New Ursa 12k sensor tech

The place for questions about shooting with Blackmagic Cameras.
  • Author
  • Message
Offline

Jean Paul Sneider

  • Posts: 264
  • Joined: Sat Jul 01, 2017 5:46 pm

New Ursa 12k sensor tech

PostSat Jul 18, 2020 12:36 pm

I can't find a reliable source that speaks about the new sensor in detail. Can someone point me somewhere? It is said to be a non Bayer sensor with equal r g and be photosite but that's are there more details out there about the tech side of it?
Someone suggested a RGBW sensor but without pointing at any source.
Offline
User avatar

Robert Niessner

  • Posts: 4946
  • Joined: Thu Feb 21, 2013 9:51 am
  • Location: Graz, Austria

Re: New Ursa 12k sensor tech

PostSat Jul 18, 2020 1:14 pm

John Brawley talks a bit about this in detail. Beyond someone from BMD directly, he is the most reliable source you will find.

Brand new sensor, 3 years in the making.

79 MP.

Native 800 iso.

14 stops (that’s probably a bit conservative, they haven’t been able to properly check it because the models are based on Bayer sensors...:-)

It’s not Bayer, but it has a very small pixel pitch of 2.2 microns. (Alexa is 8)

Instead of Bayer 2x2 grid of GRBG it has a 6x6 grid. 6G, 6B and 6R plus 18 W pixels.

The W are clear or “white” pixels. This overcomes the reduced sensitivity issue of a 2.2 micron pitch.

Combined with their codec that is profiled to the sensor, it enables other modes of operation. This what get’s you 12k, 8K 4K RAW without changing field of view. There’s also a 6K super 16 mode.


Source: https://cml.news/g/cml-general/message/6815

There is also a presentation of Image Algorithmics on RGBW color filter common misconceptions:

http://image-sensors-world.blogspot.com ... ilter.html
Saying "Thx for help!" is not a crime.
--------------------------------
Robert Niessner
LAUFBILDkommission
Graz / Austria
--------------------------------
Blackmagic Camera Blog (German):
http://laufbildkommission.wordpress.com

Read the blog in English via Google Translate:
http://tinyurl.com/pjf6a3m
Offline

Jean Paul Sneider

  • Posts: 264
  • Joined: Sat Jul 01, 2017 5:46 pm

Re: New Ursa 12k sensor tech

PostSat Jul 18, 2020 1:21 pm

Thank you!
Offline
User avatar

Steve Holmlund

  • Posts: 512
  • Joined: Fri Jun 28, 2013 11:30 pm
  • Location: Montara, California

Re: New Ursa 12k sensor tech

PostSat Jul 18, 2020 5:40 pm

Robert Niessner wrote:
There is also a presentation of Image Algorithmics on RGBW color filter common misconceptions:

http://image-sensors-world.blogspot.com ... ilter.html


Very fortuitous timing of this article.

Steve


Sent from my iPhone using Tapatalk Pro
Steve Holmlund
Hobbyist
BMPCC, vintage Rokkor lenses, Olympus 12-40 and 12-100, Panasonic 100-300 II
SmallHD Focus, i7 8700k / GTX 1080
Offline

Howard Roll

  • Posts: 2506
  • Joined: Fri Jun 03, 2016 7:50 am

Re: New Ursa 12k sensor tech

PostSat Jul 18, 2020 10:23 pm

Here's a link to the patent doc that likely relates to this sensor and it's IP.

https://patents.google.com/patent/US201 ... r&sort=new

Good Luck
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostSun Jul 19, 2020 1:37 am

Robert Niessner wrote:John Brawley talks a bit about this in detail. Beyond someone from BMD directly, he is the most reliable source you will find.

Brand new sensor, 3 years in the making.

79 MP.

Native 800 iso.

14 stops (that’s probably a bit conservative, they haven’t been able to properly check it because the models are based on Bayer sensors...:-)

It’s not Bayer, but it has a very small pixel pitch of 2.2 microns. (Alexa is 8)

Instead of Bayer 2x2 grid of GRBG it has a 6x6 grid. 6G, 6B and 6R plus 18 W pixels.

The W are clear or “white” pixels. This overcomes the reduced sensitivity issue of a 2.2 micron pitch.

Combined with their codec that is profiled to the sensor, it enables other modes of operation. This what get’s you 12k, 8K 4K RAW without changing field of view. There’s also a 6K super 16 mode.


