24f vs 24p

[Fischer Spooner]

What is the difference between 24f and 24p?

[Trenton Scott]

In a nutshell, 24F is virtual 1080 24p.

[Pasty Jackson]

Originally Posted by Trenton Scott

In a nutshell, 24F is virtual 1080 24p.

24F has nothing to do with image size. 24 Frame mode is basically when a non-progressive imaging chip achieves 24P to tape, usually by means of frame blending of some sort.

-pasty

[Andrew Olson]

I was also wondering this, thanks for the answers. Is there one that's considered better than the other? I think I've heard that 24p is better, but is that true?

[Tony Tibbetts]

Ugh... this again. I can't believe these threads still pop up. 'sigh'...

Better is subjective. You can extract 24 progressive frames from 24f. The resolution is just as good (if not better) as any of the other HDV based 24p cameras.

[Justine Haupt]

...yes, and as someone said above, 24F is an simulation of 24P.

Short answer -- 24P is better. There might be some great looking 24F out there, but there's no subtitute for the real thing IMO.

[Pete Bauer]

Not so short answer -- neither is "better."

24F is a term Canon uses for its method of creating 23.976fps video using interlaced CCDs. F-mode processes the image (by means as yet undisclosed) from an interlaced CCD to produce a progessive-frame video file. For HDV, it lays the image to the tape as a true 23.976fps file, not 60i pull-down as is done in most "affordable" 24p cameras. So editing software that handles 24F puts it on a 24fps timeline without having to do pulldown processing, but otherwise exactly like typical 24p-from-60i video. If you use HD-SDI to get full raster 4:2:2 video or use composite-out, the 24F will leave the camera as 60i 3:2 pull-down.

So 24F and 24p are just two flavors of 24fps video. One's not a substitute for the other, and neither is inherently "better." They are just two different means of creating a 23.976fps video file. That's all. The resolution, image detail, latitude, color space, color rendition/depth, etc are dependent on the hardware and software used. The 24F from the XL H1 has as good or better image detail as any currently competing camera, but each camera has it's own particular strengths and each shooter has his/her own particular likes, so I think "better" is a rather meaningless word in the endless 24F vs 24p measurebation.

[Glenn Chan]

measurebation

That's an interesting word. Sometimes we are too focused on insignificant details.... seeing the trees instead of the forest.

[Bill Pryor]

Great word!

So, from Pete's post, I think my question I was gonna ask is answered: the 24f mode of the Canon is true 24fps, not the pulldown. So if I set an in and out point a second apart and count the frames, there will be 24, not 30 with 6 of them blurred, right? And if that's so, can we assume the new A1/G1 will do the same thing?

[Nate Weaver]

24P in HDV is done differently than in any other video format (I'll leave 24F out of this discussion because 24F is just a Canon name for how 24P on tape is created).

Since HDV is more or less a computer file streamed to page (an MPEG file), there's more flexibility in what data is written. Or put another way, MPEG2 had a history of multiple framerates and frame sizes long before somebody had the idea of streaming it to tape.

Ok, so a 24p MPEG stream doesn't use pulldown. That's the news. None of the 24p HDV cameras use pulldown ON TAPE. On tape is the qualifier here. The data stream is 24 discrete frames. That's it.

But, within that stream, are hints for the MPEG decoder device as to how to display that 24fps info for 29.97 devices. The hints are called repeat flags. So one frame will have an instruction tagging along with it to display for say, 2 fields. Another will have an instruction for displaying over 3 fields. You know the rest...it's instructions for the DECODER device to create the pulldown ON THE FLY. Decoder in this context means hardware decoder in a tape deck, or something of the sort.

So anyway, what this means is that some software will read the stream verbatim and tell you the footage is 29.97 interlaced, and it'll display it in this manner. Other software will be smarter and read it all as just 24 discrete frames.

There's so much confusion on the boards about this, I just had to shout it out. I'll credit David Newman of Cineform for explaining this to me once in the HD100 forum.

