Cinematography Mailing List - CML
    advanced

Using Multiple Video Formats

I am involved in a music video to be shot on a stage using two video cameras at 24p in High Definition. That said, the Director would like to have one of the shots at 18 frames per second. We are thinking of using the Sony F900 in HDCAM format as the primary camera. As a secondary camera we are thinking of using the Panasonic Varicam in DVCPro format. Our choice of using DVCPro is based on the understanding it is capable of variable speeds (specifically 18fps). We hope to use DigiPrime lenses. I should mention that almost all the shots involve some blue screen to create the exteriors for windows. There is also one shot involving a 5-foot by 12-foot miniature set with surrounding blue screen "sky" above the horizon-- we pass over a moonlit field in which sits a cabin and pass into the cabin's window-- which will require some kind of "snorkel" lens such as the Innovision or Century. This last shot would be the one to be shot at 18 frames per second.

It is our understanding that the Panasonic Varicam in DVCPro format is limited to 720 x 1280 pixels and that the Sony F900 in HDCAM format is limited to 1080 x 1920 pixels. I gather there may be some difference between these two systems as far as color bit depth is concerned, though I am unsure what it is.

I am trying to determine the following :

1/. Is shooting in two formats (HDCAM & DVCPro) at their largest pixel dimensions is plain stupid and if so why? If not, why not?

2/. Are HDCAM systems, such as the Sony F900, just as capable as the Panasonic Varicam of shooting in progressive mode at variable speeds but specifically 18fps?

3/. Is there is a notable difference between the look of the DVCPro format and the HDCAM format (at their largest pixel dimensions) when the intended distribution is to television?

4/. Is it possible to use DigiPrime lenses with either the Sony F900 or the Panasonic Varicam? If so, are there issues with either camera and the DigiPrimes?

5/. What is the best choice of "snorkel" type lens for the shot of the night time miniature set using either the HDCAM or the DVCPro formats?

Do certain "snorkel" type lenses show deficiencies when using HDCAM over DVCPro? Are "snorkel" lens systems which have a 90 degree break in them less desirable (f-stop, resolution, etc.) than their "straight" alternatives? Are there "snorkel" type lenses which match the image characteristics of DigiPrime lenses best. (As I understand it, both the Varicam and F900 use B4 mounts, so there should be no issue with using any B4 mounted lens with either).

Please let me know if I have overstepped my bounds with all these questions. I have looked around in the CML archives, but have not found the answers.

Cheers,

Piotr Jagninski
Gaffer / New York City


>1/. Is shooting in two formats (HDCAM & DVCPro) at their largest pixel >dimensions is plain stupid and if so why? If not, why not?

I guess the question is, can you match them? Dumping the DVCPro to Sony HDCAM is no problem, so there shouldn't be any post problems there. Make sure your post house has the FRC-27 frame rate converter on hand as that's how you'll extract your 18 fps footage from the DVCProHD tape. (Or they can run it through Final Cut Pro and output the results.)

>2/. Are HDCAM systems, such as the Sony F900, just as capable as the >Panasonic Varicam of shooting in progressive mode at variable speeds >but specifically 18fps?

Nope. The Varicam is the only one.

>3/. Is there is a notable difference between the look of the DVCPro >format and the HDCAM format (at their largest pixel dimensions) when >the intended distribution is to television?

Are you talking resolution or overall look? I can tell the difference in resolution but I doubt the general public will. I see a lot more detail and tonal scale in 1080 footage. 720 footage just looks like hi-res video. It'll be interesting to see what happens when you cut them back to back.

I'm sure the differences will be more noticeable.

Make sure you've got a very good DIT on set to match those two cameras. I'm told it can be done but I'm not sure which camera ends up being compromised. I love the Varicam colorimetry and I've never seen it matched on an F900, but I'm assured it can be done.

The rest you'll have to get from someone wiser than I.

Art Adams, DP [film|hidef|video]
Mountain View, California
San Francisco Bay Area - "Silicon Valley"
http://www.artadams.net/
Local resources : http://www.artadams.net/local


Art Adams wrote:

>Dumping the DVCPro to Sony HDCAM is no problem, so there >shouldn't be any post problems there.

