Genesis for "Superman Returns"

[Guest Jim Murdoch]

It's three RGB stripes wide by two photosites high per HD pixel, or 5760 x 1636 photosites on the chip for a scope image.  That gives them 1920 complete RGB sample sets across, by 1636 vertically. 

-- J.S.

<{POST_SNAPBACK}>

Not according to John Galt! (Although personally, I don't know whether that means anything or not).

 

But he was quite emphatic that the rows of RGB triads are averaged vertically to produce 1080 lines. Which doesn't make much sense to me; because he was also quite emphatic that the Genesis chip doesn't use the alternating RGB pixel arrangement found in single-chip standard defintion cameras to cancel out colour aliasing effects.

 

So why they seem to be "wasting" the extra vertical resolution is the big mystery. As far as I could gather from what Galt said, they don't seem to be using the extra native resolution to give them better detail correction, so why didn't they just have half as many double-sized pixels?

[Guest Jim Murdoch]

Sampling, say, 1620 vertical lines and then interpolating with an algorithm that would convert three uninterpolated lines into two interpolated ones (much the way that D-20 interpolates) would yield 2.37:1 aspect ratio and full 1920 x 1080 RGB data. Probably even without infringing any Thomson's Dynamic Pixel Management patents... B)

 

<{POST_SNAPBACK}>

"Interpolation" EQUALS "guessing"

"guessing" DOES NOT EQUAL "high resolution"

Film does not guess....

[Ilmari Reitmaa]

"Interpolation" EQUALS "guessing"

"guessing" DOES NOT EQUAL "high resolution"

<{POST_SNAPBACK}>

This is a bit of a generalization. Interpolation equals educated guessing; educated, because it relies on the assumption that the majority of screentime comprises of natural scenes, in the sense that we do not need to distinguish between different images of white noise or artificial patterns specifically designed to bring about the "worst guess" for some interpolation algorithm. Such image elements are rare enough in natural scenes that interpolation is able to reconstruct the image with sufficiently small error residual.

 

If it were that for an interpolated pixel the probability distribution over the available RGB values was uniform, interpolation would indeed equal "guessing" and as such would be an illegitimate method of improving resolving power, but as it happens, in practice this never is so. In natural scenes the RGB values of interpolated pixels strongly depend on those of the surrounding pixels, therefore making it possible to extract information by interpolation. Note that interpolation is not a case of generating information that wasn't there, it's a case of not throwing away information that, for all practical purposes, is there. Color filter array is a form of fixed compression rate lossy compression, interpolation being the decompression method. Consequences of this can be seen e.g. with low-light images containing noise, for which color interpolation will produce larger error residual, as would any fixed compression ratio lossy compression scheme (or equivalently, smaller compression ratio with fixed error residual).

 

At any rate, resolution (in the sense of pixel count) is a secondary concern, provided that it is sufficiently high (which, arguably, 1920 x 1080 isn't for cinematic purposes). What is of concern is the ability to capture visual information, preferably close enough to the amount processed by the human visual system. This is why interpolating from higher resolution image down to a 1920 x 1080 will yield, dare I say, better image quality than direct 1920 x 1080 acquisition. Simple resolution analysis can be misleading.

 

Film does not guess....

<{POST_SNAPBACK}>

Agree. Also agree that the amount of visual information captured by film is higher than that of any 1920 x 1080 4:4:4 system. Just for the record.

 

As for the non-alternating RGB pattern, my take on it would be that an alternating pattern isn't necessary because the optical low-pass filter (supposing Genesis is utilizing one) will remove all aliasing artifacts before any color filtration even takes place; furthermore a non-alternating RGB pattern makes it possible to use a low-pass filter that vertically convolutes adjacent pixels, instead of having to convolute every second pixel, yielding higher vertical resolving power.

Edited April 21, 2005 by Ilmari Reitmaa

[Guest fstop]

Take a look at this footage of the steadicam track through the crowd- you can CLEARLY see the camera has a film magazine on top. Singer talks about enhancing his crowd scene in post, so is this one of the 35mm SFX plates?

 

CLICK ME!

 

Also, this marks the FIRST time a steadicam has been used in a Superman movie.

[David Mullen ASC]

I can't tell looking at the footage if it is a film camera or not, considering that the Genesis also looks like it has a physical "film magazine" because of the way the SRW1 tape drive can be attached top-load or back-load.

 

Take a look at this footage of the steadicam track through the crowd- you can CLEARLY see the camera has a film magazine on top. Singer talks about enhancing his crowd scene in post, so is this one of the 35mm SFX plates?

