Jump to content

David Fincher's The Social Network


Recommended Posts

I just had the pleasure of watching Zodiac on Blu Ray and marvelled at all the beautiful imagery from the Viper. I'm disappointed to see, via imdb.com, the his next picture was captured on the Red. Anybody here know any details about why the switch, or his workflow?

Link to comment
Share on other sites

  • Premium Member
I just had the pleasure of watching Zodiac on Blu Ray and marvelled at all the beautiful imagery from the Viper. I'm disappointed to see, via imdb.com, the his next picture was captured on the Red. Anybody here know any details about why the switch, or his workflow?

 

He's using the new M-X sensor in the Red, which is going to give him much cleaner results than the Viper, which he often underexposed to shoot in low-light and spent millions doing an expensive noise reduction pass at John Lowry's company. He's always been interested in working at very low light levels and he's using the M-X Red at 2000 ASA, which would have been difficult to do with the Viper without a lot of noise reduction in post.

 

He also complained that Thomson was not interested in upgrading the Viper -- he had problems with fan noise, for example, and because of that, had to switch to the F23 for the quiet hospital dialogue scenes in "Benjamin Button" near the end of production. I think that experience probably soured him on using the Viper further.

 

He's always been an advocate of file-based digital cameras, using the Viper tied to a S-Two recorder, so the Red fits into that way of thinking. And now he doesn't have to be tied into a large data recorder.

 

There are a lot of nice things about the Viper image, especially in terms of dynamic range, but I think the new M-X sensor and new color management software has greatly improved the Red One in that regard -- and obviously Fincher must feel the same way.

Link to comment
Share on other sites

  • Premium Member

What I don't quite get is that Red has really not made Viper or F23 that much cheaper, which is at least one positive side-effect you'd hope to see.

 

Is the new Red sensor actually released and available to rent? Becuase if it isn't, it really hasn't done anything yet.

 

P

Link to comment
Share on other sites

  • Premium Member

The upgrade process has been going on now for a month or so, in camera number order, so some private owners and rental houses have gotten their M-X Red's back already. So it's a bit of a search but every week, more and more become available to rent.

 

Yes, you'd think that a Viper would be cheaper to rent, though there is no reason for the recorder (SRW1 or whatever you choose) to have dropped much in price. F23's are in constant demand though so I don't see a drop in those prices for awhile.

 

Unlike Fincher, a lot of TV producers are uncomfortable outside of an HD tape origination and workflow, hence why these HDCAM-SR cameras are in high demand for television work. I don't see that changing for a few more years, not until enough post houses are out there promoting their Red workflow capabilities to producers and showing them how affordable, reliable, and painless it is. Trouble with having just a few post houses that are "experts" in Red footage is that producers like to be able to take their footage anywhere for the cheapest deal, which is why they like HD tape so far. I still run into many shows shooting on F900's.

Link to comment
Share on other sites

The upgrade process has been going on now for a month or so, in camera number order, so some private owners and rental houses have gotten their M-X Red's back already. So it's a bit of a search but every week, more and more become available to rent.

 

Yes, you'd think that a Viper would be cheaper to rent, though there is no reason for the recorder (SRW1 or whatever you choose) to have dropped much in price. F23's are in constant demand though so I don't see a drop in those prices for awhile.

 

Unlike Fincher, a lot of TV producers are uncomfortable outside of an HD tape origination and workflow, hence why these HDCAM-SR cameras are in high demand for television work. I don't see that changing for a few more years, not until enough post houses are out there promoting their Red workflow capabilities to producers and showing them how affordable, reliable, and painless it is. Trouble with having just a few post houses that are "experts" in Red footage is that producers like to be able to take their footage anywhere for the cheapest deal, which is why they like HD tape so far. I still run into many shows shooting on F900's.

 

In Toronto, I received a quote on a Viper that dipped by 25-30%, but it still doesn't go out for a price comparable to what I can get a Red package for.

