. So be it, I do not have anything else than can be offered.

[Guest]

C'mon Joe, it's simple. The F1 and F2 fields are defined and so is the means of combining them into a frame.

Can't you simply tell us why you said the signal cannot specify that, or admit that you misspoke?
To make a mistake doesn't mean everything you think and do is wrong. It just means you made a mistake.

Here's what I do when I make a mistake. First I admit it. Then I make an excuse for why I made the mistake. :)

http://forum.doom9.org/showthread.php?p=1248205#post1248205

[Cyberdave]

I would like to thank everyone for their insights, this has certainly been a fascinating discussion which ensued from what I thought was a simple question! Given I've done a bit of research already I understand a lot of it but there is still more to learn and understand. I wanted to respond to a couple comments.

Better would be to figure out what you have before doing the encoding.

For my current project, it is analog NTSC tv broadcasts via satelite (DISH) in the US recorded on a DVR of a variety of programs which could be films like from IFC or Sesame Street on PBS for my kid or anime. I'm also capturing a few old VHS tapes. This is then channeled through the mini-dv cam converting the analog signal to digital and captured as DV-AVI. So the final format to work with is interlaced 30fps DV-AVI. In the future my projects will be converting mini-dv tapes. The rest is subject to answer below on which device it will viewed on.

Yeah, this is the common advice. But is it still true these days when CRT TV sets are quickly entering the endangered species list?

Today it is most likely that your DVD will be shown on an LCD TV or a projector, and these devices can only display full frames. Which means that your interlaced sources must be deinterlaced somewhere in the chain (either the display device itself or the DVD player does the deinterlacing).

Cheers
manolito

I would say basically yes that SDtv's are endangered but not in my house or the other members of my family whom we visit and will use the created DVDs there. Although I would love a huge 1080p flatscreen, that's just not going to happen for some time. We have a really nice big Sony Trinitron CRT SDtv which these DVD's will be viewed on. There will be times they may be viewed with a portable dvd player on a flat screen in a hotel room and that will be interesting to see how it turns out on those. Also interesting to see how they would work on an upconvert dvd player hooked up to an HDtv. If it works nicely then I won't need to save the captured files to reconvert them when I eventually do get an HDtv.

[hkazemi]

The reason is that "29.97i" implies that there are 29.97 of something per second and that these things are interlaced together.

This is the source of the confusion. The convention, at least in many forums, is the number is FRAMES per second followed by an indicator whether these frames have interlacing or are progressive. It isn't x# of somethings per second. This probably comes from the fact that DVDs store content as frames, which may originally been video fields or film frames.

I'm not sure if the i/p nomenclature is derived from a similar use with HDTV signals, however the numbers are totally different. For HDTV, the number means the nominal vertical resolution, e.g. 720p = 720 pixels vertically, 1080i = 1080 pixels vertically. The i/p means interlaced or progressive content. To the eye they're pretty much equal when a 1080i image is deinterlaced compared to a 720p image. One trades temporal resolution (60 at frames/second) of crisp individual images for spatial resolution (at 60 fields/second). While 1080i may have a theoretical advantage for near stationary content, how much TV content is actually stationary? The best deinterlacing methods (like MCBOB) may be able to give the advantage to 1080i. I've read that deinterlacing 1080i to 1080p and then downscaling that 1080p to 720p results in a image that is essentially the same to the eye. The original amount of information presented to the eye (pixels/second) remains about the same in either case.

Interlaced NTSC video does not have anything that repeats at 29.97 times per second, because an interlaced NTSC video stream does not have frames. There is no such thing as a "frame" in interlaced video. Interlaced video is a repeating sequence of fields that are not grouped together in any way. If fields A, B, and C arrive at the display device, one every 1/59.94th of a second, fields A and B don't "go together to form a frame" any more than fields B and C go together. The video is simply a sequence of fields that alternate in spatial location and have a 1/59.94th of a second difference in temporal sampling. So, with 59.94 fields per second, that are interlaced together, the proper notation is 59.94i.

Interlaced content at 59.94 fields per second is the same thing as 29.97 interlaced frames per second. Hence 29.97i.


A simple number and letter is completely insufficient to describe this process.

You illustrated that "59.94i" can be used to describe anything on DVD and that it's an ambiguous notation. That is exactly my point.

When someone talks about a "29.97i" stream, I have no idea at all what they're talking about. To me, this could be referring to:

1. Actual 59.94i pure interlaced material that has been encoded in an MPEG-2 stream using frame picture structure.
2. 29.97p material (DV camcorder progressive mode) that will be played back on an interlaced display device (2:2 pulldown).
3. 29.97p material encoded in an MPEG-2 stream using field picture structure.
The first one is the most common interpretation, particularly when talking about commercial DVD source material (of video source content). Turning this into progressive content without losing image data requires spatial and temporal interpolation (which will double the frame rate to 59.94 full frames per second, because each field is from an independent point in space and time).

The second one is rarer, unless talking about DV camcorder source. Probably not going to be found much on commercial DVDs, except possibly documentaries or (low budget) independent films. Essentially 59.94 half frames per second; each pair of fields is from an independent point in space and time.

The third one sounds like something you might get by feeding progressive content into an mpeg-2 encoder set to encode as interlaced. Seems similar to the second case.

