Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

The CMOS Smearing

Status
Not open for further replies.
Im not saying this can not be a problem.

Every operator and and camera manufacturer would love for CMOS smear to go away. It's not. It's there, it turns up, it rears its ugly head and it spoils shots.

Its not reds fault, arris fault, Sonys fault, it's just CMOSs fault.

The best thing to do is look out for it and make adjustments when you can. There are going to be situations when you get boned by CMOS smear. Dragon is more sensitive to this, because it's just generally more sensitive.

When CMOS rears its ugly head, you need a strategy. Phil has provided plenty and shown options.

In ieach case each person needs to adjust accordingly. It's not a predictable thing (like black sun) that can easily be fixed by one thing in the signal chain.

Sorry to to be a party pooper, but the best advice is to be aware of it and adjust.
 
interesting. never heard of cmos smear. CCDs suffer from vertical smear ..yes.. not normally CMOS ..(although dodgy drive cables with the old red drives could emulate it)..

..the fact that pixels are read out one by one normally side steps the problem.. it is more common on global shutter systems/ CCD sensors

cmos sensors tend to have noisier characteristics..we seem to be getting the best of both worlds
 
David Battistella said:
Its not reds fault, arris fault, Sonys fault, it's just CMOSs fault.
Click to expand...

I'm not saying it's anyones fault, but I've never had any problems with this type of artifact. I've been filming with digital cameras for over 15 years and I've never even heard about CMOS smear or seen it so severe. My point is, yes, it's there, but why is it so visible with the Dragon sensor? And if this is such a common issue, why aren't there anything done in processing to counter it?

We can prepare a shot and try to minimize it, but I've never needed to do that with the MX. It's like I'm more limited now than before my $9500 upgrade.
Shooting with light sources in frame is very very common. If this limits what you can shoot it limits the whole camera as a creative tool and that is not good.

This is why I urge to look into this, especially on the Dragon sensor. There has to be a way to counter it in processing.
 
Christoffer Glans said:
We can prepare a shot and try to minimize it, but I've never needed to do that with the MX. It's like I'm more limited now than before my $9500 upgrade.
Shooting with light sources in frame is very very common. If this limits what you can shoot it limits the whole camera as a creative tool and that is not good.

This is why I urge to look into this, especially on the Dragon sensor. There has to be a way to counter it in processing.
Click to expand...

Same with the gate shadow! I never had to deal with it on MX, and I've seen it nearly every single day I've had the Dragon out.
 
And of course, this is not something the use of a motion mount (or a camera with true global shutter) would make less dominant, would it?
 
Red has responded officially the last time this came up. I'm not sure anything has changed. It's all about exposure, seeing it and adjusting. In the shot christoph posted a half a stop less could eliminate it for example. Then process that shot one stop brighter.

how else can I say this. It's a known problem, it will be a problem.

"but they should just fix it" is not a response to that problem. We all have to deal with it. I've seen it at different times on every red I've owned.


Here are a few "official words" for you all.

Battistella

Jarred Land said:
It doesn't happen all the time... a certain combination of things need to happen just right ( or just wrong depending on how you look at it ) It did happen on Epic MX... and Epic M, and R1 and it happens on the other guys cameras as well.

It is hard to guess but the shots above from last year it might not have the right combination of dark enough dark right next to super clipped to happen.

Calibration isnt going to fix that... its a hardware thing that im sure will be fixed in a few sensor generations down the road.
Click to expand...

Jarred Land said:
Gavin and Rob are correct.. this is a relatively normal CMOS thing.. kinda like the black sun dot all cameras used to have ( i think some cameas from other manufacturers still do ) its been there since the R1 since we do use CMOS technology like almost everyone else. If you take a small source highly clipped a bunch of stops against almost black and it crosses the edge of the frame or just off the edge of frame ( most cases ) you will see what looks like a streak. People call this the "phantom blinds" effect because the easiest way to replicate this is to open slit blinds into daylight in a pitch dark room. Of course you will see this at higher ISOs, and most post houses have developed a post fix from dealing with it over the years from many cameras. Its a silicon thing.

