Welcome to our community

Be a part of something great, join today!

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 :)

Issues

HOLLYWOOD SCHMOLLYWOOD said:
That looks like a reglular old lens flare to me. I dont think any camera would resolve that much differently. Its right on the edge.
Click to expand...

That's not a lens flare, it's a "sensor protect flare" - it's been [in my opinion] the worst camera-related problem the RED has had from day one. It should be a huge priority to get fixed.
 
Rudi Herbert said:
It does if you light the scene based on that particular ISO
Click to expand...

Again people. Let's eliminate the confusion of the ASA and talk relected footcandles to communicate luminace/noise levels. Or just talk strictly 320 which all this camera really is. The other ASAs are just brightness/ gamma adjustments. Not actual amplified sensor sensitivity.
 
The black vertical line is not a lens flare (flares are brighter than the background and this is darker). But I'm not sure it is a "sensor protect" mechanism, (eg. the black spot that appears on the sun, or an arc welding spot etc.)

Remember the difference between "total sensor pixels" and "effective pixels"? Normally in any CCD or CMOS sensor, some edge pixels are optically masked off and used to set the dark level. However, if you have a super-powerful light at the very edge of the sensor so some light leaks in there (directly through the mask, or diffusing around the edge) then it looks to the camera like that pixel column has a high black level, and compensating for it in the output gives you a darker line than the normal background black.

I'm not saying that's what's happening here- just one possible explanation.

Come to think of it, you would have a similar effect even without any light leakage. If the light is strong enough, the area on the chip where it is focused becomes hotter. That increases dark current of the reference dark pixels and you get the same effect, overcompensation for that pixel column in frame. Only solution I can think of: do not put powerful lights at the very edge of the frame. Maybe there is a firmware solution to disregard any dark reference pixels that are X standard deviations hotter than average.
 
Anders Holck said:
He isn't talking abou the lens flare, but abuot the dark vertical column overflow.

(Edit: Ok, Brook IS faster...)
Click to expand...

This is absolutely new to me. Looks like a kind of inverted smear of old IT-CCDs. I am concerned. Very distracting.

Sorry, cannot contribute anything more to this.

Can anyone shed light on this?

Hans
 
Wow. What a rough week. This, the noise, the delivery schedule, George Lucas. But I have faith. Its all part of the public testing, feedback, and REDs commitment to getting these things right. We live in an amazing time.
 
jbeale said:
The black vertical line is not a lens flare (flares are brighter than the background and this is darker). But I'm not sure it is a "sensor protect" mechanism, (eg. the black spot that appears on the sun, or an arc welding spot etc.)

Remember the difference between "total sensor pixels" and "effective pixels"? Normally in any CCD or CMOS sensor, some edge pixels are optically masked off and used to set the dark level. However, if you have a super-powerful light at the very edge of the sensor so some light leaks in there (directly through the mask, or diffusing around the edge) then it looks to the camera like that pixel column has a high black level, and compensating for it in the output gives you a darker line than the normal background black.

I'm not saying that's what's happening here- just one possible explanation.

Come to think of it, you would have a similar effect even without any light leakage. If the light is strong enough, the area on the chip where it is focused becomes hotter. That increases dark current of the reference dark pixels and you get the same effect, overcompensation for that pixel column in frame. Only solution I can think of: do not put powerful lights at the very edge of the frame. Maybe there is a firmware solution to disregard any dark reference pixels that are X standard deviations hotter than average.
Click to expand...

I had thought is was possibly extremely bright light refracting down the sensor between the sensor and glass protection filter of the sensor but your theory makes a lot more sense.
 
Hans von Sonntag said:
Hi Mac,

Quick answer.

Yes, this was adressed http://www.reduser.net/forum/showthread.php?t=5091&page=5 Jarred comment on this. No concern anymore.

Thanks a lot,

Hans

PS: Still some concern, since the flare in the thread mentioned above was white and not black. Well... May be the same causing.
Click to expand...

Not so fast. I have seen this problem with new cameras and new firmware builds.

It is an issue. Any extremely bright light that hits the top area of the sensor and wham big black line that remains until you tilt down or up. There must be some kind of fix though.

Anyone have any idea if it happens when bright light hits the side or bottom of the sensor?
 
