Questions about digital video in standard definition and analog television

[Xanthipos]

The full title is supposed to be Questions about digital video in old standard definition and filming T.V shows towards the end of analog television but the title won't fit.

Let me start by noting that digital video exists in old standard definition, D.V.Ds being the classic standard definition digital format. Other digital video formats were used by professionals late in the era of analog television, when nearly all television broadcasts in the world were 480i60 or 576i50.
Digital video in 576p25 and 480p30 can be converted respectively to 576i50 and 480i60 simply by storing each frame in a memory and re-arranging the lines. If the video signal is from a camera scanning progressively, then this conversion, if you think about it, does not introduce combing - an interlacing artifact.
Is this conversion what D.V.D players do on playback using a T.V set or video monitor programmed for interlaced scanning? Is this also what television stations did to what they broadcast once fully converted to digital video internally?
Did a lot of television content from the late 1990s, if not shot on film, originate at 25p or 30p at a higher resolution than even 576 lines and get downconverted to 480i or 576i for broadcast?

[emcodem]

I joined broadcast industry early 2000's so i only have limited knowledge about the details how they worked analogue. 3:2 pulldown and telecine are techniques i didnt really face honestly, even if they are valid up to today.

Anyway, i think throughout the 90's we worked on MII or Betacam entirely so nothing was shot at a different resolution than it was broadcasted later on. This is of course how we work today as well. Changing video formats was always a hassle, so we try to avoid it whenever possible.
Not sure what was the most common techique for rescaling at these times, but even up to today, broadcasters themselves usually are not equipped to do super-high-quality rescaling. Of course every broadcaster works different but i think in general they usually shy away from remastering. If we rescale, we usually just use some standard techiques but those are not used for movie masters because the quality is not high enough for cinema use.
Today we receive movies ready to broadcast, so the contributor (e.g. the producers) took care about all the standards conversion.

Have in mind that most things at broadcaster's need to be done in realtime for the live usecase.

[DB83]

Ok. This is just my understanding of things. Some detail may be 'cloudy'.


Firstly, the vast majority of dvds are interlaced. You see the terms 576i25 and 576i50 but they tend to mean the same thing for this medium.


That being so, a dvd-player does not do any progressive to interlaced conversion. But what the player does is to convert the digital stream to analogue. And it is then sent as analogue over the composite/scart. The CRT sees the analogue, interlaced stream and just uses it. Only with the advent of digital tvs is that same interlaced stream converted by the tv to progressive.


And, surely, tv cameras were shooting interlaced since that was the medium that the tv set expected.

[Quint]

If the video signal is from a camera scanning progressively, then this conversion, if you think about it, does not introduce combing - an interlacing artifact.

It shouldn't, because the two fields fit very well together (not 100%, because of minimal changes caused by compression), but in PAL you often have field delays (which lead to even less well fitting together caused by compression) that have to be corrected. So you often have combing without real interlacing, sometimes scenewise.

Did a lot of television content from the late 1990s, if not shot on film, originate at 25p or 30p at a higher resolution than even 576 lines and get downconverted to 480i or 576i for broadcast?

If not shot on film: No. At least not in PAL countries.

[Xanthipos]

Anyway, i think throughout the 90's we worked on MII or Betacam entirely so nothing was shot at a different resolution than it was broadcasted later on. This is of course how we work today as well. Changing video formats was always a hassle, so we try to avoid it whenever possible.

Digital Betacam was one of digital video formats used by professionals at this time.

It shouldn't, because the two fields fit very well together (not 100%, because of minimal changes caused by compression), but in PAL you often have field delays (which lead to even less well fitting together caused by compression) that have to be corrected. So you often have combing without real interlacing, sometimes scenewise.

I was asking about uncompressed digital video.

[Skiller]

If you've ever seen a standard definition analog signal on an oscilloscope you will find that analog video is an endless stream of fields. There exist no frames. Manipulation of the sync pulses makes the position of the fields alternate between top and bottom fields.
At that point, it does not make any difference if these fields originally come from a progressivly stored frame on a digital medium or not. A progressive frame is read out field by field and that's really all there is to it. 25p is the exact same amount of data as 25i (in uncompressed form).

So technically, it is not required that the two digitized fields are paired like the original progressive frame was, so that the resulting frame is free of combing, although in most cases that would happen incidentally.

[Cornucopia]

You are vague with your timeline, which muddles the correct answering of the question. Transition from analaog to digital HOW, and WHERE? This started in some form in the early/mid 90s and ended around 2015. In the US, broadcasting changed (by law) in Feb. 2009 (in some sections delayed until June of that year), but that date is different for different countries, and for other mediums.
Since this also coincided with the overlap of transioning from SD to HD, there is a bunch of weird overlaps, so it is hard to say unless you get more specific. Remember that there was also a change in cinema production and distribution at this time from analog (film) to digital, though that is still not complete (usually for artistic reasons).

