Strange backlight phenomenon

[T_Freundlich]

Hi all,

ran into a strange artifact the other day, shooting R1MX 4.5K WS. The scene was shot in available light with strong daylight coming in from a dirty window behind the performers. As you can see, there's a weird dark bar going across the performers where they are in front of the backlight - in this screen grab you can see it especially with the dancer at frame right. The lens was a Samyang T2.2 16 mm, don't remember the stop but probably around 4. No filters were on the lens.

Any ideas what this might be? Glass-related or digital? Looks somehow optical to me (the dark band followed the line of the bright window even as the camera tracked and panned) but I can't figure out how the part of the performer that's blocking the backlight can be darker than the part in front of the wall. And what's with the green tint? :w00t: The effect was consistent in two clips with the same setup but wasn't seen in other shots with tighter lenses, even with the performers crossing the light.

R3D snapshot in dropbox: https://db.tt/71RGFpZH

The R3D and JPEG screenshots are ungraded and with settings as shot.

 

[Donald MJ Anderson]

Welcome to the very real world of CMOS smear.

 

[Adam Carboni]

Glad you posted this, Thomas. I'm currently compiling tons of test footage I shot on DRAGON, and I'm finding some very discouraging results. To me, the Dragon system still feels like such a science experiment (CMOS smears, OLPF differences, light halo issues, red-channel noise, etc.)

I've tested six lens sets, (Ultra Primes, Leica Summilux, Uncoated Super Speeds, Canon K35, Leica R, and Zeiss ZF), in identical setups, and I wanted to share some stills from a couple shots that are exhibiting the same artifacting like you show above. Crazy vertical flares, "x-ray lines" shining through foreground objects, and strange digital artifacts on a complete column of vertical pixels.

Uncolored (and poorly compressed) examples from my tests are attached. If anyone desires full-res R3D examples, please let me know.


This wide Night/EXT was shot on the Ultra Prime 16mm, 6K FF, SkinTone/Highlight OLPF.

"X-Ray Streaks" noticeable on assistant's body. 6KFF Canon K35 18mm Lowlight Priority OLPF

"X-Ray Streaks" slightly less noticeable the closer the talent is to the horizontal light sources. 6KFF Canon K35 18mm Lowlight Priority OLPF

5KFF Zeiss UltraPrime 16mm. Lines from railing visible through talent's midsection. Highlight/Skin Tone OLPF

6KFF Zeiss UltraPrime 50mm. Lines from railing visible through talent's shoulders. Highlight/Skin Tone OLPF. *ALSO, notice vertical yellow pixel streak that occurs in right third. Seems like a sensor flaw in the rented DRAGON body. It appears in most footage, even full-res R3D.*

6KFF Zeiss Uncoated Super Speed 50mm. Lines from railing AND horizon visible through talent's body. Highlight/Skin Tone OLPF

Crazy vertical flare (looks like CMOS smear). Uncoated Zeiss Super Speed 35mm 6KFF Highlight/Skin Tone OLPF. (This lens flares beautifully and organically on MX and Alexa.)

Now, a lovely image displaying the appreciated extra latitude in the curve's shoulder that DRAGON+Skin/Tone OLPF can provide in Daylight/EXT. 6KFF Lecia-R 19mm V1.

Are any folks out there experiencing/accepting these flaws on set with Dragon? I've shot in countless conditions like this with MX sensor, and haven't run into CMOS issues like this. Any feedback/advice appreciated.

-Thanks,
Adam

 

[T_Freundlich]

Welcome to the very real world of CMOS smear.

Well, there ya go... learn something new every day.

Going back and searching for "CMOS smear", looks like plenty of people are having this streaking problem on Dragon, but seems to be pretty rare on MX. I've shot something like 1,000 hours on this camera over the last two years and this is the first time this artefact was noticeable enough that I actually paid attention to it... might have seen it a handful of times now that I think about it. But good to know what it is, if and when it shows up the next time.

 

[Donald MJ Anderson]

I`m hoping a firmware update will help reduce the smear to MX levels.

From a picture fidelity stand point I`d have thought this was high up on REDs to do list.

 

[T_Freundlich]

Right, so from what I gather, this issue

1) is a "feature"/bug of CMOS chips in general, and that all CMOS's will/may display it under suitable circumstances
2) occurs at sharp boundaries between areas of extreme/out-of-range brightness and dark
3) can be mitigated to some extent by in-camera processing and OLPF design (?)
4) seems to be relatively more frequent on Dragon at the moment.

Can anyone actually explain what exactly is going on in the chip when this horizontal line thing happens? Couldn't find much via Google. If I understand this correctly, it seems to have something to do with the charge from the bright pixels leaking onto successive pixels on the same horizontal row in the sensor. (Pardon my possibly inaccurate terminology, I'm not an electronics engineer...) But would be interesting to understand this on a technical level in a bit more detail.