NTSC HDTV is usually broadcast at 23.976fps, yes?

[Gew]

Hi!

I live in a PAL country, therefor this is more or less a mystery to me (I can't just zap and check). You see, I was wondering. I've noticed, that 99.999% of all "pirate rips" of NTSC HDTV stuff is encoded at 23.976fps. Not only tv series, but talk shows, well, for most part, everything. So I've come to wonder, does ie. Fox/USA broadcast their HDTV in this framerate, or is it broadcasted as 29.97fps (like old days SDTV), and then IVTC'd ie. by scene groups..? I've come to believe that NTSC HDTV is for the most part broadcast as progressive scan, even tho (source: Wiki: HDTV) I've read that the 60Hz system is capable of a wide range of distributions rates/scan modes.

I'm more or less wondering, ie. is Jay Leno Show at broadcasted as progressive 23.976 at NBC HDTV, or 29.97 (interlaced HDTV, errr??) and then IVTC'd by cappers?

Forgive me if acting confused, it's just how I am.

[jagabo]

NTSC over-the-air HDTV is broadcast as 30i (3:2 pulldown with film based material) or 60p (3:2 frame repeats with film based material). I'm pretty sure Leno shot on video is broadcast at 30i (60 different fields per second).

[Gew]

Ty for a fulfilling answer. Makes sense. Just about what I suspected. Quite funny that I couldn't dig that valuable one-liner answer up in the Wikipedia HDTV article though.

[jagabo]

ATSC supports other frame rates but I don't know how often they're used:

http://en.wikipedia.org/wiki/ATSC_%28standards%29#Resolution

[Gew]

Oh. This may be kind of a stupid followup-question, but I couldn't help it came to me. Where I've learned from another thread how there's "soft" pulldown and "hard" pulldown, where you can simply Force film in DGIndex to invert soft pulldown, but need to do TFM().TDecimate() (Avisynth) when there's hard pulldown.

So, my question is simply this, I presume that you cannot "capture" the "pulldown flags" (which more or less 'metadata') that a soft 3:2 pulldown means, so that you will always need to do TFM().TDecimate() with captured HDTV material. Is this a correct assumption?

[manono]

So, my question is simply this, I presume that you cannot "capture" the "pulldown flags" (which more or less 'metadata') that a soft 3:2 pulldown means, so that you will always need to do TFM().TDecimate() with captured HDTV material. Is this a correct assumption?

No, not always. If it's 29.97fps and hard telecined, then yes, you IVTC. If it was shot on video, then no. And if it's progressive 59.94fps and 720p and film sourced, you get it back to 23.976fps using other methods (I would SelectEven().TDecimate()). If it's 720p and video sourced, you just SelectEven().

[Gew]

A good thing to att to my own knowledge database, that 59.59p path.
However, still, there's no (never) such thing as soft telecining being applicable on captured HDTV material..?

[manono]

As far as I know, no.

[txporter]

Thanks for this thread. I didn't know before that 720p material had so many repeat frames. Are sports actually captured and broadcast with ~60 unique frames/sec? I guess I don't understand what benefit there is in a 59.94fps stream that doesn't contain any information that a 23.976fps contains.

What about 1080i broadcast TV shows? Are they also built from 23.976fps progressive film?

[jagabo]

Are sports actually captured and broadcast with ~60 unique frames/sec?

Yes.

I guess I don't understand what benefit there is in a 59.94fps stream that doesn't contain any information that a 23.976fps contains.

Movies are shot at 24 fps. So they have to repeat frames to make 60 fps. Repeat frames don't take much bitrate in MPEG streams. The encoder simply says "repeat the last frame again."

What about 1080i broadcast TV shows? Are they also built from 23.976fps progressive film?

Movies go through 3:2 puldown to make 60 fields per second, packaged as 30 interlaced frames per second. Live video is 60 fields per second, packaged as 30 interlaced frames per second.

[txporter]

And if it's progressive 59.94fps and 720p and film sourced, you get it back to 23.976fps using other methods (I would SelectEven().TDecimate()). If it's 720p and video sourced, you just SelectEven().

Is there anything wrong with using SelectEven().TFM().Tdecimate() on 720p film sourced? My 720p material are broadcast captures where I remove commercials with VideoRedo. I was wondering if I need to process to look for best matches or something like that.

[jagabo]

720p should already be progressive and should not have any interlaced frames to field match. Using
SelectEven().TFM().Tdecimate() on it -- TFM() is a waste of time and can only lead to possible artifacts (where it mistakenly creates an interlace frame from two progressive frames).