FRC-27... got it. I will pass that on to the Producer. Does this device go by a more elaborate name or do all post houses know it by that name?

> Nope. The Varicam is the only one.

Could you explain to me why this is so? A DIT once told me slower and faster than 24p frame rates could be achieved with HDCAM... never having heard of it before I must admit to wondering if he was not mistaken.

>Are you talking resolution or overall look?

To explain what I was after further, knowing that HDCAM has greater pixel dimensions (and I think greater color bit depth--am I correct here?) than DVCPro, I would assume one would perceive a difference between their images. Being that we plan to shoot any scene with both cameras at the same time, my concern is that such a difference would be obvious to the point of being unacceptable.

>Make sure you've got a very good DIT on set to match those two >cameras.

Do you happen to remember who said it could be done? I would love to talk to that person about the ins and outs of matching the two systems.

> The rest you'll have to get from someone wiser than I.

Thank you very much for answering as much as you did!

Cheers,

Piotr Jagninski
Gaffer / New York City


Hi Piotr-

1/. The FRC-27 is the Panasonic frame rate converter. It'll take all of your "overcranked" and "undercranked" footage and convert it to 24 fps (29.97/30/60 too?) so that you can see the effects of your fast/slow motion footage. As Art mentioned, you can also load the footage into Final Cut Pro over firewire with the Panasonic HD1200A deck. There's a free plugin available on the Panasonic Varicam website that will do the frame rate conversion in Final Cut.

2/. I had heard a rumour about an intervalometer for the F900, but I don't think it ever made it into production. The F900 was not designed for variable-frame rate capture, so any device specifically designed to do that would be "tricking" the Sony HD cameras into shooting variable frame rates. I couldn't even begin to go into the logistics behind this, as I don't know how this would be done, and it shouldn't matter because you'll want to use the Varicam anyway.

3/. DVCPROHD(100) has better color sampling (4:2:2 vs. 3:1:1 on HDCAM).

4/. Don't know anything about shooting miniatures with HD cameras. Waste of virtual paper...

Hope this helps!

Henry Gretzinger
NYC


>4/. Don't know anything about shooting miniatures with HD cameras. >Waste of virtual paper...

The only thing I know is that you need to make them of a fire retardant material, as you'll be lighting them up very, very brightly.

Art Adams, DP [film|hidef|video]
Mountain View, California
San Francisco Bay Area - "Silicon Valley"


Henry Gretzinger wrote :

>3/. DVCPROHD(100) has better color sampling (4:2:2 vs. 3:1:1 on >HDCAM).

Henry Gretzinger says DVCPro (100) has better color sampling than HDCAM, yet Art Adams says that in a side by side viewing of 1080 footage versus 720 footage he can see greater tonal range in the 1080 (HDCAM) footage. Wouldn't color sampling directly affect tonal range?

As yet, I do not understand terms like 4:2:2 and 3:1:1...perhaps this is where my lack of understanding lies?

Cheers,

Piotr Jagninski
Gaffer / New York City


Piotr -

Color sampling ratios certainly affect the tonal quality of an image. What I heard Art say is that the Varicam generates a more natural or true image, but that it lacks detail (when compared to the F900). Fun fact: the DVCPROHD codec is simply four DVCPRO(25) codecs combined to form one large image (except that the final image is 4:2:2 vs. 4:1:1). The color sampling ratio is important because it represents the camera's physical ability to acquire an image.

Chroma subsampling still doesn't make sense to me (several engineers have tried to drill it into my head) and I can't say that I care, but basically, with a 4:1:1 sample ratio, you're getting four times as much luma information as you are color information. With chroma subsampling, the camera takes those 3 missing samples and rounds them up or down (quantization) according to the next chroma sample. Not to get too long-winded but essentially it doesn't matter because the human eye can only process half the color information that it receives. My head hurts, I'm taking a nap :-)

Henry Gretzinger

P.S. For what it's worth, I agree with Art about the Varicam. It's a pretty image, but lacks the resolution of the CineAlta cameras (particularly the F900/3).


>>Wouldn't color sampling directly affect tonal range?

My interpretation of what's going on is that the reason I see more subtle shades in the F900 vs. the Varicam is that the F900 sees 50% more resolution. I'm not sure that the increased number of shades of gray has to do so much with color sampling (in this instance) as it does with the actual resolution of the camera being able to see more shades of gray in whatever it is shooting.

