24F LAID TO REST: The honest answer about 24F

[Barlow Elton]

Originally Posted by Joseph H. Moore

Of course, the look of the H1's 24F motion isn't dead on film-like either

I've thought the same thing about HDCAM F900 at times. There are many examples of it that look rather slurry and electronic, (Fahrenheit 911, Collateral) but then I've also seen wondrously dead-on filmic stuff too.

Maybe they'll come up with a smart deinterlacer for 24F material. (jk) Now THAT would be some serious voodoo.

[Joseph H. Moore]

It is possible that the best results will be from shooting 1080/60i at 1/48 and deinterlacing in post. In post you can put some serious iron behind the process that just isn't possible in camera.

What would be awesome is for someone to shoot the same scene in 24F and 60i and post the raw files so that we could try a variety of deinterlacers and see what yields the most natural, detailed results.

[Barlow Elton]

Originally Posted by Joseph H. Moore

It is possible that the best results will be from shooting 1080/60i at 1/48 and deinterlacing in post. In post you can put some serious iron behind the process that just isn't possible in camera.

What would be awesome is for someone to shoot the same scene in 24F and 60i and post the raw files so that we could try a variety of deinterlacers and see what yields the most natural, detailed results.

I'll try that soon, Joseph, and if I can get my G5 portable set up going, I'll try it in different codecs too. Actually, at least with HDV, --in theory-- if one were to get the PAL modification and shoot 50i and deinterlace, that might be the ultimate option. There might be a little more bit rate benefit as you're dealing with a lower temporal sampling.

Curiouser and curiouser.

[Joseph H. Moore]

You're definitely right about the PAL modification being the best starting point to end in true 24P ... forgot about that option.

I'm really amazed at how little MPEG artifacts any of these HDV cameras are producing, especially the H1. I think the big advantage of capturing via SDI will be the increased chroma resolution. Is that how you're going to go into the G5? AIC seems to mangle the pic a bit, but there are some other lossless and near-lossless codecs worth trying.

[Barlow Elton]

Gonna try them all. I think the SDI contenders will be 10 and 8 bit uncompressed (I need to build a minimum 4 disk SATA RAID for this) and DVCPRO HD. I might even have the option of QRez with the Kona, which I believe is 4.2.2/4:1 compression, but will probably require their software.

The Kona SDI will be interesting because, in theory, I may be able to capture 24p DVCPRO HD from a live 24F signal. I've captured HDCAM 24p into a basic Decklink HD card this way, with success. It was cool because the pulldown removal was automatic.

We shall see what we shall see...

[Joseph H. Moore]

Tou may want to try Sheervideo, as well. Back in the day it was killer, but I honestly haven't tried it with HD and G5's.

[Pete Bauer]

Hey Joseph, I'm all for someone trying a de-interlace in post just to see what happens. But my prediction is that no matter what software one uses, it won't exceed the detail found in 24F/25F/30F. I think this because the purpose-built DIGIC processor has the raw, unprocessed pixel-shifted 1440x1080 stream off the CCD block with which to work its magic. Any de-interlace software is only going to have the processed 50i/60i final output to work with. No doubt it'll still look good, but I doubt it'll actually look better. But, if someone wants to give it a go, would sure be interesting to see.

FWIW, I'll get around to it eventually because it occurred to me that due to the slight reduction in vertical resolution that 24F gives compared to the 60i setting, carefully done 40%-speed slow motion derived from 60i ought to match VERY nicely to my 24F projects. But I personally probably won't get around to that for a while, so must leave the de-interlace question to someone with more time or a more immediate need.

[Barlow Elton]

Originally Posted by Joseph H. Moore

Tou may want to try Sheervideo, as well. Back in the day it was killer, but I honestly haven't tried it with HD and G5's.

I think the best codec is going to be PhotoJPEG at 75%, (1/10th file size of uncompressed, I think) and I'll try 100% if my drive can handle it. I think the data rate of this codec will be similar to DVCPROHD, but better quality because it's not based on the DCT DV codec.

