24F mode on tape

[Barry Green]

At ResFest I was able to get some footage from the XL H1 in 24F mode, recorded on tape. I was also fortunate enough to have Nate Weaver along (a very talented shooter well-known from his posts about the HD100, as well as an ex-XL1 owner) so I let Nate do the shooting.

I was going to post the clip here for all to see, but there's a hitch -- a big one. It turns out that apparently, Canon has implemented their own, completely incompatible recording format! 24F footage won't play on any existing HDV equipment -- won't play on an HD100, won't play on an HD1, won't play on an FX1, won't play on an HC1. Won't play on the cameras themselves, and it won't play through firewire to HDV Rack from those cameras. On the Sony cameras the timecode display updates, but there's an error message on the LCD screen that says (IIRC) "incompatible tape format."

It seems like the only device that can play 24F footage is: an XL H1.

This also helps explain why, at RESFest, Canon was showing their footage off of DVCPRO-HD tape. There is no HDV deck that could play 24F footage! It's a new, third incompatible format of HDV. They had to use a non-HDV deck in order to be able to play it at all.

I talked at length with the Canon representative at the show, and he predicted that this would be the case -- that 24F won't play on existing HDV equipment. He'd seen the file coming out of the firewire port and he said it was a 24-frame file (which, of course, Sony HDV isn't -- all Sony HDV is recorded as a 60i or 50i file). And a 24-frame file doesn't make sense within the context of a 15-frame GOP either -- how do you fit 24 frames into a 15-frame GOP? You can't. He also said the CCD is clocked to 24fps, which CF24 isn't.

So, based on what the Canon rep told me, and what has been reported here, plus my firsthand experience with 24F, I think we can now make some educated guesses as to what 24F actually is, and what it isn't.

We know it isn't actual progressive scan, again verified by the Canon rep. And we know it isn't CineFrame 24 (thank goodness), because if it was CF24 it'd play in a Sony camera or deck. And, although I was only able to observe the 24F footage in the Canon's viewfinder (they didn't have a monitor at RESFest while I was there), the motion rendition didn't have the CF24 mode's characteristic herky-jerky movement. It looked consistent on consistent-speed objects. That's not a scientific observation, because I can't view the footage on a monitor or frame-by-frame, but that's a casual observation made of what was in the viewfinder.

Based on all this, I feel confident in making an educated guess as to what 24F actually is. I believe CF24 will likely prove to be a 48hz version of Sony's CF25. I think they clock the CCD at 48hz (instead of 50hz like CF25), and they capture one field (540 lines). They probably use the Digic II chip to synthesize the missing field, which will probably lead to a vertical resolution actually a bit higher than CF25. And, I am guessing that this stream is then encoded using a 12-frame GOP without pulldown -- a 12-frame GOP would work quite nicely with a 24-frame data stream.

Clocking the CCD at 48hz and using one field would give you a motion rendition identical to 24P shot at 48hz. It'll be only half the theoretical maximum vertical resolution, but then again the news isn't all bad: the Sony has about 775 lines of vertical res, but CF25 mode skips the low-pass filtering that lowers interlaced-footage res, so CF25 delivers a legit 540 lines rather than the 380 you'd expect from straight de-interlacing. So if the Canon employs the same process, it should offer at least 540 lines, and maybe a tad more depending on how good Digic II is at synthesizing/up-rezzing the "missing" field.

So, while Nate and I thought we would have a nice "scoop" for DVInfo readers, by posting the first 24F clip to the world, it turns out that it just isn't possible. It can't be captured (from any other HDV device). Maybe Kaku will be able to capture some footage if he tries to digitize directly from the XL H1, although I'm pretty sure that if he's using FCP it'll probably be a no-go (FCP doesn't support JVC's 24P mode, I'm guessing it won't support the Canon mode either without an update). I'll be sure to ask him to try this.

[Gary McClurg]

So Barry based upon your guess.

Maybe Nate knows this.

Does that mean the JVC has a true 720 lines of resolution or not in 24p mode?

Thanks Gary

[Barry Green]

Haven't shot a res chart yet but I would expect that yes, the HD100 should offer a full 720 lines.

[Gary McClurg]

Thanks Barry,

If I knew the Canon was going to be there yesterday I would have dropped.

But I have to finish painting by house in Orange County so I can sell it.

And I spent the morning at EVS checking out the JVC and Z1 and other wonderful goodies that I can spend all my money on.

But once again thanks for your good answers.

Gary

[Stephen van Vuuren]

Originally Posted by Barry Green

So, based on what the Canon rep told me, and what has been reported here, plus my firsthand experience with 24F, I think we can now make some educated guesses as to what 24F actually is, and what it isn't..

Very interesting guess. That would completely explain while they cannot call it 24p but why it would look like 24fps motion rendering and much better than CF24. It really is a 24p version of Frame Movie Mode with new tech (Digic II) to help make it better.

It will be interesting to see how 1080 24F compares to the HD100 720 24p. Of course, comparing while require a table full of cams. The incompatible HDV formats really stretches "format".

[Obin Olson]

that would make it shoot SD images if that is the case, UNLESS they use extra resolution from the sensor that we don't know about, this could be done if the chip had more then 1080lines high, 540 is not enough to bother with for output to a large screen, I would guess the jvc will beat the canon in resolution.

[John Trent]

Thank you, Barry Green. Thanks for the early information and for going out on a limb with your educated guess. What you say sounds right and would answer a lot of questions I've been having about 24F.