Scott

[Quint]

It shouldn't, because the two fields fit very well together (not 100%, because of minimal changes caused by compression), but in PAL you often have field delays (which lead to even less well fitting together caused by compression) that have to be corrected. So you often have combing without real interlacing, sometimes scenewise.

I was asking about uncompressed digital video.

Even DigiBeta has a slight lossy compression, and that was one of the main digital formats used.

[Xanthipos]

If you've ever seen a standard definition analog signal on an oscilloscope you will find that analog video is an endless stream of fields. There exist no frames. Manipulation of the sync pulses makes the position of the fields alternate between top and bottom fields.
At that point, it does not make any difference if these fields originally come from a progressively stored frame on a digital medium or not. A progressive frame is read out field by field and that's really all there is to it. 25p is the exact same amount of data as 25i (in uncompressed form).

There are frames, each frame consists of two fields, one of odd and the other of even numbered lines. 25p is actually the same as 50i. So for example 576p25 would be the same amount of data as 576i50. The only difference is the order of the lines.

So technically, it is not required that the two digitized fields are paired like the original progressive frame was, so that the resulting frame is free of combing, although in most cases that would happen incidentally.

I don't understand the first clause of this sentence but the similarity, say between 480p30 and 480i60 is why I said that converting digital video in from the former to the latter is a simple matter of storing each frame in a memory and reading out the lines in a different order.
I would think all D.V.D players have frame buffers, this would be what allows freeze-frame, and with perfect clarity. So if what is stored on the disc is 576p25, it would a simple matter to use the same sort of random access memory to convert this to 576i50 for display on a television set or video monitor programmed for this number of lines and refresh rate.
Or the video on the disc may just be 24p, the disc instructing the player to repeat certain lines on playback to fit the 60i raster, is a frame buffer involved?

In fact, 576p25 could even be converted to 576i100 if displaying on a T.V set or video monitor that is 100Hz ready, each frame is now shown twice.

[emcodem]

Or the video on the disc may just be 24p, the disc instructing the player to repeat certain lines on playback to fit the 60i raster, is a frame buffer involved?

Usually the "player" would not convert it at all but output the original signal format. If you got a box that supports internal conversion (force output standard), some hardware chips will take care about the conversion. Likely that there are frame buffers involved, depending on the types of conversion they built into your player.
Actually i just said the same as @DB83 above

[Cornucopia]

No, Xanthipos, your statement is flawed and incomplete. 25p material stored in a 25p structure is the same as 25p material stored in a 25i (aka 50i) structure. It is just a matter of interpretation of the flagging or intent of the structure.

However, 25p material in 25p structure is NOT the same as 25i/50i material stored in a 25i/50i structure, as the former has both fields of a frame existing from the same time slice but the latter has consecutive, different time slices, which would be exbitited most clearly in sequences of fast motion.


Scott

[Skiller]

There are frames, each frame consists of two fields, one of odd and the other of even numbered lines.

Yes and no. Read again what I wrote.

In fact, 576p25 could even be converted to 576i100 if displaying on a T.V set or video monitor that is 100Hz ready, each frame is now shown twice.

576i100 does not exist. 100 Hz TVs, no matter if modern or the 100 Hz CRT kind, bob-deinterlace analog standard definition video to the native field rate (50 or 59.94 fps) and then show each frame (former fields) twice.

[Xanthipos]

Scott misunderstood my statement. I meant that 25p material can be stored in a 25p or 50i structure, the only difference between the technical standards is the order of the lines. 25i would not be 25 pairs of fields per second, it would be half that. In fact, 25p material includes anything from a film source which in turn was recorded at 25 frames per second.

However, 25p material in 25p structure is NOT the same as 25i/50i material stored in a 25i/50i structure, as the former has both fields of a frame existing from the same time slice but the latter has consecutive, different time slices, which would be exbitited most clearly in sequences of fast motion.

But what I meant is material from a 25p source (including but no longer limited to anything shot on film at 25 fps) in a 25p structure or a 50i structure differ only by the order of the lines.

Yes and no. Read again what I wrote.

I did read what you wrote, but you said 25p is the same as 25i, but the former is really the same as 50i. The latter (25i) would not be 25 pairs of fields per second, it would be half that.

576i100 does not exist. 100 Hz TVs, no matter if modern or the 100 Hz CRT kind, bob-deinterlace analog standard definition video to the native field rate (50 or 59.94 fps) and then show each frame (former fields) twice.