Source: https://cml.news/g/cml-general/message/6815

There is also a presentation of Image Algorithmics on RGBW color filter common misconceptions:

http://image-sensors-world.blogspot.com ... ilter.html


That article didn't even cover the Kodak pattern, which just replaced one of the Bayer greens with white.
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline

Howard Roll

  • Posts: 2506
  • Joined: Fri Jun 03, 2016 7:50 am

Re: New Ursa 12k sensor tech

PostSun Jul 19, 2020 2:54 am

tldr

Patent.png
Patent.png (62.82 KiB) Viewed 12931 times


Good Luck
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostSun Jul 19, 2020 3:13 am

Thanks. That's similar to what I figured out.
Last edited by Wayne Steven on Sun Jul 19, 2020 6:44 am, edited 1 time in total.
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline

Mark Grgurev

  • Posts: 802
  • Joined: Fri Nov 03, 2017 7:22 am

Re: New Ursa 12k sensor tech

PostSun Jul 19, 2020 4:32 am

Thanks, Howard! Think I might do a little simulation in Fusion to see how well that pattern recreates the underlying image. :-)
Offline

nikolayatanasov

  • Posts: 13
  • Joined: Mon Nov 17, 2014 12:01 pm

Re: New Ursa 12k sensor tech

PostWed Jul 22, 2020 7:01 am

With so small pixel size there should be a lot of diffraction
Also what lens will support such a resolution?

Sent from my SM-N910C using Tapatalk
Offline
User avatar

Pawel Achtel

  • Posts: 15
  • Joined: Fri Jul 06, 2018 4:29 am
  • Real Name: Pawel Achtel

Re: New Ursa 12k sensor tech

PostSun Oct 25, 2020 8:30 am

With so small pixel size there should be a lot of diffraction
Also what lens will support such a resolution?


To answer this, one has to understand sampling theory and go through the patent application with a fine comb. This camera samples full colour in a cluster of 6 x 6 (36) photosites - 9 times more than that for Bayer pattern CFA (2x2 = 4). It means that it will not resolve colour detail finer than an equivalent approx 3K Bayer pattern CFA or else you will end up with colour aliasing.

The new BMD RGBW 12K CFA pattern was designed to extend the dynamic range where White (in patent referred as wideband) photosites are more sensitive than aggregated R, G, B cluster of photosite, which is used to handle highlights. This explains why we also see significant noise in mid-tones, but not as much in the shadows as they are derived from White photosites.

The bottom line is that this camera does indeed have measured 14 stops of dynamic range, but it also has "muddy" and noisy mid-tones and sharpness of approx. 3k ~ 3.5K Bayer pattern camera. It is a relatively low resolution camera with high-resolution sensor, if that makes sense.

So, to answer your questions: soft lenses would work fine :)
Pawel Achtel ACS B.Eng.(Hons) M.Sc.
“Sharp to the Edge”

www.achtel.com
Offline

nikolayatanasov

  • Posts: 13
  • Joined: Mon Nov 17, 2014 12:01 pm

Re: New Ursa 12k sensor tech

PostSun Oct 25, 2020 8:37 am

Pawel Achtel wrote:
With so small pixel size there should be a lot of diffraction
Also what lens will support such a resolution?


To answer this, one has to understand sampling theory and go through the patent application with a fine comb. This camera samples full colour in a cluster of 6 x 6 (36) photosites - 9 times more than that for Bayer pattern CFA (2x2 = 4). It means that it will not resolve colour detail finer than an equivalent approx 3K Bayer pattern CFA or else you will end up with colour aliasing.

The new BMD RGBW 12K CFA pattern was designed to extend the dynamic range where White (in patent referred as wideband) photosites are more sensitive than aggregated R, G, B cluster of photosite, which is used to handle highlights. This explains why we also see significant noise in mid-tones, but not as much in the shadows as they are derived from White photosites.

The bottom line is that this camera does indeed have measured 14 stops of dynamic range, but it also has "muddy" and noisy mid-tones and sharpness of approx. 3k ~ 3.5K Bayer pattern camera. It is a relatively low resolution camera with high-resolution sensor, if that makes sense.

So, to answer your questions: soft lenses would work fine :)
Thanks for the info Pawel Achtel! I should go trough that pattent explanation to get better understanding but overal what you're saying makes sense. Thanks!

