Thanks. This isn't going out to film, I just want to treat it as if it was film in the first place. There will be a lot of visual effects work done on it, so naturally I don't want to deal with fields in any way.

Ideally this will be broadcast on cable/pay cable (probably not HD) as well as DVD. I want to have it in HD for the future rather than have to prepare another version.

If that's what you're looking for, I'd shoot CF25, and then slow that by 4%, or rather, just let Vegas do the conversion on its own.
You'll be most happy with it, I'd wager.

I've been thinking about this a lot lately and for SD or even 720p output one could probably pull off a neat trick. 1080i has each field with about the same res as 720 and more than enough for SD. So technically one could just ditch every second field and convert the remaining fields to frames to give 720p at 25fps which as said is easy enough to convert to 24fps.
Only issue would be shutter speed, you'd still be missing the motion blur of the slow shutter at real 24/25 fps.
And just to confuse us it looks like the EBU will go for 720p at 50fps for HD.
Bob.

yeah, who knows what the EBU is really thinking...? That's a strange one indeed. There aren't many cams that even can shoot 720p60, let alone anything else. Worse, you gotta wonder if they are considering the market at all, with most displays soon coming off at 1080p, and games like PS3 being 1080p, and BD being 1080p60 ready. Something seems afoul. One of the EBU members made a really (IMO) stupid comment about "1080 is just too much resolution." Huh? Maybe if you're in a small room, seated 1.5 metres away...but other wise....

The EBU is thinking that interlaced is dead technology, and that the European customers would be far better served by progressive. So they've announced that of the existing formats that are available today, 720/50p is their choice. And they're keeping a strong eye towards the future, where they want 1080/50p. They believe the 1920x1080 frame is ideal, but they want nothing to do with interlace scanning or display.

They acknowledge that interlace production will probably occur, and they said they have no interest in prohibiting it, but that they strongly endorse progressive-scan acquisition and progressive-scan "emission" (aka broadcasting).

To read an editorial from Philip Laven, Director of the EBU Technical Department, click here:
http://www.ebu.ch/departments/technical/trev/trev_301-editorial.html

Yeah, I'd read that article from when you posted it on DVInfo, but it seems strange to establish a lower grade standard when it only hurts the consumer in the grand scheme of things.
For short term, a 720p standard is fine. But for long term, I'm just glad I don't live there.

I was kind of surprised too, I do like the idea of 50p though, better temporal resolution is the way to go. What I do find odd is the 720 res, PAL at 16:9 has enough resolution for free to air broadcasting (as opposed to home cinema) but if you want a 'wow' factor I just don't think 720 is a big enough step up, coming from NTSC to 720 is a bigger step though.
What they're wisely factoring in is ongoing research by the BBC about viewing distances and image size, their results show viewer preference hasn't changed since the early days of TV broadcasting.
I'd make much the same argument for radio broadcasting, how many listeners would benefit from broadcasting in 24/96, one could well argue the benefits for listening as opposed to just something in the background, which is how most radio and TV is viewed.

Of course I'm a bit over the hill, maybe the next generation will have a different view but so far the 'progress' seems to be downhill, with everyone quite happy with horrid mp3 compression!

Yes I'd agree, the market is apparently trying to go in different directions at the same time.

For web video purposes, we are looking for ever smaller file sizes and lower bit rates, while at the same time, 720p is considered by some to be insufficient for home viewing.

On the audio side, the SACD and DVD audio camps are promoting their incompatible "HD" audio formats, while many users consider MP3 playback on DVD players to be very attractive. This is hardly a sign that users are interested in the best quality, HDV's MPEG1 Layer 2 audio acquisiiton is even worse than MP3, so I suppose it's clear where the HDV consortium are aiming.

What to do?

Richard Hunter

I keep wondering whether HDV, being in MPEG format, is problematic in applying FX and color correction in post? That's why I bought a DVX100, to use progressive frames.

Which reminds me, the HVX200 is hitting the market soon, no?

On original topic and a little product plug. If you are to shoot in either CineFrame 24 or (preferrably) CineFrame 25, the CineForm tools for capturing has 24p extraction (to the best of the cameras image.) Here is a link that explains the CineFrame modes : http://www.cineform.com/products/SonyHDVSupport/CineFrame.htm

David Newman
CTO, CineForm

Dan,
what's the advantage of shooting CF25, processing that through your tools versus shooting 50i and using de-interlacing with motion compensation?
Bob.