The term "29.97i" is just as ambiguous as referring to all DVD streams as "59.94i". There is simply more to it than what this notation can represent.

29.97i is being used to refer to interlaced content being stored 29.97 frames per second.

The number is referring to the frames per second on the DVD storage medium, and the i/p to indicate whether those frames have interlacing between them or if they're just independent progressive frames.

By convention, '59.94i' would mean 59.94 frames per second, with interlacing between them, not 59.94 fields per second.

On top of all that, when dealing with camcorders, there's even more oddities. There's the Panasonic 24p Advanced mode, that delivers a 2:3:3:2 pulldown stream. There's Adobe Premiere Pro's alternate 24p mode that repeats every 4th frame. (I've actually seen this encoded on a commercial DVD directly. The stream required Decimate(cycle=5) to recover the 23.976p material, but did not require Telecide() ).

Here the DVD would have been encoded as an interlaced 29.97 frames per second. The reconstruction technique (to recover 24 frames per second progressive) is different from the common IVTC technique, but 29.97i still describes how it is stored on the DVD.

My point is just to be clear and accurate when talking about streams <SNIP>...<SNIP> with all the oddities out there and multi-step processing/decoding/display chains

I'll agree with being clear and accurate as practically possible when describing what one is working with.

[Jay Bee]

The most useful notation IMO:

"number of lines" followed by "encoding mode" followed by "temporal resolution".

So a PAL movie DVD is 576i25, a HDTV sports broadcast is 1080i50, a PS3 game is 1080p60 etc.

But even this notation isn't perfect. Most German HDTV channels broadcast 720p50, even for movies. So really there's an element missing. A full notation would need vertical pixels, horizontal pixels, encoding mode i/p, encoded frames per second and temporal resolution.

[2Bdecided]

But a correctly encoded PAL movie DVD is 576p25!


Anyway, the entire industry can't agree on 60i vs 30i nomenclature, so why should we try to sort it out for them? ;) Just know that both mean the same thing, and that there's no real confusion. (It's a good job there isn't a 30 or 120 fields per second video format, or we would be in trouble!).

FWIW, I think SomeJoe has a point. Analogue video, which is where digital video came from, is a series of fields. The indication of which field is which (such as it is*) tells you which fields sits at the top, and which sits at the bottom. That's all.

There's nothing in the analogue video signal to tell you which pair of fields to combine to generate a frame. There are conventions on how to do this, and they pre-date digital (at least in the PAL world - I have references!) but they're not always consistent - e.g. BFF PAL DV forces you to either pair different fields or vertically shift the image by one line compared with TFF PAL MPEG-2. Even today, people sometimes break these conventions (e.g. field shifted video) - and in the early days, there were no such conventions.


* - the analogue signal has sync pulses which, if used to drive certain analogue circuits in a specific way, force the electron beam to be deflected such that the lines from one field sit between (above) the lines from another - the traditional CRT TV doesn't "do" anything to position the fields correctly - it just gets driven by the signal. It's a very clever piece of 1930s design.

Cheers,
David.

[Guest]

The analog signal identifies the fields and the requisite standard defines how to assemble the fields into frames. That refutes the statement that "there is nothing in the signal...".

Of course it is possible to ignore the standards but that is something else entirely, and not what SomeJoe originally asserted.

[manono]

But a correctly encoded PAL movie DVD is 576p25!
So why are the vast majority of movies on PAL DVD encoded as interlaced? Are you saying that's incorrect? Even interlaced PAL DVDs with shifted fields are correct. We may not like them much, and we may realign the fields when reencoding them so we can reencode them as progressive, but they're perfectly compliant.

[2Bdecided]

The analog signal identifies the fieldsYes - both their temporal order and relative spatial location are defined by the signal itself.

and the requisite standard defines how to assemble the fields into frames.Yes and no...

Page 10 of SMPTE 170M shows how field pairs relate to frames for analogue NTSC.

There's no equivalent pairing of fields to frames in ITU-R BT.470-6 for PAL (see Page 23) - but of course in all standards the pairing is implied by numbering lines 1-525 and 1-625 - i.e. two fields as one frame.

...but the signal is still a sequence of fields, and there's no need (in analogue video) to pair them up into frames following these standards.

...and when digital video came along, what happened? ...


TFF NTSC and BFF PAL completely break this convention.

Have you ever seen TFF digital NTSC? ;) (hint: most DVDs).

Have you ever seen BFF digital PAL? ;) (hint: all DV).


I don't think this proves we should count fields instead of frames - but it should at least give people who think it's 100% correct to count frames some pause for thought. Fields are the smallest temporal unit, no frames - and anyone who assumes they can always go back into frames in a specific way is going to face serious problems!

[2Bdecided]

So why are the vast majority of movies on PAL DVD encoded as interlaced? Are you saying that's incorrect? Even interlaced PAL DVDs with shifted fields are correct. We may not like them much, and we may realign the fields when reencoding them so we can reencode them as progressive, but they're perfectly compliant.

"compliant" does not equal "correct".

Interlaced encoding means interlaced chroma - a completely sub-optimal way of representing a progressive image.

Of course you can use either - the point was Jay Bee said "So a PAL movie DVD is 576i25" - well, not necessarily, or even ideally!

(and there are those who would call it 576psf25 in that case anyway!)

Cheers,
David.