Sent from my SM-N910C using Tapatalk
Offline

Howard Roll

  • Posts: 2506
  • Joined: Fri Jun 03, 2016 7:50 am

Re: New Ursa 12k sensor tech

PostSun Oct 25, 2020 3:21 pm

Pawel Achtel wrote: It means that it will not resolve colour detail finer than an equivalent approx 3K Bayer pattern CFA or else you will end up with colour aliasing.


I think you mean 6K Bayer. Why is arithmetic the bane of sensor wizards?

Good Luck
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostSun Oct 25, 2020 4:41 pm

They calculate missing colour too, and while it may not be authentic all the time, it works well enough, but I do pick up on things here. A number of samples the colour looks sparse, and when they boost up it losses more colour definition. 4:4:4 is still great, but we have been let down in vertical colour sensors.
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline

John Brawley

  • Posts: 4267
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: Los Angeles California

Re: New Ursa 12k sensor tech

PostSun Oct 25, 2020 9:18 pm

Howard Roll wrote:
I think you mean 6K Bayer. Why is arithmetic the bane of sensor wizards?

Good Luck


And the way I read the patent, W pixels can also be used in the colour difference / interpolation as well. Colour isn't only generated from the RGB pixels, it's from RGBW.

JB
John Brawley ACS
Cinematographer
Currently - Los Angeles
Offline
User avatar

Robert Niessner

  • Posts: 4946
  • Joined: Thu Feb 21, 2013 9:51 am
  • Location: Graz, Austria

Re: New Ursa 12k sensor tech

PostSun Oct 25, 2020 11:31 pm

Pawel Achtel wrote:
With so small pixel size there should be a lot of diffraction
Also what lens will support such a resolution?


To answer this, one has to understand sampling theory and go through the patent application with a fine comb. This camera samples full colour in a cluster of 6 x 6 (36) photosites - 9 times more than that for Bayer pattern CFA (2x2 = 4). It means that it will not resolve colour detail finer than an equivalent approx 3K Bayer pattern CFA or else you will end up with colour aliasing.

The new BMD RGBW 12K CFA pattern was designed to extend the dynamic range where White (in patent referred as wideband) photosites are more sensitive than aggregated R, G, B cluster of photosite, which is used to handle highlights. This explains why we also see significant noise in mid-tones, but not as much in the shadows as they are derived from White photosites.

The bottom line is that this camera does indeed have measured 14 stops of dynamic range, but it also has "muddy" and noisy mid-tones and sharpness of approx. 3k ~ 3.5K Bayer pattern camera. It is a relatively low resolution camera with high-resolution sensor, if that makes sense.

So, to answer your questions: soft lenses would work fine :)


Sorry Pawel to spoil your beans:

This is a RGBW algorithm from 2018:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5982632/

That's about "Sensitivity and Resolution Improvement in RGBW Color Filter Array Sensor" through a clever post processing mode based on texture decomposition.

Comparison:
RGBW CI W = demosaiced clear (white) pixel only

IMG_0682.JPG
Fig. 11
IMG_0682.JPG (54.73 KiB) Viewed 12077 times


IMG_0683.JPG
Fig. 13
IMG_0683.JPG (74.85 KiB) Viewed 12077 times
Saying "Thx for help!" is not a crime.
--------------------------------
Robert Niessner
LAUFBILDkommission
Graz / Austria
--------------------------------
Blackmagic Camera Blog (German):
http://laufbildkommission.wordpress.com

Read the blog in English via Google Translate:
http://tinyurl.com/pjf6a3m
Offline

Howard Roll

  • Posts: 2506
  • Joined: Fri Jun 03, 2016 7:50 am

Re: New Ursa 12k sensor tech

PostSun Oct 25, 2020 11:41 pm

John Brawley wrote:And the way I read the patent, W pixels can also be used in the colour difference / interpolation as well. Colour isn't only generated from the RGB pixels, it's from RGBW.

JB


Interpolated color is stored in the w pixel position, however the w pixels are zero weighted in the 7x7 chrominance image filter kernel.

Good Luck
Offline
User avatar

Pawel Achtel

  • Posts: 15
  • Joined: Fri Jul 06, 2018 4:29 am
  • Real Name: Pawel Achtel

Re: New Ursa 12k sensor tech

PostMon Oct 26, 2020 12:21 am

Robert Niessner wrote:Sorry Pawel to spoil your beans:

This is a RGBW algorithm from 2018:


You are not "spoiling my beans", you are just changing the subject :D

I'm talking of actual performance of the camera, not what theoretical experimentation might have been during research and development of it.

