One thing I don't understand

[Juni Zhao]

Originally Posted by Chris Hurd

No that's not how frame mode works. It's not 540 vertical pixels each frame.

You need to get your head around this resolution thing. That's not what makes or breaks an HD camcorder.

Well, I have to admit that I am confused. Pixels are the most fundamental elements in digital imaging, how can I get my head around pixels? I am from graphics background, sorry if I am asking ignorant questions here about video cameras.

[Scott Jaco]

Originally Posted by Jad Meouchy

Steve, I agree completely. I've got an HD100 and it's amazing how many people chew me out for not shooting in 24p. Even after I explain that the per-frame quality of 30p is the same

I thought that shooting 24p gives better quality per frame since the camera uses repeat flags rather than 2:3 pulldown. Can someone explain this once and for all?

[Chris Hurd]

Originally Posted by Juni Zhao

Pixels are the most fundamental elements in digital imaging, how can I get my head around pixels?

Videography has an awful lot in common with still photography. In both applications, the number of pixels involved has little to do with actual "image quality" than any number of other, far more important factors. For example, if we compare a compact little digicam having a 1/6th-inch image sensor with ten megapixels on it, and a Digital SLR with a nearly 35mm image sensor with eight megapixels on it, which one will take a "better" picture? The one with more pixels, or the one with a larger sensor, larger and better glass and a better processor? Obviously the one with fewer pixels... lower resolution, if you will... takes a much better picture because other overriding factors (sensor size, quality and size of lens, etc.) make more of an impact upon image quality than do the number of pixels on the chip.

The same holds true for HD camcorders. It's not about resolution.

Again: 720p and 1080i HDV camcorders are recording the *same* amount of information to tape. The choice between 720 progressive vs. 1080 interlace has less to do with the number of pixels than it does with the motion signature and other image aesthetics plus the camera's form factor, ergonomics, price range, application, workflow, feature sets and a number of other considerations. Hope this helps,

[Steve Mullen]

Originally Posted by Scott Jaco

I thought that shooting 24p gives better quality per frame since the camera uses repeat flags rather than 2:3 pulldown. Can someone explain this once and for all?

It's true that 24p has few frames per second than 30p so yes the math says less compression per frame. But, 30p is half the rate of HDTV at 19Mbps so there is a whole lot of extra bandwidth.

60p, of course, doubles the frame rate and the data that needs to be encoded. But, the GOP length doubles from 6 to 12 which compensates. The encoder is also much better in the HD200HD250.

Bottom-line -- it isn't going to make any difference what frame rate you choose in terms of quality. It's a matter of the look you want. This is a matter of taste.

[Thomas Smet]

The other thing to think about is the limitations of 1/3" chips and lens optics at this price point. There seems to be a certain limit as to how much detail can be resolved from the current crop of HDV cameras even if they can do progressive scan video. Take the Panasonic HVX200 for example. This is a true 1080p camera but the chips use pixel shift on 960x540 pixels. While pixel shift is a great way to get extra detail it is not perfect and you will never get a solid 1920x1080 pixels of detail.

What number of pixels a format records is different then what the camera itself can resolve. Think of a flatbed scanner for example. You could scan a 4x6 photo at 300dpi or 1200dpi. While the 1200dpi will have more pixels there isn't really any more detail to the image then the 300dpi version.

So while a 1080p lower cost camera may have more pixels the extra pixels might end up being a little overkill and end up resolving the same amount of detail in the image.

All of this is of course in theory and somewhat true depending on the camera. The newer 1080 cameras that have true 1080p do seem to have a tiny edge in detail over 720p. It is not a lot however and you have to ask yourself if the extra processing for 1080p video is worth that very tiny edge in detail. Now if we were talking a Cinealta camera then maybe there would be a lot of extra detail but any 1/3" camera I have seen is only going to have a tiny edge in detail if any at all. If your focus is not dead on then any edge the 1080 camera had is out the window.

[Pete Costanzo]

Originally Posted by Chris Hurd

Incorrect. The output is indeed 24p. After all, Final Cut Pro and other major NLEs capture it as 24p.

The "f" does not stand for "fields." It stands for "frame." Hope this helps,

Chris, the following is an excerpt from this article ->http://dvinfo.net/canonxlh1/articles/article06.php. The Canon cameras are indeed interlaced. Correct about the frame/field notation tho.

