Technically Speaking, is a "1080p" Telecine Transfer a Misnomer?

[Karl Lee]

I’m trying to learn a little more about the nuances of telecine transfers (specifically, 24 FPS film to 30 FPS video), and as I’ve been reading more about 3:2 pulldown, I’ve started thinking about HD telecine transfers, many of which are referred to as “1080p”.

 

Considering that a fundamental of 3:2 pulldown is creating 5 frames of 30 FPS video from every 4 frames of 24 FPS film, a process in which 2 of every 5 video frames are interlaced even and odd fields from two different frames of film (the Wikipedia article does a nice job of illustrating this), then is a “1080p” telecine transfer technically a progressive scan transfer? In other words, what would be the difference between a “1080p” and “1080i” telecine transfer? In the video domain, is a frame of video consisting of odd and even fields from two separate frames of film technically considered a “progressive” frame of video?

 

While I understand that moving images captured entirely in the digital domain using digital cameras can be done so using progressive video from the start and thus remain a true progressive signal throughout the editing process, an inherent part of adapting film for video is the 3:2 pulldown, in which some degree of interlacing is an integral part of the process. Regardless of whether 3:2 pulldown interlacing is introduced in a telecine transfer or interpolating is accomplished through some type of smoothing algorithm for a scanned film transfer, it seems that there’s always going to be some degree of interlacing when adapting 24 FPS film to 30 FPS video.

[David Mullen ASC]

HDCAM-SR tape records 24P at 1920 x 1080 but stores it as two fields that are recombined back into a P frame on playback, they call it the PsF format. It's not interlaced-scan 1080i. However broadcast HD is 1080/60i so 24P or 24PsF material would get a 3:2 pull down for display even if not recorded that way. So an HD transfer to HD tape can be progressive, though if you want to avoid PsF, you'd probably get the transfer on a hard drive instead of tape.

[Keith Walters]

HDCAM-SR tape records 24P at 1920 x 1080 but stores it as two fields that are recombined back into a P frame on playback, they call it the PsF format. It's not interlaced-scan 1080i. However broadcast HD is 1080/60i so 24P or 24PsF material would get a 3:2 pull down for display even if not recorded that way. So an HD transfer to HD tape can be progressive, though if you want to avoid PsF, you'd probably get the transfer on a hard drive instead of tape.

The main deficiency of PsF (Progressive Segmented Frame) is that each frame is effectively assembled by interleaving two JPEG-like images, each with half the line count of the original frame. As far as image compression goes, that does not work as well as directly compressing the full frame in one go.

The reason Sony used that system was that in the 1990s they didn't have the technology to make full-frame DCT/Huffman compression work in a battery operated recorder. It works OK for Interlaced scanning because they simply have two image compression "engines" running at the same time. For progressive scanning, the best they could do was splitting the progressive frames into an interlaced format and then afterwards treating them as an interlaced scan.

 

The original HDCAM format (as used on Star Wars 2) only has a horizontal resolution of 1440 pixels, because that is the horizontal resolution of 1080i. Effectively there is very little difference between 1080i and 720p on modern LCD TVs. Most of the processing TVs have to do is converting interlaced material back into a watchable progressive format, which is hellishly difficult sometimes.

 

It is not generally known that the reason interlaced scan formats only have about two-thirds the resolution of the equivalent progressive format, is that it is essential to reduce the vertical resolution by about a third, to avoid edge flicker on interlaced scan displays. The assumption was that interlace flicker would determine the distance that viewers sat from the screen. and that any more than about two-thirds the actual horizontal resolution would be a waste of bandwidth, because most viewers would not see it!

 

Well, that was one assumption; the main assumption was that HDTV was mainly going to be watched on CRT TVs or projectors. So in a lot of cases we're still stuck with video standards tailored to match the deficiencies of 1930's TV technologies....

[David Mullen ASC]

My point is that recording a HD signal to tape doesn't necessarily mean you have to record 60i with a 3:2 pulldown for 24 fps material. You could store it as 24 PsF, which is progressive for all intents and purposes. And the compression of HDCAM-SR in HQ mode is 880 mbits/sec with a 2:1 compression, which is pretty low, good enough for most TV shows to archive to HDCAM-SR even today.

 

If Karl wants to telecine 24 fps film material to HD for a final release in 1080P on blu-ray, then he could use HDCAM-SR tape if he didn't want to work with data files directly. Just because he masters onto HDCAM-SR doesn't mean his project will be 60i with a 3:2 pulldown.

 

However, HD broadcast is either 1080i or 720P so at some point, the only people who would get to watch his project in 1080P would be on blu-ray or on the internet.

[Keith Walters]

And the compression of HDCAM-SR in HQ mode is 880 mbits/sec with a 2:1 compression, which is pretty low, good enough for most TV shows to archive to HDCAM-SR even today.

