Jump to content

Black Magic Design 12k Digital Cinema Camera


Recommended Posts

Hey all,

I’m interested in hearing your thoughts on this new camera https://www.blackmagicdesign.com/products/blackmagicursaminipro. I am a previous owner of the Ursa Mini Pro which was a decent camera however a little unreliable. The timing on this for me couldn’t be more perfect for me as I just finished rewatching Yedlin’s resolution Demo when they announced it. Would such high resolution at this point make any difference other than compromising the latitude and quality of said pixels? Their latitude going from 15 stops (already through my tests some pretty generous advertising) to advertised 14 stops. But still 12k resolution 12,228 x 6480. The numbers impressive at least. 

Gabe

Edited by Gabriel Devereux
Link to comment
Share on other sites

  • 1 year later...
  • Premium Member

I have yet to use the camera and I had thought about breaking the piggy bank to purchase one. However, messing around with Blackmagic's own samples, I feel that they are less cinematic than even the BMPCC 4k. I guess it all depends on what work you do with the camera but the built in NDs seem like a nice surprise for me (coming from no-budget land) but I am turned off by the size of the body. Some wont care (likely the more professional peoples around here who are used to massive rigged out cameras).

Is there any benefit to this over the 6k Pro?

  • Upvote 1
Link to comment
Share on other sites

  • Premium Member

I think the idea is that it has such overwhelming resolution that the sharpness compromises of a mosaic sensor are overcome and even 8K material is so oversampled that it approaches the Nyquist limit of the frame dimensions in terms of sharpness and detail. In may circumstances, the factor limiting resolution is not the sensor nor the quality of the lens, but the diffraction limit of the lens, which is geometrically determined and therefore exactly the same for a low-cost stills zoom as it is for a Leitz Cine Prime. The point isn't really 12K. The point is it has much more than you need for any reasonably foreseeable circumstance.

Yes, smaller photosites are noisier, which compromises dynamic range, but that's sometimes a rather misunderstood metric. If you take an HD area from the 12K sensor, it will be physically smaller than the HD area of a 4K sensor, and it will be noisier. But if you take the whole 12K sensor and scale it to HD, you gain back much of the loss. Not all of it, of course, because more pixels have more gap around them, so your fill factor becomes lower and you're using less sensitive area overall, but this isn't quite as simple as is often assumed.

The big problem with the 12K is that using it to its full potential locks you into a Blackmagic Raw workflow. This, again, is not quite as simple a problem as we might think; generating 12K ProRes would be an impractically-huge processing load even if anything supported it, and scaling that vast sensor down to saner resolutions is hard work too. I do think they could have made slightly more effort in this regard, though; I think this is what puts most people off. Blackmagic Raw or bust, really.

  • Like 1
Link to comment
Share on other sites

1 hour ago, Phil Rhodes said:

I think the idea is that it has such overwhelming resolution that the sharpness compromises of a mosaic sensor are overcome and even 8K material is so oversampled that it approaches the Nyquist limit of the frame dimensions in terms of sharpness and detail. In may circumstances, the factor limiting resolution is not the sensor nor the quality of the lens, but the diffraction limit of the lens, which is geometrically determined and therefore exactly the same for a low-cost stills zoom as it is for a Leitz Cine Prime. The point isn't really 12K. The point is it has much more than you need for any reasonably foreseeable circumstance.

Yes, smaller photosites are noisier, which compromises dynamic range, but that's sometimes a rather misunderstood metric. If you take an HD area from the 12K sensor, it will be physically smaller than the HD area of a 4K sensor, and it will be noisier. But if you take the whole 12K sensor and scale it to HD, you gain back much of the loss. Not all of it, of course, because more pixels have more gap around them, so your fill factor becomes lower and you're using less sensitive area overall, but this isn't quite as simple as is often assumed.

The big problem with the 12K is that using it to its full potential locks you into a Blackmagic Raw workflow. This, again, is not quite as simple a problem as we might think; generating 12K ProRes would be an impractically-huge processing load even if anything supported it, and scaling that vast sensor down to saner resolutions is hard work too. I do think they could have made slightly more effort in this regard, though; I think this is what puts most people off. Blackmagic Raw or bust, really.

