8K Weapon: The New Super Sensor

[David Mullen ASC]

The Primo 70 lenses looked pretty sharp on "Ride Along 2" and that was only on the normal-sized Dragon sensor...

 

[Peter Lyons Collister ASC]

Ah Michael Cioni... When I bought my first Red One and went to Irvine to listen to you talk you made digital acquisition understandable to an old film DP like myself. And you continued to help me understand through the past 7 years. And now this opus about 8K. Was just arguing on set with my operator today about resolution and when is enough enough. I will use your words.

And thank you Panavision for your amazing lenses and now some rental modules for Weapon. Smart. Really smart.

Cheers,

Peter

 

[Blair S. Paulsen]

I indeed hope that the days of Panavision resisting the RED Rebellion are over. The inevitability of digital replacing photochemical process for 90+% of motion imaging work has become fact. There's still rumors that PV is going to have a rental only digital camera available this year - but until that becomes actual fact, the Alexa65 and Weapon8K would seem to be excellent capture devices for that terrific glass.

Cheers - #19

 

[Steve Sherrick]

Ah Michael Cioni... When I bought my first Red One and went to Irvine to listen to you talk you made digital acquisition understandable to an old film DP like myself. And you continued to help me understand through the past 7 years. And now this opus about 8K. Was just arguing on set with my operator today about resolution and when is enough enough. I will use your words.

And thank you Panavision for your amazing lenses and now some rental modules for Weapon. Smart. Really smart.

Cheers,

Peter

The argument that is made by what I'll describe as the "anti-resolution" crowd is that all it does is make actors look terrible, too much detail, etc. Basically, the same arguments that followed High Definition's inception. Having more resolution doesn't necessarily have to translate to showing every pore on an actor's face. It does require learning how resolution benefits you and how it could have a negative impact and then creating the images you want through the usual means of lighting, filtering and post processing. In other words, resolution is but one of many components to the image and blanket statements that there's too much resolution are just that - blanket statements. There are also those that will argue the eye can't tell the difference under a lot of circumstances. But that's missing the point on a lot of levels.

In my opinion, RED will continue to improve many aspects of IQ, resolution being one of them. It's not like they are going to ignore dynamic range, color science, etc. But if they can improve all aspects of the IQ they will.

 

[Björn Benckert]

just saying. If weapon is said to have 16.5 stops then weapon 8k naturally has plenty more DR as the noise floor is considerably lowered for delivery resolution at 4k and massivly so at 2k. And from what I read on this forum and other places people seam, just like me, to be thrilled about having greater dynamic range and lowlight sensitivity. simple math says weapon 8k deliveres more than red and most other camera manufacturer has ever done on that point.

Another thing sensor noise is a bigger problem than film noise as sensor noise is static over frames to some degree where as film noise does not stay still over frames and for that reason is presived as showing more detail even if individual frames might show less detail. So even if you like texture capturing it all in a digitial camera is to me usually less apealing than adding texture in post as the post grain has that same randomness per frame as film has.


So even if recording 422 2k proress files in camera the weapon 8k is still in a class of its own, for low noise reason, the larger sensor look aside.

Sidenote. Recorded at 6k r3d + proress 2k today... Honestly have to say that was the most slick proress files I seen coming out of any camera. So its absolutly nothing wrong with the internal down sampling in weapon. Something I find quite amazing... Camera is so small and still they managed to fit somthing like the redrocket x and what ever else computing power in there that handles high framerate 6 or even 8k stuff and bake it as clean proress file on the same mag as the r3d...Its nuts if you think of it.

first I was playing with it and forgot it was on, later at night when in edit it was like a litle happy surprise... look we got transcodes for all the files...just searching for .mov down in your capture folder and there they all are. Only little oddity i found was that clips looked a frame short but starting on the sane frame/tc. That does not matter much tome but I guess it can cause "missing media" issues when using full length takes in xmls etc.

 

[Tom Gleeson]

Bjorn,

I would assume and sincerely hope that the Prores files and R3d files are frame accurate with identical start and stop frames? If you are certain they are not identical I can set up a test on my camera to see if it is particular to your machine or the same on all cameras?
Maybe somebody with more post expertise can clarify how important the start and stop frame accuracy is across the file types assuming timecode is identical?

