georg lamshöft

I finally saw it in Berlin in 15/70 - they got a print after two months... It was a little bit softer throughout (in comparison to TDK-15/70-scenes) and I couldn't tell the large aquisition formats as well. But otherwise it was pretty impressive! The detail was astounding, it really shows what even 35mm is capable of! Lightyears ahead of "the social network" in digital cinema (despite a 4 times smaller screen!). I don't think the 35mm-sequences of TDK had the same quality in IMAX - did they change the IMAX DMR-process (different scanner, new grain-reduction?). It was nearly absolutely grainless at 28m wide sitting in the front 1/3 of the theatre!

As an engineer I was a little bit disappointed not to see more about component manufacturing (e.g. machining and in-house casting processes!) and more about assembly - but as said it's not an production manual. I think it becomes clear that this is not a simple product that cannot be manufactured by any sweatshop.

Thank you very much! It's nice to see a German involved in good looking cinema. They had to shoot a Michael Caine-Movie on a F35 (who much does a working set cost? 3000€/day?) for budget-resons? Congrats to "the American" - interesting film, very classy look and style and yes, it is clearly technical superior to "the social network" - why does Fincher compromise his work for no reason?

Depth of field is a physical effect which cannot be changed by lens design. Aspherical elements are just a technology to widen possibilities in optical design (smaller, better correction...). Although sharper lenses show a stronger subjective "drop" in sharpness from in-focus to out-of-focus areas.

www.fdtimes.com They have a quite comprehensive article about ALEXA.

SEPARATE, sorry...

I've just stumbled over the article on provideocoalition (was it already posted?). So it is regular HDR - two SEPERATE exposures, one normal and one for highlight recovery. Great, they have trouble with motion rendering already (like all shutter-less cameras) and now they put two exposures taken at different times over each other just to claim superior DR?

They could use log - just like the ALEXA in ProRes but I thought the question was regarding linear TIFF 16bit? And it still doesn't resolve the "mystery" of doing HDR in motion at all.

With marketing...

Alexa would have been called D-22 with on-board-uncompressed-recording and an optical viewfinder and would cost as much as the D-21, no arridigital-message-board and no shiny brochures - that's what changed. Let's see if that's a good thing. They're working since 10 years (!!!) on digital cine-style-cameras! Back on-topic: So the final dynamic range is limited to the range captured within the sensor, period. Here's the technical data for a quite new interline-CCD used in the Ikonoskop or other single-chip-2/3"-HD-solutions: http://www.kodak.com/global/plugins/acrobat/en/business/ISS/productsummary/Interline/KAI-02150ProductSummary.pdf On one hand you have the read noise of 12 electrons - that's some kind of noise floor - everything that's below 12 electrons cannot be seen due to the internal processing noise within the sensor. On the other hand you have the charge capacity of 20000 electrons - the sensels are like buckets collecting photons - once they're full, they clip, no highlight information whatsoever. 20000 electrons capacity-limit are 1666 times "brighter" than the 12 electrons from the read noise - it results in less than 11 stops (that would be a 2048 times ratio) of DR. This is a new sensor-design with a pixel-pitch similar to RED (5.5µm instead of 5.4µm), the charge capacity is directly related to the size/area of the photosites, like small buckets can contain less water/photons. They try to increase the fill-rate (the amount of light-sensitive area) but they're already beyond 70% in modern CMOS-designs (so less than twice the area/charge capacity is possible). Given a similar fill-rate, ALEXAs photosites are 2.3x times larger than RED - that's the trade-off between number of photosites on a given sensor size and "per-pixel"-quality! But that's the theory, in reality much more effects lower the usable DR (electric interferences, even the temperature of the sensor affecting the noise floor) in the processing chain. One of them are the amplifiers and the ADCs. That's where the dual-gain-architecture steps in. By merging two outputs with different gains (and separate ADCs) they can increase the amount of usable DR within the theoretical DR (charge capacity/read noise) of the sensor itself! NEVER more! Not in one single exposure (crucial for video of course). But this is really basic understanding, what happens in detail is the secret of a few state-of-the-art sensor-designers and manufacturers.