To expand and question, now that we've had time to somewhat dissect and look at the image and its pro's and cons.
I find the ideology - potentially the wrong term; of a sensor out resolving a lens flawed. As you say above the photosite pitch, the total pixel count crammed onto the sensor exceeds the resolving power of near, if not all-optical systems especially in realistic situations even with a hypothetical perfect lens. 
The main issue argued with the above is low capacity photo sites, easily saturated, total camera latitude etc. What you say makes logical sense about noise - I can understand to a certain extent assuming the singular white/clear photo site in each 'matrix' acts as a higher gain channel to interpolate shadow information to improve the dynamic range but the compromises are endless, yes all imaging and optical systems are a series of compromises. However, with the rather sudden and rigid highlight roll-off that is especially apparent with RGB clipping, the lack of workflow infrastructure around the total resolution and even with that, they haven't necessarily made a 12k camera as to somewhat even compete with image fidelity downsampling is a must. The selling point of the camera seems redundant.
To return to my initial statement about a sensor out resolving a lens. Maybe I'm being a fool, but, from my rather limited understanding of the engineering of an imaging sensor the ARRI with its 8.25um (diagonal) photo size 'size'/pixel pitch is in near-perfect harmony with the majority of its optical systems, a perfect example being the master prime. I believe achieving MTF(74) at 70lp/mm resolves an airy disk around a similar size to the photo site. It's relatively balanced so why compromise further on photosite size for pixel count. - it almost seems if they're attempting to reproduce the qualities of multiple channels (with different gains) from each photosite but in a two-dimensional sense which compromises, as you say above the total area to sample from. Which realistically results in a lower fidelity image (concerning SNR and other aberrations) even downsampled to 8k at a 1:1 pixel ratio with its competitors. 

 

Edited by Gabriel Devereux
Link to comment
Share on other sites

  • Premium Member
8 hours ago, Matthew W. Phillips said:

I have yet to use the camera and I had thought about breaking the piggy bank to purchase one. However, messing around with Blackmagic's own samples, I feel that they are less cinematic than even the BMPCC 4k.

The samples kinda suck from BM's website, it's a shame. 

I have a friend who is a pretty good DP, who swears by the 12k for reasons explained below. 

8 hours ago, Matthew W. Phillips said:

I guess it all depends on what work you do with the camera but the built in NDs seem like a nice surprise for me (coming from no-budget land) but I am turned off by the size of the body. Some wont care (likely the more professional peoples around here who are used to massive rigged out cameras).

Is there any benefit to this over the 6k Pro?

As you know the 6k pro has built-in ND's as well. 

The main thing to think about is that the 6k pro has horrible rolling shutter effect when using the full 6k imager. It's also not a S35mm imager, it's a "normal" 35mm imager. So unless ya use a speed booster, you're not getting the S35mm look. There is an "internal" speed booster made for the non-pro model, but I don't think they have anything for the pro model yet. 

The 12k is an amazing design, you can use the full imager to capture at any frame size. So even though it does have a color pattern like a Bayer pattern, it does not however force cropping. So you can shoot at 8k for instance, using the full imager and get a full RGB image, rather than half res on the R and B channels like a bayer pattern imager. In fact, the imager at 8k looks much better than it does at 12k. Being able to get a 12 bit full 444 image from the camera without sacrifice is huge. From my knowledge, nearly all other cameras at full resolution are using a Bayer pattern which uses half resolution on the R and B channels. 

The 12k has very good rolling shutter, but it's not global shutter. It's as good as the Red and Alexa cameras in that department. 

Of course,  it has other benefits over the 6k; Real S35mm imager, 120fps at DCI 8k, Better dynamic range, larger body = much more usable for professional settings, 8k capture which means it's a lot easier to re-frame in post for 4k finish, proper viewfinder position, lots of in's and out's without having to worry about damaging connectors, 4 card slots, interchangeable lens mounts.