 

[Christopher*Seguine]

The Primo 70 lenses looked pretty sharp on "Ride Along 2" and that was only on the normal-sized Dragon sensor...

That's pretty awesome that you would watch an Ice Cube film and I didnt know any features but the dreadful Gods of Egypt used these.
2k mastering, so I don't think sharpness was a big priority, and from the horse's mouth:

Panavision VP of Optical Engineering Dan Sasaki said the lenses aren't "too scientifically sharp," noting that the Primo 70s "share the Primo family resemblance" with the original Primos that became popular in the 1990s.

I also didn't realize Panavision bought up Michael's Light Iron, thread makes sense now.

Would like to know if the Primo70s do actually cover 8k FF or only 8k 2:1

 

[Björn Benckert]

Bjorn,

I would assume and sincerely hope that the Prores files and R3d files are frame accurate with identical start and stop frames? If you are certain they are not identical I can set up a test on my camera to see if it is particular to your machine or the same on all cameras?
Maybe somebody with more post expertise can clarify how important the start and stop frame accuracy is across the file types assuming timecode is identical?

in flame one showed up one frame longer than the other... tc in was the same. can check in PRemere tomorrow. To me it makes no difference. But I know for certain work flows it would be crucial...

Like if all clips are laid back to back onto new tc in a long source reel or such then that missing /extra frame between online and ofline files will for sure create a little hurdle.

 

[Tom Gleeson]

Bjorn,

So only one clip out of many clips was one frame longer than the R3D?

 

[Björn Benckert]

Bjorn,

So only one clip out of many clips was one frame longer than the R3D?

I have to check tomorrow and in different nle it was more than one clip, did not check to see if it was on all. Was to happy about having those prores files playing like nothing straight from the mag reader on my shity old laptop... so at that moment I did not care But have two mags full that I can check tomorrow.

 

[Tom Gleeson]

Bjorn,

It is worth understanding these issues as I am selling to clients the potential for convenience in certain workflows for the camera generated Prores files and I look bad if they end up in a world of hurt. The QT file issue probably should be moved off on to this thread?

http://reduser.net/forum/showthread.php?140767-Importing-mov-into-Premiere-and-Resolve-issue

 

[Björn Benckert]

Bjorn,

It is worth understanding these issues as I am selling to clients the potential for convenience in certain workflows for the camera generated Prores files and I look bad if they end up in a world of hurt. The QT file issue probably should be moved off on to this thread?

http://reduser.net/forum/showthread.php?140767-Importing-mov-into-Premiere-and-Resolve-issue

Checked my files and sent a note to Graeme Nattress about it. r3d´s are a frame short in the tail end. Understand that can cause some problems in editing but in reality if it´s known not so much until the editor really wants to use that last frame... which then does not exist in r3d 6k glory but only in low res proxy prores format....
The opposite would better the r3d longer then the offline file...then we can happily bring out that extra frame when needed

Again dont know if this is jus current firmware, just my camera, or something that is and will be on all cameras. No matter what its not really a big hurdle, is it? those proress files are still very impressive as I see it.

[/url]PastedGraphic-9 2 by Björn Benckert, on Flickr[/IMG]

 

[Christoffer Glans]

Checked my files and sent a note to Graeme Nattress about it. r3d´s are a frame short in the tail end. Understand that can cause some problems in editing but in reality if it´s known not so much until the editor really wants to use that last frame...

It can create problems. I've used the definite ins and outs of a clip sometime because it was the best take and the camera stopped recording in the middle of it and you want as much as possible out of the image before a cut. So yeah, it creates a problem.

 

[jake blackstone]

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

Resolution of the film stays the same regardless of the size of the negative. Increasing the size of the negative at the same resolution will result in the increase of MTF as a whole. In fine art photography there was never need to create an image for a particular resolution -HD, 4K, 8K etc. It was always about the resulting image, making it beautiful, regardless of resolution. Incidentally, large format lenses were notorious for it's low resolution, which didn't stop anyone from producing the most iconic images known to man.