So do 100Hz T.V use 576p100, at least internally? Whatever they use, any digital video in 576p25 can be converted to this standard on playback through a 100Hz T.V.

[Cornucopia]

How could I misunderstand something I stated up front wasn't clear? Which was why I was qualifying what I said, so that if you meant one thing it would be in agreement but if you meant the other it would not.

25i === 50i. It is just a quirk of inaccurate nomenclature, bolstered by marketing hype, that caused this strange equivalency.

Maybe this will clear things up for you:

FPS = Frames per second.
fps = fields per second.

25i = 25FPS, Interlaced
50i = 50fps, Interlaced
Same thing

And Data-wise,
50p = 2x 50i(aka 25i)
50i = 2x 25p

And to be clear,
There is legitimately no such thing as:
50i FPS (which some might want to call 100i fps)
or
25i fps (which others might want to call 12.5i FPS)
Only that middle format actually has overlap, and only because of legacy compatibility.
You might not like that nomenclature, but that is what it has evolved as.

There might be a file built as 100p FPS, etc, but there are as yet NO consumer gear built to handle it. It would likely be experimental in the lab.

Note: there are no SD 100Hz TVs, so a 100Hz tv is only going to be HD or UHD. Which means that, yes, internally the data rate is 2x 50p FPS, and upconverted from 576 to 720/1080/2160. But as Skiller has already said, ALL 100Hz TVs frame double (equivalent to films' double flash projection).

(All discussion refers to PAL-land specs, but similar NTSC-land specs can be inferred).


Scott

[Xanthipos]

25i === 50i. It is just a quirk of inaccurate nomenclature, bolstered by marketing hype, that caused this strange equivalency.

So in that case it should be called 50i and 25i could be interpreted as half that many pairs of fields per second.

FPS = Frames per second.
fps = fields per second.

I had no idea this acronym stood for something different according to letter case.

25i = 25FPS, Interlaced
50i = 50fps, Interlaced
Same thing

Confusing, let's just call it 50i because 25 fields per second would be 25i, not less.

And Data-wise,
50p = 2x 50i(aka 25i)
50i = 2x 25p

These can't both be right, if 576p50 is 2*576i50 then 576i50 is the same as 576p25 apart from line order.

And to be clear,
There is legitimately no such thing as:
50i FPS (which some might want to call 100i fps)
or
25i fps (which others might want to call 12.5i FPS)
Only that middle format actually has overlap, and only because of legacy compatibility.
You might not like that nomenclature, but that is what it has evolved as.

It's not that I don't like it, it's just that there is a reason why it is confusing for the same technical standard to be referred to as both 25i and 50i, I thought only 50i referred to that. Best to just call it 50i and understand there is no such thing as 25i, though the latter is technically possible.

Note: I do recall 100Hz T.Vs from the late 1990s when televisions broadcasts were still in what was then standard definition.

[Cornucopia]

Whoops, you got me in a mistake.

50i fps IS the same data rate as 25p FPS.

That was a case of missing text in copy/paste.

What I should have kept in was
50i fps = 2x the MOTION SMOOTHNESS of 25p FPS.

Don't call 25i FPS half as many, because it is equal to 50i.
Don't call 25i fps half as many, because it doesn't exist.

If you want to avoid confusion in your discussions, feel free to use the term 50i exclusively, but don't be surprised that others say it as 25i. That actually is the older of the 2 terms, back when TV engineers came up with interlacing.


Scott

[Xanthipos]

50i fps IS the same data rate as 25p FPS.

That's what I have tried to tell others all along. Converting the latter to the former should be as simple.

What I should have kept in was
50i fps = 2x the MOTION SMOOTHNESS of 25p FPS.

If it is from a T.V camera that scans like that, indeed the temporal resolution is greater, but at the expense of combing. But 50i (fps) converted from 25p is the same in motion as 25p, and may well be the same in motion as anything shot on film at 25 F.P.S.

Don't call 25i FPS half as many, because it is equal to 50i.

25i FPS is equal to 50i (fps). Not only do I use the terms 50i (fps) and 60i (fps) but I would like to encourage others to do the same, at least in threads I have started. Indeed 25i (fps) doesn't exist, but I didn't say it is half as many but it would be if it did exist.

[hello_hello]

But what I meant is material from a 25p source (including but no longer limited to anything shot on film at 25 fps) in a 25p structure or a 50i structure differ only by the order of the lines.

There's probably lots of progressive material encoded as interlaced on DVDs.