I'm not David, but let me throw something out there to think of...

I think CF25 is better for two main reasons: 1) compression, and 2) predictable performance.

First, CF25 is a consistent 540 lines at all times. It's one field of the interlaced video, so yes your resolution is substantially lower, but with MPEG-2 that helps the compression out as well. You'll likely encounter fewer MPEG compression artifacts from CF25 footage than you would from 50i.

Second, predictable performance: again, tied to the same thought above, CF25 is always 540 lines. With software de-interlacers you can get "smart de-interlacing", which will only de-interlace (and thus "progressive-ize") moving subjects. So the background can be displayed at the 775 lines of resolution that a full "frame" of interlaced HDV is capable of, but a moving subject would have half that resolution. What happens if you have a stationary object that starts to move? Instantly it'll become half-resolution. If it stops moving? It'll "sharpen up" to full resolution. And while it's moving, it'll be substantially lower resolution than it would have been had you just shot in CF25 in the first place.

CF25 is not a great solution for filmlike looks; you'd be a lot better off with a progressive-scan system that takes advantage of the full resolution of the frame (1080/24p could have 1080 lines of res, vs. the 540 max for CF25). But if you're going to be using an interlaced camera, and want to try for the filmlike look, I think CF25 is probably the best way to go. And it's massively superior to CF24!

David, thanks. I took a look at the Cineform article and it's very interesting.

CF25 is 540lines? I can't find that in searching the CF site and documentation. Can you give me a pointer please?

Lets examine that in some detail.
Lets assume for the sake of argument that our video system is capable of 1440x1080. Shooting progressive each frame contains a full set of 1440x1080 pixels. Shooting interlaced each field contains an effective 1440x540 pixels, that's 777K pixels. Compare that to 720p at 921K and we're not looking too sad. Smart interpolation could make good use of the horizontal res to give a result somewhat better than 540 lines.
Now all this assumes we're just ditching every second field. As Barry pointed out we can for static objects blend the second field to up the res in my simple example to 1080 lines. Now as he rightly points out when things move we have a problem. Now in a simplistic approach all we can do is ditch those bits and interpolate between lines so yes res goes down on moving objects.
However isn't there a solution to this, adaptive motion compensation. Calculate a motion vector and move those bits back where they belong. From the little I know of this technology it's the preserve of rather expensive hardware boxes but it can and is done and as sure as the day is long it's becoming more affordable.
But there's another issue, quite regardless of de-interlacing issues. Anyone shooting for a 'film look' is using slow shutter speeds, result, moving objects get blurred anyway.
Now for my money the whole 'film look' is one of the greatest cons of the last century but if that's your thing I can't see that it matters much if your moving objects acquire an unnatural blur due to cheap de-interlacing or due to slow shutter speeds.
Bob.

Hi Bob. Interesting. Barry's description was clear enough but I'm unable to find that among the CF literature (where it apparently is). Motion blur is actually desirable, otherwise motion looks jittery. I guess you're not advocating using high shutter speed for freezing motion? There's not much difference between 1/50 and 1/60 in terms of motion blur. On "film look" I substantially agree with you that the search is often misguided, as you'll have gathered from my various comments on the topic. Generally the contrast ratio of outdoor scenes exceeds the range of 8bit CCDs, so unless one is happy with artifacts like burnt out highlights then either you reduce the scene dynamics to fit video capabilities, or extend the capabilities. So much depends on how the data is processed in the camera. Film has dynamic range adequate for most outdoor scenes. But I don't think you're arguing that a small dynamic range is a good thing; otherwise read Bergeron's paper. I use 16bit CCDs in astronomy where we have a very large contrast ratio and are really looking to extract a lot of detail at both ends of the range, hence we're processing with extremely non-linear characteristics. 8bits in video is really pretty inadequate but that's what we've got.

cheers,
Serena

Thanks Barry,

That would have been my argument precisely. I would also add the workflow simplification, by shooting CF25 you directly have your source material ready for editing. If you shoot 50i or 60i you have to generate 24p output before you can edit -- the tools for do this are incredibily time consuming, and the quality performance will vary. CF25 is a good compromise when shooting a interlace camera.

