Weapon 8k Queue

[Stacey Spears]

Some of us have half a dozen 8k's coming

You have half a dozen coming? Nice! VR?

Yeah I can't wait to get my Zeiss Otus on this sensor size.

As soon as someone can verify that the 28mm Otus covers the 8k, at least 8k HD, I will order one. I love the Otus lenses. I am sad when I have to fall back to something like a Canon 70-200mm MK II, which is a nice lens. I hope they go more telephoto on the next one, something in the 100mm range would be nice.

 

[Vladimir Eugene]

Tonaci is the VR guy

You have half a dozen coming? Nice! VR?

 

[roryhinds]

The Otus 28mm covers my Sony A7S full Frame so I would think it would have no issues covering Red's Full Frame (which is smaller than true full frame)
Phil will know as he's a Zeiss man

 

[Jeff Kilgroe]

RED 8KFF sensor area is approximately 44mm diagonally or just a fuzz larger diagonally than 35mm FF. It's 40.5mm wide, but quite a bit shorter with its 1.9:1 aspect ratio, so still fits within a comparable image circle.

I want to say all the Otus primes cover fine, but I'm not certain about the 28mm and Phil would be the man to ask.

 

[Phil Holland]

The Otus 28mm covers my Sony A7S full Frame so I would think it would have no issues covering Red's Full Frame (which is smaller than true full frame)
Phil will know as he's a Zeiss man

RED 8KFF sensor area is approximately 44mm diagonally or just a fuzz larger diagonally than 35mm FF. It's 40.5mm wide, but quite a bit shorter with its 1.9:1 aspect ratio, so still fits within a comparable image circle.

I want to say all the Otus primes cover fine, but I'm not certain about the 28mm and Phil would be the man to ask.

Much, much, much of this has been on my mind for too long to even remember. So I made a handy thing:

The reality based on those numbers (and more recently my personal experience) is damn near everything will cover 8K HD, probably 8K WS as well. All of the lenses I used on Forged covered 8K FF just fine and did very well. I suspect the Otus 28mm will cover, but I had to return the prototype lens literally a few days before we shot.

I went 3 routes with glass. Super 35mm designed lenses that are optimal for Dragon 5K, FF35/VistaVision friendly glass that is optimized for those formats, and medium format glass that is suitable for an even larger format.

Some manufacturers have really narrow tolerances and I suspect there's a couple lenses that won't perfectly cover. I have started thinking about the optical path of some lenses in relationship to their PL or EF mount designs as well, which may in fact cause some slight vignetting. It also might not.

I need to do more aggressive coverage tests and that's actually going to happen soon-ish.

Much like Tyrion Lannister, I've begun hearing heavy Dragon breathing from within the darkness.

 

[Blair S. Paulsen]

The 8K Dragon sensor is just a hair less than 21x41MM vs the 24x36MM "full frame" spec. That extra width, 40.96 vs 36, does require a larger image circle to avoid vignetting or port-holing. Most FF35 lenses have enough "extra" coverage in the horizontal to make the difference moot - but not all. Phil's tests are the best info I've seen and his methodology is solid.

Cheers - #19

 

[Stacey Spears]

Phil, love the chart(s). Hopefully one day you can make as a webpage where the user can check/uncheck the options they care about to reduce some of the circles and rectanges to make it easier to read. If you sold an app that did this, I would buy it.

 

[roryhinds]

Phil don't you have the 8k already?You should get a 28mm its amazing.I just de-clicked my Otus 55, 85 and 28mm. One step closer to cinema perfection

 

[Phil Holland]

Phil don't you have the 8k already?
You should get a 28mm its amazing.

I just de-clicked my Otus 55, 85 and 28mm. One step closer to cinema perfection

No. Not yet. The camera I used early February was Jarred's prototype 8K Weapon.

However...... There are Dragon whispers in the darkness. Just waiting to see that glimmer of light so I can unchain them Dragons. (don't read this if you aren't watching GoT right now)

 

[Milano Questa]

With the yield rate at the moment being 1-3 good sensors out of every 300-500 potential sensors it's damn near impossible to provide estimates as it's literally a day by day news report. If that gives you any indication, I'd say at the moment it's a slow moving boat with very small batches coming out of the oven.