The images from the production camera exhibit significant colour aliasing and they do so at less than 4K (2000 line pairs per picture width) spatial frequency response, meaning: this sensor cannot produce 4K or smaller detail without significant colour aliasing (if you increase the spatial frequency, the aliasing will become worse)

Image

Here is the link to source test frame published by Geoff Boyle at CML: https://cinematography.net/BMD-12K.html

This sample frame does not have any detail passed approx. 3,500 LW/PW (1,750 line pairs per picture width), meaning: a 3.5K Bayer sensor camera can produce sharper images than this.

If you have actual test images from BMD Ursa 12K that would show finer detail, please post.
Pawel Achtel ACS B.Eng.(Hons) M.Sc.
“Sharp to the Edge”

www.achtel.com
Offline
User avatar

Pawel Achtel

  • Posts: 15
  • Joined: Fri Jul 06, 2018 4:29 am
  • Real Name: Pawel Achtel

Re: New Ursa 12k sensor tech

PostMon Oct 26, 2020 12:54 am

John Brawley wrote:And the way I read the patent, W pixels can also be used in the colour difference / interpolation as well. Colour isn't only generated from the RGB pixels, it's from RGBW.


John, it is quite clear that the algorithm uses a cluster of 6 x 6 (36) photosites to determine full colour. This is defined by pattern "51" in the patent and described as:

"Filter array 51 also has 16 . 7 % green photosites . This enables
full color information to be derived from the CFA "


https://patentimages.storage.googleapis.com/88/08/39/428dbbc9e5dfca/US20190306472A1.pdf

This is nine (9) times larger than that of Bayer pattern CFA 2x2= 4.
Indeed, this is exactly what I'm measuring in sample test images published by Geoff Boyle. Those images do not have any detail (spatial frequency) passed about 3.5K Bayer pattern equivalent.

Image

Do you have any test samples (say ISO 12233 Test Chart) that would have such frequency? If so, I'd like to see them.
Pawel Achtel ACS B.Eng.(Hons) M.Sc.
“Sharp to the Edge”

www.achtel.com
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostMon Oct 26, 2020 2:09 am

Pawel Achtel wrote:
John Brawley wrote:And the way I read the patent, W pixels can also be used in the colour difference / interpolation as well. Colour isn't only generated from the RGB pixels, it's from RGBW.


John, it is quite clear that the algorithm uses a cluster of 6 x 6 (36) photosites to determine full colour. This is defined by pattern "51" in the patent and described as:

"Filter array 51 also has 16 . 7 % green photosites . This enables
full color information to be derived from the CFA "


https://patentimages.storage.googleapis.com/88/08/39/428dbbc9e5dfca/US20190306472A1.pdf

This is nine (9) times larger than that of Bayer pattern CFA 2x2= 4.
Indeed, this is exactly what I'm measuring in sample test images published by Geoff Boyle. Those images do not have any detail (spatial frequency) passed about 3.5K Bayer pattern equivalent.

Image

Do you have any test samples (say ISO 12233 Test Chart) that would have such frequency? If so, I'd like to see them.


No need to pick on John, my earlier post on the same subject was directly above his. These things don't matter as much as one might think. If you apply a sophisticated enough view of colour science, you can pick up colour information from the shifts between pixels related to the overlap in colour spaces. That might be techno babel to those without ability, but gold to those with. Colour tends to physically pool in real life compared to detail. So, this gives more opportunity to work out colour within the pixel space. So these effects allow for missing colour to be more accurately predicted than in interpolation. However, the configuration of colour information is sparse here, requiring the more advanced techniques described, maybe that will help with what I'm seeing too. So, yes a reasonable colour in the white pixels "could" be calculated. I don't know why it's coming out the way it does, or any product does, I don't work there.
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline

lost_soul

  • Posts: 239
  • Joined: Fri May 31, 2019 12:30 am
  • Location: Vancouver
  • Real Name: Shawn Metheny

Re: New Ursa 12k sensor tech

PostMon Oct 26, 2020 7:52 pm

Pawel Achtel wrote:
Robert Niessner wrote:
The images from the production camera exhibit significant colour aliasing and they do so at less than 4K (2000 line pairs per picture width) spatial frequency response, meaning: this sensor cannot produce 4K or smaller detail without significant colour aliasing (if you increase the spatial frequency, the aliasing will become worse)

Image

Here is the link to source test frame published by Geoff Boyle at CML: https://cinematography.net/BMD-12K.html



Looking at the video I see when it switches to "Zero with filter 1" that the moire disappears. Is filter 1 a low pass filter?

