FX1 and simple FX in Cineframe24. My experiences

[Ignacio Rodriguez]

> CF24, CF25 and CF30 all reduce the frame resolution by half.

Only vertical resolution, not frame resolution. And if the built in deinterlacer is as good as some hope, only when there is motion.

[Alex Raskin]

Got it... so there's no point for in-camera frame manipulation in Fx1 if it all results in halving the reolution.

Instead, it seems like post-production ways of changing the footage from interlaced to progressive, plus possibly changing the frame rate from 30 to 24, is the way to go?

Anyone has a proven and true workflow on this?

[Ignacio Rodriguez]

> seems like post-production ways of changing the footage from
> interlaced to progressive, plus possibly changing the frame rate
> from 30 to 24, is the way to go?

It would appear from all the preliminary info available that yes, that would be the way to go. Almost everybody seems quite upset by CineFrame24.

But since the camera is not very sensitive to light, CineFrame30 might be ok for getting some more sensitivity out of the camera and assuming NTSC delivery. Or with the PAL FX1 or Z1 there is CineFrame25, which would seem to be the best if you need film out and can live with a 4% time shift.

[Dylan Pank]

<<<-- Originally posted by Ignacio Rodriguez : Only vertical resolution, not frame resolution. And if the built in deinterlacer is as good as some hope, only when there is motion. -->>>

Re: resolution - Ignacio, you're right, that's what I mean to say.

As for the built in de-interlacer, I don't think there is one - that's the point - it simply discards one entire field (which I guess is a dinterlacer of sorts, but not one that's worth a great deal).

To get adaptive, or "smart" deinterlacing, you'd have to go for converting it in post.

That or wait for 25p 1250 cameras, which will probably create images as good as deinterlaced 1080i

[Alex Raskin]

Ignacio, you see, I'm aiming for SD distribution on NTSC DVDs, and possibly digital projection (1280x720?) in theaters.

CF24 is out as it's way too jittery for my taste.

CF30 I'm unsure about. Previous poster indicated that CF30 loses half of the resolution. You however said that most likely the loss only occurs in the areas where the movement is present.

So bottom line is, I'm still unsure what exactly is CF30 mode doing.

Quick camera test showed that CF30 actually produced visually SHARPER images of the fast moving objects! So far all this is rather unclear to me.

[Ignacio Rodriguez]

> As for the built in de-interlacer, I don't think there is one -
> that's the point - it simply discards one entire field (which I
> guess is a dinterlacer of sorts, but not one that's worth a
> great deal).

It's what I imagined at first too. But according to http://home.earthlink.net/~dvcnyc/Sony%20HDR-FX1.htm, It would appear that resolution loss will only happen when there is motion. This would seem different to what happens with the PDX10 and PD170, where you can see a visible resolution loss even when there is no motion in 1/30 (NTSC) or 1/25 (PAL).

Quote from there: 'In CineFrame 30 (FX1/Z1) and CineFrame 25 (FX1e/Z1), "smart" deinterlacing is used to create video that has a temporal resolution of either 30fps or 25fps, respectively. Diagram 10 shows how the FX1Õs 2:1 pulldown accomplishes deinterlacing of 1080i60 video.' Of course, Steve might just be thinking that this is how it works, without actually knowing it. But already people have shot res charts and report they can't see resolution loss on an HD monitor when running the cam in CineFrame mode.

There is a whole thread on the subject of resolution loss and low light response: http://www.dvinfo.net/conf/showthrea...threadid=36132

[Dylan Pank]

I did an eyeball test on Kaku's footage posted on HDVinfo - since taken down - and on static images (ie non-moving video - not still frames), 60i was sharper than CF30. CF30 and CF24 were about the same, but in CF24 there was a very slight "shivering" on lines and edges which did reduce the overall quality.

So if CF30 does use smart de-interlacing, it's not that smart. Also reading through Steve Mullen's article, I'm not sure how much of his discussion of CF24/30 is fact and how much is hypothesis (the qualification "...an intriguing possibility " came up a couple of times.)

