Search the Community

I wish I had discovered this before RED cameras vanished from high end productions. Nevertheless, considering there's still a niche community of filmmakers shooting small gigs and personal projects on old DSMC2 cameras or antique DSMC1, like my Epic Dragon - which I've owned for 10 years now, having bought it way too early in my film making journey - I thought I'd post about my discovery, to deter potential buyers of these cameras on the used market. Recent work with the camera involving panning shots on long lenses, drew my eye to faint artifacts lurking beneath the image. At first I wasn't even sure what I was looking at, or that I was even seeing anything at all, but ruling out debayering, scaling and playback issues during editing, all it took was an extreme curve to confirm my suspicion: there were indeed faint vertical lines in my material - and they were not a byproduct of improper black shade calibration. In fact, I black shaded my camera multiple times before taking it to the air show the below frame is from. And even material I shot many years back, exhibits the same identical pattern when pushed similarly in a side by side comparison. For some perspective, the attached is a 320ASA shot with the STANDARD OLPF at high noon, so pretty much as clean an image as the notoriously noisy Dragon sensor can deliver. Alas, drawing a curve reveals the sensor's fixed pattern noise, which algorithms like Neat Video can't remove. Granted, in the normal version of this particular shot, it's pretty much invisible to the viewer, however there are instances in similar material of dimmer, afternoon and dusk landscapes captured with panning movements on long lenses, where the more pronounced lines in the pattern are just visible enough to distract the viewer's eye and make it evident that something is there. It would however require seeing the actual shot in motion to truly grasp the problem, hence my choice to exaggerate a brighter still frame with a curve instead, if only to isolate and show the pattern off. I've applied a similar curve to a sample Dragon R3D from RED's website and the FPN emerges as expected, so this isn't my camera being faulty. I did also come across a 2015 article titled "3 Things I learned Shooting the Red Dragon" where the author identifies the same exact problem of faint FPN being all over the footage. I should have done more research back in the day... I'll admit the fact it took me a decade of owning & shooting the camera to discover this issue, probably makes it inconsequential in the final imagery. It certainly wasn't a problem in the major productions shot on this system. I never spotted it in Terrence Malick's "A Hidden Life" (which was all landscapes!) when I saw it at a 4K theatre seated in the third row from the front. RED Dragons are pretty much forgotten, near-worthless "e-waste" cameras, only suitable to hobbyists these days, so I'm not even sure there's any educational purpose to this post. The very section of this forum dedicated to RED cameras hardly sees any activity. I am however, at a point in my artistic journey, where I pour so much mental energy in passion projects that I truly can't tolerate such defects in my the footage, so based on the attached example, I'd love to know which (digital) cameras are free of this fatal sensor flaw - beyond Alexas, from which I've stretched sample clips and they are indeed free of imperfections, and the Venice 2, which again, based on some clips from it I played with, is free of FPN, but for the time being out of my price range.

Hello everyone , what is "Fixed Pattern Noise " and "Temporal Noise " in CMOS imagers. how do they control or reduce it .. Thank you , Sundar.