 

I wonder if Fincher used a digital recorder, since the data rate of the .r3d's is probably nowhere near that of the Viper's data. Or does the M-X up the data-rate and 'raw-ness' of the .r3d's?

Link to comment
Share on other sites

  • Premium Member
In Toronto, I received a quote on a Viper that dipped by 25-30%, but it still doesn't go out for a price comparable to what I can get a Red package for.

 

I wonder if Fincher used a digital recorder, since the data rate of the .r3d's is probably nowhere near that of the Viper's data. Or does the M-X up the data-rate and 'raw-ness' of the .r3d's?

 

I don't recall if anything above RedCode 36 is available yet (i.e. anything with less compression.) I heard rumors about a RedCode 42.

 

I'm not really savvy on data rates, but isn't 4K RAW RedCode 36 mean 36MB/sec or 288 Mbit/sec? HDCAM-SR normally is set to record at 440Mbit/sec, so that is a bit more but on the other hand, it's having to record RGB, not a monochrome RAW signal.

 

Anyway, I doubt that Fincher is recording to anything but Red's recording products (CF, RedDrive HDD, RedRam SSD).

 

I see now that Redcode 42 is available, that would be 336 Mbit/sec. Don't know what Fincher is using.

Link to comment
Share on other sites

  • Premium Member
What I don't quite get is that Red has really not made Viper or F23 that much cheaper, which is at least one positive side-effect you'd hope to see.

Is the new Red sensor actually released and available to rent? Becuase if it isn't, it really hasn't done anything yet.

 

So far this pilot season, we have two on Red MX, two on F-35, one on D-21, and four that I don't know yet. The current numbers are RedCode 42 and Build 30.

 

The Viper, alas, was ahead of its time. It's a discontinued product now, IIRC.

 

 

 

 

-- J.S.

Link to comment
Share on other sites

I don't recall if anything above RedCode 36 is available yet (i.e. anything with less compression.) I heard rumors about a RedCode 42.

 

I'm not really savvy on data rates, but isn't 4K RAW RedCode 36 mean 36MB/sec or 288 Mbit/sec? HDCAM-SR normally is set to record at 440Mbit/sec, so that is a bit more but on the other hand, it's having to record RGB, not a monochrome RAW signal.

 

Anyway, I doubt that Fincher is recording to anything but Red's recording products (CF, RedDrive HDD, RedRam SSD).

 

I see now that Redcode 42 is available, that would be 336 Mbit/sec. Don't know what Fincher is using.

 

"36 MB/sec or 288Mbit/sec". I feel like a moron for not knowing the difference MB vs Mbit... ?

Link to comment
Share on other sites

  • Premium Member
"36 MB/sec or 288Mbit/sec". I feel like a moron for not knowing the difference MB vs Mbit... ?

 

MegaBytes vs. MegaBits

Typically 8 Bits = 1 Byte

http://en.wikipedia.org/wiki/Bytes

 

When it comes to computers, I don't feel like a moron... I am a moron. But I'm good at looking things up, my only saving grace.

Link to comment
Share on other sites

  • Premium Member
I'm not really savvy on data rates, but isn't 4K RAW RedCode 36 mean 36MB/sec or 288 Mbit/sec?

 

Yes, although it doesn't really compare to an uncompressed recording of a dual-link HD-SDI feed, at about 1440Mb.

 

Actually, this is an interesting numeric comparison. A 4K image, for instance, would ideally be four times the data of a Viper, so let's say the best part of six thousand megabits per second. Red are recording under a couple of hundred. These numbers are why flinty-hearted cynics such as myself tend to raise a hand and say "er, just a minute, here..."

 

P

Link to comment
Share on other sites

Guest Stephen Murphy

:P

David Fincher and Emmanuel Lubezki on RED.

 

Maybe MX?

 

Pretty sure that was the M sensor - it was shot about a year ago before the MX sensor was testing.