One way to test it would be side by side tests of footage that has been shot CF30 and 60i with a discard de-interlacer applied, but since thaty would mean re-encoding to a new file, maybe that would also affect the results.

[Barry Green]

FX1 60i is definitely sharper than CF30. I know there's been speculation about how CF30 is just as high resolution, but it isn't. Then there's been speculation about whether it's a "smart de-interlacer", which would mean not losing resolution on portions of the frame that don't move. That speculation appears unfounded as well.

I have shots of a CamAlign chart, that includes vertical resolution information. It's easy to see that in 60i the resolution is much higher than in CF30, and that's on a still shot. I don't have a definitive way to test the resolution loss on a moving shot, but it definitely loses vertical resolution even on still shots when you switch to CF30.

Here's an extraction, blown up to 300% in PhotoShop so you can see the details:

http://www.icexpo.com/FX1-60i-vs-CF30-res.jpg

Here's the raw frame extraction from both frames, for those who want to examine the pixel-accurate version without any potential JPG artifacts (100k BMP version of the same picture above):
http://www.icexpo.com/FX1-60i-vs-CF30.BMP

It looks like in 1080i mode the camera delivers about 750-770 lines of vertical resolution (which is surprisingly close to Steve Mullen's estimate of 820). In CF30 mode, it's more like 575, which pretty much would be what you'd expect from a straight field-drop de-interlace.

*edit* to be clear, when I say straight field-drop, that's not saying that resolution is cut in half when you go to CF30 (or, presumably, CF25)! Obviously the res would have to drop to 380 lines or so for that to be true. Instead, it looks like CF30 is foregoing the filtering and field-blending that normally gets put on interlaced video to smooth out interline flicker, so you're getting the full raw resolution of one field (which should presumably be 540 lines, which is certainly close enough to the observed 575 lines to fall within the margin of error).

[Ignacio Rodriguez]

Thank You Barry. It's clear then... well, no pun intended at all but... uhm... rather: It's conclusive. CineFrame loses vertical resolution, there seems to be no "smart" de-interlacing going on.

[Steven White]

"basically the footage is not really HDV anymore, though it may still be High Definition (as the case with Steve's footage: it's Hi Def quicktime, not HDV)."

As if there's anything sacred about HDV... My workflow is designed to allow me to rotoscope on progressive frames without losing any information of the original HDV footage until the final render. If I were so inclined I could (hypothetically - I don't have the software to do this) re-interlace the final 24p video to a 2:3 HDV signal for future display on a 60i television.

My plan is to save a final uncompressed render of this footage to DVD-R using HuffyUV, so that when an acceptable high-def delivery format comes out, I can re-master my flick.

I don't understand why anyone would shoot in Cineframe24 and then edit on a 60i stream without appropriate pulldown.

Further to this: I would not shoot any documentary-type event using Cineframe24... it takes too much effort to set up individual shots for run-and-gun type shooting.

-Steve

[Davi Dortas]

Who is ever to think that it uses smart de-interlace is either uninformed or crazy. I tell you whys.

1. To do smart de-interlace, the camera need to have a frame buffer memory of AT LEAST three 1440x1080 pixel images. You looking prolly around 4-6MB of pixel data that need to be buffered.

2. Secondly, you need a chip to process the images, looking for changes between frames, and it need to be done all in real times. Such processing chip, haha, would be more expensives then just to put in a real progressive CCD.

3. Thirdly, because video never completely clean and noise free, a smart-deinterlace would not give you perfect results each time. Deinterlace would be subjecting to many artifacts, such as twitter, pulsing of fine detail, etc.

So given these objections, why would Sony use smart-deinterlacing and get shabby results, and prolly also more expensive to implement? They could save money just by switching to progressive CCD instead. Now you understand why CF is fake progressive and reduce whole resolution.

Think people! OK?

Why they not use progressive CCD? Prolly cuz they use interlace CCD to get better low light performance, and/or they is want to keep it 1080i to not eat their higher-end offerings.

[Ignacio Rodriguez]

> you need a chip to process the images, looking
> for changes between frames, and it need to be
> done all in real times. Such processing chip,
> haha, would be more expensives then just to
> put in a real progressive CCD

That's already in there there for the MPEG compression.