I've worked with some 12k material shot by a decent cinematographer and the stuff is stellar. Working with the raw image really opens your eyes to the potential, something I didn't really see in the many pocket 4k and 6k shoots I've worked on. I just never was happy with the image from those cameras. Like the original pocket which suffered heavily from rolling shutter, the 4k and 6k pockets, both struggle with the same thing AND worse, the color science is troublesome. BMD have done a good job resolving SOME of the color science issues in the 6k, but they've struggled to get the 4k in the same ballpark in my opinion. We know they of course don't use similar imagers at all, the 4k is an off the shelf Sony imager and the 6k is a custom product. So that could be why the two cameras look so radically different. But the 6k is absolutely a step above the 4k in color science and I say the 12k is an ever bigger step above both of the 4k and 6k in the same department. They all however, have the same limited working range, kind of in the 800 - 1000 ISO, much outside of that and you're destroying your dynamic range AND/OR adding too much noise. 

Remember tho, even though the price is lower on the 12k, BMD are not done releasing cameras. There are rumors about a 12k Gen 2 and a 8k pocket coming out soon. So if you're going to buy, ya may wanna wait until they release something new because they do address many of the issues on the new releases. I'm also sadly not a fan of the URSA Mini body design, they really need to make a pro version of it, that's more robust. Things like the screw threads, are very easy to strip out and damage. I've seen several, all be it rental, UMP's that are totally trash from people pretending the cameras are an Arri Alexa. So that fragility does play a bit on my mind personally. Another thing to think about is that they aren't well sealed like an Alexa, the ventilation that runs through the center of the camera is on a heatsink, but also on the boards themselves. So any dust can unfortunately get into the electronics, much like the Red cameras. But for cost? I mean I think it's hard to beat, especially with the Komodo being so expensive and the alternatives in that price bracket, kinda being typical Sony or Panasonic stuff. The FX6 is an outstanding camera. The EVA1 is a pretty decent camera as well. There are other options, but I'm kind on board with BMD's workflow and no other camera brand has that dialed like BMD does, so that's the major tossup. 

In the end, the camera you use makes little to no difference. The biggest thing is how you light, compose your shots, block your scenes and of course, the written word and how it's spoken. You can shoot a shitty script with an Alexa Mini LF, with Arri Master Primes, Fisher 11 dolly, using a 10 ton grip truck and full crew and it will still be a bad movie. Just watch the junk on Netflix! lol ?

 

Edited by Tyler Purcell
  • Upvote 1
Link to comment
Share on other sites

1 hour ago, Tyler Purcell said:

The 12k is an amazing design, you can use the full imager to capture at any frame size. So even though it does have a color pattern like a Bayer pattern, it does not however force cropping. So you can shoot at 8k for instance, using the full imager and get a full RGB image, rather than half res on the R and B channels like a bayer pattern imager. In fact, the imager at 8k looks much better than it does at 12k. Being able to get a 12 bit full 444 image from the camera without sacrifice is huge. From my knowledge, nearly all other cameras at full resolution are using a Bayer pattern which uses half resolution on the R and B channels. 

Correct me if I'm wrong -
The Bayer CFA exists with twice the amount of green photosites - 'sensors' because our eyes - our photopic vision is more sensitive to the green (550nm or about, frequency band). 


With this, let's look at the sensitivity and the collection of data in an incredibly simplistic sense from a Bayer CFA. Looking at colour science and colour rendition in a similar simplistic fashion.
My eye is perceiving an incoming wavelength with a certain amount of value in each RGB frequency band, an amount of energy/photons in each. This theoretical ray of light has a value of 20 in the red channel, a value of 20 in the blue channel and a value of 20 in the green channel (emphasising this is an extreme hypothetical). 
For this example, attributes such as the density and efficiency of each filter on the CFA pattern will be ignored, each filter is even in its respective frequency band and as efficient as possible (the same with the QE of the photo-sites bellow). 
So with this, the data collected from the imager should be R20, G20, B20, G20 and after demosaicing (I'm tempted to attempt a calculation using a relatively basic Bicubic interpolation algorithm however I think it'd be unnecessary with such simplistic numbers) the general gist is the two green channels would be more 'sensitive' a higher value, than all the others. The hypothetical pixel created would be around R20 G40 B20.  The goal/hope would be my eye - my photopic response would achieve similar with this increased sensitivity in the same frequency bands. If I took the same ray of light, the same amount of photons and somehow got an RGB readout like a camera from my eye the values would be similar. 


