Why is the 24p option buried in the 1080i video?

[thisiswells]

How is it possible to yield native HD progressive and native HD interlaced output from the same CCD's??

I am certain Barry will give (for the umpteenth time on this forum) a thorough explaination of "pulldown" in an NTSC signal. But, in an attempt to save him some words, I offer a simplified, practical, non-technical explaination.

DVD's are recorded in 24P. But, your TV doesn't understand progressive. So, the DVD player automatically adds an interlaced "pulldown" which is a way of distributing the 24 frames across 60 interlaced fields in a way that is invisible to the viewer; you're unaware of it and you never notice it. For a "normal pulldown" the distribution of the full frames into fields follows a repeating pattern like this: 2-3-2-3 etc.

The DVX captures 24P on the chips, but, in the same way that a DVD player must add "pulldown" to display on a TV, so does the DVX in the way it records to tape. It absolutely must be recorded as interlaced so the tapes playback in any camera. Otherwise, it would be proprietary and frankly, it would have never caught on, in my opinion.

This is nothing at all like how Sony is doing CineFrame24. Instead Sony is capturing 60 fields on the camera, and then randomly tossing a bunch of them away until they end up with only 24. But, the motion is NOT smoothe at all. In fact, it looks like the cat's job.

All I say is, buy Barry Green's book and save yourself a lot of time asking questions!

 

[thisiswells]

But how does the HVX produce both native interlaced and native progressive images?

According to a translated Japanese article, everything on the HVX is sampled at 1080p/60 and downconverted, for lack of a better word, to 720p60 or 1080i60 and others. It was a translation, which is far from being a definitive source on the subject. Maybe this answers your question?

 

[Jim Arthurs]

The fact that they "won't" break the specs is one thing that makes this camera so attractive. The inclusion of a 24 frame 1080P option would essentially be a new codec, or at least a new version of the DVCProHD codec.

This is a question I don't know... does the P2 stored 720 24fps format (as well as the other non 60fps 720 speeds) break or go beyond the original specs? Does ANY NLE on the market today handle them without an upgrade? I've asked this before to no answer... FC, Avid, etc...

Regards,

Jim Arthurs

 

[Jim Arthurs]

According to a translated Japanese article, everything on the HVX is sampled at 1080p/60 and downconverted, for lack of a better word, to 720p60 or 1080i60 and others. It was a translation, which is far from being a definitive source on the subject. Maybe this answers your question?

I agree it would be very simple DSP work to take the two progressive frames that together equal 1/30th of a second and store them as fields, with potential better subsampling than just "normal" fields from interlaced chips, remember sampling down gives great results, and the vertical would be downsampled by half....

That translation also got me thinking, and here's the only bit of speculative conjecture I'll do on the HVX... if we take as given that the chips are some form of progressive 1080 at 60fps (I'm betting on 960 by 1080) then...

...why can't the HVX record 1080 60p? No, hear me out, no codec breaking necessary... all that would be required is that the 60p be recorded to P2 at 2X data rate and wrapped into the standard DVCPRO 100 29.97 header. It's basically the reverse of making fields from the progressive chips we just talked about. Each 1/60th of a second progressive frame would be used for two fields of 29.97i. We know the P2 media is plenty fast enough to write at 2X.

Talk about ground breaking! What other camera can do 1080 60p at this point? More likely this is what is planned for the next Varicam, or the next generation HVX.... but I bet there's no reason other than feature restriction that this couldn't have been done to the HVX...

Food for thought,

Jim Arthurs

 

[Emanuel]

DVD's are recorded in 24P. But, your TV doesn't understand progressive. So, the DVD player automatically adds an interlaced "pulldown" which is a way of distributing the 24 frames across 60 interlaced fields in a way that is invisible to the viewer; you're unaware of it and you never notice it.

And in 25p/50i - if it's not necessary the 2:3:3:2 pulldown process - how is it works?

 

[Luis Caffesse]

This is a question I don't know... does the P2 stored 720 24fps format (as well as the other non 60fps 720 speeds) break or go beyond the original specs? Does ANY NLE on the market today handle them without an upgrade? I've asked this before to no answer... FC, Avid, etc...

Regards,

Jim Arthurs

DVCProHD 720 is a purely progressive codec.
DVCProHD 1080 is a purely interlaced codec.

So, it becomes a problem to try to add a native 24P recording option in 1080.
But 720 is no problem because the codec already includes the ability to record any framerate from 4fps to 60fps progressively. Granted, on tape the signal is padded with duplicate frames. But these duplicate frame are removed in post, leaving you with a native 24P file.

My point was that editing systems can handle 720 24P footage today.
I can tell you that FCP 5 has the options for DVCProHD 720P 24 in it's capture presents. Recording to P2 merely removes the step of having to remove the duplicate frames on capture (in fact, it removes the step of capturing all together).
But being able to work with 720P 24P files in DVCProHD is nothing new.