"The Canon XL H1 CCD block is interlace, not progressive, therefore the 30fps and 24fps frame rates cannot be referred to technically as 30P and 24P. However, 30F and 24F from the XL H1 appear almost indistinguishable from 30P and 24P, as they are basically the same results as progressive scan, but produced by different means. When the XL H1 is set to Frame recording, the CCDs are actually clocked at 24 frames per second. The video signal remains at 24fps as it is passed from the CCD block to the baseband LSI, and through the HD Codec LSI. Only when it reaches the recording output stage is it resampled to 60i via a 3:2 pull-up method."

[Jad Meouchy]

Originally Posted by Steve Mullen

It's true that 24p has few frames per second than 30p so yes the math says less compression per frame. But, 30p is half the rate of HDTV at 19Mbps so there is a whole lot of extra bandwidth.

That's right, the container is 720p60, so 24p has 36 repeated frames while 30p has 30. That doesn't exactly add up to a 20% difference in quality as some may initially believe. I remember seeing some real-world tests a while back and there was no difference in quality between JVC HD100 24p and 30p that couldn't be explained by shutter rate. At the same shutter, the picture was the same. Does anyone know where to find that test, or other tests that disagree?

[Chris Hurd]

Originally Posted by Pete Costanzo

Chris, the following is an excerpt from this article ->http://dvinfo.net/canonxlh1/articles/article06.php. The Canon cameras are indeed interlaced. Correct about the frame/field notation tho.

No need to quote that article to me, Pete -- after all, I'm the one who wrote it!

Since the time that I first put that article together, Canon has released more information detailing how Frame mode works. In the very near future, DV Info Net will host a once-and-for-all in depth explanation of Frame mode, but for now, what's important to understand is what I said in my post above:

Canon's 24F Frame mode does indeed provide 24p output. Period.

Those who doubt this are urged to capture 24F video with the latest version of Final Cut. Guess which capture setting is used -- 1080 HDV 24p! Final Cut cannot distinguish any difference between 24F and 24P. As far as it's concerned, and for all intents and purposes, 24F is 24P.

It doesn't matter that the CCD block is native 1080i. The CCDs are actually clocked at 24 frames per second and what comes out of them is the same results as progressive scan.

There is a slight cost of vertical resolution in this process which is negligible and a complete non-issue for all but the most chronic and habitual of measurebators. Hope this helps,

[Pete Costanzo]

Originally Posted by Chris Hurd

No need to quote that article to me, Pete -- after all, I'm the one who wrote it!

Ha! That's hilarious. I need to check my sources before I try to pull out the ace card ;)

Ok, thanks for the info. Makes more sense now. I read some articles a few months back on the canon frame mode, less in depth, but still walked away thinking interlaced or interlaced+. What's the reason with Canon not just labeling it as progressive then, if the camera outputs progressive...?

[Chris Hurd]

Canon didn't disclose how Frame mode works until November 2006, so anything you read that was written prior to that time has only been speculation. If they had actually called it "progressive," it probably would have caused quite a stir since it is first and foremost a 1080i camera. But in the process I think they created more confusion than they've prevented.

[Steve Mullen]

Originally Posted by Chris Hurd

But in the process I think they created more confusion than they've prevented.

Yes they did. But, your comments are also confusing.

"The CCDs are actually clocked at 24 frames per second and what comes out of them is the same results as progressive scan." Plus, "There is a slight cost of vertical resolution."

If they clock the chips at 24fps and all 1080-lines are output, so output is "the SAME results as progressive scan" then how do they/you explain the "slight cost of vertical resolution." What causes this loss?

The explanation for the loss of V. rez. -- which you agree exists -- is that only 540-lines are output at 24fps. One 540-line field is one field -- and is not a frame. And, this is the reason Canon didn't call it 24p.

No matter how often you deny it -- 24F is type of "field-doubled" video and not true 1080p24. There is no need for measurebating. It's been measured and the numbers published for all to read. And, it's measured resolution shows it to not be 1080p24. It's about 540-TVL/ph on "dynamic" video.

NOTE: "field-doubled" -- contrary the negative reactions by Canon owners is not a negative description. It is a factual description. There are many ways it can be done. Some are crude "line-doubling" while others are very sophisticated and can increase apparent V. rez to more than 540-lines. Canon, obviously does it very well. (However, these processes work best on still images, not moving images. And, video is mostly motion.)