The company that makes the dynamic range charts we use actually have added a little flag that goes over the clipping square now, because if your shooting a CMOS camera and you get that flare it can actually trick you into thinking you have more DR... since the streak replicates the shape of the highlight and lifts everything with it.
Click to expand...

Jarred Land said:
snip...

People just need to figure out what OLPF they want, and if they want their stop to be at the top or at the bottom and select the OLPF that is right for them. The vast majority have chosen the right one ( the new OLPF ) and are very happy with it but there are a few vocal people that want to see that OLPF back and we are happy to put it in their cameras.
Click to expand...

Jarred Land said:
Ivan, the upside down OLPF was our mistake.. that one we accept. When we do dumb things like that it doesn't really help our credibility so I can understand the wink.

But the new OLPF is not a botch. As much as there is a handful of people here adamant to keep pressing the issue over and over and over again, There are alot of very high caliber cinematographers and directors out there shooting the biggest films that tested both and have chosen the new OLPF for the right reasons.

And the new OLPF does not make the camera less sensitive than the old MX. Yes, it is less sensitive than the old OLPF+Dragon, but you gain a stop in the highlights, you get better color , better skintones and it eliminates the IR leak and flares. The old OLPF since it didnt do the best at stopping IR it boosted the image in the bad kind of light.. IR color contamination all over.

That's why we made the new OLPF, which has more green and less RED/IR in it ( much just like the Alexa OLPF ) . One side effect was the red channel spiked up in the noise floor so you would see that red chroma noise in the shadows and it would look much worse than it was.. Thats what Graeme is taking care of right now. Even with the new OLPF, Dragon has lower luma noise than the MX, and of course the Dragon has the extra stops in the highlights, as well as an extra stop on top of the sensor that the new OLPF gives.

We are not going to add noise reduction by default... that's something best left to people that want it and its a bit subjective in application. I know everyone else does it.. much like sharpening... but that doesn't mean it's right.

I do agree with you that having 2 OLPFs out there isnt the best idea.. but we are giving people the choice anyways. There are some people out there that turn the camera to ISO 800 and point to a scene that has 6400 ISO worth of light and think they have noise at 800.. they just don't get it so it is hard to compare to.. but to be fair everyone did that with our last camera as well. What we can do is compare things to used to be and make sure we do as much as we can better and move forward.. sometimes you need to make hard decisions and trade something for something else.

For the last 7 years we have heard Skin Tones and IR and highlights over and over and over again.. I am happy that with the Dragon one OLPF and alot of work from Graeme took care of those 3 things, at the expense of a bit of light optically which we feel is worth the tradeoff.

Anyways... Graeme should have some images to share tomorrow... he is almost done.
Click to expand...
 
I say yes. But it comes down to exposure.

When you have a hot row of pixels you have to take steps to limit that exposure.

If you frame a window, expose for an interior and have say, Venetian blinds half open, you can recreate the problem in two seconds. It generally happens when you are blowing out pixels near or close to black or in the widest highest contrast situations.

But it is hit and miss.

Since the the sth allows more room at the top stops maybe it's a bit worse because it clips later than the LLO. I'd have to test
 
Plus. When I blow out windows I like to add a very light diffusion. 1/8 digicon is a goto.
 
With LL and SKH I have rarely seen this prob. But reviewing all the images sent in on at least two threads I have noticed (often not all the time) that the CMOS is helped immensely by composing for subject in focus and say a lamp or blown out window a little out of focus. Shane Hurlburt did night tests with a subject with streetlights etc behind them. No streaks on either OLPF. In part because he exposes well obviously, but also because the highlights are in soft focus. That would explain also in part why the use of digi con helps as it achieves a similar thing. The discussion gets complex because people also submit other artifacts caused by other things and discuss it like it is the same thing. I find frankly the single biggest thorn of this type I have to adjust for on set is light creep causing flares on uncoated lenses. Which is way more troublesome to manage than this.
 