[txporter]

Movies go through 3:2 puldown to make 60 fields per second, packaged as 30 interlaced frames per second. Live video is 60 fields per second, packaged as 30 interlaced frames per second.

So I would just do a standard TIVTC to get back to the original stream? The fact that they are interlaced frames confuses me.

720p should already be progressive and should not have any interlaced frames to field match. Using
SelectEven().TFM().Tdecimate() on it -- TFM() is a waste of time and can only lead to possible artifacts (where it mistakenly creates an interlace frame from two progressive frames).

Excellent. Thanks!

[manono]

So I would just do a standard TIVTC to get back to the original stream?

You would if the original stream was telecined film and not video.

[txporter]

So I would just do a standard TIVTC to get back to the original stream?

You would if the original stream was telecined film and not video.

Ok, great. So am I left with 23.976fps frames that are interlaced, or are they progressive? Do I set the --interlaced flag on x264 if I am sending the avisynth output into that?

[manono]

If you're IVTCing to restore it to film then, by definition, it's progressive.

[poisondeathray]

yes, if you ivtc for content that was telecined (not interlaced) , you will be left with 23.976 film rate and progressive content

the --interlace switch in x264 is for encoding to interlaced content

[edDV]

While ATSC permits 18 different digital broadcast formats (resolution/frame-rate) only three are commonly used for broadcast (i.e. 1920x1080i/29.97, 1280x720p/59.94 and 704x480i/29.97) and each of those use different techniques for live and film presentation.

I'll assume for now that you understand classic live NTSC 480i and 480i with pulldown (telecined). 1920x1080i/29.97 behaves the same way. Only the resolution differs.

The unusual format is 1280x720p/59.94 as used by the following networks:
ABC, FOX, Disney, ESPN (sports), National Geogragphic and PBS* (as broadcast in many communities).
(Edit: Also A&E, History, Biography and other "cable networks")

For live broadcasts (including sports, performance, news, chat shows and "reality" programming) 59.94 unique progressive frames are transmitted. Some source is native 59.94p, other is smart bobbed from 1080i/29.97 source. Conversion of 1080i/29.97 to 720p/59.94 is near lossless. The reverse is not true due to the need for resolution upscale.

24p (23.976p) film based source is common for movie, TV series, advertisng and often for show intros and/or bumpers. For 1280x720p, 23.976 frames are nominally repeated 3,2,3,2 although some networks alter the stream adaptively to motion. This is often done for animations. I don't know if their motivation is increased motion accuracy or to make capture/copy difficult. One must analyze the captured source for frame repeat patterns.

There are a couple of exceptions to the above. While most Fox local stations broadcast center cropped 480i/29.97 on their SD subchannels, some broadcast network programming as 16:9 480p/59.94 which displays as telecined letterbox on 480i/29.97 televisions but as 16:9 480p/59.94 on wide digital sets. This is mostly done by Fox owned local stations.

The other exception is 720p networks or stations might broascast over the air in one standard, but feed cable/sat systems a separate format optimization. Since sat systems require a setbox tuner/decoder, the uplink feed can be dynamically switched between 1280x720p/59.94 (unique frames) and 1280x720p/23.976 (unique frames) with flags. This eliminates uplink of repeat frames for greater transmission efficiency. The repeat frames are added back at the set top box but some brands of tuners will allow passing 1280x720p/23.976 to the TV. The SD feeds can also be dynamically switched between 704x480i/29.97 and 704x480p/23.976 for transmission effeiciency.


*The PBS networks offer 1920x1080i/29.97, 1280x720p/59.94 and 704x480i/29.97 sat feeds to their stations. In many communities, the local station will offer the primary channel in 720p and add two or three 480i subchannels. In other communities, a 1080i primary channel is offered with one 480i sub.

[txporter]

I guess my confusion stems from the fact that it was a 1080i broadcast. Doesn't that mean it is interlaced? Don't I need to de-interlace somehow? Or are 30 interlaced frames (as opposed to fields) basically the same thing as 30 progressive frames?

[txporter]

While ATSC permits 18 different digital broadcast formats (resolution/frame-rate) only three are commonly used for broadcast (i.e. 1920x1080i/29.97, 1280x720p/59.94 and 704x480i/29.97) and each of those use different techniques for live and film presentation.

I'll assume for now that you understand classic live NTSC 480i and 480i with pulldown (telecined). 1920x1080i/29.97 behaves the same way. Only the resolution differs.

The unusual format is 1280x720p/59.94 as used by the following networks:
ABC, FOX, Disney, ESPN (sports), National Geogragphic and PBS* (as broadcast in many communities).