At plus 4.5 stops the magenta explodes? Is that jacking the ISO to the second gain stage and that stage is producing all that magenta? Is it clipping one of the color channels or is some thing else at play?
Centos 8
RX 580
AMDGPU-pro
Studio (What ever the latest is)
Offline

ChristopherSeguine

  • Posts: 89
  • Joined: Thu May 02, 2013 5:00 pm
  • Location: California

Re: New Ursa 12k sensor tech

PostMon Oct 26, 2020 8:26 pm

Pawel Achtel wrote:it is quite clear that the algorithm uses a cluster of 6 x 6 (36) photosites to determine full colour.


Red channel has an exponential amount of noise even though there is an equal amount of r,g,b pixels - compared to bayer cfa. why?

Diffraction limit on a bayer sensor is first visible in softness of green channel, with this sensor r,g,b would be equally effected by the diffraction limit? how do the the W pixels effect the diffraction limit?

Can't judge sensor resolving power by this, test metadata says 32mm cooke mini s4i - aka panchro
I doubt the coated version of that lens at 5.6 can resolve more than 1300 LW/PH in the center.
Its a fine lens if you like soft warm 70s look, not if you like sharp and clean.
I think your seeing a lens limit, not a sensor limit here.

Metadata also says Compression: Q0
Didn't someone do a test on the old camera that showed Q0 was softer than the comparable constant bit rate when it came to static objects - like test charts?
Still waiting for the test with an Outs 85 or sigma art 105 at f4 of a chart in both 12k q0 and 5:1
John - isnt Canada kinda booring during quarantine? Cant you knock that out between takes?
Primary: TRX40 Xtreme (3970x), Decklink Extreme 12G, 2x RTX3090, Areca Raid
Primary source: Red 8k .r3d, Phantom .cine, secondary footage: ArriRaw, DJI X5 CDNG
Offline
User avatar

Pawel Achtel

  • Posts: 15
  • Joined: Fri Jul 06, 2018 4:29 am
  • Real Name: Pawel Achtel

Re: New Ursa 12k sensor tech

PostMon Oct 26, 2020 8:49 pm

ChristopherSeguine wrote:
Red channel has an exponential amount of noise even though there is an equal amount of r,g,b pixels - compared to bayer cfa. why?


The algorithm uses White (wideband) photosites for reproduction of shadows, this is why shadows are relatively clean. But, it uses R, G and B (narrowband) photosites for mid-tones and highlights. Since Narrowband (r, G and B) pixels are quite small and sparse, their low light ability is somewhat compromised. This is why, I think, all mid-tones appear "muddy" and noisy and are actually noisier than the shadows.

ChristopherSeguine wrote:Can't judge sensor resolving power by this, test metadata says 32mm cooke mini s4i - aka panchro
I doubt the coated version of that lens at 5.6 can resolve more than 1300 LW/PH in the center.
Its a fine lens if you like soft warm 70s look, not if you like sharp and clean.
I think your seeing a lens limit, not a sensor limit here.


The fact that we are getting Moire pattern on the zone chart tells me that it is the sampling that is limiting the sharpness, not the lens. This is how Moire forms.
Pawel Achtel ACS B.Eng.(Hons) M.Sc.
“Sharp to the Edge”

www.achtel.com
Offline

John Brawley

  • Posts: 4267
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: Los Angeles California

Re: New Ursa 12k sensor tech

PostTue Oct 27, 2020 3:35 am

ChristopherSeguine wrote:John - isnt Canada kinda booring during quarantine? Cant you knock that out between takes?



I’m actually shooting a series in LA. Bit busy.

Best I can offer this very hastily shot test done with a pre-production 12k.

Two very well known 135 format lens brands featured. Very different results.