The Ultra Primes and Superspeeds are decades apart! They don't share any similarities with the Superspeeds, neither in mechanical or optical design. The Ultra Primes are modern lenses, they are sharp and contrasty even wide open, while the Superspeeds show typical aberrations of old, fast lens designs (softer, less contrast, coma...). Even stopped down to T1.9 they will be softer and less contrasty than the Ultra Primes at the same (open) aperture. I cannot tell you much about the differences in handling, but I guess the Ultra Primes are the overall superior choice. The Superspeeds are better if you want the distinctive "look" - but not as an universal lens.

I just found an article about a company from Nürnberg named Solectrix who designed electronics for the HD-IVS and the Alexa. http://www.eue24.net/pi/index.php?StoryID=253&articleID=176131 The interesting part: They currently develop an on-board-flash-recorder (most likely to replace the DTE card module) for ARRI for uncompressed data up to 5.6Gbit/s (700MB/s)!

"Up in the air" was great, "Juno" a masterpiece - I`m looking forward to your collaboration with Mr. Reitman any time, even if you would use an Iphone... I`m impressed with the possibilities of ALEXA but have you talked to the guys from ARRI? As far as I know, the engineers (marketing is a different story) still see it as an alternative to other digital systems and S16 while the "big screen" is still 35mm + 4k+/ARRISCAN territory for them. Didn`t you prefer Panavision anyway? As long as you don`t want to use extreme low-light work (1000ASA+) I would still prefer 35mm, don`t you have all the possibilities with such a project? Maybe even more exotic formats you wouldn`t have the chance to work with in other projects? Maybe it`s just my naive outsider perspective, I`m sorry... :rolleyes: ARRIRAW is the only possibility to capture all the sensor information and propably isn`t too different from well-made internal processing (I`m sure the guys in Munich know what they`re doing) - as long as you don`t push it. What`s happening in a critical situation where lighting/exposure wasn`t right and you want to heavily manipulate the look in post? I guess this makes the difference, Isn`t it nearly ten times the data rate? Anyway it`s your work, you`re the cinematographer, you gonna decide and I`m looking forward to it!

Film isn't obsolete IMHO, because it's still technical superior for cinematography as an aquisition medium. Even when superior digital technology might be available within the next decade, film will still offer a distinctive "look" - that's why I agree that Kodachrome despite it's technical inferiority compared to modern stocks should have been kept alive. I also agree on the huge management mistakes that destroyed most of the once impeccable Kodak-company and destroyed know-how, technology and high-quality jobs. Building 38 is doing a great job - don't let them suffer for the mistakes of some stupid managers. But at least theoretical (we don't know yet how well Kodak executed this technology) Laser projection is THE projection technology, way superior to LCOS, LCD, DLP, 2k, 4k, 35mm, 65mm, IMAX - it's capable of impeccable resolution and the laser light sources can create an unique colour spectrum. When something should replace film projectors in our beloved cinemas that offers an unique experience (something that HDTV can't) than it's laser projection - from the technical perspective. @Karl Borowski Bravo!

I've been curious as well and the story seems interesting. But the cinematography? Honestly, I would expect more from Mr. Aronofksy & Libatique than shacky "un-cinematic" (some of the early facial shots in the trailer look very "TVish" to me) 16mm. I'm not a big fan of the "rough & dirty"-look but as long that's exactly what they wanted and not some stupid producer/budget-thing ("what? 35mm? in this economic crisis!?") I'll have to accept their vision, I guess ;-)

It was quite obvious to me: "Normal" scenes were shot in a very natural way, available-light even under boring lighting conditions to give the dreams (those created for the victim for example) a "natural" feeling - only special scenes (like the 4th layer or the train-suicide) stood out.

Don't watch it in 70mm! I've got a response from our IMAX theatre in Berlin (they currently show Toy Story 3 3D on 2x70mm because it had the same starting date as Inception) and they still try to get an 70mm-copy from the States, so when too many people in the US watch it on 70mm, we won't get our copy! ;-) :-( I've seen it on a decent 35mm-projection and the different formats were indistinguishable (aside anamorphic effects). The filt itself was good, maybe even a masterpiece, but I'll have to watch it again - I overlooked to many details the first time. "One weird thing I did notice was that the VistaVision helicopter shots seemed soft with more visible grain then the 35mm anamorphic shots" Did they use some kind of optical process to copy the 8perf-shots within the anamorphic 35mm-print?