<{POST_SNAPBACK}>

[Tenolian Bell]

I zoomed in on it as OSX allows you to zoom your screen.

 

 

The camera on the steadicam is black. It looks more like an Arri 35-3 to me.

It also has the silver strip on the side like old Arri mags have.

 

Would Panavision have painted the Genesis black?

[David Mullen ASC]

Would Panavision have painted the Genesis black?

<{POST_SNAPBACK}>

 

Looks gray to me here:

http://thetechnofile.com/index.php?m=200503

[Guest Jim Murdoch]

This is why interpolating from higher resolution image down to a 1920 x 1080 will yield, dare I say, better image quality than direct 1920 x 1080 acquisition. Simple resolution analysis can be misleading.

<{POST_SNAPBACK}>

No mystery there! To prevent aliasing artifacts, a CCD or CMOS sensor HAS to have an optical low-pass filter that prevents any detail finer than two pixels wide from reaching the sensor surface. Now, strictly speaking, this isn't necessary for ALL images, and the resolution would be doubled if the filter was removed, but the trouble is, the camera or its operator has no way of knowing in advance what sort of images are going to come through the lens. So in the real world, a practical 1980 x 1080 sensor can only ever be allowed to produce 960 x 540 pixels, period. Finito. That's it!

 

BUT, if you capture the same image on film and scan it at 4k (or more), and then downconvert to 2K, with modern image processing equipment there will be a massive improvement in apparent resolution.

 

Why? Because the downconverting computer can compare the original 4K image with the downconverted version and look for precisely the sort of aliasing artifacts that the optical low-pass filter is needed to prevent in a video camera. These can be removed automatically by software, in most cases leaving the vast majority of the image intact.

 

It's like having a custom-designed optical low-pass filter for every frame, which only applies the low-pass filtering to the areas of the image that actually need it.

 

As for the non-alternating RGB pattern, my take on it would be that an alternating pattern isn't necessary because the optical low-pass filter (supposing Genesis is utilizing one) will remove all aliasing artifacts before any color filtration even takes place;

<{POST_SNAPBACK}>

 

The Genesis definitely uses one (again, according to John Galt). But that's only part of the story.

 

Consider a horizontal light-to-dark transition that covers several RGB "macropixels". If the pattern is RGBRGBRGBRGB surely you can see that in all cases the "R" pixel of each triad is going to receive more illumination than the "G" and "B" ones of the same triad. In other words, the light-to-dark edges will have reddish cast, and conversely the dark-to-light ones a bluish cast.

 

By alternating the pixel arangement and then vertically averaging adjacent rows this effect should be largely cancelled out, which is why all the standard definition CCD RGB cameras I've ever seen do precisely that.

 

This is what I don't understand: they do that for non-broadcast SD cameras, but deem it unecessary for something that's going to appear on a fifty-foot screen!

[Guest Jim Murdoch]

Looks gray to me here:

http://thetechnofile.com/index.php?m=200503

<{POST_SNAPBACK}>

That's a pretty good item, although I wonder where they got their information from! I notice that (well according to them anyway) PV are still claiming that the Genesis has an "11-stop" overload headroom. That's a pretty cool trick considering it only has a 12-bit analog-to-digital converter!

 

So if you were trying to capture a scene with, say a "practical" window that is two thousand time as bright (11 stops) as a living room, if the window pixels were just fitted inside the 12-bit linear capture range of the ADC, the pixels representing the living room would have only ONE BIT to play with. In other words, they could be either 100% black or 100% white!

 

Apparently later versions of the Genesis are supposed to have a 14-bit ADC, which would then give them THREE bits (ie 8 levels) to play with . Oh, that's all right then:-)

 

Frankly my dear, I still don't get it....

[Mark Allen]

If this thing on the back is presumed to be the magazine (or deck) - in this shot it's curved in a way that the genesis' deck does not curve. The genesis deck is, in fact, quite square and the specular light reflecting on it would seem pretty much like a 90 degree angle.

 

I don't have a vested interest - just making an observation.

 

click to make larger

 

 

 

Seems like we could also just ask someone what they're using.

Edited April 22, 2005 by Mark Douglas

[Sam Wells]

Applying the classical "Figures On The Grassy Knoll" scientific analysis I conclude Superman is using a Genesis AND a Millennium II, as well as three HDV-Z1's, a Kinor - which oddly enough seems to read "Property Of MIR" on the side although my Russian's a bit rusty (then again, so's the Kinor....) - along with what looks suspiciously like a late model Pixelvision (the one with the anamorphic adaptor) and a vintage Debrie Sept.

