8K Weapon: The New Super Sensor

[Björn Benckert]

What add a lot of this smoothness Michael describe is the sensors size relative to the object shot. Resolution also play a part but the sensor size in relation to whats in front of the camera and the distance to it is what creates the big difference. Same goes for DSLR´s vs iphone Facebook. The downsampling on the other hand very much neutralize noise. But no matter how much noise you ad to a large sensor picture it will still look smoother then a Iphone pick in a lot of ways.

 

[Michael Hastings]

I'm not questioning the value of 8K or oversampling, but wondering about this idea of smoothness when down sampled.

Seems to me (as Michael pointed out) the roundness of film is the random distribution of both orientation and size of the grains on the film, while the "unroundness" of digital is the perfectly rectangular and evenly spaced grid pattern of the pixels and even when you downsample to 4K the grid pattern is still rectangular and evenly spaced so not much different than if it was down sampled to 4K from say a 5K source (so you don't lose actual resolution to the debayering process.)

Bjorn's explanation below makes more sense to me intuitively for the source of the "roundness".

 

[Marcus Friedlander]

Thank you for this write up! Will share with my professional network ASAP!

 

[Andreas Mendritzki]

Great post Michael

But the first question I'm left with is:

Is Panavision going to SELL that Weapon module!?!

 

[Martin Stevens]

Awesome post Michael.

Thanks.

 

[Brett Harrison]

Not a mathematician - but I think the idea is that once you pass that 1:1 demosaiced pixel : screen resolution ('superresolution') you begin to approach (or resemble?) the benefits of randomness of the halide crystal sampling. I understand that film can resolve a fair way past HD, too, so it's also a supersampler with respect to HD.

I definitely recognize that 'rough' texture of lower MP cameras.

I'm not questioning the value of 8K or oversampling, but wondering about this idea of smoothness when down sampled.

Seems to me (as Michael pointed out) the roundness of film is the random distribution of both orientation and size of the grains on the film, while the "unroundness" of digital is the perfectly rectangular and evenly spaced grid pattern of the pixels and even when you downsample to 4K the grid pattern is still rectangular and evenly spaced so not much different than if it was down sampled to 4K from say a 5K source (so you don't lose actual resolution to the debayering process.)

Bjorn's explanation below makes more sense to me intuitively for the source of the "roundness".

 

[mike zorger]

Michael, you killed it. What lights did you use?

 

[Mark Toia]

There are already a number of people chatting about how unnecessary 8K is and that the "race for resolution" is a pointless contest. Let me explain why this way of thinking is flawed:

One of very few people on this forum that is miles ahead of the curve.

Everything you say and have said over the years Michael in regard to camera tech, editing, audio and colour grading programs etc etc resonates so deeply with my own thinking.

Your logic is so refreshing, because so many people tell me I'm doing everything the wrong way... but in fact, everything you just wrote is my thinking as well. I'm glad I'm not alone.

The industry needs people like you to tell it straight.

Well done man.

 

[Michael Cioni]

Andreas-
Hot Swap Module is for rental only. Here is a more detailed diagram of what it delivers in a tiny, ultra lightweight package.

m

 

[Terry VerHaar]

Michael Cioni, Phil Holland and Mark Toia on the same page, here (almost literally). That says it all. Love these guys.

 

[Christopher*Seguine]

8K is not about sharpness, it's about smoothness.

Good segue from sharpness to smoothness, Primo-70 lenses were designed with Panavision's failed 65mm digital camera in mind, larger 70mm image format with a 6.25 micron photo-site. On Weapon 8k 5 micron pitch they would be far from sharp, a FF35 lens at 6k image is probably as sharp or sharper for a 4k deliverable.

Did Panavision build their own thicker 7mm OLPF for Weapon8k or do you still need to use the additional goofy glass insert like you do on Dragon and other cameras?


Nice that they cranked out a PV70 mount for weapon8k quickly, I guess thats how Guardians of the Galaxy II was able to commit, does the PV mount on weapon provide lens metadata to the R3d?

 

[Jarred Land]

Standard OLPF here in the Weapon 8k BTW.

 

[Nick Morrison]

Standard OLPF here in the Weapon 8k BTW.

Awesome. Great to know!

best

 

[Andrei Goaga]

Will the Primo 70 series ever be available to buy? Or rent only, as anything Panavision? Maybe under another brand's name?

 

[Blair S. Paulsen]

What you describe as "super samples" process usually implies taking a higher resolution image, like in this particular case an 8K Bayered image and downsampling it to UHD or 4K. By doing so you do end up with resulting image of higher quality.
On the other hand, I'm a bit puzzled about your example of the importance of super samples with use of large format photography by Ansel Adams. Like many other fine art photographers, he had used a 5x7 or 8x10 view cameras, resulting in the same size negatives. Most of his prints are in the neighborhood of 30x40 all the way to 40x60. I believe, those viewers visceral reactions to Ansel Adams images you're describing, actually derive from the superior MTF response from large format viewing camera photographic process and not from super sampling. In my opinion, super sampling doesn't come into play in this particular example, as the image actually end up being much bigger in size unlike the process of downsampling an 8k to 4k you describing. The equivalent process to the one you're describing would be for Ansel Adams to dupe his 8x10 negative to say medium format negative and then print from that. That wold be an example of super sample.