For live broadcasts (including sports, performance, news, chat shows and "reality" programming) 59.94 unique progressive frames are transmitted. Some source is native 59.94p, other is smart bobbed from 1080i/29.97 source. Conversion of 1080i/29.97 to 720p/59.94 is near lossless. The reverse is not true due to the need for resolution upscale.

24p (23.976p) film based source is common for movie, TV series, advertisng and often for show intros and/or bumpers. For 1280x720p, 23.976 frames are nominally repeated 3,2,3,2 although some networks alter the stream adaptively to motion. This is often done for animations. I don't know if their motivation is increased motion accuracy or to make capture/copy difficult. One must analyze the captured source for frame repeat patterns.

There are a couple of exceptions to the above. While most Fox local stations broadcast center cropped 480i/29.97 on their SD subchannels, some broadcast network programming as 16:9 480p/59.94 which displays as telecined letterbox on 480i/29.97 televisions but as 16:9 480p/59.94 on wide digital sets. This is mostly done by Fox owned local stations.

The other exception is 720p networks or stations might broascast over the air in one standard, but feed cable/sat systems a separate format optimization. Since sat systems require a setbox tuner/decoder, the uplink feed can be dynamically switched between 1280x720p/59.94 (unique frames) and 1280x720p/23.976 (unique frames) with flags. This eliminates uplink of repeat frames for greater transmission efficiency. The repeat frames are added back at the set top box but some brands of tuners will allow passing 1280x720p/23.976 to the TV. The SD feeds can also be dynamically switched between 704x480i/29.97 and 704x480p/23.976 for transmission effeiciency.


*The PBS networks offer 1920x1080i/29.97, 1280x720p/59.94 and 704x480i/29.97 sat feeds to their stations. In many communities, the local station will offer the primary channel in 720p and add two or three 480i subchannels. In other communities, a 1080i primary channel is offered with one 480i sub.

Thanks for all of the detail, edDV. Up to this point, I have really only dealt with manipulation of DVD material (SD). It very interesting to learn about what's going on with HD transmissions. I was surprised (although I guess I shouldn't have been) to find that there were so many repeated frames in my 720p recordings. I guess either Time Warner doesn't utilize the repeat flags for transmission efficiency or my Tivo records the repeat flags as full frames?

I have seen discussions before about temporal resolution and such when folks were debating the merit of 720p vs 1080i. I assume that since I am just stripping out repeat content, that I am not really giving anything up when I take a 720p 59.94fps stream and return it to its 720p 23.976fps original?

[poisondeathray]

I guess my confusion stems from the fact that it was a 1080i broadcast. Doesn't that mean it is interlaced?

Nope; it might be either. It can be progressive content "packaged" in an interlaced stream , an example of that is telecine. The 23.976 progressive frames are essentially "packaged" in a 30i container to make it compatible with NTSC systems

Don't I need to de-interlace somehow?

Not necessarily, if it's progressive content (but just packaged in an interlaced format for transmission), you usually want to recover the progressive frames

If it's interlaced you might want to leave it interlaced, or deinterlace depending on what your target goal was

Or are 30 interlaced frames (as opposed to fields) basically the same thing as 30 progressive frames?

Nope. "30i" is the same thing as "60i" , just different naming conventions. They both mean 30 frames per second, or 60 fields per second. And neither is the same as 30p. Just to confuse you more, you have to differentiate the signal from the actual content. It's the actual content that matters. You can have 23.976 progressive frames packaged in that 30i stream - again that's telecine

[txporter]

I guess my confusion stems from the fact that it was a 1080i broadcast. Doesn't that mean it is interlaced?

Nope; it might be either. It can be progressive content "packaged" in an interlaced stream , an example of that is telecine. The 23.976 progressive frames are essentially "packaged" in a 30i container

Don't I need to de-interlace somehow?

Not necessarily, if it's progressive content (but just packaged in an interlaced format for transmission), you usually want to recover the progressive frames

If it's interlaced you might want to leave it interlaced, or deinterlace depending on what your target goal was

Or are 30 interlaced frames (as opposed to fields) basically the same thing as 30 progressive frames?

Nope. "30i" is the same thing as "60i" , just different naming conventions. They both mean 30 frames per second, or 60 fields per second. And neither is the same as 30p

Thanks for all of the help. I think I understand what I need to do now, even if I can't totally wrap my mind around packaged progressive content, et al.

One additional question, if I use the same constant rate factor (21) to encode a 1080i mpeg2 file (run through TIVTC) once with the --interlaced flag and once without, the final bitrate varies by 10-20%, with the --interlaced encode being smaller. Does anyone understand why that is?

