Cinematography Mailing List - CML
    advanced


23.98 - The Archival Recording Frame Rate Of Choice


>I've read that, when shooting HD, 23.98 fps is the best way to record >historic events, as opposed to 30 fps or other frame rates.

Can someone explain to me why that is?

Many thanks,

Tom Kaufman
Washington, DC based DP


Thomas Kaufman wrote :

>Can someone explain to me why that is? [shooting HD, 23.98 fps is the >best way to record historic events, as opposed to 30 fps or other frame >rates]

Its progressive - more spacial resolution at the expense of temporal - but less interlace issues such as "dot crawl" and "jaggies".

It converts well to NTSC & PAL - that's due to the frame rate and the progressive capture. It also allows a better theatrical release (24/25fps)... better than had you shot 30 or 60i.

Its conceivable that ultimately a 60p or 72p system might downconvert well to other frame rates, but you'd lose exposure due to the shorter shutter cycle, and I'm not certain how stroby things would look downconverted to 24/25fps. The pans that looked great at 72fps could look lousy at 24p (again, due to the faster shutter speed at 72fps).

Plus who's got the bandwidth for all that ? So for many reasons (including video camera marketing) we're at 24p.

Still, it is strange that 24p is thought of as "superior" to, say, 30p. The latter could look great - many commercials are shot 30fps/xferred at 30. You get less strobing and more image info per sec. Its the conversion to PAL/25 and cinema that's not so great.

Mark Doering-Powell
LA based DP


Mark Doering-Powell wrote :

>...Its progressive - more spacial resolution at the expense of temporal...

I once asked Marc Schubin to define "spatial" and "temporal" resolution. He said he was trying to come up with a good definition, but I haven't seen one so far.

How do you define them?

Wade K. Ramsey, DP
Dept. of Cinema & Video Production
Bob Jones University
Greenville, SC 29614


Spatial resolution measures how finely details can be resolved in space.

Temporal resolution measures how finely events can be resolved in time.

Andreas Wittenstein
BitJazz Inc.
http://www.bitjazz.com/sheervideo/


Wade Ramsey wrote :

>I once asked Marc Schubin to define "spatial" and "temporal" resolution. >He said he was trying to come up with a good definition, but I haven't >seen one so far.

>How do you define them?


Good question - I've never heard of an official definition myself.

The usual example of NTSC as 60i, but 30 full frames per second come to mind as a great "cheat" in resolution and capture rate back in the day when cramming all that signal into our airwaves was a bandwidth challenge. Plus the slow decay of CRT phosphors probably helped make that a good solution as well.

But the same can be said for a progressive format shown on an LCD screen or DLP.

Some oversimplified math :

60p at 640x480 is the same bandwidth as 30p at 1280x960, but the former has higher "temporal" resolution while the latter has higher "spacial". 60p would resolve motion better while 30p gives you more pixels per static image. That's the only way I could define it, besides looking at a side-by-side comparison.

Mark Doering-Powell


Tom Kaufman writes:

>I've read that, when shooting HD, 23.98 fps is the best way to record >historic events, as opposed to 30 fps or other frame rates. Can >someone explain to me why that is?

I'm not familiar with the preferred 23.98 HD frame rate being limited to "Historic Events" per se. As I understand it: "24 Frame HD Video" is actually recorded in the 23.98 frame mode in the USA and other NTSC standard countries 99.9% of the time.

Reason :

Because it is cross-compatible with broadcast and off-line editing standards. In NTSC countries, most 24P originated material gets dubbed down to Beta, VHS, or some other 29.97 based medium for editing, distribution, transfer, color correction, etc. It is technically difficult, and expensive, to transfer "24P HD" (at 24.00 fps) unless it is recorded at 23.98.

You make brief mention of the so called "30 fps" standard: Similarly, "30 fps" TV is actually recorded at 29.98 fps. This is/was done to make color TV compatible with black & white TV in NTSC countries, without having to scrap the entire TV infrastructure that had been built until the advent of color TV in NTSC. The "24P" 23.98 HD standard follows this path.

Lew Comenetz
Video Engineer USA


Behalf Of Wade Ramsey :

>I once asked Marc Schubin to define "spatial" and "temporal" resolution. >He said he was trying to come up with a good definition, but I haven't >seen one so far.