Cameras are weird. The goal from my understanding is not to achieve a true RGB value from incoming light but match or achieve near similar results if it'd be our photopic vision interpreting said information. With this, I do wonder what kind of nifty amplification has been applied to the green channel on the UMP12k or other RGBW CFA cameras in an attempt to match this. I do believe if they were to capture all RGB values at 'full resolution', sampling all three values at the same sensitivity, the image produced would look a lil weird. 

Link to comment
Share on other sites

Does the P4K really have worse color than the P6K? I'm very, very impressed by the P6K's image both technically and aesthetically but the sensor is a bit too small for my S35 and FF lenses and too big for my S16 lenses. To me a P4K (and a speed booster XL or regular depending on the lens set) seems more viable. Also 6K is too heavy for the post work I do, though of course I could downscale. What, if any, other cameras in this price bracket have excellent color? Is the P6K just the thing to get?

29 minutes ago, Gabriel Devereux said:

I do believe if they were to capture all RGB values at 'full resolution', sampling all three values at the same sensitivity, the image produced would look a lil weird. 

I have a DP2 Merrill from Sigma, which has a Foveon (stacked) sensor. The tonality of the images is gorgeous and incredibly sharp and crisp. It's night and day compared with Bayer and there is no aliasing despite the lack of a low pass filter. But there are quirks with color and sensitivity, to be polite about it. But for certain landscapes the images are beautiful with the Foveon rather than Bayer sensor. 

However film is not really that sharp. Isn't the red layer softest? I always liked how skin looks with the Alexa (when it comes to digital cameras) despite the Alexa having relatively few red photo sites and I don't like the look of over-sharpened skin.

Link to comment
Share on other sites

  • Premium Member
44 minutes ago, M Joel W said:

However film is not really that sharp. Isn't the red layer softest? 

This is a good point. If I recall correctly, the red was the lowest level of the emulsion above the clear and the remjet on the bottom. Fast and slow blue was on top with yellow filters between, and the green all before the red. Much of the digital footage I see seems like the red is the strongest for some reason. I am not going to pretend to understand the algorithms of how digital cameras process data but the red (and in the case of Sony, the magenta) shift is real.

Link to comment
Share on other sites

  • Premium Member
7 hours ago, Gabriel Devereux said:

Correct me if I'm wrong -
The Bayer CFA exists with twice the amount of green photosites - 'sensors' because our eyes - our photopic vision is more sensitive to the green (550nm or about, frequency band). 

Yep exactly, that was the idea. But sadly the Bayer pattern has a lot of issues like the zipper effect and false color artifacts, which I call aliasing because it's basically the same thing. 

Edited by Tyler Purcell
Link to comment
Share on other sites

  • Premium Member
6 hours ago, M Joel W said:

Does the P4K really have worse color than the P6K? I'm very, very impressed by the P6K's image both technically and aesthetically but the sensor is a bit too small for my S35 and FF lenses and too big for my S16 lenses. To me a P4K (and a speed booster XL or regular depending on the lens set) seems more viable. Also 6K is too heavy for the post work I do, though of course I could downscale. What, if any, other cameras in this price bracket have excellent color? Is the P6K just the thing to get?

Yep it sure does! The 4k imager is an oddball because it has a lot of weird issues the 6k just doesn't have and the 6k has a far worse rolling shutter effect. So it's a tossup really. If the pocket 6k had a better rolling shutter than the 4k AND was a true S35mm imager, I'd own one. But the rolling shutter issues kill me because I love to do hand held work and it's horribly noticeable, especially in full-imager capture, which is the whole point of buying a 6k in the first place. 