 

[Michael Hastings]

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 understand the concept, just not sure I'm buying it. Seems to me your explanation would be workable if we were scaling from say 10x resolution but here we are only talking an 8/5 (or 5/8 depending on how you look at it) difference in relative horizontal or vertical pixel distribution which to me wouldn't create that big of a difference in "roundness". Increased MTF and noise reduction, yes but this roundness idea not so much - i.e.it is there but only maybe a 60 percent difference - or for example like going from a pentagon to a hexagon of the same size.

But I have to admit, figuring it out is a head scratcher.

 

[Robert L. Mickles]

Well said and well done, Michael. I hope to own this camera and get to use these lenses one day.

 

[Karim D. Ghantous]

Thanks, Michael. Everything you wrote makes sense. And I like your choice of subject matter. Very much! I suppose 8K is to 4K as Hasselblad is to Leica. ;-)

so many people tell me I'm doing everything the wrong way

Okay, let me guess the BS that gets thrown your way (if I'm right, I get a lollipop):

- Shooting RAW is overkill

- 4K is a waste of time - until Arri says so

- VFX are rendered in 2K today, therefore it will always be this way

- Most people have HD displays, therefore you should only shoot 2K

In the case of 35mm film, most slower stocks have been proven capable of exceeding 4K in optimal pipelines.

Super 8 7203 is amazing - technically it's about 720 lines worth of detail (you do need a good lens, obviously). Super 16 7203 is about 1080. 3-perf 5203 is about 4K. I suspect that 5207 & 5213 would not be far from 4K.

 

[Graeme Nattress]

Yes, it's never been resolution that makes actors look bad, but false electronic detail to make up for the fact you don't have actual resolution!

Graeme

The argument that is made by what I'll describe as the "anti-resolution" crowd is that all it does is make actors look terrible, too much detail, etc. Basically, the same arguments that followed High Definition's inception. Having more resolution doesn't necessarily have to translate to showing every pore on an actor's face. It does require learning how resolution benefits you and how it could have a negative impact and then creating the images you want through the usual means of lighting, filtering and post processing. In other words, resolution is but one of many components to the image and blanket statements that there's too much resolution are just that - blanket statements. There are also those that will argue the eye can't tell the difference under a lot of circumstances. But that's missing the point on a lot of levels.

In my opinion, RED will continue to improve many aspects of IQ, resolution being one of them. It's not like they are going to ignore dynamic range, color science, etc. But if they can improve all aspects of the IQ they will.

 

[Christopher Probst]

Yes, it's never been resolution that makes actors look bad, but false electronic detail to make up for the fact you don't have actual resolution!

Graeme

Boom!

Which is also why no actors would complain about being shot in 65mm film! Increased resolution has been proven to reduce eye fatigue and increase pleasure during the view experience.

 

[Blair S. Paulsen]

+1 to Graeme's note about why most HD cameras got a reputation for accentuating skin blemishs, wrinkles, etc. Due to color sub-sampling, limited DR, over compression, mediocre optics, aggressive noise reduction, etc - many HD cameras struggled to look appreciably sharper on typical displays than good SD cameras could manage (which had been through many generations of refinements). Concerned that it would be hard to sell people on the pricey new HD gear if the difference didn't jump out at you, they simply dialed up the sharpening (cum aliasing) to make it "pop".

This over-sharpening would often find a contrast transition on human skin that it would "enhance" as if it were an edge, essentially outlining and accentuating minor skin variations. Mr. Nattress has consistently chosen smooth over false sharpness in his algorithms, letting the author of the image decide how much sharpening to add when needed. Especially in the first few years of the RED Rebellion some producers/directors actually considered RED 4K images too soft after initial processing! Their eyes having become conditioned to the artificial "pop".

It's quite easy to sharpen the bejeezus out of any RED footage if that's what you desire. It's extremely difficult to undo over-sharpening/aliasing in post without damage. My analogy is that you can't un-cook the steak, but you can cook it more if it's too rare. BTW - my insider tip is to experiment with scaling filter selection (Mitchell, Lanczos, Bi-Cubic, etc) first, then add just enough sharpness to get the look you want. IMO, this approach leads to more natural looking imagery than just cranking up a sharpness slider. YMMV.

Cheers - #19