I'm dying to find out which sensor fab has such bad yields. Seriously...Red 6K sensors use pixel mapping to get rid of bad pixels. Why not do the same on 8K sensors that aren't 100% perfect?If you're gonna make an excuse, atleast make it a believable one. Jeez

Adding a few milimeters to an existing sensor design doesn't suddenly make the yield go down the pooper like Jarred is telling us.

For reference: Sony's 8K a7rii sensor yield is about 70%

 

[Antony Newman]

Interesting statistic - although I think 70% might be a bit optimistic. ;-)

http://www.digitaltrends.com/business/japan-quake-disrupts-sony-cmos-plant/

AJ

 

[Phil Holland]

I'm dying to find out which sensor fab has such bad yields. Seriously...Red 6K sensors use pixel mapping to get rid of bad pixels. Why not do the same on 8K sensors that aren't 100% perfect?If you're gonna make an excuse, atleast make it a believable one. Jeez

Adding a few milimeters to an existing sensor design doesn't suddenly make the yield go down the pooper like Jarred is telling us.

For reference: Sony's 8K a7rii sensor yield is about 70%

Hi Milano,

Without divulging too much information here regarding sensors I can in short say a few things.

- It's not simply "yield" from the fab
- There is a VERY thorough QC process that is necessary
- The pixel qualification is not a linear scale, it exponentially grows for the larger the sensor
- The demands of the 8K Dragon sensor are extremely high dynamic range and a successful readout of information from those pixels at high frame rates
- RED's pixel quality demands are different than a typical DSLR in terms of what qualifies as a "good and usable" pixel.
- RED's camera needs to maintain a consistent image quality throughout the shooting day.

You are comparing a "slow sensor" with a fair bit of image processing to solve various pixel issues with a very, very fast sensor with more dynamic range and an entirely different degree of pixel qualification.

You're asking somebody why it's hard to move a rock that you have never seen, but because you've held a rock before you think all rocks weigh the same and take the same level of effort to move. The difference here is the rock you are familiar with weighs 100lbs and is fairly easy to move and the other ways 1200lbs and is significantly more difficult to put in it's place.

And yes, RED sensors do indeed deploy pixel mapping for pixels that remain within a QC limit. But pixels themselves have different states and then there's the concept of bad clusters, which isn't ideal. Worse yet, to have "good" looking footage with consistent dynamic range and color with a broad range of integration times/shutter speeds the voltage variation and characterization is rather complicated. Profiling a high resolution sensor such as this becomes a larger chore than you'd think. In the example of Dragon 6K versus 8K for instance, we are talking 55% more pixels which in turn makes it about 219% more difficult to do.

I won't bash another company, but there is one recent camera on the market that if you have a shutter speed of faster than 1/250th of a second you begin to see patterns in flat colors due to the sensor level of QC, how they've deployed it in camera, and on the boards. Worse yet the way they've decided to deploy calibration breaks up into 4 quadrants and can eventually result in a cross shape image tear across the 4K sensor. I don't own that camera. No movies are being shot with that camera (that I know of). That is not what RED is trying to produce with their Dragon and future sensor technologies.

Also, just to add another wiggle to your pickle party, as a happy Sony A7RII owner I am pleased with much of what the camera can do. HOWEVER. The damn posterzation that gets draw in the bloody raw files at times is so damn stupid. It's the number one reason I can't use it as my main timelapse camera on studio level productions and they need to address it in a big, big way. It can induce visible blocking and worse yet jagged edges that aren't a result of aliasing, but rather how the camera "draws and renders". It's possible to create this artifact on any contrasty edge straight or round. It's a pain in the ass.

 

[Elsie N]

I'm dying to find out which sensor fab has such bad yields. Seriously...Red 6K sensors use pixel mapping to get rid of bad pixels. Why not do the same on 8K sensors that aren't 100% perfect?If you're gonna make an excuse, atleast make it a believable one. Jeez

Adding a few milimeters to an existing sensor design doesn't suddenly make the yield go down the pooper like Jarred is telling us.

