Hi Everyone
I conducted a little test to see into the depths of the Spark sensor, RED Raptor & Komodo Sensors.
This post this is purely an exploration of sensors and not a direct comparison of sensors. The last few days we all have chatted a bit about the Spark and the sensor stitching, So far after the update I haven’t seen a real world issue with my stitching. Please keep in mind this is such an edge case of extreme under exposure however I believe in exploring even the extremes since we all want to know where that line of do not cross is.
I took a room temp Spark, Raptor 8K VV & OG Komodo 6K covered their sensors with caps and covered the camera body with DUV so zero light leek could be an issue.
-
The Komodo was a control, as a mature global shutter with a single unified sensor.
-
The Raptor is a mature non global shutter sensor however its is a stitched sensor.
-
The reining king of FPS, The Spark as the sensor we all wanna stress test.
All cameras were set to the max sensor size (6K,8K,4.6K) 23.98 / 24 FPS, ISO 1600, shutter 360, WB 5500 with the body caps on. Then I took the files and used Davinci 21 and applied the Rec 709 lut to the RED footage, Imported the DNG files for the spark. Captured stills of the sensor at 0 gain. Then lifted the shadows and highlights 100% on each clip to expose the most extreme noise floor you could ever get…
Let me reinforce this is such an edge case it’s almost ridiculous. but it revealed the heart of the sensors. I also found it interesting that on the spark the fixed pattern noise changes when looking 24 FPS vs 670FPS on 4.6K… (Young has recently commented that the camera does not have dead pixel correction yet so this could be why.) I didn’t get a chance but I am curious of how this changes when cropping into the sensor… I will try that soon.
From what I know of sensors, and I am sure Young and many here likely know infinitely more than me. I love the uniformity of the fixed pattern noise on the RED sensors, they are masters at this but they have been at it a LONG time. I also note that there is in my opinion no visual stitch line on the Raptor sensor unlike the Spark, I even tried boosting well past this and still couldn’t see a line on the Raptor.
The Spark shows me three things, Yes there is still a slight stitch line there however I have to go well into the blacks to see it as you see in the 0 gain photo its invisible, maybe this can be refined in time and with more calibrations. I am also noticing the High Gain ADC’s on the left and right doing what I assume is Sensor Glow… I am surprised this isn’t more calibrated with adjustment columns but maybe we’re just way beyond that, honestly I am not worried about seeing this, I believe this is somewhat normal for a Turbo Charged sensor… Young could tell us more I am sure.
The changes in the pattern noise is something as creators we will want to keep an eye on and help guide Pixboom on what we like to see in future sensors that they will develop, thus helping them to define the LOOK of a Pixboom image just as RED, Arri and Sony has with their cameras. This fixed pattern noise deeply affects the look and feel of our imagery as it truly is the foundation of which our image is built.
As always I will upload the raw date here for y’all to look at.
I’ll check these out.