JB

John Brawley ACS
Cinematographer
Currently - Los Angeles
Offline
User avatar

Pawel Achtel

  • Posts: 15
  • Joined: Fri Jul 06, 2018 4:29 am
  • Real Name: Pawel Achtel

Re: New Ursa 12k sensor tech

PostTue Oct 27, 2020 3:48 am

John Brawley wrote:Best I can offer this very hastily shot test done with a pre-production 12k.


Thanks, John!
Any chance you could pull one BRAW frame from the lens "A"?
I can quite easily derive MTF chart from it.
Pawel Achtel ACS B.Eng.(Hons) M.Sc.
“Sharp to the Edge”

www.achtel.com
Offline

lost_soul

  • Posts: 239
  • Joined: Fri May 31, 2019 12:30 am
  • Location: Vancouver
  • Real Name: Shawn Metheny

Re: New Ursa 12k sensor tech

PostWed Oct 28, 2020 9:59 pm

That is really eye opening. Is the noise always there and we only perceive it when the chart gets washed out or is one lens actually contributing to the noise?

Come on John, don't leave us hanging. Who is lens A and who is lens B?
Centos 8
RX 580
AMDGPU-pro
Studio (What ever the latest is)
Offline

John Brawley

  • Posts: 4267
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: Los Angeles California

Re: New Ursa 12k sensor tech

PostWed Oct 28, 2020 10:02 pm

lost_soul wrote:
Come on John, don't leave us hanging. Who is lens A and who is lens B?


It wouldn’t be fair to say which is which. This wasn’t a “controlled” test. Both are well known “brands”

JB
John Brawley ACS
Cinematographer
Currently - Los Angeles
Offline
User avatar

robedge

  • Posts: 2617
  • Joined: Tue Dec 24, 2019 1:24 am
  • Location: U.S.
  • Real Name: Rob Edge

Re: New Ursa 12k sensor tech

PostFri Oct 30, 2020 5:25 pm

ProAv:


Online

Andrew Kolakowski

  • Posts: 9207
  • Joined: Tue Sep 11, 2012 10:20 am
  • Location: Poland

Re: New Ursa 12k sensor tech

PostFri Oct 30, 2020 7:57 pm

Robert Niessner wrote:
Pawel Achtel wrote:
With so small pixel size there should be a lot of diffraction
Also what lens will support such a resolution?


To answer this, one has to understand sampling theory and go through the patent application with a fine comb. This camera samples full colour in a cluster of 6 x 6 (36) photosites - 9 times more than that for Bayer pattern CFA (2x2 = 4). It means that it will not resolve colour detail finer than an equivalent approx 3K Bayer pattern CFA or else you will end up with colour aliasing.

The new BMD RGBW 12K CFA pattern was designed to extend the dynamic range where White (in patent referred as wideband) photosites are more sensitive than aggregated R, G, B cluster of photosite, which is used to handle highlights. This explains why we also see significant noise in mid-tones, but not as much in the shadows as they are derived from White photosites.

The bottom line is that this camera does indeed have measured 14 stops of dynamic range, but it also has "muddy" and noisy mid-tones and sharpness of approx. 3k ~ 3.5K Bayer pattern camera. It is a relatively low resolution camera with high-resolution sensor, if that makes sense.

So, to answer your questions: soft lenses would work fine :)


Sorry Pawel to spoil your beans:

This is a RGBW algorithm from 2018:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5982632/

That's about "Sensitivity and Resolution Improvement in RGBW Color Filter Array Sensor" through a clever post processing mode based on texture decomposition.

Comparison:
RGBW CI W = demosaiced clear (white) pixel only

IMG_0682.JPG


IMG_0683.JPG


This would explain all issues in URSA 12K with moire. Looks like it requires new approach to debayering, so BM has to make it happen in BRAW SDK.
Offline

Dan Cotreau

  • Posts: 103
  • Joined: Wed Aug 12, 2020 8:43 pm
  • Location: Los Angeles
  • Real Name: Dan Cotreau

Re: New Ursa 12k sensor tech

PostTue Nov 03, 2020 9:39 pm

I thought I would share this article discussing the new sensor in the 12K Ursa Mini Pro.

https://nofilmschool.com/blackmagic-ursa-mini-pro-12k-review
Offline
User avatar

rick.lang

  • Posts: 17173
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 2:08 pm

Thanks, Dan. Very positive but balanced review.
Rick Lang
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 2:59 pm

The Germans couldn't find a lens that was good enough.
https://www.slashcam.de/artikel/Test/Der-Sensor-der--Blackmagic-URSA-Mini-Pro-12K---Demosaicing-und-Rolling-Shutter.html

Carsten will give you the google translate version ;).
Offline

John Brawley

  • Posts: 4267
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: Los Angeles California

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 3:15 pm



Interesting that they also did not see moire / aliasing issues on a test chart designed to provoke moire.