Hi Serena,
high shutter speeds and low frame rates are bad, local sports stuff is always shot this way, I think just for the benefit of good slomo.
No, the answer in my humble opinion are faster frame rates, I think the EBU have it right going for 50p but I don't understand why they didn't go for 60p, we could have finally had one universal standard.
Totally agree with your comments about lack of latitude in video, that's the biggest single factor contributing to the 'film look'.
Something I'm going to try when time and money permits, there's a camera built for the security industry that pulls an interesting trick to get a huge latitude, take the same frame twice an different shutter speeds and process that in a DSP. Result is you can see the interior of an offfice and what's outside the window isn't blown out, the part of the frame that's the office interior is shot at say 1/50th, the exterior at maybe 1/100,00th. The smarts in the DSP composites the two together.
I'm wondering if the same thing but at higher quality and res could be done using two genlocked cameras sharing the same lens through a splitter and running the two cameras at different shutter speeds, I've seen guys doing this with digital still cameras, use the auto bracket function and composite all the shots together.
Bob.

Bob, that sounds like an interesting project! I guess we're wanting CCDs with much greater well depth (16bit is enough -- a contrast ratio of 16 stops) and prefereably on all RGB channels (although 16bit illuminance with lesser data on RGB is enough). Put that together with high frame rates (is 60 progressive enough?) and the bottle neck is in CCD readout and data recording. Got to be the end of tape! The camera will contain a "wow" of a processor, but we could do all that off-line (be like the old days of film processing). This all sounds "beyond blue", but where were we 10 years ago?

Serena

>>Can you give me a pointer please?<<
My point of reference was shooting a resolution chart. It delivers what looks to be 575 lines, which is close enough to 540 to say that yes, it's simply doubled fields.

Raw 1080i delivers 775 lines, which is close enough to the theoretical maximum of 820.

>>Shooting interlaced each field contains an effective 1440x540 pixels<<

Ah, but that's where the logic goes south. Interlaced does *not* deliver a discernible 1440x540. The very process of shooting interlaced involves a vertical field blending pass, row-pair summation, which lowers effective resolution. The most you'll get out of an interlaced field of 1080i is a theoretical maximum of 420 lines, and the Z1 delivers an observed maximum of 775 lines, which puts the real, true raw resolution (per field) at 392 lines. Not 540.

The fields get blended together to minimize interline flicker -- since this is a Vegas forum, I can describe it like this: basically, whenever you shoot interlaced video, the camera is applying a "reduce interlace flicker" filter on every frame you shoot. Progressive doesn't do that, but interlace has to or the footage would look terrible on an interlaced monitor, very flickery etc.

Where CineFrame 25 (and CF30) get their benefit is that they bypass the row-pair summation and simply double the fields to reduce interlace flicker. That means that instead of the compromised 392 lines, they deliver the full 540.

So, you're still ending up with only as much vertical resolution as you'd get from a standard-def PAL camera (shooting progressive), but you're getting double the horizontal resolution, better/more effective compression, and no field fiddling.

Barry,
This is where i get confused. What difference does CF30 do than just shooting 1/30 shutter speed? Doesn't 1/30 just drop a field and double the remaining one? Or does it also undergo the row-pair summation process? I personally can't tell them apart.

What about the old "deinterlace blend" technique that some use to use where 60i footage was copied, placed on top of each other in the time line, and each clip was deinterlaced using the opposite field. One clip contained a doubled upper field and the other contained a doubled lower field. The two dienterlaced clips were then composited with the top clip set at 40-50% opacity. Resolution-wise how does this compare to CF30 or 1/30th?

What I'm waiting for is for someone to test out the various shooting modes (60i, 60i >30p, 60i> 24p, CF24, CF25, CF 30, 1/30 shutter speed) using the same or similar scenes on a high-end digital projection system, like the Sony Qualia 004 SXRD or 3 chip DLP. No one has seemed to have reported on this yet. While I know there is much more to creating the elusive "film look" than shooting modes, it would be nice to know which mode is preferable and why.

My projector is only 1/4 HD so I cannot tell a huge difference between the different modes in resolution, only cadence and motion blur. CF24 is too strobey for my tastes. CF30 and 1/30th shutter look just alike to me. Plain 60i looks crisp and clear but not very cinematic. 60i > 24p or 30p cut a lot of resolution out. Before I shoot my next short, I would like to know subjectively and objectively how much resolution is really lost and its real-world impact on its appearence on good projection equipment. I will likely shoot using just plain 60i, as this seems to give the most flexibilty in post.