The "easy" definition that I often give students is that Spatial Resolution is that which we can physically measure and Temporal Resolution is that which we think we see...

Tom Tcimpidis


Andreas Wittenstein wrote :

>Spatial resolution measures how finely details can be resolved in >space. Temporal resolution measures how finely events can be >resolved in time.

Well, yes. That's like defining daytime as the time during the day. Using the word to define the word is not very illuminating. How about applying these terms to the question at hand?

Mark Doering-Powell was saying that 23.98fps vs. 30fps amounts to more spatial resolution at the expense of temporal resolution. I read that as meaning that temporal resolution is less, in the sense that you are recording fewer slices per period of time. But how is spatial resolution benefiting from this? More pixels per screen because of the slower frame rate? Finer detail is resolved? How can horizontal resolution be affected by these frame rate differences?

Mark Doering-Powell wrote:

>...Some oversimplified math: 60p at 640x480 is the same bandwidth as >30p at 1280x960, but the former has higher "temporal" resolution while >the latter has higher "spacial"....

That makes sense, but I thought the question assumed the same number of pixels per screen, the only variation being 23.98 vs. 30 (actually 29.97). So I guess the original post was actually comparing 1080p with 1080i, interlace will have only 70% of progressive's spatial resolution .

Wade K. Ramsey, DP
Dept. of Cinema & Video Production
Bob Jones University
Greenville, SC 29614


Lew Comenetz wrote :

>You make brief mention of the so called "30 fps "standard : Similarly, >"30 fps" TV is actually recorded at 29.98 fps.........The "24P" 23.98 HD >standard follows this path.

True, although for the purposes of temporal vs. spacial rez, its all the same. And when we say 24p, we almost always mean 23.98, and that's actually running 23.976, no matter what the camera menus tell ya'.

I know I've always shot 23.98p, and never 24.00p. Same with 60i really being 59.94. If we were talking sync sound and telecine speed settings and so on, then its a more important distinction to make.

That darned colourburst and subcarrier signal messing up all those nice, round numbers. Actually engineering genius when you think of it. Did I just say NTSC was genius ?

Mark Doering-Powell
LA based DP


Wade wrote :

>Mark Doering-Powell was saying that 23.98fps vs. 30fps amounts to >more spatial resolution at the expense of temporal resolution.

This is only true if "all else is equal" in this case bandwidth, or bit rate or recording capacity. Think of the situation with PAL and NTSC-- same general bandwidth, but NTSC favours temporal resolution at the expense of scanning lines (vert resolution), while PAL borrows bandwidth from frame rate five frames less, and uses it to add 100 more lines. (The Imagevision 24fps SD video system went a little further, and added 30 more lines at the expense of one less frame, in the same bandwidth space.)

It's not a truism though. It's just that within a specific format space, you push in one area and you pull in another.

To think of it another way, in any one second of SD video, there are 15,750 lines of video, including active blanking lines. (525 x 30) If you think of this as a long strip of stacked lines with no dividing frame lines, you can divide it any way you want. So a temporal resolution of 25 divided into a spatial resolution of 15750, gives, voila!, 630 lines, approximately PAL. You could divide the one second strip of lines into 20, and get 20 frames with 787.5 scanning lines a piece.

So to get better temporal and better spatial resolution, you have to increase your recording capacity, or bandwidth or bit rate, what have you.

Steven Bradford
Film HD Dept Chair
Collins College
Phoenix AZ


Tom Kaufman writes:

>I've read that, when shooting HD, 23.98 fps is the best way to record >historic events, as opposed to 30 fps or other frame rates. Can >someone explain to me why that is?

I'm not familiar with the preferred 23.98 HD frame rate being limited to "Historic Events" per se. As I understand it : "24 Frame HD Video" is actually recorded in the 23.98 frame mode in the USA and other NTSC standard countries 99.9% of the time. Reason:

Because it is cross-compatible with broadcast and off-line editing standards. In NTSC countries, most 24P originated material gets dubbed down to Beta, VHS, or some other 29.97 based medium for editing, distribution, transfer, color correction, etc. It is technically difficult, and expensive, to transfer "24P HD" (at 24.00 fps) unless it is recorded at 23.98.

You make brief mention of the so called "30 fps" standard : Similarly, "30 fps" TV is actually recorded at 29.98 fps. This is/was done to make color TV compatible with black & white TV in NTSC countries, without having to scrap the entire TV infrastructure that had been built until the advent of color TV in NTSC. The "24P" 23.98 HD standard follows this path.