Moreover, the fact Apple named THEIR preset "24p" means nothing about the process by which the video was obtained from the CCDs. It also says nothing about the actual resolution of the video from the Canon.

[David Scattergood]

Today, viewing high quality images are no longer associated with film -- they are associated with 720p60 video they see on ESPN, FOX, and ABC. When NASA wants to see reality -- they don't shoot film. They shoot 720p60. And, in the next few years they will be shooting 1080p60.

The idea that narrative productions must be shot on film is a belief that is challanged every evening in Japan and Korea where drama is shot and viewed using 1080i60. Good drama is the story, sets, costumes, and lighting -- not 19th and 20th Century frame-rates.

In fact, one could argue that putting people in the "reality" of the situation increases emotional involvement. Why is Aliens far more scary on a huge screen than on an iPod? Why was scope and Cinerama created? Why does IMAX use high frame-rates?

Sorry to bring this back up Steve - these high quality images used on ESPN, FOX et al and shot on 720p60 - are we talking about news, current affairs etc or drama's? On another post 30p has been referred to as 'live' as opposed to the filmic 24 (and 25 - is there actually a great deal of difference between these two formats...I've heard, somewere, 25p is being used over 24p in the US now?).
I'm guessing then that 720p50 would not merely be the 25p film look but with "adequate progressive temporal resolution" but more akin to interlaced footage....which is where I'm missing something.
You obviously know the hd100 inside out and are a fan, but I personally can't see a reason to choose this camera if it were not for the true 24p/25p true formats?
Has there been any 720p50 footage posted on these boards? I'd be very interested in seeing this.

Our soaps over in the UK are all shot interlaced - they altered one to a film look once (it looked like they had de-interlaced the footage) - it just didn't work at all (probably alienated tons of viewers and they soon changed it back). Same if the news were shot 'film style'.
Blair Witch works very well because of the 'live' camcorder format used (David Lynch's Inland Empire works well in this format also) but then something like the Exorcist, shot on film, really gets under my skin...it had to be shot in film.
There are some terrible judders in films/tv shows shot in film (was watching a bit of ER at breakfast the other morning - if you take your eye of the subjects the background is pretty dizzy).
I presume real film has to be 24p but the likes of Varicam, HDCAM and say a viper (Zodiac being the most recent film I can think of) - will they continue to use a 24p rate of more the likes of 60p?

If I ever need to look for a second camera then I'd just have to retain the film look but also have the flexability to shoot for different visual requirements (HD250 and a 1080i would be probably tick those boxes perhaps).

Forgive me if I've got the wrong end of the stick here, but this intrigues me very much.

[Chris Hurd]

Originally Posted by Steve Mullen

The explanation for the loss of V. rez. -- which you agree exists -- is that only 540-lines are output at 24fps.

There's only one published result I'm aware of that puts Frame mode at 540 lines, and that's the camera test that I helped organized last April. As much faith as I have always put in Adam over the years, I think our limited Frame mode testing was a bit rushed, as we were working hard to get a lot accomplished in a relatively short period of time. If you ask him, I think Adam will readily agree that certain parts of what we did are worth doing over, including the Frame mode tests. After all, if it were only 540 lines then the image would be quite noticeably soft, and nobody I'm aware of who uses the XL H1 on a regular basis has ever claimed it to be. The 24fps output is worth looking at again more closely, in a testing environment that wasn't as ambitious and as rushed as ours. So I wouldn't rely so heavily on 540 lines if I were you.

No matter how often you deny it -- 24F is type of "field-doubled" video and not true 1080p24.

I am not the one in denial here Steve, and I'm not sure why you're still confused about Frame mode. I know that Canon has tried to help you understand it more than once. 24F is definitely NOT a type of "field-doubled" video. It is not combining fields. It is not doubling fields. It is a full frame capture. And you of all people should know better than to push the term "true" in this business, when you know as well as I do that nothing has been more closely associated with marketing hype than that one word. "True" is a sales pitch which seldom has little bearing on actual real-world results. "True" is a desperate fall-back term that is resorted to when separate technologies compete closely for dollars. But you should know that already.

We're striving for technical accuracy here. Frame mode does not work the way that you seem to think it does. It works instead the way it was explained in detail in New York last November. The publication for which you write, Digital Content Producer, was invited to attend that press conference. You most certainly would not still be clinging to the false notion that it's "field-doubled" had you been there.