What is the resolvable resolution of a square inch of 100 ISO negative film? I'd posit the answer is far higher than the nominal dots/pixels/lines per inch of 99.99% of display media - projected/printed/fixed panel/etc. In the case of 35mm film, most slower stocks have been proven capable of exceeding 4K in optimal pipelines. Shooting larger negatives provides more square inches of original information per frame that I believe is properly considered super-sampling. If a 5x7 view camera creates a 35 square inch negative and a 35mm negative is at most 4 square inches...

Cheers - #19

 

[Meryem Ersoz]

listening to Michael talk tech is like listening to Bach. always breaking new ground but staying true to the beauty of invention.

thanks, always a pleasure to hear your thoughts.

 

[David Mullen ASC]

Good segue from sharpness to smoothness, Primo-70 lenses were designed with Panavision's failed 65mm digital camera in mind, larger 70mm image format with a 6.25 micron photo-site. On Weapon 8k 5 micron pitch they would be far from sharp, a FF35 lens at 6k image is probably as sharp or sharper for a 4k deliverable.

I don't think Panavision designed the lens so that it would only be sharp on a 6.25 micron sensor but unsharp on a 5 micron sensor... I don't think that's really how lenses are made...

 

[Michael Tiemann]

Good segue from sharpness to smoothness, Primo-70 lenses were designed with Panavision's failed 65mm digital camera in mind, larger 70mm image format with a 6.25 micron photo-site. On Weapon 8k 5 micron pitch they would be far from sharp, a FF35 lens at 6k image is probably as sharp or sharper for a 4k deliverable.
[...]

Too much speculation for me. Here's a good article from Canon that puts a lot of optical principals and criteria into once place. Of course we're talking 8K across a mini-VV image circle rather than 4K across an S35 image circle, but I think one can see how some properties translate and how others scale from one format to another. In response to the above, I quote from page 7:

Recognizing the impossibility
of achieving a totally constant MTF across the image plane, the optical designers define two circular
zones – based upon years of collaborative experiences with cinematographers – where they make every
attempt to maintain close to constant MTF.

Specifically, one can very much expect that the of the central region of a 65mm image circle will indeed be sharper than the overall averaged MTF that passes for the lens's MTF. For that one reason I would judge false the statement that Panavision Primos would be judged a failure using 5 micro pixels. I suspect that Panavision Primos have "excess" sharpness in the middle region in order to provide adequate MTF falloff toward the edges--edges that the 8K sensor won't see, because it doesn't have photosites out there.

There are many other reasons that I think the above statement about 5 micron vs. 6.25 micron pixels don't hold. Read that paper, or many other papers on MTF if you care to do the research.

 

[Patrick Tresch]

goofy glass insert

Is this the new name for standard Olpf? The Goofy olpf?

 

[Christopher*Seguine]

I don't think Panavision designed the lens so that it would only be sharp on a 6.25 micron sensor but unsharp on a 5 micron sensor... I don't think that's really how lenses are made...

I was inferring they were designed for a much larger sensor size, so their resolving power per mm would be diminished on a much smaller sensor. They target 90% MTF at 40 lp/mm

I thought I remembered these being designed for 65mm 5-perf - 52.5mm x 23mm , but apparently they won't cover the Alexa 65.

So I looked back and the panavision camera they were designed for was 7680 x 3240 @ 6.25 mircrons, so their design target was to cover 48mm x 20.25mm, significantly wider than Weapon 8k's 40.96mm, but its actually shorter than weapon's 21.6mm.

So now I have the opposite question, Michael Cioni, did you shoot this Erin Gales piece 8k FF and crop to 16:9, and the Primo 70's covered the full height of the 8k sensor? Or did you shoot some form of cropped 8k and then crop to 16:9 ?

Is this the new name for standard Olpf? The Goofy olpf?

No I wasn't calling the "Red Standard OLPF" goofy. One problem is these lenses were designed for a 7mm sensor stack thickness, pretty thick, so on modern cameras you have to stick in a piece of glass inline to compensate, Dragon is around 3mm olpf, not sure if this changed on Weapon? John Marchant?

No other camera has had an easily changeable OLPF, F55 doesnt really count - so it would have been pretty cool if Panavision made their own OLPF that is the correct thickness and less aggressive filtering in the H direction to make up for the fact the lenses were designed for a wider sensor. But looks like that is not the case.

Michael Tiemann, you can look up the old SMPTE papers on the Panavision camera for more information on their intended sensor target.
The using the "sweet spot" of the lens argument works ok when the lens's intended target is slightly larger, but not when there is a large disparity in size - say a Medium Format film lens on a Weapon 8k sensor, the MF lens does not need to resolve very much per mm since the film plane/sensor is so large, when your only using a small piece of it on a smaller sensor your sweet spot is now a soft spot.


Interested if they do work with Red to add lens metadata, seems like a VFX heavy Guardians of the Galaxy would demand it, but I guess they can still record it externally with the Arri UMD4