Looking forward to your report on the JVC HD100.

[David Newman]

"Maybe Kaku will be able to capture some footage if he tries to digitize directly from the XL H1, although I'm pretty sure that if he's using FCP it'll probably be a no-go (FCP doesn't support JVC's 24P mode, I'm guessing it won't support the Canon mode either without an update). I'll be sure to ask him to try this."

All need is a transport stream. I believe he is using the beta of LumiereHD. With one stream all can be explained. A non-pulldown 24 frame mpeg stream will has a bit-rate advantage (20%), but yes that will break HDV compatibility. As for modify software to support it, that should be even easier than support the JVC's 24P (which is HDV standard, and was still pretty easy.) Just looking for the data.

[Obin Olson]

David think it's 1/2 the resolution of 1080i = 540lines ? they would not do that would they? and then upsize it to 1080?

arrgg i Hope not! :)

BTW David what type of computer can you run for recording LIVE HD-SDI from VariCam or the new Canon? has anyone you know built a small system that could go out of the studio with your capture card?

David do you know of any HD-SDI cards that work with premiere AND your codec for display out HD-SDI and or DVI from Premiere Pro?

I understand Apple and BlackMagic have a working solution that allows FCP to use ANY video codec and spit it out the HD-SDI jack as long as its a valid Quicktime codec....would that mean if CineForm had support under apple you can edit with CineForm and still have a live preview out HD-SDI?

SOrry for all the questions but I am trying to get a setup going that uses your codec for editing, we have the dvx100 HD camera, now we need a good way to edit that footage...one thing needed is a way to convert from tiff jpg etc to CineForm WITHOUT using After Effects to do it...way to hard when you have folder upon folder and hundereds of gigs that need to be converted! ;)

[John Jay]

this is my take on the 24p malarchy - pretty much as Barry Green says above

however its not just straight de-interlace. I got the panning clip from Kaku and dumped it into Procoder and tried the 3 types of deinterlacing
1 adaptive
2 interpolate from dominant field
3 discard non dominant field (field double)

the winner is option 2 so it looks like a simple interpolation in the DSP

take a look below especially at the motor cycle mirror, the original is on the left



http://s49.yousendit.com/d.aspx?id=2...93R46COEYEED9U

[Nate Weaver]

Originally Posted by John Jay

the winner is option 2 so it looks like a simple interpolation in the DSP

Er, what does this prove? I can't see how what some program chooses as some sort of automatic feature as indication of anything.

[John Jay]

Originally Posted by Nate Weaver

Er, what does this prove? I can't see how what some program chooses as some sort of automatic feature as indication of anything.

Consider this:

IF the left panel original was Full vertical resolution then the right panel; after deinterlacing though interpolation; would be exactly as you see, since it would be interpolating every other field line. Under these conditions there would be a marked difference between left and right panels with the right panel looking obviously softer.

However the left panel is 'very similar' to the right panel which suggests that the processing is similar.

This is known as Proof by Induction. OK?

[Obin Olson]

That looks like "video" I guess the Canon is going to be just a new option for the video look. After working with the HD dvx it is so hard for me to find anything else at ANY resolution that looks "good" to my eye(cept a film scan)! ;)

Uncompressed raw files are so unlike anything else, even working with the VariCam is hard. We just did a greenscreen shoot in the studio for a comp in a VariCam shot feature film, the DVCPRO HD looks like it has 1/2 the resolution of the dvx for keying the green! all jaggy edges of the kid that was to be keyed, and the same thing with the little dvx HD has almost no jaggy edges at all. Night and day. the shot would not have worked if it was not for the fact that we need the kid at 25% resolution for the shot, the down size made it workable. I guess I could have used the HD-SDI out the side of the camera but I don't have a HD-SDI uncompressed recorder here.

I am glad we have people like Juan at reel-stream.com working on this type of mod, it really saves the day when you need CLEAN raw files in post.

I hope this Canon will have a good image out the HD-SDI port.

[Nate Weaver]

Originally Posted by John Jay

Consider this:

IF the left panel original was Full vertical resolution then the right panel; after deinterlacing though interpolation; would be exactly as you see, since it would be interpolating every other field line. Under these conditions there would be a marked difference between left and right panels with the right panel looking obviously softer.

However the left panel is 'very similar' to the right panel which suggests that the processing is similar.

This is known as Proof by Induction. OK?

I see. I also misread your method, I was thinking that your software auto picked method 2 as (best).

BUT, I still don't agree with "proof". An indication yes, proof no. Also the processed side IS softer, but only barely so. Not enough to prove you wrong.

I think your argument holds water, but don't necessarily agree.

I can say this: I've had my HD100 for a month now...the Canon is putting many more real pixels to tape no matter how it's doing it.

[David Newman]

Originally Posted by Obin Olson

David think it's 1/2 the resolution of 1080i = 540lines ? they would not do that would they? and then upsize it to 1080?

arrgg i Hope not! :)

It currently looks like that is what they they did it.

BTW David what type of computer can you run for recording LIVE HD-SDI from VariCam or the new Canon? has anyone you know built a small system that could go out of the studio with your capture card?

Expect an announcement towards the end of the month. We are working with a company to do extactly that.

David do you know of any HD-SDI cards that work with premiere AND your codec for display out HD-SDI and or DVI from Premiere Pro?

AJA Xena works great.

Obin, could you post the remainer of your questions to the CineForm section; I don't what to take this further off topic.