Lew Comenetz
Video Engineer USA.


Mark Doering-Powell wrote :

>60p at 640x480 is the same bandwidth as 30p at 1280x960

I'm not a math guy, but I don't think you are right.

1280 x 960 is 4x the number of pixels of 640 x 480, but 60P is only 2x the number of pixels of 30P.

It's 307,000 pixels vs. 1,228,800 pixels.

So they are not the same.

Jeff "used a calculator" Kreines


So, will the HDCam SR studio deck record at true 24P, or just 23.976 (or as Sony likes to call it, imprecisely, 23.98)?

Are the 4:4:4 cards shipping yet? How does the compression look at 4:4:4, and at 4:2:2?

Is this a good enough tape format for 24P HD work, or is something better in the pipeline? (I know D6 is "better" but it's not economically feasible for most.)

I know the portable deck will record dual stream 4:2:2 for 3D and apparently has twice the bandwidth of the studio deck in this mode.

Will the studio deck be able to play these tapes?

Finally, what sort of street price are people seeing on these decks? How overpriced is the Digibeta card?

Jeff "maybe it's time to sell the Digibeta deck?" Kreines


Wade Ramsey wrote :

>...temporal resolution is less...But how is spatial resolution benefiting >from this? More pixels per screen because of the slower frame rate?

Horizontal resolution is not really affected when you consider it in a vacuum (separate from the rate of capture, and so on). But remember we're talking about moving images here. If you are panning or filming a fast moving object, the increased rate of capture, and the increased "number of slices" per sec will give you a better, "more resolved" image. The issue of movement, shutter speed, capture & display rate, motion blur, interlace vs. progressive, all of it influences the perception of the total image.

A different, absurd example - but I think its of interest :

True slow motion of, say, a rock splashing into the water (say you shoot it at 360fps and view it at 24fps). Compare that to the blurry, choppy image you'd see if you shot it at 24fps and step printed it down to make the displayed water droplet event take the same period of time.

Like watching a blurred slide show.

Mark Doering-Powell


Jeff Kreines wrote :

>I'm not a math guy, but I don't think you are right. It's 307,000 pixels vs. >1,228,800 pixels. So they are not the same.

Jeff,

You are correct, since they're both progressive capture they're not the same bandwidth - I made the mistake of doubling both H & V Rez to make the point (and not using a calculator) and therefore it ended up at a factor of 4x, when it should have been 2x resolution to make it "apples to apples".

I think it should've been something like this:

60p at 640x480 is the same bandwidth as 30p at 896x672 (or damn close to it), but the former has higher "temporal" resolution while the latter has higher "spacial", etc.

Which brings me to a math question :

What is the factor to multiply by to double the resolution from the 60p example I gave above to solve for 30p, while keeping bandwidth equal? I used 1.4x just like f/stops to get close, but its not dead on exact. Using 50% increase or 1.5x is too high. Can you tell I'm ready for the new math ? Or cml-math !

In any case, I know the difference when I see it.

>What we all really want is High temporal sampling like 59.94 FPS

The higher sampling rates work great almost all the time for everyone except for cinematographers who might find it more desirable to use a lower sampling rate (24-30fps) for increased exposure for night/ext work, or available light situations.

60fps is 1.33 stops less sensitive.

The ultimate frame rate is elusive depending on who you talk to, and what you are filming, what look you're after, and how you're displaying it, and to what other formats it must convert. Ultimately it would be great if the display rate could change depending on what's embedded in the signal

Mark "definitely not a math guy either" Doering-Powell


>what is the factor to multiply by to double the resolution from the 60p >example I gave above to solve for 30p, while keeping bandwidth >equal?

Not really new maths, but the correct answer is the square root of two, aka :

1.4142135623730950488016887242097

1.4 usually does the trick, 1.414 is closer if you need it.

It's one of those magic but inexplicable numbers like pi, e, and the golden mean.

Dominic Case
Atlab Australia


Thanks Mark, Lew, Steven, Jeff for illuminating the matter nicely!

Wade K. Ramsey, DP
Dept. of Cinema & Video Production
Bob Jones University
Greenville, SC 29614



Copyright © CML. All rights reserved.