IDK why the imager would be too small for S35 and FF lenses, that seems very odd. 

Too heavy for post work? You mean the file sizes are too large? I mean the raw codec is much smaller than the "industry standard" pro res 444 codec that people shoot with the Alexa and honestly, even Red's raw codec isn't much smaller. I think BMD have done a great job with their own compressed codec, all be it, with the handicaps they were given through the Red patent issues. I will take BMRaw any day of the week over anything .h264/.h265 which is what nearly all of the other cameras shoot natively, just gussied up with names like XAVC. 

The Pocket 6k PRO is kind of the camera to get. The non-pro, wouldn't get near with a 10 foot pole. The PRO is a much better camera over-all, fixes many of the issues. 

Link to comment
Share on other sites

  • Premium Member
7 hours ago, M Joel W said:

However film is not really that sharp. Isn't the red layer softest? I always liked how skin looks with the Alexa (when it comes to digital cameras) despite the Alexa having relatively few red photo sites and I don't like the look of over-sharpened skin.

People consider film soft because A) they're using shitty lenses and B) their camera/scanner aren't calibrated. A high resolution scan of 35mm, is very crisp. Looks just like an Alexa honestly. The original Blackmagic pocket isn't crisper than 16mm properly captured. 

Yes the final layers are red. They actually use UV filters between the layers, which is why the layering system is kind of funky. They capture the most difficult colors first; blue and green, then they capture red last, with special UV filters between each layer. The red layer may seem like it's far away, but the entire emulsion is .00003" thick. So no, the red layer is not necessarily "softer" then the other layers. If anything the color red in general is one of the good things about Kodak color negative. 

I agree, the Alexa is probably the only camera that matches film in terms of skin tones. Arri achieved this by using huge photosites and unbelievably accurate color science. It's amazing what they've achieved, considering nobody else has gotten close, yet. They've also figured out how to make the image soft, without losing resolution. It's quite amazing working with Alexa images one day and then Red images the next, how entirely different the smoothing is between the two imagers. I shoot a lot of BMD cameras as well and it's one of the things I really like about the BMD cameras in general, they do offer some very good smoothing, it helps make the image less crisp. Ya just don't see that in lower end cameras much. 

Link to comment
Share on other sites

Maybe I phrased that poorly. The lens sets I have are S35, full frame, and S16. The 6K sensor lacks an S16 mode and has a narrower field of view with the others than the P4K with an appropriate speed booster. I meant 6K render times are slow.

I might be totally off base on the red channel/film thing. I just toggled between a 4K S16 scan (presumably 7219) and some log Alexa footage and of course the Alexa footage is cleaner but the noise texture of the film is similar for each channel with the red channel being cleanest whereas the Alexa footage has a cleaner green channel and finer noise texture in green, too. 

What issues are there with P4K color specifically? I'm still leaning toward that. In theory to be used on its own and occasionally to mix with S16 and Alexa footage, so I want something where there are transforms to get into Log C.

I find Alexa footage very green overall but also prefer Alexa skin tones to other digital cameras. But the Venice and P6K impress me too. I liked the OG C300's skin tones a lot for caucasian skin, less so for darker skin tones. Red Dragon I think has a fine look too.

Edited by M Joel W
Link to comment
Share on other sites

  • Premium Member
3 minutes ago, M Joel W said:

What issues are there with P4K color specifically? I'm still leaning toward that. In theory to be used on its own and occasionally to mix with S16 and Alexa footage, so I want something where there are transforms to get into Log C.

It just looks like a Sony camera. If you like that look, go for it. But I don't own Sony cameras for that reason. They always tend to be very super hyper white, they don't resolve skin tones properly. 

3 minutes ago, M Joel W said:

I find Alexa footage very green overall but also prefer Alexa skin tones to other digital cameras. But the Venice and P6K impress me too. I liked the OG C300's skin tones a lot for caucasian skin, less so for darker skin tones. Red Dragon I think has a fine look too.

I think many things look green when they're ungraded. When you grade Alexa footage it comes out very nice. 

Edited by Tyler Purcell
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...