I'll watch it tomorrow, so propably it's a bit early to judge, but let's not forget that Nolan's Batman-movies and "Inception" share a similiar "atmosphere". They're dark, near/alternative-future visions - while Prestige was playing in a different time, hence was allowed an entirely different look!?

But a 6k DI of the 65mm source material blown-up to IMAX should still hold plenty of information and barely any noticable grain - I'll see it next week, let's hope our print in Berlin looks better.

I read somewhere thtat they used the Phantom for 1-2 scenes left in the film, the other high-speed stuff was partly shot on a 1000fps 65mm photosonic! Batman 3? I fear he limits his writing/storytelling-abilities by sticking to comic-adaptions. Dark Knight was great - by comic-standards but it lacked the original story and storytelling of his original movies (Memento/Prestige). But as long it's 65mm...

The Ikonoskop A-cam Dll uses a modern 2/3"-Kodak-CCD and records UNCOMPRESSED (unlike RED) RAW 1080p. It should be available soon. I don't know if it's up to Hasselblad-like Swedish-craftmanship but it's made in Sweden and not Foxconn...

"... "Flesh coloured" to mean "Brown" ..." Somebody saw the new Fincher-Trailer? ;-) Remember when Fincher was once visually/ technical top-notch, just like Mann, or Coppola :( ... But we have the new young guys now :lol: , like Nolan - I'm eager to see "Inception" on 70mm 15perf WITHOUT brownish skin or clipped highlights and forget these digital cinema discussions (I admit, I'm pushing them) at least for 148min... Who's with me? B)

@Thomas James Looks like a serious difference to me (2k vs. 4k scan from S35 source: ARRI), but you're right - every medium becomes more "inefficient" at higher resolutions because MTF goes down and artifacts (interpolation, grain...) become more obvious - that's why 4k 65mm will look way superior to 4k 35mm - although both are capable of 4k actual resolution. But I think the "4k" derived from compressed bayer-sensor-data hits a limit - you will loose information by downsampling to 2k, but it's very unlikely that the trade-off in actual photosite-size (and therefore DR - something that doesn't exist in the film-world, a 4k-scan doesn't cost you DR) is worth it in comparison to a 3k-sensor. @John Sprung I don't think oversampling saves you the OLPF, ARRI still uses one, it just compensates for the loss in MTF at high frequencies (when the OLPF sets in to reach 0% at Nyquist). It's a strange variety of concepts - ARRI uses oversampling but doesn't fully compensate color interpolation and Panavision goes the opposite route by not using oversampling but compensating color interpolation. What surprises me is their very ambitous new sensor design with 37MP on a S35-sizes sensor with 2,9µm pixel-pitch! I wonder how they plan to handle DR and noise or if they plan to make it the users-choice (hardware-binning - 4k with high noise, low DR or 2k with low noise and high DR) @Keith Walters What is the difference between Genesis and F35 anyway? Different ergonomics?, different firmware?, one can be bought, the other one has to be rented - but it's basically the same camera, isn't it? I'm curious - the Genesis is practically unknown in Germany. By the way, I think Mr. Tattersall does a fabulous job on "House" - I think the 35mm-episodes are a proof that even on the small screen, the "cinematic look" pays off - it stands out in comparison to Genesis/D21/F950-material when switching channels (I think that's really important in this market). Maybe the 5D was the right choice for this specific episode, but it was nowhere near 35mm - even my parents noticed that! I remember I was blown away by how great the first minute of "Castle" looked - until I've read it's 35mm... "Good enough" is not "good enough" in my eyes and saving a few thousand dollars (doesn't the F35+extras cost a few thousand dollars per day?) on high-profile TV-series over the head of the DP/producer/director is hideous.

Germans still have to wait two more weeks, I'm über-jealous! According to this source: http://www.hollywoodnews.com/2010/07/13/cinematographer-wally-pfister-gives-an-inside-look-at-inception/ about half of the film was made traditionally 35mm anamorphic for handheld-work and the other half either 5perf 65mm (6k DI) or Vistavision (aerial shots).