 

I'm not saying people shouldn't do that, but they should be able to make an informed choice. No matter how you slice it, they're still be going to be archiving only a small part of the available detail from the camera negative.

 

To cut to the chase, I don't know why anybody wouldn't just have the film scanned at 1080p at whatever the original frame rate was. You don't need to deliver 1080i of any sort, all modern TV stations can do the necessary conversion. You can get freeware packages now that can edit and render full 1080p! Once again Interlaced deliverables are an outdated 1990s concept,

 

the compression of HDCAM-SR in HQ mode is 880 mbits/sec with a 2:1 compression, which is pretty low,

 

Two points need to be made here

1. HDCAM has a fixed data file size per field because the digital data "package" simply replaces the FM video carrier of an analog recorder. The notion of 2:1 compression is somewhat misleading, because amount of data to be stored depends on the particular image. A completely white screen for example has virtually no data in the compressed version, so in that

case it would be vastly more than 2:1. The recorded file size is always the same, but most of it is "filler" data packets.

 

2. The modest compression rate is more a reflection of the limitations of 1990s compression technology than any consideration of actual image quality. Just because it's only 2:1 does not guarantee high image quality, in fact the early F900s were notorious for falling over when hit by data bottlenecks (as seen on the originasl release of Star Wars 2)

 

The bottom line here is that I think Karl is confusing the best Sony could do two decades ago, with what is best practice today.

 

 

[Robert Houllahan]

You can also record direct to disk as Uncompressed or with light compression like ProRes444 or HQ as 1080P 23.98 or 24.00 P

[Karl Lee]

Thanks to everyone who has replied and for all of the information...I think I know more now about the topic than I would have ever expected!

 

Going back to my original post, though, I guess my main question was that unlike a film scan which deals with discrete scans of each individual frame, can an HD film telecine (at "24", or 23.976 FPS) destined to HD 30 / 29.97 FPS video be technically considered a "progressive scan" end product? The reasoning behind my question is that while the telecine camera/sensor/optics certainly would be capable of producing a 1080p video conversion of individual film frames, in 3:2 pulldown, the (video) frames composed of fields from two different (film) frames are likely a product of digital processing that synthesizes and combines the two fields into a single frame of video, as opposed to being a true, optical 1080p video representation of an individual frame of film. With that in mind, I'm just curious if these "combined" video frames are technically considered a "progressive" video product. It's certainly possible that since the "combined" video frames are generated from individual film frames that were captured in 1080p format, then the resulting combined video frames are still regarded as being "progressive" video.

 

Does this make any sense? I do have a habit of over-analyzing things :)

[David Mullen ASC]

I thought I answered that. Yes, you can store film transferred to progressive-scan HD video on tape without a 3:2 pulldown, it would just be 24P or PsF (or 23.976). It would usually get a 3:2 pulldown added from the deck if the recorded 24PsF signal were sent to a 59.94i monitor.

 

If you want to store it as 59.94i, then yes, 24 fps material would need a 3:2 pulldown.

 

You can release 24P HD on a blu-ray and play it back on a flatscreen monitor and it would stay progressive throughout the chain.

 

PsF is not the same as interlaced scan. Interlaced scan fields are sequential images and thus when combined into a frame, movement causes a sawtooth edge along objects. PsF merely stores a progressive frame as two fields containing alternate lines and then recombines them back into a progressive frame on playback.

 

Pulldown is a separate issue and relates to showing 24 fps material at 60i. You don't have to record pulldown onto tape if you don't want to, you can store 24 fps material at 24P instead of 60i.

 

But I thought I said most of this in my first reply so obviously something is not sinking in...

[Keith Walters]

PsF merely stores a progressive frame as two fields containing alternate lines and then recombines them back into a progressive frame on playback.

It's not exactly the same as true 24p though.

In PsF, every 8x8 pixel block in the 24p output is actually made up of four rows of 8 pixels from one "JPEG" (A) interleaved with four rows of 8 pixels from a different "JPEG" ( B ) like this:

 

A-A-A-A-A-A-A-A

B-B-B-B-B-B-B-B

A-A-A-A-A-A-A-A

B-B-B-B-B-B-B-B

A-A-A-A-A-A-A-A

B-B-B-B-B-B-B-B

A-A-A-A-A-A-A-A

B-B-B-B-B-B-B-B

 

Each 8 pixel block of the "JPEG"s on the other hand, is derived from eight rows of eight pixels from the original image but the eight rows are actually "every other row", because the original progressive scan is split into two fields, making a "pseudo-interlaced" scan.