Link to comment
Share on other sites

  • Premium Member
Yes, although it doesn't really compare to an uncompressed recording of a dual-link HD-SDI feed, at about 1440Mb.

 

Actually, this is an interesting numeric comparison. A 4K image, for instance, would ideally be four times the data of a Viper, so let's say the best part of six thousand megabits per second. Red are recording under a couple of hundred. These numbers are why flinty-hearted cynics such as myself tend to raise a hand and say "er, just a minute, here..."

 

P

 

Well, you have to remember that a bayer-filtered sensor produces a monochrome signal, so 4K RAW would naturally be one-third the data of 4K RGB. That's not compression, it's just the nature of a single-sensor design. So you'd take that 4X figure of 4K over the 3-CCD output of an HD camera and divide by 1/3.

 

Or to put it another way, three HD sensors, 2MP each, is 6MP of data to record, versus roughly 10MP of a 4K RAW sensor. So yes, the 4K sensor camera should still be more data to record, but it's not 4X the data due to the natural "compression" of a single-sensor with a color filter array.

 

If the difference between 6MP and 10MP is about 1.6X, and dual-link HD-SDI is 1440 Mb/sec, then uncompressed 4K RAW should be about 2304 Mb/sec, or 288 MB/sec. That's a total guess though, I'm sure someone has the real figure out there.

Link to comment
Share on other sites

Well, you have to remember that a bayer-filtered sensor produces a monochrome signal, so 4K RAW would naturally be one-third the data of 4K RGB. That's not compression, it's just the nature of a single-sensor design. So you'd take that 4X figure of 4K over the 3-CCD output of an HD camera and divide by 1/3.

 

Or to put it another way, three HD sensors, 2MP each, is 6MP of data to record, versus roughly 10MP of a 4K RAW sensor. So yes, the 4K sensor camera should still be more data to record, but it's not 4X the data due to the natural "compression" of a single-sensor with a color filter array.

 

If the difference between 6MP and 10MP is about 1.6X, and dual-link HD-SDI is 1440 Mb/sec, then uncompressed 4K RAW should be about 2304 Mb/sec, or 288 MB/sec. That's a total guess though, I'm sure someone has the real figure out there.

 

David, Phil. Thanks guys, I've learned quite a bit from this thread.

Link to comment
Share on other sites

  • Premium Member
4K RAW would naturally be one-third the data of 4K RGB

 

Well, isn't that the entire point at issue?

 

Does that 10MP bayer image contain as much information as a hypothetical 4K cosited RGB image? No, it doesn't, and nor can it ever.

 

P

Link to comment
Share on other sites

  • Premium Member
Well, isn't that the entire point at issue?

 

Does that 10MP bayer image contain as much information as a hypothetical 4K cosited RGB image? No, it doesn't, and nor can it ever.

 

P

 

Yes, but isn't that being a bit unrealistic and impractical? No one is ever going to build a 4K 3-sensor 35mm camera, it would be huge -- not to mention, to avoid aliasing at 4K measurable resolution, it would actually have to have more than 4K pixels per sensor. It would actually have to be a 4.5K 3-sensor camera. Plus a prism block would probably be involved. So it's a bit academic, this "ideal" 4K RGB camera.

 

The only reason why the term "4K" is thrown around so enthusiastically is that it's considered some digital benchmark to match 35mm, which can be scanned at 4K per color channel. But that doesn't mean every color layer in film stock resolves 4K even if that much data is assigned to record it.

 

You mention advertising overall pixel count instead, but the trouble with that is that aspect ratio affects pixel count, and cameras that combine photosites for the final image, like the Viper or the F35/Genesis does, actually have a very high number of photosites on the sensor, yet by their nature, are only designed to deliver a 1080P image.

 

I don't think there is any real-world value in "discovering" what a "true" 4K RGB digital image would look like, not in terms of filmmaking in general. Which is why I think it's a bit academic. May be fun to take a monochrome 4K camera and take three images in a row through filters to create "true" 4K RGB, but even that would be a bit inaccurate since you couldn't have any movement in the frame.