...how do they/you explain the "slight cost of vertical resolution." What causes this loss?

Obviously they're not explaining it, but apparently it's the cost of the process, the price that's paid for producing a progressive image from 1080i CCDs. The important thing -- what really matters here -- is that for those who are actually using the camera on a day-to-day basis, it's negligible and entirely inconsequential to the camera's capability to produce viable 24p material that sells. That's what matters most.

NOTE: "field-doubled" -- contrary the negative reactions by Canon owners is not a negative description. It is a factual description.

Incorrect -- it is not at all a factual description, and I think you're well aware of that. While there is truth to the statement that there are different ways to accomplish field doubling, it's entirely irrelevant to this discussion because there is no "field doubling" in Frame mode.

Moreover, the fact Apple named THEIR preset "24p" means nothing about the process by which the video was obtained from the CCDs.

Actually it means everything, because the entire point here is that if the output wasn't progressive, then it couldn't be captured as progressive. And yet that's just exactly what happens -- not only with Final Cut Pro, but also with Canopus / Grass Valley's Edius and every other major NLE (except Avid, which is sadly behind the curve). Video shot in Frame mode is captured as 24p. There's no separate capture setting. There's no separate capture preset. No distinction at all is made between 24F and 24p. The software can't tell the difference because there is no difference. The output is indeed progressive.

If it wasn't, then you couldn't use a 24p capture setting to get Frame mode video into the editor. But that much should be quite obvious.

[Steve Mullen]

Originally Posted by Chris Hurd

The 24fps output is worth looking at again more closely, in a testing environment that wasn't as ambitious and as rushed as ours. So I wouldn't rely so heavily on 540 lines if I were you."

As the only multi-camera tests, I must rely on it. And, it seems only Canon and Pana owners -- who don't like the results -- suggest there were problems with both tests. (These are also the only two companies that tried their best to hide their capture process.)

Obviously they're not explaining it {the loss}, but apparently it's the cost of the process, the price that's paid for producing a progressive image from 1080i CCDs.

Chris, there are interlace CCDs -- see my books -- that can output ALL lines in one pass as long as the output rate is half their maximum (60Hz) interlace rate. Thus, were Canon to use these CCDs -- they could capture 1080-lines at 30p or 24p.

Why then, is resolution lost? As I remember, these CCDs do Row-Pair summation INSIDE the CCDs. Which is perfect when outputting interlace. But, if it can't be turned OFF for "progressive" mode -- then there will be about a loss of 25% V. Rez.

If I remember, that's about the number you have stated. So, I can think of an explanation that fully matches your description. So, I'm not saying you are wrong!

HOWEVER, I can't explain why Canon tried to hide this and why they were willing to create heat by calling it "F" since, if I'm correct, it is "P." Why would they have done this? Waiting on a patent?

I know that Canon has tried to help you understand it more than once.

Canon has never contacted me about 24p. In fact, when I wrote my review they refused several times to explain the process to me. Everytime I asked, the USA group got told by Japan they could say nothing!

I have not a single email or document from Canon -- other than their several refusals to explain 24F. Nor did they invite me to their NYC meeting. In fact, searching the DCP website, the only comment I can find is from Dave Leitner's H1 review: "Canon has chosen not to divulge how the camcorder captures 24F and 30F." So, perhaps, Canon never invited ANYONE from DCP! Mike has Dave's email and he has mine. If Canon really cared about getting the facts out -- they've had over a year to do so.

Actually it means everything, because the entire point here is that if the output wasn't progressive, then it couldn't be captured as progressive.

Nooo! All types of HDV is encoded as "Frames." It is recorded as Frames. It is transferred via FireWire in Frames. It's stored on disk as Frames. The only point at which it becomes interlace or progressive is AFTER it is decoded. A flag tells the decoder's output what they frames are.

The way the sensors work before encoding has nothing to do with MPEG-2 encoding, recording, NLE capture, disk storage, NLE decoding.

[Steve Mullen]

Originally Posted by David Scattergood

I'm guessing then that 720p50 would not merely be the 25p film look but with "adequate progressive temporal resolution" but more akin to interlaced footage....

1080i60 or 1080i50 = 540-lines of vertical information every 1/60th second.

720p60 or 720p50 = 1080-lines of vertical information every 1/60th second.