> Why they not use progressive CCD? Prolly
> cuz they use interlace CCD to get better low light
> performance, and/or they is want to keep it 1080i
> to not eat their higher-end offerings.

I totally agree. That's probably it. It's already terrible for low light compared to similarly priced SD equipment. If it were real proscan, they would have to include larger sensors, and then they would not be able to sell Betacams again. Ever.

[Dylan Pank]

<<<-- Originally posted by Steven White : "basically the footage is not really HDV anymore, though it may still be High Definition (as the case with Steve's footage: it's Hi Def quicktime, not HDV)."

As if there's anything sacred about HDV... -->>>

Nothing sacred about HDV? KILL! PERSECUTE! BURN THE HERETIC! ;-)

<<<--- My workflow is designed to allow me to rotoscope on progressive frames without losing any information of the original HDV footage until the final render. If I were so inclined I could (hypothetically - I don't have the software to do this) re-interlace the final 24p video to a 2:3 HDV signal for future display on a 60i television. -->>>

Only if you make VHS/Tape versions - any decent mpeg2 encoding software would be able to make 23.976 fps mpeg2 files with 2:3 flags in there that would play in a DVD player without. I assume that DVB receivers do the same; does any one know if digital broadcast of film material in 29.97Hz land is 2:3 interlaced or 23.976 progressive?

<<<--- My plan is to save a final uncompressed render of this footage to DVD-R using HuffyUV, so that when an acceptable high-def delivery format comes out, I can re-master my flick. -->>>

This is an expremely good plan. If I may ask, how big are your huffYUV compressed files, minute by minute? and what resolution are you holding them at? I also have a long form project in mind and would like to think in terms of shooting HDV, and huffYUV looks like a good solution ot storage problems of the final Master.

<<<-- don't understand why anyone would shoot in Cineframe24 and then edit on a 60i stream without appropriate pulldown. -->>>

There are reasons to do this, such as if you're heading to a broadcast out, one may want to retain a 29.97 timeline and timecode for convenience sake, but I think effects work like yours would always be done on uninterlaced progressive frames, for the reasons you mentioned before.

[Toke Lahti]

<<<-- Originally posted by Davi Dortas : Who is ever to think that it uses smart de-interlace is either uninformed or crazy. I tell you whys.
1. To do smart de-interlace, the camera need to have a frame buffer memory of AT LEAST three 1440x1080 pixel images. You looking prolly around 4-6MB of pixel data that need to be buffered.
2. Secondly, you need a chip to process the images, looking for changes between frames, and it need to be done all in real times. Such processing chip, haha, would be more expensives then just to put in a real progressive CCD.
So given these objections, why would Sony use smart-deinterlacing and get shabby results, and prolly also more expensive to implement? They could save money just by switching to progressive CCD instead. Now you understand why CF is fake progressive and reduce whole resolution. -->>>

Well, the camera has a chip that makes more complex calculations than deinterlacer and it has 18 frames memory, it's called mpeg encoder.
Like Steve Mullen suggested, that if sony was smart, they could use mpeg encoder to enhance cineframes deinterlace.

Big question still is: why no progressive?
It just remains the most asked feature from this camera.
Mullen also said, that basically all ccd's are progressive, it's just a matter does it have big enough buffer to hold the whole frame.

Maybe sony underestimated the success of progressive dv cameras back in 2003 when they started to design this ccd.
And now they might be in a hurry to grow that buffer to hold a whole frame and camera with this renewed ccd might come to market very soon.

[Chris Hurd]

In the HDV format, you have a choice of either 1080i or 720p and Sony decided to go with 1080i for these particular camcorders. As more HDV camcorders become available in the marketplace, you'll have some that are 1080i and some that are 720p. If Sony recognizes a demand for true progressive scan, then I wouldn't be surprised if they offer a 720p camcorder eventually. My own guess is that Canon will go with 720p, but that's just a guess.

One thing you won't see is 1080p, because that's not part of the HDV specification.