[Lauri Kettunen]

I've also spent time trying to figure out whether to shoot in 25f or 50i. In case of XL2, I end up using only 25p. In case of XL H1, have not been able to make up my mind yet.

I do agree with Shannon's point that in the end of the day all what matters is the output. However, I still find myself thinking, what's going on there inside the camcorder. This thread forced to ask, why I'm one of those who would like to know, what' s going on there within the XL H1. And the answer is, in sense of logic there's a difference between

i) something which works for all data,
and
ii) something for which one finds data everything works out well.

This is to say, the issue of 25f leaves a justified question, are there conditions in which the 25f failed where the 25p did not?

Compare, there are conditions in which the mpeg2 codec fails. So, having material for which the mpeg2 is just fine does not imply mpeg2 were fine for all possible data.

[Daniel Patton]

Originally Posted by Pete Bauer

Hey Joseph, I'm all for someone trying a de-interlace in post just to see what happens. But my prediction is that no matter what software one uses, it won't exceed the detail found in 24F/25F/30F. I think this because the purpose-built DIGIC processor has the raw, unprocessed pixel-shifted 1440x1080 stream off the CCD block with which to work its magic. Any de-interlace software is only going to have the processed 50i/60i final output to work with. No doubt it'll still look good, but I doubt it'll actually look better. But, if someone wants to give it a go, would sure be interesting to see.

FWIW, I'll get around to it eventually because it occurred to me that due to the slight reduction in vertical resolution that 24F gives compared to the 60i setting, carefully done 40%-speed slow motion derived from 60i ought to match VERY nicely to my 24F projects. But I personally probably won't get around to that for a while, so must leave the de-interlace question to someone with more time or a more immediate need.

Hope I'm not opening any old wounds here, seriously not my intention. This thread looked to get a little heated from what I was reading, but Peter you mentioned something I have been wondering while I'm running the H1 through it's paces... did you ever get to test what process yielded better results? In camera de-interlacing with it's reduced resolution, or as a post process better done in an NLE?

If this has been covered in a more recent post (anyone) please feel free to point me in that direction.

Thanks and Peace!!

.

[Herman Van Deventer]

A Word of thanks.

Just a small word of thanks to everyone that contributed to Shannon's
topic / Output - also taking all technical factors involved into serious
consideration.

In Africa, we follow a different approach - First shoot... Then talk.....
By the time of making the purchase decision i dit not have access to this
forum - 6 months ago / and i had to rely on output perceived to make the
final decision / Panasonic-Canon.

The H1's, F-mode produced vertical resolution loss that introduced slight vertical aliasing, yet it popped out glorious images footaged.

Relying on my eyes / output I made the final decision.

CANON XLH1

[Daniel Patton]

No one care to comment on my question regarding de-interlacing? Still wondering.

[Tony Tremble]

Adam Wilt explained the process, it is not de-interlacing.

In F modes the green CCD is clocked half a field out of sync. So the scanlines would look like this;

RBRBRBRBRBRBRB
GGGGGGGGGGGGG
RBRBRBRBRBRBRB
GGGGGGGGGGGGG

The first line is just data from red and green CCDs and the second line is just from the green CCD etc etc. The DSP does the rest i.e. work out from surrounding data what pixels should be what colour based on adjacent real samples. Likewise with luminance.

It is very similar to CCD offsetting used by other manufacturers but in this case the CCD is not offset in position but in time.

There is a slight drop in resolution over Interlace mode but as Canon use 1440x1080 CCDs to begin with the resolution drop merely brings the camera in line with others that use lower res CCDs to begin with.

TT

[Daniel Patton]

Thanks Tony.

So what I'm getting by this is that the end result should be roughly the same, by either process.

[Tony Tremble]

Originally Posted by Daniel Patton

Thanks Tony.

So what I'm getting by this is that the end result should be roughly the same, by either process.

Yes, but don't take my word for it just view the posted footage.

I made a mistake above, the green CCD is clocked a field out of sync not half a field. That didn't make sense.

F modes work and they work well as anyone who has used the cameras will testify.

TT