 

You're basically saying that until someone makes a 30MP RAW uncompressed camera, that aren't allowed to call it 4K.

 

You notice that nobody gets up in arms over the use of the term 2K... like the SI2K camera, it has a single 2K Bayer sensor, so does the Phantom, and none of them can deliver the same amount of data as a 2K RGB scan would.

Link to comment
Share on other sites

David Fincher and Emmanuel Lubezki on RED.

 

Maybe MX?

 

Wait! That commercial was shot on the red? I had no idea and it's just about my favourite commercial ever. Fincher, Lubezki and Morricone together!

 

Also, love the concept of the one guy always moving left and the other right until they collide - it's pretty much perfect, and I don't even like football.

Link to comment
Share on other sites

  • Premium Member
You notice that nobody gets up in arms over the use of the term 2K

 

I think I'm on record as having done exactly that.

 

It was the first thing I said about Kinetta, when I was told what the resolution was - or rather what the resolution they were going to claim in advertising was, and compared it to what they were talking about using to record it. At which point my deep and lasting respect for mathematics kicked in and said "well no, because you can't actually do that".

 

The difference, as I've said many times before, is that Dalsa, when confronted with this unfortunate reality, said "well, OK, yes, but we still think it's a great camera", whereas Red's reaction is to whine and make excuses and take offense. I think getting pissed off that someone caught you in a - let's be nice - elasticisation of the truth is just about the most wretched, pathetic, and frankly just hopelessly childish thing anyone can possibly do, and suggests a complete lack of confidence. This sort of thing offends me in exactly the same way that the Catholic church offends me for having nearly tortured Galileo to death for pointing out that the earth orbits the sun, and then having taken three hundred years to apologise for their mistake.

 

I really don't know how many more ways there are to say the same thing, but at the end of the day there should be no dishonor in this. Red haven't made a 4K camera with a 4000-pixel bayer chip, because it is fundamentally impossible to do so, but then again that does mean nobody else will ever do it either.

 

P

Link to comment
Share on other sites

At 2K, a 35mm frame is generously oversampled. Each pixel has discrete information. Comparing the 2K scan to a "2K" Bayer image, you should find that the film is slightly noisier (grainier) overall, but that the digital image is not quite as sharp or pleasing.

 

4K is another story, because you're starting to be grain-limited in the scan. Depending on the camera you're comparing to film, you may find that the red channel is clearer and sharper in the scan, while the blue channel is much nicer in the digital image.

 

Ultimately, at the pixel level, it will come down to taste. Bayer interpolation means you're dealing with a characteristic set of artifacts, just as with film you're dealing with grain. Having implemented a few de-bayer algorithms, I'm very sensitive to the Bayer "look," which is simultaneously edge-enhanced and interpolated. Perhaps a result, I find film to look much cleaner despite the grain.

 

To wheel back around on-topic, as a huge Fincher fan I'm not yet convinced by his switch to digital. To my eye, the low light scenes in The Game, Fight Club, Panic Room, etc, have a more dynamic look than the low light stuff in Zodiac or Benjamin Button. I can't put my finger on why. The digital stuff just feels "flat" or monochromatic.

Link to comment
Share on other sites

  • Premium Member
No one is ever going to build a 4K 3-sensor 35mm camera, it would be huge --

 

The Lockheed-Martin "Blue Herring" camera of the 1990's was a three chip prism block design, and yes, it was huge. IIRC, the chips were about IMAX size. The prototype that I saw was smaller than a mini-cooper, but not by a lot. There doesn't seem to be much on the web about it.

 

http://www.allbusiness.com/services/motion.../4860211-1.html

 

 

 

 

-- J.S.

Link to comment
Share on other sites

  • Premium Member

I'm not sure a 4K 3-chipper would have to be that big; remember, Viper's sensors (which were noisy as hell, to be fair) use four subpixels per pixel.

