A 12K Sensor Isn’t Necessarily a 12K Camera
We recently wrote an article highlighting the availability of Blackmagic Design’s 12K camera for $6K. Renowned cinematographer and camera developer Pawel Achtel wrote to push back on calling this device a 12K camera. Below, we summarize his argument, which questions the camera sharpness claims. This assessment is based on independent testing done by Members of the Technical Committee of the Australian Cinematography Society.
Achtel says that the number of sensors on the imager is one thing to measure, but the resolving capabilities of the camera is another thing. It is not intuitive to think that a sensor with 4,448 x 3,096 photosites (ARRI Alex LF) creates a sharper image than a sensor with 12,288 x 6,488 (Blackmagic Design URSA 12K) – but that is precisely what this team found.
The Alexa LF uses a traditional Bayer color filter pattern covering a 2×2 matrix of photosites with RGGB color filters. This is the most basic unit (Nyquist sample) that can deliver a full-color pixel.
The BMD URSA 12K camera uses a much different set of color filters, including red, green, blue, and white (clear) color filter elements, as shown in the graphic below.
Now, a 3×3 block of pixels is needed to capture full-color information, but Achtel believes that a 6×6 matrix is used to determine the color of a pixel.
The Australian cinematographers evaluated four cameras using several different tests. The zone plate test was good for looking at the resolving power of the cameras and gave a good indication of artifacts that show up as Moire patterns. The Imatest SFR process calculates MTF by analyzing the image of a slanted edge chart. The MTF (Modulation Transfer Function) is a measure of contrast at various spatial frequencies. This allows the determination of the sharpness of the camera/lens/electronics system at different contrast levels (30% or 50%, for example).
To be sure the lens did not meaningfully contribute to the MTF measurement, Achtel used a Sigma ART 135mm T2.0. “This is probably the sharpest lens I ever handled (and I handled a lot of sharp glass, including Master Primes, Signature Primes, etc.). This lens resolved 18K comfortably with plenty of contrast left.”
The MTF 50 horizontal calculation results for the URSA 12K was 1309 line widths/picture height (LW/PH), as shown in the chart below. The ARRI Alexa LF measured 1857 LW/PH. This says the image for the Alexa LF is sharper in the horizontal direction than the URSA 12K. 2000 LW/PH represents 4K resolution, so the URSA 12K does not reach this level, while the Alexa LF almost does.
According to Achtel, “Due to differences in aspect ratio between the two cameras (BMD being 17×9 and ARRI being 4×3), this difference [in sharpness] is even higher vertically with ARRI LF being approximately 60% sharper (830 cycles per picture height for BMD 12K versus 1,325 cycles per picture height for ARRI Alexa Mini LF).”
“Despite having only a 13.7 Megapixels sensor, the ARRI Alexa LF delivers nearly twice the amount of detail (121% horizontally and 160% vertically) than the 79.7 Megapixels BMD Ursa 12K,” explained Achtel. “If you have a sharper lens and if you focus precisely, as we did, the [BMD 12K] camera produces severe Moire patterns and aliasing, meaning: the input spatial frequency vastly exceeds the actual sampling frequency of the camera. The sharpness didn’t improve with the sharper lens. Only aliasing got worse. Most 4K Bayer sensor cameras comfortably out-resolve Ursa 12K and with fewer artifacts. We have not encountered any aliasing or Moiré on sample images taken with ARRI Alexa Mini LF,” concluded Achtel.
For further details, see these references below.
A quick visual (subjective) comparison can be watched here: https://vimeo.com/528541990.
- CAM A is Ursa 12K
- CAM B is 9×7
- CAM C is RED Monstro
- CAM D is ARRI LF
The results, including MTF figures, can be downloaded from here: https://drive.google.com/drive/folders/14Y7Bgb-LrYsYv4-6MWohZly4McU2sisW?usp=sharing.
RAW camera files can be downloaded from here: https://drive.google.com/drive/folders/1vvVvGIOlzP1mXVZWuED5_DklsfjLInPE?usp=sharing
BTS how the tests were performed: https://drive.google.com/drive/folders/1fnx9s8jSzQGUST57gx4ma1BKIDB0JYGM?usp=sharing
A white paper explaining some of the reasons this camera cannot match even a 4K Bayer sensor camera in terms of sharpness or aliasing / Moire artifacts: https://achtel.com/BMD12K/
The Ursa to my eye is sharper > LF. 9×7 is mind blowing. Monstro next after the 9×7. I see the calculations but that’s what I see.
appreciate your thoughts