 

So the 8x8 pixels of "JPEG A" are actually derived like this, the numbers (1-1-1 etc) represent 8-pixel segments of the rows of original image pixels, starting from the top:

1--1--1--1--1--1--1--1

3--3--3--3--3--3--3--3

5--5--5--5--5--5--5--5

7--7--7--7--7--7--7--7

9--9--9--9--9--9--9--9

11-11-11-11-11-11-11-11

13-13-13-13-13-13-13-13

15-15-15-15-15-15-15-15

 

And the 8x8 blocks of "JPEG" B are derived like this:

2--2--2--2--2--2--2--2

4--4--4--4--4--4--4--4

6--6--6--6--6--6--6--6

8--8--8--8--8--8--8--8

10-10-10-10-10-10-10-10

12-12-12-12-12-12-12-12

14-14-14-14-14-14-14-14

16-16-16-16-16-16-16-16

 

That does NOT give the same result as directly encoding the full progressive scan, regardless of what George Lucas told us!

The DCT processing can't incorporate the pixel values of the"off-site" pixels into the derived coefficients, so there will be encoding innacuracies when dealing with "busy" images, which is exactly what happens. (Two glaring examples were rippling water and a grassy field in the original release of SW2).

 

Certainly it works well enough non-critical TV production, but it was nowhere good enough for large-screen production.

[David Mullen ASC]

SW2 was shot in HDCAM though, not HDCAM-SR. PsF but HDCAM is only 3:1:1 and a lot more compressed.

 

I shot a TV series for three years on the Genesis and I can tell you that PsF doesn't look like interlaced-scan video, there are no interlaced-scan motion artifacts.

 

To answer the original question, a telecine does not have to store HD to tape as interlaced scan with a 3:2 pulldown.

[cole t parzenn]

"United States of Tara?" I'm in the middle of binge watching that. It was a good looking show, for whatever my opinion's worth.

[David Mullen ASC]

Thanks, I only wished that it had come out on blu-ray...

 

It's crazy that in this day and age when some distributors are mandating 4K origination now that some movies and TV shows are only be distributed in standard def video, we live in a world where people are watching on 50" HDTV screens and iPhone screens... I shot this indie movie "Big Sur" in 5K on the Epic because it was a landscape movie, we only had the budget to finish it in 2K, and the distributor only released it on 480P DVD because that's what WalMart told them to do!

[cole t parzenn]

If it makes you feel any better, I'm seeing it in HD, albeit compressed HD, on Netflix. Again, for whatever my opinion's worth, UHD is silly; the problem with Rec. 709 is the colors, not the pixel count; it makes more sense to just use 2K. But I think that Super Technirama 70 should be the standard format for 2.39 material, so...

[Keith Walters]

I shot a TV series for three years on the Genesis and I can tell you that PsF doesn't look like interlaced-scan video, there are no interlaced-scan motion artifacts.

Of course not, because with a progressive scan camera the two "interlaced" fields are taken from the same image. It's exactly equivalent to scanning 30fps film to produce 30 frame (29.97 to be precise) interlaced NTSC.

 

I wasn't talking about motion artifacts, it's DCT encoding/decoding innacuracies I was referring to.

 

It's certainly vastly easier to turn a progressive scan source into a signal that can be displayed on an interlaced-scan CRT TV than the other way round. But there is the great irony: How many Interlace Scan CRT HDTVs are there? Virtually none. But most of the current Digital TV transmissions in the world are interlaced 576i/50 or 480i/59.94, with no real need for either...

 

Actually it's quite instructive to use a basic Freeware such as Avidemux, which makes no attempt whatever to correct the interlace-to-progressive issue on the editing screen.

[Keith Walters]

Thanks, I only wished that it had come out on blu-ray...

 

It's crazy that in this day and age when some distributors are mandating 4K origination now that some movies and TV shows are only be distributed in standard def video, we live in a world where people are watching on 50" HDTV screens and iPhone screens... I shot this indie movie "Big Sur" in 5K on the Epic because it was a landscape movie, we only had the budget to finish it in 2K, and the distributor only released it on 480P DVD because that's what WalMart told them to do!

Not that uncommon, but at least they have the option of re-finishing in 4K.

That's always the problem with electronic orignation: Whatever resolution you shot it in; that's all the resolution it's ever going to have. Meanwhile we get digitally restored re-releases of Technicolor movies from over 70 years ago that come up looking and sounding far better than the original audiences ever saw them...

 

It's crazy that in this day and age when some distributors are mandating 4K origination now that some movies and TV shows are only be distributed in standard def video, we live in a world where people are watching on 50" HDTV screens and iPhone screens

 

Or people buying UDTVs so they can mostly watch over-compressed video with essentially the same resolution that was being broadcast in 1941. Or illegal downloads of current release movies shot with a low-end consumer video camera.

Or Google Hangouts where you have to get people to wave so you can identify them :rolleyes:

 

Go figure.