For reference: Sony's 8K a7rii sensor yield is about 70%

This is new ground for RED and in order to get it right, I'm sure they have their Facebook beta testers shooting or preparing to shoot all over Hollywood right now. In time, if there are no problems I suspect they will start trickling out to us forum guys. If there are problems, that's bad news for us because they will likely keep pushing out new characterized sensors to the beta testers.

 

[Milano Questa]

Hi Milano,

Without divulging too much information here regarding sensors I can in short say a few things.

- It's not simply "yield" from the fab
- There is a VERY thorough QC process that is necessary
- The pixel qualification is not a linear scale, it exponentially grows for the larger the sensor
- The demands of the 8K Dragon sensor are extremely high dynamic range and a successful readout of information from those pixels at high frame rates
- RED's pixel quality demands are different than a typical DSLR in terms of what qualifies as a "good and usable" pixel.
- RED's camera needs to maintain a consistent image quality throughout the shooting day.

You are comparing a "slow sensor" with a fair bit of image processing to solve various pixel issues with a very, very fast sensor with more dynamic range and an entirely different degree of pixel qualification.

You're asking somebody why it's hard to move a rock that you have never seen, but because you've held a rock before you think all rocks weigh the same and take the same level of effort to move. The difference here is the rock you are familiar with weighs 100lbs and is fairly easy to move and the other ways 1200lbs and is significantly more difficult to put in it's place.

And yes, RED sensors do indeed deploy pixel mapping for pixels that remain within a QC limit. But pixels themselves have different states and then there's the concept of bad clusters, which isn't ideal. Worse yet, to have "good" looking footage with consistent dynamic range and color with a broad range of integration times/shutter speeds the voltage variation and characterization is rather complicated. Profiling a high resolution sensor such as this becomes a larger chore than you'd think. In the example of Dragon 6K versus 8K for instance, we are talking 55% more pixels which in turn makes it about 219% more difficult to do.

I won't bash another company, but there is one recent camera on the market that if you have a shutter speed of faster than 1/250th of a second you begin to see patterns in flat colors due to the sensor level of QC, how they've deployed it in camera, and on the boards. Worse yet the way they've decided to deploy calibration breaks up into 4 quadrants and can eventually result in a cross shape image tear across the 4K sensor. I don't own that camera. No movies are being shot with that camera (that I know of). That is not what RED is trying to produce with their Dragon and future sensor technologies.

Also, just to add another wiggle to your pickle party, as a happy Sony A7RII owner I am pleased with much of what the camera can do. HOWEVER. The damn posterzation that gets draw in the bloody raw files at times is so damn stupid. It's the number one reason I can't use it as my main timelapse camera on studio level productions and they need to address it in a big, big way. It can induce visible blocking and worse yet jagged edges that aren't a result of aliasing, but rather how the camera "draws and renders". It's possible to create this artifact on any contrasty edge straight or round. It's a pain in the ass.

My pickle party wants you to update your a7rii. This was fixed a million years ago.

http://www.sonyalpharumors.com/dpreview-tests-the-new-sony-a7rii-14-bit-uncomrpessed-raw-files/

As far as i know the A7rii sensor is capable of outputting 60fps at full res. It just lacks the hardware to process and store all that information. At 14stops of (usable) DR compared to 14.5stops (usable) on Dragon.... the difference ain't big.

for comparison: 4K smartphone sensors can easily output Raw 60fps 4K video. But due to hardware and size limitations it gets compressed to h264. and all smartphones max out at 30fps at 4k.

Sony obviously has the advantage of higher volume in regards to production cost. But that doesn't really explain the bad yields on 8K VV. Red might have run into production problems, who knows.

 

[DJ Meyer]

For reference: Sony's 8K a7rii sensor yield is about 70%

How do you know that? Did you read it on the internet?

 

[Phil Holland]

My pickle party wants you to update your a7rii. This was fixed a million years ago.

http://www.sonyalpharumors.com/dpreview-tests-the-new-sony-a7rii-14-bit-uncomrpessed-raw-files/

As far as i know the A7rii sensor is capable of outputting 60fps at full res. It just lacks the hardware to process and store all that information. At 14stops of (usable) DR compared to 14.5stops (usable) on Dragon.... the difference ain't big.

for comparison: 4K smartphone sensors can easily output Raw 60fps 4K video. But due to hardware and size limitations it gets compressed to h264. and all smartphones max out at 30fps at 4k.