 

BTW the camera you see on the Steadicam is just a decoy.

 

-Sam

[Rik Andino]

Applying the classical "Figures On The Grassy Knoll" scientific analysis I conclude Superman is using a Genesis AND a Millennium II, as well as three HDV-Z1's, a Kinor - which oddly enough seems to read "Property Of MIR" on the side although my Russian's a bit rusty (then again, so's the Kinor....) - along with what looks suspiciously like a late model Pixelvision (the one with the anamorphic adaptor) and a vintage Debrie Sept.

 

BTW the camera you see on the Steadicam is just a decoy.

 

-Sam

<{POST_SNAPBACK}>

 

Why yes the decoy camera! Should've thought of that.

 

Do you figure they got the Millennium and the Kinor by the white picket fence?

 

And the three Z1u's around the grassy knoll?

 

 

:)

 

Doesn't these all seem like a tad bit ridiculous to the rest of you?

I was staring at a blown-up blurry picture for 20 goddamn minutes!

It's just a camera and just a movie--we might be becoming too obsessed

Maybe a breath of fresh air will do us all some good.

[Guest Jim Murdoch]

Seems like we could also just ask someone what they're using.

<{POST_SNAPBACK}>

Damn! (Smacks Forehead) Why didn't I think of that?!

That's the weird thing: Panavision won't tell anybody anything!

If they reply to any questions at all, all they ever do is refer you to their useless Web Site.

[Guest Jim Murdoch]

It's just a camera and just a movie--we might be becoming too obsessed

Maybe a breath of fresh air will do us all some good.

<{POST_SNAPBACK}>

Well you see, that's the trouble; it's not "just a camera" it's supposedly a massive technological breakthrough, at least according to Panavision last year it was.

 

So what happened? Was the long awaited Pana-Prince-Regent born with a hare-lip?

 

Where's Oliver Stone when you need him :rolleyes:

[Mark Allen]

I was staring at a blown-up blurry picture for 20 goddamn minutes!

<{POST_SNAPBACK}>

 

LOL. It only took 40 seconds to snap, blow it up and post it. :)

 

I was just happy I found Waldo in the original picture, that was the tough one.

[Max Jacoby]

Doesn't these all seem like a tad bit ridiculous to the rest of you?

It's just a camera and just a movie--we might be becoming too obsessed

Maybe a breath of fresh air will do us all some good.

 

Quite right!

 

I mean the Genesis is just an HD camera, it's not the holy grail of filmmaking. Sure it is probably an improvement over current HD cameras, but still it is no match for 35mm. And as long as this is not the case, I see no reason to get excited over it. This whole thread reminds me of when the F900 was first used on 'Star Wars' and we all know how that one turned out...

 

And while we're at it, I must say that I find the whole 'HD only' forum to be the most useless part of www.cinematography.com simply because it is low on actual facts and full of speculation. There are interesting discussions here every now and then, but most of the time there is very little here that actually increses my knowledge of HD.

 

That being said, I think the manufacturers of these cameras, chief among them Panavision and Dalsa are also party to blame because they do not seem to want to give an honest assessment and specs of their cameras. ARRI are the only ones who give you an honest answer about their D20. Because really, if a company likes to shroud its camera in mystery, then that leads me to think that their camera isn't that good to begin with.

[David Mullen ASC]

I don't think Dalsa is particularly secretive. I went to a presentation at their open house and they were showing us details like cross-section electron microscope photography of their sensor sites, which bordered on more information than I wanted to know... It was a pretty thorough presentation on the nature of their Bayer filter, how it is processed, etc.

 

They'll tell you pretty much anything you want to know, unless they are aren't sure. Things like optimal ASA rating, color-balance, etc. their attitude is: you test it and tell us what you think. Because they really want the feedback. It's still somewhat a work in progress. Their claims that the camera is "4K" make sense from their logic so I stopped arguing with them about that point, but I think practically it is more like 3K. But what really matters are the results and the comparison tests I saw were compared to 4K scans of 35mm 5212 (100T), blown-up several times over to look at small portion of the 35mm frame and the Dalsa frame to compare edge sharpness, grain (or lack of), etc.

 

As for the Genesis, the specs of the HDCAM-SR recording format that it uses are pretty well-documented so we all know what's getting recorded, which is 4:4:4 HD, not 4K or 3K, etc. I don't know why everyone thinks the camera is so mysterious other than the fact that they aren't using a Bayer pattern. Again, what matters are results. My gut instinct tells me, from the design, that essentially you've got the same quality as 4:4:4 HD coming out of the Viper or F950... except that you also have the depth of field from using a 35mm-sized sensor. Maybe some improvement in dynamic range -- that's harder to know for sure. And Panavision is not denying that this is an HD-resolution camera so I'm not what's the big deal here because they are vague on the RGB striping of the chip!