These urban myths in digital cinematography are hard to destroy... So I'll try it again: Typical test charts are problematic because they operate under unrealistic conditions (high-contrast, black/white, alignment along photosites) and easily "pretend" high resolution-numbers (2k, 2.4k, 2.8k, 3.2k...28k!) http://provideocoalition.com/images/uploads/ResDetail.jpg That's a sharpened RED-sample @"4k" wih an excellent lens and not strictly horizontal/vertical patterns - the only one I've found (hopefully it's properly done) and it barely reaches 2.5k ("1250" on the chart) without alaising (just because you see lines, doesn't mean they're actual detail). But this is only half the picture, actual resolution-numbers (color, lower contrast, less obtrusive sharpening) should be even lower - especially compared to images not derived from bayer-filtering, compression or OLPF (film for example). But that's still just resolution, what actually matters in the resolution/sharpness-debate for our perception is MTF at a given frequency - resolving a certain amount of detail at a given contrast and even regular testing equipment can be "tricked" by adding excessive sharpening) . That's the basic theory. Actually defining what pixel count you want is not a matter of production technology or development skill in the sensor department (sensor cost is mostly R&D and some size-related production-costs and has nearly nothing to do with pixel count) - it's a trade-off, a compromise: Higher pixel-count (with lower pixel-pitch) results in higher MTF (also because of weaker OLPF) but lowers (given similar technological standards in sensor-design - ALEXA has >70% fill-rate, that's already top-notch for CMOS - I don't expect more from RED) DR (saturation - read noise) and increases noise (lower photon count due to smaller photosite-area). A specific problem in cinematography is the fixed output size (1080 or 2k or 4k - but not 3k or 5k) and the data-rate. 3k (16:9 aspect ratio -EPIC is a bit wider) @16bit and 24fps is 2Gbit/s, 5k is 5.6GBit/s - impossible to manage with todays technology. ALEXA has a pixel-pitch of 8.25µm (= 68µm²) which results in 3072 actual pixels @ 25,3mm sensor wideness (the sensor itself is nearly 29mm wide). EPIC has a pixel-pitch of 5.4µm (= 29µm²) which results in 5120 actual pixels @ 27.6mm sensor wideness (the sensor itself is ~30mm wide). ALEXA as a real 4k-camera (although even 1.5x oversampling is not perfect) would need to have a 6k-sensor and RAW-data - theoretically that would result in a 2 stops loss in DR and noise (although the higher pixel counts compensates for that to a certain degree) with enormous data-rates - they decided against it for a good reason, just like Panavision generated 1080p from a 12MP-sensor and not 4k like RED. Judging from RED's 4k-image, a 4k-image downsampled from native 5k EPIC-image (compressed as well) will only gain very little MTF (and complete suppression of alaising in comparison to a downsampled 2k-image - hard to justify given a four times higher data rate and high costs. So they're basically both cameras for 1080/2k-output but because ARRI never intended to create 4k+ (that's what 35mm is for), they could use more than twice as large photosites (given similar fill-rate) on the sensor! They can record uncompressed RAW (is 2k not ready yet, just ARRIRAW with 2880 pixels?). That's what it's about - actual IQ (not just resolution, but MTF, color, rendition, DR, lack of artifacts...) in a 2k workflow, reliability and usability in everyday work - not 1080p, 2k, 4k, 5k or 28k or other marketing claims. When we don't appreciate these sometimes difficult to communicate design-decisions, the digital cinematography-market will become a shareholder-value and marketing oriented business! More gimmicks, lower standards, bugs and less real innovation. Worse for everyone - just like NASA has to use cheap foxconn-made notebooks which have be thrown away every few weeks, because the market lacks high-end alternatives. Epic is made in the US? So they have trained (3+ years) technicians for machining and assembly like ARRI now? That would be really surprising in a good way - or is it just the prototypes? ALL previous products were not even assembled in-house, not even in the US. Don't get me wrong, major economic value is supplied by international specialists in all digital cameras (sensors made in US or Israel-fabs with German/Dutch technology, for example) - but the majority of the ALEXA is not made with slave-labour.