For example, there's one shot I've seen done with both cameras that illustrates the difference: sunlight coming through patchy clouds and illuminating small waves on a dark sea. On the F900 I see all sorts of shades and textures in the sea and I can see the sunlight rippling off of very small waves. It really is a magical image. On the Varicam I see what looks like a really nice video image: the shades and textures that I saw in the sea on the F900 footage are now half of what they were, and the whole shot loses that "WOW!" feel. It's still nice, and the color is probably more accurate and pretty, but the subtlety is gone. I'm firmly convinced that's because I'm seeing 360 fewer lines of resolution.

In theory the Varicam is 4:2:2 and the Sony is 3:1:1. I've heard people debate those numbers saying they don't apply in the same way to HD technology. I don't know technologically what is going on but I know this for sure : The Varicam gives me truer color more easily than the F900 does, but the F900 gives me much greater resolution and creates a picture full of subtlety that the Varicam won't. And if you shoot blue screen or green screen on either HDCAM or DVCProHD formats you're probably going to have trouble pulling nice keys because both formats are heavily compressed.

Your issue in combining the footage might be less about matching the cameras as it may be that you should record the images from both cameras onto an uncompressed 4:2:2 deck. I'm told that the camera heads can be matched fairly easily but once the image hits the tape you've moved into a totally different ball game. I'd take this over to CML-HDTV and ask this question there, but my guess is that if you have a good DIT who can match the camera heads and if you feed the footage from both (via SDI, presumably) into a deck that records 4:2:2 (preferably uncompressed) you'd be in very good shape.

>>As yet, I do not understand terms like 4:2:2 and 3:1:1...perhaps this is >where my lack of understanding lies?

That just talks about how much information gets thrown away when the image is laid down onto tape. My crippled understand of how that nomenclature works is :

The first number is the luminance. A full luminance signal is usually denoted by "4". I also think of this as representing the green channel since the luminance signal is derived completely from the green chip.

The other two numbers represent the other two colors, blue and red, and basically tell you how much information is being tossed away on tape and made up later. 4:2:2 means you have a full luminance channel but you're only devoting half the bandwidth to storing blue and red. You're essentially sampling every other pixel for blue and red as opposed to sampling every pixel for luminance. There's an equation by which that information is stored and then reconstituted later.

Here's a more technical explanation :

http://forum.matrox.com/rt2000/Forum8/HTML/000052.html

I await the waves of corrections.

Art Adams, DP [film|hidef|video]
Mountain View, California
San Francisco Bay Area - "Silicon Valley"


Here's a copy of an article I wrote for Showreel magazine :

I'm donning the flame proofs...

What is real HD?

The answer to that question should be obvious, unfortunately it isn't.

This is partly because of confusion caused by the FCC in the US and partly because of misinformation disseminated by manufacturers to make it easier to sell their products. Hype in other words.

To a purist it’s very simple, HD is 1080 * 1920, is progressive, is 4:4:4 and is recorded uncompressed, I'll get back to what all that means later.

Unfortunately, according to the FCC, HD can have as few as 480 active lines as long as they are progressive scan.

Equally unfortunate is the contention of some manufacturers that system’s that only store 109 pixels of the original 1920 of colour are HD.

I certainly wouldn't be happy if someone paid me a seventeenth of my daily rate but I'm supposed to be happy with a seventeenth of the colour in a picture!

Let’s get to the definitions of the terms:-

1080 * 1920 should be fairly clear; the picture is 1080 lines or pixels high and 1920 pixels wide. The Varicam works at 720 * 1280 but can run at twice the frame rate of other cameras as a result. Normal PAL TV, or SD, is 576 * 768; widescreen SD is 576 * 1024.

Progressive, normal TV works in an interlaced way, the 25 pictures you see every second are actually shown in 2 parts, they are sliced into alternate lines and the odd lines are transmitted followed by the even lines. These are called fields and are transmitted 50 times per second to give you 25 complete frames per second.

A major problem with this is that there can be movement between the odd and even lines and this gives an odd feel to the pictures. I believe that this is one of the reasons that people prefer the look of film on TV.