 

I'm not necessarily proposing a prism block as the be-all of this, though, just a really big bayer chip that's demosaiced in camera and recorded as RGB. I think once you get Bayer chips beyond 6-7K wide, the resolution argument comparative to 35 becomes largely moot, although specific implementations will have to be judged on their merits. Then, if we haven't got the noise down low enough, we can start talking about prism blocks to increase dynamic range, though I suspect that the most that'll ever get you is a couple of stops unless there's fundamental advances in sensitivity.

 

Canon's current DSLR range may represent a first step toward that advance, because, good grief, they're quiet as the grave.

 

P

Link to comment
Share on other sites

To justify the effort for 4k resolution in post you need a source that is capable of "filling" these large amounts of data with "useful" information (few artifacts, high contrast beyond Nyquist-limit for 2k), otherwise it's better to stick with 2k and it's advantages in the first place.

 

To achieve high-quality 4k data without adding unnecessary artifacts (especially alaising or at least moire) you have basically three different options:

 

1. 35mm-sized sensors with half the pixel-pitch of "2k"-cameras - the area of an otherwise unchanged photosite is therefore quartered, reducing sensitivity and DR by about 2 stops! And we just start to get decent DR with large photosites...

 

2. Large sensors - beyond the obvious reasons (cost, size, lenses, DoF) the production process of sensors becomes the limit. I think 20x30mm is the max. size given by the production equipment (and even Canon with their own equipment don't change that for their 24x36mm-sensors), otherwise you'll have to stitch multiple sensors - not a good idea for professional cinematography.

 

3. Three sensors - even if the added weight and cost (+new lenses) is not an issue, color rendition/accuracy (even more so than with regular filters for bayer sensors) is. But the beam-splitters are more efficient (you'll save about one stop in comparison to color filters) and you have three times the effective photosite-area.

 

I don't think option 3 is very likely, industry much rather waits till sensor-technology improves and option 1 becomes viable - but certain companies much rather scream 4k/5k/6k to sell new cameras instead of working in favor of the needs of professional film production...

 

But why think so complicated? What about using a 3k-sensor with large photosites, making 4 exposures with slightly changed positions to get 6k, using "HDR" by making two exposures to create a true 16bit-file and instead of using inaccurate (it's always a trade-off between transparancy and accuracy) color filters use LEDs. So you'll merge 24 (!) digital images in one pristine 16bit oversampled 4k-image - hmm, that would make sense, I think it exists, I think it even got an Academy Award (not given by actors and costume designers but cinematographers/engineers) but I forgot it's name... :P What about using it to make new digital masters from Alien3-Panic Room and show them to Mr. Fincher? Does he really prefer the output of Zodiac/Button?

You can even use the most robust, reliable and cheapest cameras, "the look" is not so much dictated by the camera but the artist and if you want to, you can even put a little 1080p-single-sensor camera into the viewfinder if you want "HD" on set and if you don't like grain, you can make it just as unnoticeable as digital noise. Isn't 2010-technology great? Why stick with cameras never designed for the big screen and limit your artistic skills (Mr. Fincher is a better artist than I will ever be) by technology? Oh well, he cannot afford the film stock in these rough times.... :blink:

 

I think digital cinema cameras will have their time but right now, we should mix technologies to achieve the best possible quality - am I the only one who thinks it's strange that Mr Finchers 1992-2002-films have a higher IQ than his work from 2005/2007? Isn't that the wrong direction?

Link to comment
Share on other sites

I'm sorry that I came up with that again, but I was just looking for some technical information on the ARRI Alexa (because I wish for some technology transfer into the photographic world) and several hundred results within the last week came up stating: "...will FINALLY kill film..."/"...end of celluloid..."... I look at the movies shot on film and I look on the recent work done with digital and wonder: why!? To cut the budget by 0,1%-0,5% by saving film stock?

I think we should use/support the best technology and not the one which helps the sharholders of film studios and camera manufacturers by selling a new cam every 2 years...

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...