Sony obviously has the advantage of higher volume in regards to production cost. But that doesn't really explain the bad yields on 8K VV. Red might have run into production problems, who knows.

I have and it still exhibits it. Specifically in the world of architecture. Yes, even with the 14 bit raw. I sent Sony files showing the artifact in a controlled environment. Same subject, camera tilted .5 degrees and the lines on metal tile look entirely jaggy with blocks. All good. I'm happy to use this camera for what I use it for.

The core thing you are missing is the difference between what these sensors are, the differences between them, and how they are used in their respective systems.

That makes up a fair bit of the difficulty at what it's taking to do this. Sony and Canon are making 8K sensors for their premiere cameras at the moment as well. Both have been working on them for about 2-3 years. Both as far as I know are over a year away from hitting the market. The criticism is valid in terms of the wait for the larger sensor Weapon. It's been a year really since pre-orders were made. But RED is first on this front and even Canon mentioned how much they were impressed with the 8K Weapon at NAB.

 

[Blair S. Paulsen]

RED is smaller and younger than Sony/ARRI/Canon/etc. That said, when it comes to high resolution, high dynamic range, high speed digital sensors I'd suggest they've done their homework and then some these last 10 years. Furthermore, Phil's notes point to what I would imagine is the crux of the biscuit - the extremely demanding specifications.

If I had to guess, I'd wager RED has stuck with their sensor production partner because nobody else was willing to take on such a difficult task. Perhaps I'm all wrong, perhaps Jarred and the RedTeam are just dupes who are getting taken for a ride by some half assed sensor fab... :svengo:

Cheers - #19

 

[Stacey Spears]

the crux of the biscuit

pickle party,

I am getting hungry with all this talk of biscuits and pickles. Okay, not really pickles!

 

[Milano Questa]

RED is smaller and younger than Sony/ARRI/Canon/etc. That said, when it comes to high resolution, high dynamic range, high speed digital sensors I'd suggest they've done their homework and then some these last 10 years. Furthermore, Phil's notes point to what I would imagine is the crux of the biscuit - the extremely demanding specifications.

If I had to guess, I'd wager RED has stuck with their sensor production partner because nobody else was willing to take on such a difficult task. Perhaps I'm all wrong, perhaps Jarred and the RedTeam are just dupes who are getting taken for a ride by some half assed sensor fab... :svengo:

Cheers - #19

Maybe they should ask Sony to produce their sensors for them. :-D

I'm sure Red will get around to volume production soon. It's just the yield explaination sounds a bit funny to me.
VV is a ground breaking camera, so i'll cut them some slack. 75fps 8K in a 2kg package is bananas.

@Meyer i visited their sensor fab last year.

 

[Elsie N]

I have and it still exhibits it. Specifically in the world of architecture. Yes, even with the 14 bit raw. I sent Sony files showing the artifact in a controlled environment. Same subject, camera tilted .5 degrees and the lines on metal tile look entirely jaggy with blocks. All good. I'm happy to use this camera for what I use it for.

The core thing you are missing is the difference between what these sensors are, the differences between them, and how they are used in their respective systems.

That makes up a fair bit of the difficulty at what it's taking to do this. Sony and Canon are making 8K sensors for their premiere cameras at the moment as well. Both have been working on them for about 2-3 years. Both as far as I know are over a year away from hitting the market. The criticism is valid in terms of the wait for the larger sensor Weapon. It's been a year really since pre-orders were made. But RED is first on this front and even Canon mentioned how much they were impressed with the 8K Weapon at NAB.

Phil, I think what people are curious about is the fact Jarred has had a working camera (as you are well aware) for quite some time now. That and the fact that a big production has cameras (even an 8k crash cam) and new multiple 8k specialty cams are being fabricated for Hollywood superstars. These things they are aware of may be making people wonder if they can get some of these to work, how come they can't replicate that process in at least acceptable numbers?

Also, I think if DARPA and NASA would post their timestamps in Stacey's thread, that would give the queue members a better idea how long before 8ks will move into the civilian world.