When film is scanned the odd and even lines of a frame are identical so you get no inter line blurring.

This is also how progressive scan TV works, like film it creates 25 original complete frames every second and they are displayed/transmitted as 25 complete frames every second.

The different formats are identified with a p or an i to differentiate them, so to me a 1080p image is better than a 1080i one.

4:4:4 what the hell does that mean and why should you care?

It means simply that you have 4 quarters of the monochrome image, 4 quarters of colour picture A and 4 quarters of colour picture B.

Why it’s described this way is irrelevant ?

Normal SD is generally recorded at 4:2:2 this is because research has shown that people are more tolerant to less detail in colour than they are in black & white, so TV systems were designed to save bandwidth by losing half the colour information.

Of course whilst your eye may be tolerant of that loss keying, colour correction and effects programs aren't. So for critical application you need to work in 4:4:4.

What do HD systems work in? well they vary from 3:1:1 via 4:2:2 to full 4:4:4. I'm going to deliberately avoid naming companies here but how anyone can expect you to make pictures with a full range of post production possibilities when they throw away a quarter of the monochrome picture and three quarters of the colour picture is totally beyond me!

Finally we come to compression, there are two different approaches to judging compression, mathematically lossless which has a maximum limit of three times and visually lossless which has whatever maximum the person designing it thinks is visually lossless.

Now bear in mind that the visually lossless decision is made by an engineer, and they are not famed for their visual sense, I mean, look at how they dress! They are also under pressure from an employer to get the data rate down.

Personally I'm only interested in uncompressed although the new Sony HDCam SR portable is only 2.5 times compressed at 4:4:4 so I guess I'd be happy with that! If I could get it that is, it’s not due for delivery for at least another 6 months.

A comparison chart :-

Resolution Colour Bit depth Compression
Film 4,000+ 4:4:4 20+ None
Viper 1080 4:4:4 10 None
HDCam 1080 3:1:1 10 4.7
Varicam 720 4:2:2 10 6.7
DigiBeta 576 4:2:2 10 2
MiniDV 576 4:2:0 8 5

Now some of the manufacturers may decide to show those numbers in a different way, using much larger number for the HD images but the chart is what they break down to. Anything else is designed to confuse you and obfuscate the issue.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based
www.cinematography.net


Geoff Boyle wrote:

>> Viper 1080 4:4:4 10 None

Since the Viper is a camera, not a recording format, you really can't accurately make this statement, IMHO.

Especially since the Viper is often recorded in various formats -- some of which aren't 4:4:4 and some of which may be compressed.

It's a recorder-agnostic camera...

Jeff "but it has a prism, which gets a best-supporting nod for it's appearance in Collateral" Kreines


Jeff Kreines writes :

>> Since the Viper is a camera, not a recording format, you really can't >accurately make this statement, IMHO.

I knew someone would come up with this!

When it was launched the Viper was intended solely as a Filmstream device.

It was only later that HDstream was added.

So it's perfectly legitimate to call it a 4:4:4 uncompressed camera.

If someone decides to record the Kinetta output to DV does that stop the camera being a 4:4:4 camera and make it a domestic one?

Don't reply that was a rhetorical question.

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based


Geoff wrote:

>>Here's a copy of an article I wrote for Showreel magazine....

Thanks, that's a keeper!

I trust there was a re-writer for literary and grammatical improvements...

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


Geoff Boyle wrote:

>> Unfortunately, according to the FCC, HD can have as few as 480 active >lines as long as they are progressive scan.

The FCC has never referred to anything less than 720p as HD. They have allowed for 18 or so digital broadcast formats, but it would only be through the general public's misinterpretation of the nomenclature that 480p would be referred to as an HD format. Even when Fox was using it, they referred to it as "high resolution," not high definition (which they do use in reference to their 720p programs that are now being delivered).

Mike Most
VFX Supervisor
IATSE Local 600
Los Angeles


Mike Most writes ;

>> The FCC has never referred to anything less than 720p as HD.

I stand corrected.

However, you know that was coming, I really thing high resolution as opposed to high definition........

Cheers

Geoff Boyle FBKS
Director of Photography
EU Based



Copyright © CML. All rights reserved.