JB
John Brawley ACS
Cinematographer
Currently - Los Angeles
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 4:13 pm

John Brawley wrote:


Interesting that they also did not see moire / aliasing issues on a test chart designed to provoke moire.

JB


Not really, their lens worked as an OLPF with these kind of resolutions.
Offline

John Brawley

  • Posts: 4267
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: Los Angeles California

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 4:23 pm

MishaEngel wrote:
John Brawley wrote:


Interesting that they also did not see moire / aliasing issues on a test chart designed to provoke moire.

JB


Not really, their lens worked as an OLPF with these kind of resolutions.


They didn’t seem to specify which lens and which aperture and what kind of crops those images were.

Somehow others have invoked moire so they must have have found a lens “good” enough, though I’d be surprised if slashcam couldn’t manage it.

By the way, I have also shot log / zone plates with “good” lenses within diffraction and not seen major moire so I’m just trying to figure out how some are and some aren’t.

JB
John Brawley ACS
Cinematographer
Currently - Los Angeles
Offline
User avatar

Robert Niessner

  • Posts: 4946
  • Joined: Thu Feb 21, 2013 9:51 am
  • Location: Graz, Austria

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 10:45 pm

The question is, if the internal NDs do play a role in provoking moire or not.
Saying "Thx for help!" is not a crime.
--------------------------------
Robert Niessner
LAUFBILDkommission
Graz / Austria
--------------------------------
Blackmagic Camera Blog (German):
http://laufbildkommission.wordpress.com

Read the blog in English via Google Translate:
http://tinyurl.com/pjf6a3m
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 10:52 pm

MishaEngel wrote:
John Brawley wrote:


Interesting that they also did not see moire / aliasing issues on a test chart designed to provoke moire.

JB


Not really, their lens worked as an OLPF with these kind of resolutions.


+1 you read my mind Misha and I haven't even been there yet. :)
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostWed Nov 04, 2020 10:53 pm

Robert Niessner wrote:The question is, if the internal NDs do play a role in provoking moire or not.


Good point Robert. How are they structured these days?
Last edited by Wayne Steven on Thu Nov 05, 2020 7:16 am, edited 1 time in total.
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline
User avatar

Uli Plank

  • Posts: 21278
  • Joined: Fri Feb 08, 2013 2:48 am
  • Location: Germany and Indonesia

Re: New Ursa 12k sensor tech

PostThu Nov 05, 2020 1:11 am

Those Slashcam guys have tested quite a few cameras and they wrote:
"Eine besseres 4K-Debayering haben wir noch nicht gesehen." (We have not yet seen any better 4K-Debayering, but with a grammar error in German.)

Well, they should have tried a 60mm f2.8 S-Planar by Zeiss @ f5.6, I could provoke chroma moiré with that on a Red Dragon.
But they mention "handelsübliche EF-Objektiv" (ordinary EF lens), which is probably what they tried.
No, an iGPU is not enough, and you can't use HEVC 10 bit 4:2:2 in the free version.

Studio 18.6.5, MacOS 13.6.5
MacBook M1 Pro, 16 GPU cores, 32 GB RAM and iPhone 15 Pro
Speed Editor, UltraStudio Monitor 3G, iMac 2017, eGPU
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: New Ursa 12k sensor tech

PostThu Nov 05, 2020 1:07 pm

Uli Plank wrote:But they mention "handelsübliche EF-Objektiv" (ordinary EF lens), which is probably what they tried.


The Sigma 135 f1.8(~ $1.400) at f4 is pretty good.
https://www.lensrentals.com/blog/2017/07/experiments-for-ultra-high-resolution-camera-sensors/
Offline

Howard Roll

  • Posts: 2506
  • Joined: Fri Jun 03, 2016 7:50 am

Re: New Ursa 12k sensor tech

PostThu Nov 05, 2020 8:29 pm

Can this Sudoku resolve more than 6K?