Chroma Upsampling Error (or CUE)

The Chroma Upsampling Error and The 4:2:0 Interlaced Chroma Problem

[Skiller]

Note: there are no SD 100Hz TVs, so a 100Hz tv is only going to be HD or UHD.

I have to correct you there. I know this was very different in North America, but in Europe, in the high-end segment of TVs, we had tons of standard definition – some even widescreen – 100 Hz CRT TVs in the late 90's and early 2000's. Popular makes of these were Loewe, Metz, Bang & Olufsen, Grundig... There was (although very few) analog SD widescreen broadcasts as early as 1994 in PALplus.

I did read what you wrote

I was referring to this

If you've ever seen a standard definition analog signal on an oscilloscope you will find that analog video is an endless stream of fields. There exist no frames. Manipulation of the sync pulses makes the position of the fields alternate between top and bottom fields.

In an analog SD signal itself, there is no distinction between frames and fields, there are only fields and – consequently – the receiving device (TV, capture card, etc.) has no means of knowing which fields would originally, in case of a 25p source, belong together.


As Cornucopia explained, 25i and 50i are, unfortunately, two different names for the same thing.

[Cornucopia]

@Skiller, Well, you learn something new every day! Noted.

Converting the latter to the former should be as simple.

Since there is no interpolation of frames, yes it is SIMPLER (than all the other forms). It still isn't simple, as when capturing analog material sourced from progressive masters one would have to still work out which 2 fields composed the original progressive frame (Even + Odd, or Odd + Even), but there isn't any flagging or metadata inherent in the analog stream to make this easy so it is still a guess.
Nor is there any flagging or metadata that clues in the operator that the master source WAS progressive, except visually (or possibly labelling). And while that stream can be analyzed, that is really only reliably determinable (interlace or progressive source) after it has been digitized, using tools such as AVISynth. Which means it may require a encoding/muxing to fix if the original guess was wrong.

Scott

[Sharc]

There's probably lots of progressive material encoded as interlaced on DVDs.

At least all PAL DVD's I know of are always encoded as interlaced even when the content is progressive film or progressive video.

[Skiller]

Pretty much, but more precisely, most of the time commercial PAL DVDs with progressive content are encoded like this:

Scan Type: Zig Zag
Frametype progressive flag: yes
Top Field First flag: yes
Progressive Sequence flag: no

For example, this is the demuxed video stream of the German PAL DVD "The Hunger Games: Mockingjay Part 2":

[Attachment 62087 - Click to enlarge]



We can see that the actual encoding used Zig Zag scanning which is optimal for progressive content. So there's no loss of efficiency in encoding, it's just that the flags are set in a rather conservative manner.

To flag this truly progressive we would change the flags to this:

Frametype Progressive: yes
Top Field First: no
Progressive Sequence: yes

But some people say this can cause issues on few very old players (like 2001 ish) and that's probably why it's not used. I've never had any issues with this. In fact it makes progressive scan players completely disable any of their deinterlacing attempts, which is ideal.

[hello_hello]

Pretty much, but more precisely, most of the time commercial PAL DVDs with progressive content are encoded like this:

Scan Type: Zig Zag
Frametype progressive flag: yes
Top Field First flag: yes
Progressive Sequence flag: no

I won't argue yet because it's not something I've looked at much, but I still had a ripped DVD on my hard drive. It's an Australian TV series filmed in 2016. (Xanthipos, I'm in Melbourne too. Nice and hot suddenly, isn't it?)
https://en.wikipedia.org/wiki/Wanted_(2016_Australian_TV_series)
It's definitely progressive (I didn't need to de-interlace while re-encoding).

DGIndexProjectFile16
1
E:\VTS_01_1.VOB

Stream_Type=1
MPEG_Type=2
iDCT_Algorithm=6
YUVRGB_Scale=1
Luminance_Filter=0,0
Clipping=0,0,0,0
Aspect_Ratio=16:9
Picture_Size=720x576
Field_Operation=0
Frame_Rate=25000 (25/1)
Location=0,0,0,f3585

d00 5 0 2048 0 1 1 92 b2 a2 b2 b2 a2 b2 b2 a2 b2 b2 a2
900 5 0 30720 0 1 1 32 32 92 b2 b2 a2
900 5 0 43008 0 1 1 32 32 92 b2 b2 a2 b2 b2 a2
d00 5 0 94208 0 1 1 92 b2 a2 a2 b2 a2 a2 a2 b2 b2 a2 a2
900 5 0 487424 0 1 1 32 32 92 a2 b2 b2 a2 b2 a2 b2 b2 a2
900 5 0 872448 0 1 1 32 32 92 b2 a2 b2 b2 a2 b2 b2 a2 a2
900 5 0 1255424 0 1 1 32 32 92 b2 b2 a2 b2 b2 a2 b2 b2 a2
.
.
.
.
FINISHED  100.00% VIDEO

[Attachment 62088 - Click to enlarge]


[Attachment 62089 - Click to enlarge]

[Skiller]

Yep, that was both encoded and flagged as if it was interlaced video, shame. That's just plain laziness.