It's not possible to add this functionality to the 1080 variant of DVCProHD as easily because the codec is purely interlaced. So it cannot be recorded at purely progressive frames to the P2 card. Therefore Panasonic took the same route as they did with the DVX, namely add a pulldown to the 24P signal and record it as 60i, thus staying in the interlaces parameters of DVCProHD 1080. The frames can still be removed in post, leaving you with a native 1080P24 file.

It just became a lot easier with 720, that's all.

if I'm wrong on any of this I'm sure someone like Barry will put me in my place

 

[Emanuel]

I shot in Romania a couple of months back and had to deal with some really awful low frequency magnetic ballast flourescents, and YES, I can verify that a 1/50th shutter speed eliminates the flicker.

Regarding your own words, you shot with a DVX @ 24p. Did you note any other problem about shooting 24p/60i in PAL land?

 

[Barry_Green]

Technically gifted people tell me that camera CCD sensors are usually read out in either interlace or progressive. Not both. When Panasonic describes reading out a 1080i image in 24P, I get confused.

No confusion necessary... Panasonic's been doing it for over three years with the DVX100, and also with the SDX900 and the SPX800.

Sony apparently reads out the Z1 CCD's at 24fps.

Nope. Sony reads the Z1 CCDs at 60i, and then convolutes the fields together into CineFrame 24 ( you can read a detailed examination of how it actually "works" on Adam Wilt's site).

The resulting Cineframe (if that's what it's called) seems to get universally poor reviews.

Deservedly so.

How is it possible to yield native HD progressive and native HD interlaced output from the same CCD's??

I'm still researching into that and how to explain it. For a start, you can look at Steve Mullen's article "Progressive: What You Need To Know", available as a downloadable PDF on Panasonic's site.

However, I can assure you that scanning interlaced and progressive off the same CCD definitely does work, as we've seen from years of experience with the DVX, SPX, SDX, and XL2 cameras, all of which do it.

 

[contentlab]

for all those who are pulling out their hair trying to find the steve mullen article barry mentioned:

ftp://ftp.panasonic.com/pub/Panasonic/Drivers/PBTS/papers/Progressive-WP.pdf

 

[Luis Caffesse]

Thanks for the link Contentlab

 

[makal]

And in 25p/50i - if it's not necessary the 2:3:3:2 pulldown process - how is it works?

PAL DVDs are 25fps. Movies shot on film or on digital 24p are 'dubbed' (along with audio) 4% faster to get 25fps. The DVD player outputs each frame twice alternating odd and even lines, thus feeding the tv with 50i. If you shoot a feature with a PAL 25p video camera, your theatrical film-printed release will be 4% longer.

 

[Emanuel]

Thanks makal! (a good other native or resident italian as the majority that I know)

As I could see, at least, in your first post, you sound like a technology guy, allora sei benvenuto! What's your opinion about the debate carried on by me in some threads about the 24p (60i) --> 25p (50i) and vice-versa?

 

[makal]

My understanding is that it is a non-issue technically. Personally I would never buy a NTSC camera only to be forced to do pulldown removal AND speedup for every single work that must go PAL (in my case, 100% of them). Speaking of the HVX, it would be feasible only in 720/24p recorded to P2, since there is no pulldown involved. Yet it is possible and easy, only time and space consuming. So if you want to spare some bucks buying in the US... but I would NEVER do that :evil:

The only thing that makes the NTSC version more attractive is the 60p instead of 50p for slowmo, but anyway I don't know the specs for the PAL version... maybe it is capable of 60p since there isn't (in HD 720) any resolution or color sampling difference. But I really dunno :shocked:

As for the 24p<->25p conversion, it's just a matter of forcing a 24p file to be interpreted as 25p in a 25p project and vice-versa. I guess all NLEs will take care of audio, or you can use any cheap or even freeware waveform editing software.

 

[Emanuel]

My understanding is that it is a non-issue technically. Personally I would never buy a NTSC camera only to be forced to do pulldown removal AND speedup for every single work that must go PAL (in my case, 100% of them). Speaking of the HVX, it would be feasible only in 720/24p recorded to P2, since there is no pulldown involved. Yet it is possible and easy, only time and space consuming. So if you want to spare some bucks buying in the US... but I would NEVER do that :evil:.

Why? Do you think that pulldown in 1080i (60i 2:3:3:2 process to achieve 24p) will decrease the image quality (any resolution or color sampling difference)?
I asked here and they told me that it isn't any degradation... And about the speed up or slow down? Neither?

As for the 24p<->25p conversion, it's just a matter of forcing a 24p file to be interpreted as 25p in a 25p project and vice-versa. I guess all NLEs will take care of audio, or you can use any cheap or even freeware waveform editing software.