Good Luck

Sudoku.jpg
Sudoku.jpg (423.48 KiB) Viewed 10963 times
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostThu Nov 05, 2020 10:29 pm

One can resolve an image without any pixels, by imaging and estimating it. Funny maybe, but that's the sort of process of resolving missing colours here. So, yes, it could resolve 12k, it is not always going be an accurate pixel. A lot of the time, there maybe deviations.
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline
User avatar

Uli Plank

  • Posts: 21278
  • Joined: Fri Feb 08, 2013 2:48 am
  • Location: Germany and Indonesia

Re: New Ursa 12k sensor tech

PostFri Nov 06, 2020 12:55 am

Who cares if the pictures are so pretty? As Steve Yedlin already said: "We don't need more pixels, but better pixels."
Don't confuse sensels with pixels.
No, an iGPU is not enough, and you can't use HEVC 10 bit 4:2:2 in the free version.

Studio 18.6.5, MacOS 13.6.5
MacBook M1 Pro, 16 GPU cores, 32 GB RAM and iPhone 15 Pro
Speed Editor, UltraStudio Monitor 3G, iMac 2017, eGPU
Offline

John Brawley

  • Posts: 4267
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: Los Angeles California

Re: New Ursa 12k sensor tech

PostFri Nov 06, 2020 1:57 am

Uli Plank wrote:Who cares if the pictures are so pretty? As Steve Yedlin already said: "We don't need more pixels, but better pixels."


Indeed.

Resolution is but one metric that goes into the creation of an image.

JB
John Brawley ACS
Cinematographer
Currently - Los Angeles
Offline

Wayne Steven

  • Posts: 3362
  • Joined: Thu Aug 01, 2013 3:58 am
  • Location: Earth

Re: New Ursa 12k sensor tech

PostFri Nov 06, 2020 6:57 am

Well, that means more accurate or more artistic.
As for accurate, not this setup, as I just said. If only 3 colour technology, like foveon X3, had gone forwards, it might be descent. Video use had been neglected. And I imagine that is something they were trying to correct in the lmount sensor that got scrapped. They might have not put in enough work to catch up and retain a good image in both stills and video at the same time. Sad they had to scrap it.
aIf you are not truthfully progressive, maybe you shouldn't say anything
bTruthful side topics in-line with or related to, the discussion accepted
cOften people deceive themselves so much they do not understand, even when the truth is explained to them
Offline

MishaEngel

  • Posts: 1432
  • Joined: Wed Aug 29, 2018 12:18 am
  • Real Name: Misha Engel

Re: New Ursa 12k sensor tech

PostSat Nov 14, 2020 4:20 pm

Some more info https://www.slashcam.de/artikel/Test/URSA-Mini-Pro-12K---Dynamik-und-Moir-es.html

8k full sensor readout looks to be the sweet spot, let's see if we can rent it in the coming weeks.
Offline
User avatar

rick.lang

  • Posts: 17173
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

New Ursa 12k sensor tech

PostSat Nov 14, 2020 5:53 pm

Interesting with iOS 14.2, I’m supposed to be able to translate to English when I select the aA control in the URL but it doesn’t do it.

Edit, the translation to English worked when I pasted to Safari but not when I was using Tapatalk as my front end.
Rick Lang
Offline
User avatar

rick.lang

  • Posts: 17173
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

New Ursa 12k sensor tech

PostSat Nov 14, 2020 6:28 pm

I’m concluding 4K in camera with full field of view is an excellent option and what I’d likely shoot on the UMP12K.

Edit
Removed my erroneous comment.
Last edited by rick.lang on Sun Nov 15, 2020 12:40 am, edited 1 time in total.
Rick Lang
Offline

Mark Wyatt

  • Posts: 242
  • Joined: Wed Sep 25, 2013 4:36 pm
  • Location: Vancouver

Re: New Ursa 12k sensor tech

PostSat Nov 14, 2020 6:33 pm

I'm glad Slashcam has posted this, as this is exactly what I am finding as well regarding moire (ie- there isn't any if dealt with properly in post). Too bad "the youtubers" didn't fully test their process.
www.wyattvisuals.com
Offline
User avatar

rick.lang

  • Posts: 17173
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: New Ursa 12k sensor tech

PostSat Nov 14, 2020 6:34 pm

Mark, are you typically recording in 12K or something else?
Rick Lang
Next

Return to Cinematography

Who is online

Users browsing this forum: Andrew Lucas and 96 guests