[Bill Pryor]

Excellent explanation. So I can shoot 24f with the new Canon A1, should I get one when it comes out, load it into a 24f (23.97etc) timeline in FCP (when the upgrade comes out in a couple of weeks) and I'll have real 24fps. This is good and makes me more prone to liking this camera when it gets here. I guess my confusion came from the 2 different 24p modes that the XL2 does, and the DVX100 and the HVX200 for that matter, ie., the 24p and 24pn or 24 advanced, or whatever they all call it. My understanding there is that if you shoot the "PN" or "advanced" mode, you get the "real" 24fps but you get the pulldown with the other modes. Since the XL2 and DVX aren't HDV and neither is the HVX, I guess things are different there.

[Nate Weaver]

You got it right. There's two reasons for the extra fancy pulldown in HDV. One is the reason I gave above, which could be distilled into "because they can".

The other reason is because imagine if 24p was inserted into a 29.97 1080i stream the way the DVX has been doing it for years. If that was the case, you'd have to recompress ALL your footage to remove the extra fields, and we all know how painful that would be in HDV. It'd take forever, and your HDV would look worse for it...more so than if you did the same with DV codec material.

The way 24p HDV is done now, no bandwidth is being wasted on repeat fields (unlike 24P DV, or even Varicam/DVCPRO HD). It really truly is clever as hell.

[Nate Weaver]

What's more, is that the XDCAM HD cams, Canon XLH, XHA1, G1 are all codec compatible, meaning at the NLE level, they're all equals. I suspect the new Sony V1 (or whatever it's called, I'm losing track) will do its 24p mode the same also.

The odd man out (as always) is JVC :-P

[Don Donatello]

i'm missing something here ?? f , P , Cf etc

i notice the Texas shoot put list some resolution #'s for the camera's ..

canon 24f as stated in some of the post is "progressive" ( when all is said and done) i don't quite understand why the resolution is lower then interlace ? the shootout gives 800/700 for H/V resolution - and then states in frame mode 800x540?? i was always under impression that progressive has more res then interlace ?? ( or is that only on moving or shots with motion in them)

and why is the sony 350 camera 800x800 in 24p mode and then 800x540 with shutter on ? how can it drop 20% .. where'd the resolution go ??

[Nate Weaver]

Hey Don,

The Canon is a problem of semantics. Canon doesn't want to call it 24P, because technically, that is reserved for CCDs that are progressively scanned (and purpose built for that, which is technically more difficult and more expensive). So, they take a 1440x1080 interlaced chip, and do signal processing voodoo between the field scans off the chip, and end up with 24 discrete, full, frames per second.

Of course there's no free lunches, so the downside of this is that because they're doing weird things with fields, vertical resolution takes a little bit of hit. In the end, the signal processing outputs 24 full frames per second which is then passed along to the MPEG encoder, yadda yadda yadda.

So the Canon is not true progressive, but it does possess the most important part of a true 24p camera, which is 24 separate images per second, captured with an even interval between each image (Something the Sony Z1, with it's "Cineframe24" mode, doesn't manage to do)

The amount of vertical resolution compromised is up for grabs. A 1080 line system, with Kell factor accounted for, is only good for about 750-950 lines vertically, depending on who you ask. Worst case scenario for straight-up field blending on the XL-H1 would be 540 lines vertically. Canon's special secret-sauce DSP is somehow doing a bit better than that, but still south of what it could do in interlaced mode.

The Sony XDCAM is doing the same thing, but in the end comes out a tiny bit better (and probably only because of better glass). While lens shopping or my own 350, I put on a lot of lenses and the best I saw was a little better than 800 lines H, and little less than 800 lines V. The only difference is that Sony's 24p DSP trick somehow retains the interlace resolution but only when the shutter is off. Think of the CCD being scanned at 48 hertz interlaced, and all this starts to make more sense.