[Xanthipos]

In an analog SD signal itself, there is no distinction between frames and fields, there are only fields and – consequently – the receiving device (TV, capture card, etc.) has no means of knowing which fields would originally, in case of a 25p source, belong together.

There has to be a distinction between odd and even fields to make it display the right fields together. Otherwise the television set may display the wrong fields together.

[Skiller]

There has to be a distinction between odd and even fields to make it display the right fields together. Otherwise the television set may display the wrong fields together.

The sync pluses offset the fields so that they are alternating positions on the screen (upper and lower), but two fields are never displayed together at the same time but after each other (unless there is deinterlacing or field matching going on).
See a plain CRT TV, which this signal was intended for, just draws a field's lines on the phosphors where it is told by the sync pulses and then the next one which will draw it's lines inbetween the lines of the previous field which is almost completely faded out at this point. There is no start of a new frame or something like that.

Edit: Basically, the "frame" happens in your head/vision due to the speed of the alternating fields.

[hello_hello]

There has to be a distinction between odd and even fields to make it display the right fields together. Otherwise the television set may display the wrong fields together.

And when de-interlacing, most progressive players/TVs convert each field into a full frame by interpolating the missing scanlines, so as long as the field order is correct, the interpolated frame order will be correct. The fields aren't put together as such.
That's one of the reasons why interlaced video is often described as having a "soap opera" look on a progressive display, due to the effective 50p frame rate after the PAL fields are converted to full frames (daytime soap operas were traditionally shot on video).

For digital video, you only need to know which field is first (top or bottom). From there the temporal order should be correct and remain the same, whether it's interlaced or really progressive.

Before support for 50/60p was common in players, for encoding it was also common to de-interlace by combining fields into a full frame. These days it's better to de-interlace 50i as 50p, as a progressive player or TV would, and encode it that way.

Interlaced video also has less motion blur than film. Motion blur is one of the reasons film doesn't look "jittery" at 24fps.
50i de-interlaced to 25p doesn't have the same motion blur and combined with de-interlacing artefacts, can make 50i de-interlaced to 25p look a bit "jittery".

This zip file contains some examples of 25p vs 50p de-interlacing.
https://forum.videohelp.com/attachments/16640-1362616661/encodes.zip (25MB)
https://forum.videohelp.com/attachments/16636-1362614581/original.mkv (18MB)

Yadif at 50fps would be roughly the same quality you'd expect from a player or TV's de-interlacing.

[Xanthipos]

The point is that any digital video in 576p25 can be converted into something any television or video monitor programmed to receive old standard definition broadcasts in 50Hz countries can display simply by storing it in a memory and re-arranging the lines.
Surely there is some way to ensure that the odd numbered lines and even numbered lines always drawn on the correct positions on the television screen, as opposed to drawing the even numbered lines of one frame in the odd positions and then the odd numbered lines of the next frame in the even positions.

[DB83]

Well I thought that everything had been explained to you. And more than once. But read on.


https://en.wikipedia.org/wiki/Digital-to-analog_converter


Also, I still would have thought that in the early days most digital broadcasts, dvd aside, would have been interlaced and not progressive.

[hello_hello]

The point is that any digital video in 576p25 can be converted into something any television or video monitor programmed to receive old standard definition broadcasts in 50Hz countries can display simply by storing it in a memory and re-arranging the lines.

Ignoring the fact that the color information (chroma subsampling) has only one quarter the resolution of the luma information.

DVD was designed to be an interlaced format. For an analog output a DVD player must convert to analog and output 50 fields per second for PAL and 59.94 fields per second for NTSC. If the video isn't interlaced or encoded with hard telecine for NTSC, it contains repeat field information so the player can create a 50i/59.94i output etc. These days, if the video is encoded as progressive, for a progressive display a progressive DVD/Blu-ray player would just ignore any repeat field information and output 23.976p or 25p. If the telecine is encoded, the player should perform pulldown removal and output 24p.

Surely there is some way to ensure that the odd numbered lines and even numbered lines always drawn on the correct positions on the television screen, as opposed to drawing the even numbered lines of one frame in the odd positions and then the odd numbered lines of the next frame in the even positions.

That'd be the sync pluses Skiller referred to earlier, I assume.