Still having the magenta shooting horizontal light that will cut across talents faces

Still having the magenta shooting horizontal light that will cut across talents faces

One of my DP asked if this is still typical. Any work around?

Lim
 

Attachments

  • image.jpg
    image.jpg
    78.7 KB · Views: 0
Just read through this thread. Pixels are looking a bit blown out there.
 
Björn Benckert said:
So I assume, I have not tested my self, the skintone olpf handles the Cmos smear better, yes?
Click to expand...

I've seen it more with the skintone olpf and never seen it on low light olpf in my own shooting. I'll see if I can dig up a still. With the skintone olpf it shows as a magenta smear.
 
CmosSmear1.jpg

CmosSmear2.jpg


Skintone Highlight OLPF, 400ISO, t2.8 on the 11-16.
 
We did some Lowlight OLPF tests, and CMOS smear seems very well controlled btw 320-800.

Basically, feeding camera a lot of light.

I could see the Skintone being more prone, as its a 320 olpf, but many are shooting it at 640/800 and up. That may be making the camera more susceptible to smear, as there'd be more chance of underexposure within high contrast scenes. Also, as Matt's 400iso sample suggests, Skintone seems particularly sensitive in the RED channel, so the smear appears more likely to go magenta (than remain neutral/clear as does on Lowlight), and maybe more visible than it would be otherwise?

Basically, from what I can tell, Dragon is "seeing" so much more than MX. That's wonderful, but also needs to be managed.

For me, the Lowlight OLPF is everything MX was, just much better. Not sure why it gets such a bad rap. The color is WAY better than MX, the DR fantastic, and has none of the beauty penalty of Skintone. Even the "orbs" are overblown. To me they look great, very natural, and now that they are not tinted red, are NOT an issue. Because the Lowlight OLPF is rated 800, I know that by shooting 320 or 640 we can bathe the sensor in light, get PERFECT blacks, no noise, and have the smear be very controlled. As we are not stressing the sensor at all - instead, we're empowering it to be as GOOD as it can. Yes, there's a bit less DR in the highlights, but for everyday shooting - more than enough. We do mostly controlled shoots anyway, and for me...perfect blacks are far more important.

We have yet to test Skintone, but the color would have to be WAY better to force us to switch. Jake Blackstone grades all our dailies anyway, so we don't need great color "out of the box". Jake IS our box, so as long as we have a good fat negative with decent color, we're in good shape.

From what I've seen, Lowlight is pretty amazing, and allows us to work comfortably in the ISO ranges we need - 320 to 1280, with little penalty.

My hunch is Skintone is the "studio" OLPF - the theoretically perfect OLPF that would be ideal for traditional studio folk who can afford to have lots of units and bathe their sets in light. Then yes...if you can work at 250/320 (maybe 640)..you'll get even better results (maybe 10%?). But again...we're shooting RAW...and even Lowlight is SO GOOD, you're still way ahead of MX. I feel like for everyday shooting, for the indie RED shooter, those with less units, more run and gun, and more of a need to shoot at the digital standard of 640/800/1280...then Lowlight should probably come back into conversations more.
 
Last edited:
Nick Morrison said:
The latest Hunger Games appears to have a ton of CMOS smear, fyi...
Click to expand...

CMOS smear or Anamorphic streak? Seeing as it's anamorphic and Alexa I imagine it was more likely the latter than the former. I've never seen cmos smear on alexa footage. I can hardly turn on my Dragon and point it at a high contrast scene and not get it.
 
I've been watching Better Call Saul and CMOS smearing rears its' ugly head a lot there. Makes me sad...

P.S.: On this frame the windows are even properly exposed and the smearing is still there

TTf0yQH.jpg
 
Status
Not open for further replies.
Back
Top