 

Again, we're not engineers here, we're filmmakers. This obsession over the tiny manufacturing details is not surprising for tech-geeks like us, but what matters is if we like the pictures it makes.

 

Also what matters is that cameras like the Genesis, Arri D20, and Dalsa represent a design-shift away from the tradition video camcorder design, for better or worse, and closer to a film camera design. It's a step TOWARDS creating digital cameras that deliver the same quality and convenience as 35mm. Dalsa perhaps coming closest in terms of picture quality, but having more to overcome in terms of convenience, cost, etc. because of the high data rates (which is why film is so impressive as a "data recorder"...)

 

This campaign to smear Panavision as somehow "lying" about the Genesis is a little too much like the current partisan politics out there these days, where you repeat some bit of misinformation as often as you can until people start picking it up and passing it along as if it were fact.

 

The Genesis is an HD camera, OK... it's attached to an HD deck, it's 1920 x 1080 -- so get over it!

[Guest Jim Murdoch]

I would think that the reason is that non-broadcast SD cameras do not necessarily have the low-pass filter (which is necessary for the correction funtion to work properly) because the resolution is so low to begin with that the apparent extra sharpness is deemed more important than absence of color aliasing artifacts, as artifacts can be controlled to an extent by the alternating RGB pattern -scheme. The non-alternating pattern in combination with a low-pass filter has the advantage of higher vertical resolving power, as mentioned earlier.

<{POST_SNAPBACK}>

This is almost exactly what John Galt said, although in a somewhat more patronizing manner.

 

Trust me, you HAVE to have the optical low-pass filter. Even cheapo colour door minder cameras have to have one. If you ever get your hands on one, take it off and see what happens. It's not an expensive part anyway.

 

The fact is, I not only have a 1980s JVC camera that uses exactly the same sort of RGB dot pattern color filter mask as the Genesis, (apart from it using the alternating RGB/BGR arrangement) but I also have the accompanying service manual which describes in loving detail exactly how the camera works, exactly what its shortcomings are, exactly why it needs an optical low-pass filter, and it even has an excellent description of how the low-pass filter actually works.

 

If that's not enough, it also has an excellent description of how a 3-chip camera works and even how a dichroic colour separation prism works!

 

They then apologize for the fact that the single-chip RGB camera is "noncompetitive in the broadcast marketplace" due to its low sensitivity and poor resolution, in fact giving only one-sixth of its "raw" pixel count in RGB mode. However the camera was only designed for taking Photo-ID images, which it does extremely well, given enough light to work with. (Remember, this predates digital still cameras by several years)

 

The thing is, I can't see how there's anything significantly different about the Genesis chip, apart from having more pixels. But Galt just blithely brushed aside any such comparisons, without ever explaining why they were irrelevant.

[Ilmari Reitmaa]

I'm a little off-topic already, so I'll be just quick.

 

What I particularly miss are descriptions of the qualities of digital cinematography systems in terms comparable to film, like modulation transfers functions and such (or actual test footage). Not necessarily that they aren't available, but that they aren't common terminology in discussing such things, and I admit my own guilt here as well. Things like sensor resolution and image resolution tell very little, and to add confusion some people use the word 'resolution' in the sense of actual resolving power and some in the sense of pixel count. Go figure. Current digital cinematography terminology doesn't yet seem established enough not to give rise to all kinds of rhetorical gimmicks.

 

That said, I do tend to agree with Mr. Jacoby on the general proportion of facts and speculation here, and further hope that you will pardon my contributions to the speculation side.

 

'Nuff said on my part.

Edited April 24, 2005 by Ilmari Reitmaa

[Guest Jim Murdoch]

[Dalsa's] claims that the camera is "4K" make sense from their logic so I stopped arguing with them about that point, but I think practically it is more like 3K.  But what really matters are the results and the comparison tests I saw were compared to 4K scans of 35mm 5212 (100T), blown-up several times over to look at small portion of the 35mm frame and the Dalsa frame to compare edge sharpness, grain (or lack of), etc....

 

.... This campaign to smear Panavision as somehow "lying" about the Genesis is a little too much like the current partisan politics out there these days, where you repeat some bit of misinformation as often as you can until people start picking it up and passing it along as if it were fact.

 

The Genesis is an HD camera, OK... it's attached to an HD deck, it's 1920 x 1080 -- so get over it!

