Vertical lines in high ASA

[Teague Kennedy]

I know I should be exposing properly, but that's not always possible in EFP. One thing that is bugging me when exposing above 1000 ASA (besides the blue channel grain) is obvious vertical lines that look like film scratches in the footage. (in dark areas of the screen). They are always in the same place. Is this a sensor issue? A build issue? Just wanna make sure I'm not doing something wrong or have an odd sensor......

 

[Kenn Michael]

I get the lines as well... Hopefully that'll be improved in Build 16.

 

[Dj Joofa]

I know I should be exposing properly, but that's not always possible in EFP. One thing that is bugging me when exposing above 1000 ASA (besides the blue channel grain) is obvious vertical lines that look like film scratches in the footage. (in dark areas of the screen). They are always in the same place. Is this a sensor issue? A build issue? Just wanna make sure I'm not doing something wrong or have an odd sensor......

Seems like a sensor issue that should be taken care of by proper calibration. No sensor is perfect and its calibration which hides sensor defects that are always there. A lot of such anomalies are seen in a high gain environment (raising ASA is tantamount to raising gain).

It appears to me that Red One is missing out on some important calibrations, since as I understand the end user only does a black level test with lens cap *on*. There are corresponding important tests that require the lens cap to be *off*, in a controlled environment, and I have always wondered that why Red One is not requiring users to do them.

I am sure they are doing some internal calibration before shipping out a camera, however, the very reason that Red is requiring users to do black level test means that there is something unique to each sensor chip, under certain conditions, which of course it is, and they are perhaps missing out on some important calibration.

 

[Paul Leeming]

The black shading is partly designed to counter this by matching the levels of each row of pixel sites (the way the sensor gets its image) which in Red's case are oriented vertically. On my Canon DSLR they're oriented horizontally but same deal.

I don't know if Red applies artificial gain when black shading to really find the most extreme line scenario and counter it but that may be down to the limits of sensor discrimination. Perhaps the black shading algorithms will continue to develop and eventually be able to perfectly align the gain of each row such that we are only left with single pixel variations, more closely emulating the randomness of film grain.

Here's hoping....

Paul

 

[Michael Totten]

I know I should be exposing properly, but that's not always possible in EFP. One thing that is bugging me when exposing above 1000 ASA (besides the blue channel grain) is obvious vertical lines that look like film scratches in the footage. (in dark areas of the screen). They are always in the same place. Is this a sensor issue? A build issue? Just wanna make sure I'm not doing something wrong or have an odd sensor......

I'm curious to see what the banding looks like. Can you post a still?

 

[Jannard]

Please make sure you properly do the Black Shading Calibration after installing any new build.

Jim

 

[Redrhoads]

we've had similar issues with a few of our cameras. We had to send em to red to get fixed as black shading did not remedy it. Came back flawless though! Good luck.

 

[Seth Larney]

I'd be interested to hear if this was an issue that was deemed 'unacceptable' by Red (as per Redrhoads case), or whether it is Normal. Either way I'm not complaining, just interested After all.. you have to dramatically push the image to see it.

I don't have my camera yet to test myself. Jim, are you saying that a proper Black Shading calibration should fix this ?

Cheers
S.

 

[zak forrest]

we've had similar issues with a few of our cameras. We had to send em to red to get fixed as black shading did not remedy it. Came back flawless though! Good luck.

Your vertical lines of noise disappeared after sending the camera to Red??

I thought this was normal, or hopefully going to be fixed with Build 16.

 

[zak forrest]

Quicktime: example of noise

Quicktime: example of noise

http://zf-films.com/mov/hamptonnoise/ellanoise.proxy.mov


here's a clip of what my camera does at high ASA. This is made from a proxy so it's extra noisy. But the noise still exists when rendered out of Redcine.

If you look at this clip in a browser it won't look the way it should, right click and download.

I have always done black calibration. Please tell me this is normal to my camera! If this could have been fixed when my cameras were both at Red getting stuff done (one was lubed, the other was put on the operating table) I will be disappointed. I've been shooting a feature at T1.3 the past couple months, a few scenes have this noise, but they are dark scenes, and when properly color corrected and darkened and rendered out of Redcine the noise isn't that bad, and I always figured on doing some sort of noise reduction later. But still, if this is something that can be fixed by sending a camera in to Red I really wish I was notified about it...

 

[Teague Kennedy]

Jim, I did black calibration. I saw the issue, so I re-performed another black calibration. Still an issue. I will post a still maybe later today if I have time.

 

[Kenn Michael]

Zak, thanks for posting the movie. This is the type of noise I'm seeing too. I've done the Black Calibration.

 

[zak forrest]

Yeah and like I said, it seemed normal to me, but that one guy's post saying he sent the camera back and then the noise was remedied??

 

[Redrhoads]

the lines we experienced were not normal. they were visible at all ASA settings. They were blue and sorta bunched together near the center of the frame and got more prominent the more the camera warmed up.

 

[Dj Joofa]

the lines we experienced were not normal. they were visible at all ASA settings. They were blue and sorta bunched together near the center of the frame and got more prominent the more the camera warmed up.

Very typical behavior. It might not be surprising if you start seeing magenta and reddish lines. Calibration is missing stuff, as I have said before.

 

[jbeale]

Very typical behavior. It might not be surprising if you start seeing magenta and reddish lines. Calibration is missing stuff, as I have said before.

Just curious, when you say blue vertical lines at all times is "very typical behavior", this is based on what?

 

[Zach Hilton]

To me it seems normal in high ASA/gain scenarios. I see the exact same thing on my HV20 and my Digital Rebel.

 

[Ricardo Mehedff]

Is that what happens at high ASA in low light situations? I gotta tell ya, the footage Zak posted looks terrible!!

 

[Dj Joofa]

Just curious, when you say blue vertical lines at all times is "very typical behavior", this is based on what?

Based upon on what I have seen personally and helped to correct it, and am also aware of many other companies who ran into similar problems.

 

[jbeale]

Here is a crop from one frame of Zak's high-ASA proxy export. It definitely has fixed-pattern noise in the form of vertical line artifacts. In general, fixed-pattern noise should be correctable with the right processing in-camera (and if you wanted to, with enough time it could also be done in post). The clip also has a lot of random, non-fixed noise, as you'd expect in dim light.

EDIT: Looking at the frame, I actually notice the vertical lines more on the brighter areas of the image. If they are reading out the image in columns and they have adjusted the black level offset of each column correctly, but have the gains (signal out per light in) for some columns slightly off, I imagine it might look like that.

As Joofa was suggesting, one way to fix that would be to measure the gain on each column by taking a "white frame", but getting an accurately flat uniform white exposure may be more difficult than a simple lens-cap-on dark frame. In the lab I think you would point the sensor at an integrating sphere (spherical cavity painted with flat white pigment inside). They're kind of pricey for home use, though.

EDIT2: Added a second improved (?) image crop. I tried averaging pixels vertically + high-pass filter to get the fixed-pattern column noise, re-expanding to full frame, taking inverse and multiplying that by the original (took gamma = 0.5 first to approximate "linear light", gamma = 2.0 after). I think the vertical line artifacts are diminished somewhat- didn't fix it all, certainly.