[poisondeathray]

One additional question, if I use the same constant rate factor (21) to encode a 1080i mpeg2 file (run through TIVTC) once with the --interlaced flag and once without, the final bitrate varies by 10-20%, with the --interlaced encode being smaller. Does anyone understand why that is?

You're not doing this correctly.

Assuming you had FILM content, and you TIVTCed it, you would have progressive content. You cannot encode interlaced from that.

You need interlaced content to encode interlaced, you don't have the temporal resolution 23.976 vs. 60 Hz.

In order to encode interlaced content properly, you need a 60p source (60 moments in time), interlace it in avisynth or other program, then encode as interlaced (or start with an interlaced source)

Interlaced encoding is inefficient for a number of reasons (not just limited to x264). e.g. predicitve motion vectors get truncated, search algorthims cannot traverse fields. Many features are disabled in x264 by using --interlaced

[edDV]

I guess my confusion stems from the fact that it was a 1080i broadcast. Doesn't that mean it is interlaced? Don't I need to de-interlace somehow? Or are 30 interlaced frames (as opposed to fields) basically the same thing as 30 progressive frames?

1920x1080i/29.97 is transmitted interlace but film source can be inverse telecined to 23.976 progressive. Many newer shows have clean telecine patterns, but quite a few have broken cadence. Software inverse telecine often fails to correct cadence breaks. HDTV sets adaptively react to breaks* and resync cadence.

* usually substituting a blend deinterlace until resync.

[txporter]

@pdr- Yeah, I figured it was wrong...I just didn't know what was happening. I will just leave it alone until I need to try encoding some Dr. Who episodes or something like that.

[poisondeathray]

@pdr- Yeah, I figured it was wrong...I just didn't know what was happening. I will just leave it alone until I need to try encoding some Dr. Who episodes or something like that.

This topic confused the hell out of me for the longest time. You won't figure it out until you actually go through some examples of true interlace vs. telecined progressive content vs. native progressive content. (there are other patterns but those 3 are the most common ones)

This mini guide shows you how to do it in avisynth
http://neuron2.net/faq.html#analysis

Just remember, it's the actual content that matters, and you determine that by separating the fields and looking for patterns.

[txporter]

@pdr- Yeah, I figured it was wrong...I just didn't know what was happening. I will just leave it alone until I need to try encoding some Dr. Who episodes or something like that.

This topic confused the hell out of me for the longest time. You won't figure it out until you actually go through some examples of true interlace vs. telecined progressive content vs. native progressive content. (there are other patterns but those 3 are the most common ones)

This mini guide shows you how to do it in avisynth
http://neuron2.net/faq.html#analysis

Just remember, it's the actual content that matters, and you determine that by separating the fields and looking for patterns.

Yes, that page is great. I do actually use that to understand the video. I was very confused initially by the 3 2 3 2 pattern of 720p episodes. edDV explained it well. When I looked at 1080i material the same way, it did look much like telecined DVD content. I was looking for the combing though, which didn't seem evident. I guess that's because it wasn't true interlaced content.

[edDV]

Thanks for all of the detail, edDV. Up to this point, I have really only dealt with manipulation of DVD material (SD). It very interesting to learn about what's going on with HD transmissions. I was surprised (although I guess I shouldn't have been) to find that there were so many repeated frames in my 720p recordings. I guess either Time Warner doesn't utilize the repeat flags for transmission efficiency or my Tivo records the repeat flags as full frames?

Most cable systems pass the repeat frames for 720p although the cable boxes can accept 23.976p and convert that to all the export formats. Same goes for ATSC off air tuners. They are required to accept 23.976p in 704x480p, 1280x720p or 1920x1080p.

I have seen discussions before about temporal resolution and such when folks were debating the merit of 720p vs 1080i. I assume that since I am just stripping out repeat content, that I am not really giving anything up when I take a 720p 59.94fps stream and return it to its 720p 23.976fps original?

Only if the source is film. "Live" 720p has 59.94 unique frames per second for improved temporal resolution.

[txporter]

Only if the source is film. "Live" 720p has 59.94 unique frames per second for improved temporal resolution.

Got it. Thanks. That's what I expected to hear.

[jagabo]

I think I understand what I need to do now, even if I can't totally wrap my mind around packaged progressive content, et al.

The second part of this post might help you understand what interlaced video actually is:

https://forum.videohelp.com/threads/316415-Check-Over-My-Options-From-DVD-to-XviD-Conve...=1#post1958565