<{POST_SNAPBACK}>

To me, "4K" means 4 thousand-odd RGB pixels across the width of the chip, that is 4,000 Red, 4,000 Green and 4,000 Blue. The Bayer pattern only really produces "Component" high definition video. If that is all that is required, explain why PV/Sony have gone to all the trouble of providing "true" RGB recording.

 

The fact is, when I started to ask these sorts of questions, Dalsa shut up like a clam, and the same thing has been reported by other people of a skeptical bent in these forums.

 

As for PV "lying" about the Genesis, the problem is that over the past decade they've run up debts of well over half a billion dollars solely on the basis of their claims that they were going to revolutionize the movie industry by developing "Digital Cinematography". All they've got to show for it is the Panavized CineAlta and the Genesis. Neither of those is likely to revolutionize anything, and certainly not to the extent that it will make even a dent in their massive bond and bank debts.

 

This may not mean squat to you, but I can assure you it means a great deal to certain people! People who pay good money for genuine technical information...

 

That said, I certainly don't dispute that cameras like the Genesis have their place. Just the other day I saw a $199 "personal data storage" thingamy which was about the size of a packet of cigarettes and boasted a 20Gigabyte Hard Drive. Just imagine a few of those drives coupled with a Genesis-type sensor fitted into a camera body that could take standard 35mm SLR lenses! You would have something about the size of an old Super-8 camera that could store a couple of hours of compressed 1920 x 1080 HD, more than good enough for a budding film maker to get something up on the big screen.

 

The trouble is, they're currently being pitched at precisely the wrong end of the market! I just wonder what kind of a deal they offered WB.

[Michael Most]

Just the other day I saw a $199 "personal data storage" thingamy which was about the size of a packet of cigarettes and boasted a 20Gigabyte Hard Drive. Just imagine a few of those drives coupled with a Genesis-type sensor fitted into a camera body that could take standard 35mm SLR lenses! You would have something about the size of an old Super-8 camera that could store a couple of hours of compressed 1920 x 1080 HD, more than good enough for a budding film maker to get something up on the big screen.

 

www.kinetta.com.

 

And it's not compressed. It's also not available yet, but that's another story.

[Mark Allen]

www.kinetta.com.

And it's not compressed. It's also not available yet, but that's another story.

<{POST_SNAPBACK}>

 

Supposedly a 16mm version will be ready by November and the 35mm version some time after that - but the 35mm will possibly be only a rental option or priced such that only renting makes sense.

 

I'm guessing that if the solution was as easy as buying a sensor and a attaching a few lenses, there would be a myriad of options available already. I think it was about six months ago that there seemed to be about 10 different people making cameras like this. As far as I know only the Drake and the Kinetta are on their way to completion. (If anyone knows of any other 'independent' cameras - let me know.) November will be an interesting month for this category of somewhat affordable HD cameras.

 

Panasonic P2 HD - $6,000

The Drake - $20,000

Kinetta 16mm version - $60,000 (last reported)

 

Note that The Drake is the only camera of these to have posted actual footage shot with the camera.

 

I have a feeling there will be a major manufacturer who may try to trump the P2 option about a month before offering an even more impressive camera available for the same price but four months later - announced 2 months prior. I wonder if they are even considering competing in CMOS camera arena.

[John Sprung]

To me, "4K" means 4 thousand-odd RGB pixels across the width of the chip, that is 4,000 Red, 4,000 Green and 4,000 Blue.

<{POST_SNAPBACK}>

The problem is that the word "pixel" is used in two different ways.

 

What you describe is what I also used to think of as a pixel. Now it has to be distinguished as a "post production pixel': all of the color and brightness data for one element of the x-y array that makes up an image.

 

The other version is what I'd call a "camera pixel". That's basically just a photosite on a chip. It may be under a red, green, or blue filter, it may be only one sub pixel out of four on one chip in the Viper. The underlying principle is that camera pixels are defined in whatever way will produce the highest number. It's just like shop vac horsepower vs. continuous industrial motor horsepower.

 

Now what we need to find out from Kodak is how many silver halide grains there are in a 1000 foot roll of 5245, and report that as the pixel count of film. ;-)

 

 

 

-- J.S.

[John Sprung]

Their claims that the camera is "4K" make sense from their logic so I stopped arguing with them about that point, but I think practically it is more like 3K. 

<{POST_SNAPBACK}>

Dividing 4046 by the square root of two, as the Arri presenter suggests, gives you 2861. That lines up very nicely with your assessment that it looks about like 3k.

 

 

 

-- J.S.

[Guest fstop]

Ladies and Gentlemen:

 

THE BOMBSHELL