Speed up (24p --> 25p) it's not the same of the opposite (25p --> 24p); the effect will not be the same
[there are different results - an "european time" would be speedy increase instead of the slowdown, if not the majority of the american audiences will say «this is even more boring than something I could think up!...» ]:

http://www.dvxuser.com/V3/showthread.php?t=33434&page=23&pp=10
Post#225

Nothing wrong with the format, but to go theatrical you have to convert. And after the conversion you'll definitely see the difference between 24p and 25p-converted-to-24p:
- There's slomo in it that makes it look less "real" (if you're going the route of slowing the footage down)
- There are sound problems: Either you get pitch-shifting artifacts (really nasty) or the voices are a quarter note deeper, which is very noticably subconsiously: Tiny men sound like Pavarotti and slender female models sound like Dame Edna (if you're going the route of slowing the footage down)
- If you're going the route of keeping the speed, at which point there are no sound issues, there are major sharpness issues or the image "jumps" every second.

That's what's wrong.

http://www.dvxuser.com/V3/showthread.php?t=33434&page=22&pp=10
Post#220

(...)And don't get me started about either mangling the sound through a pitch converter or living with a lower pitch for everything. 24p is the way to go and if the HVX doesn't has it, I'm majorly dissapointed.

http://www.dvxuser.com/V3/showthread.php?t=33434&page=29&pp=10
Post#282

Please explain me how do you think you gonna edit your 24p footage on your Pal equipment. I guess you better move to a NTSC land or just buy NTSC equipments.

With any NLE system. If the camera supports 24P, than it can be used to record the output from the NLE. As for monitoring, my plasma will do. It's pal and Ntsc compatible (1080/50i and 1080/60i compatible)

Have any of you tried slowing down 25p to 24p and actually seen the results? Well I have, and it sure ain't pretty. I did a promotional video on the Pal dvx100 in 25p and had to downconvert it to ntsc 24p. I had to pay $300 for the conversion in a pro post house, and the results even on hardware conversion where a lot worse than on the native 25P. So please try it first ntsc guys. Than you'll see what we are talking about.

Afterwards, I think that the human (emotions) perception can change if it is playedback @ 24 or 25 frames. Even unconsciously. That's for this reason the majority says: it's non-noticeable @ "conscience" level. But we must know there are a lot of hidden issues that not even the business experts know why A succeed & B not. That's why we call it as an art form and not only as business.

http://www.dvxuser.com/V3/showthread.php?t=33434&page=23&pp=10
Post#229

Yeah, that's what I'm talking about: pitch-shifting adds artifacts: either it's some "stuttering" or flanging. I've never heard a really good pitch-shifter so far. Plus this still means that you have 4% slo-mo. And we've discussed+mostly agreed before that 26fps-for-24fps adds some "otherworldy-ness" to the footage. Well imagine having this through all of your movie - it will completely change the perception. With 24fps-converted-to-25fps (the "true-and-tested method"), you have the opposite effect that everything is a bit faster, which makes the viewer more attentive - can't really compare the two. Conversions from 25p to 24p are actually much less common, and they don't look right, unfortunately. That's why it's important that the pal-hvx has 24p among other framerates, like the arri.

Cheers!

In my case, my needs are two:

[1] Big screen (and it's true that maybe the frame issue probably is a non-problem) - film-out or digital - for instance, now, in Portugal, until the end of 2005, we will have 20 screens to digital cinema (a digital finance program from the state film institute);

[2] DVD release
[2.1 more SD PAL than NTSC] &
[2.2 future formats HD-DVD or Blue-ray and so on for PAL & NTSC - why not?];
and that is the problem for a 24p/60i purchase (only possible if also PAL conversion goes OK).

For my sort of work, the PAL TV (but, of course, even less NTSC TV) is not my first concern. I'd live with the conversion if not with damage, that is, acceptable. Even more, it's a small screen not so important than where - the pro & cons will be blow up - it counts for very much. And also the price.

 

[a2zproductions]

How are some achieving such beautiful results with a Sony FX1, & the G35?
Very Filmic!
A.

 

[stokestack]

Why no recompression?

Why no recompression?

And the 24p is recorded same as 24p would be with the DVX, or a film to tape transfer would do, and no there's no degradation. This is real progressive, even if it's stored in an interlaced manner.

The question is, doesn't reassembling the entire frame require recompression? Because we're talking about macroblocks that are what, 8x8 pixels? Within them, pixels depend on their neighbors (including vertical neighbors), and thus it seems to me that you can't slice the image apart on alternate scan lines and then reassemble it without recompressing. I would think that you could split the image into top and bottom halves along a macroblock boundary and then store those halves progressively in the space that would normally hold a field in an interlaced stream.

Can anyone authoritatively weigh in on how exactly they're storing the compressed data, and how alternating lines could be reassembled without recompression?

 

[Jim Arthurs]

The question is, doesn't reassembling the entire frame require recompression?

You're absolutely right, for normal 3:2... That's why Panasonic came up with a "new" way to pad 24 frames out to 30. Use this Advanced pulldown to assure access to each progressive image without having to decompress it as fields across two video frames.

See here for details;

http://adamwilt.com/24p/#24pRecording

Regards,

Jim Arthurs