Concrete said:
Well, it most certainly looks like it. Maybe Häakon can shed some light.
Click to expand...
I really don't see a problem... I know that because of the h264 and on certain players/configurations, the gamma is sometimes out of wack and the image looks way brighter than it should, causing noise to be more apparent. But with that in mind, I don't see any crazy noise detracting from the images I shot. I wouldn't personally use the camera over 500ASA, and setting proper exposure is still as important as it always was, but if you use the tool correctly you will have good results. The images are NOT noise free. But neither is film. The question is, does the camera exhibit an acceptable level of noise for your own satisfaction, and in what way does the noise manifest itself - as subtle, more organic pixels, or in big splotchy blocks of artifacting? Perhaps your tolerance of noise levels demands more than RED can deliver at this point - and I would be respectful of that - but then I challenge you to find another camera that does it better. RED ONE is not the be-all, end-all solution in digital cinematography, but I would argue that it is probably the best choice available today and (more importantly) finally offers a level of quality that I am comfortable exhibiting.

I also think that the [h264] codec does exacerbate the problem. I was going to post a prores version because it looks better, but I didn't want to leave out a lot of viewers that are either on a PC or don't have FCS2 installed. I will upload that clip as well as a 2K version tonight for comparison.

I have long maintained that an image always looks better oversampled (especially by multiples of two) than in its native resolution off a sensor. This holds true for dSLRs as well as video images. If you want to end up with a 1MP image, shoot it with an 8MP camera and downsample it and it will look far superior to an image shot with a native 1MP sensor. RED has fantastic resolution, but will look better in 1080/2K form than in 4K. Noise is smoothed out, aliasing is reduced, apparent sharpness is increased and sensitivity is boosted. I can live with calling RED the best 1080p camera on the market right now considering the images it is capable of recording. I will post the HDX footage later this evening and then you can have a real discussion about noise.
 
macgregor said:
Workflow for the latest pics:
- shot in RED (firmware 10, 4K, ultraprime 16mm t1.9, shutter 1/24th, iso 1000, exposure seems ok in the on board histogram allthough other ISo settings are used just to compare)
- copy CFs to PC
- Open RedCine
- New project leaving all settings default (2K 2:1, camera RGB and rec 709 gamma)
- open files / select CF folder
- Shot / shot framing / fit width (image is resampled to 2K)

At this stage noise patterns are visible, Exposure bar is untouched by me, so each shot loads its own value from the metadata.

- export / tiff / 8 bits

- open tiff in irfanview / resize to 1080HD / jpeg at 90% / upload to ftp (jpg compersin does not introduce any artifacts at this stage)
Click to expand...


I'm bit confused here Mac. Here you say about the latest picture it was shot in 4K and THEN resampled to 2K,
but at the begining there was written this under it:

macgregor said:
And the most horrible one:

Shot at 2K ISO 1000, or RedCine exposure +1.64. 3500ºK. Sorry to say this, but we cannot use this image. Obviusly the 2K crop magnification makes things worse than in 4K mode. There´are blue noise dots everywhere...
Click to expand...

Are we talking about the same picture?
 
But its so close to being a viable 4k camera. We've all seen what Crossing looked like on a 50ft screen. If only we can find away to get the night int./ext. there. With reasonable low light noise if not great. Perhaps its already there but what Im hearing and seeing says almost. I wish I had mine to really make that judgement Haakon mentions.

BTW, does anyone know the Mbps of the Codex "toaster oven" recorder that takes uncompressed in but then puts a light compression on it? Jpeg 2000 er something?
 
Rocketeer said:
Has the new camera revision eliminated the sensor flare that is caused by a strong light source at the top of the sensor?
Click to expand...

This was a firmware big that was fixed a fair whole ago (I think some of Macgregors footage exhibited this also.)

Edit : I definitely read this a fair while ago. I assumed that was correct at the time.
 
Jannard said:
Since everyone is getting new cameras, it will be important to note which footage was shot with "original" cameras and which with "new" cameras. A full blown test would be good...

Jim
Click to expand...

Is it possible to have metadata in the .r3d file that shows which model shot the footage and what firmware was loaded?

(or does this already exist? I thought I remember hearing something about it).
 
Finner said:
Not so fast. I have seen this problem with new cameras and new firmware builds.

It is an issue. Any extremely bright light that hits the top area of the sensor and wham big black line that remains until you tilt down or up. There must be some kind of fix though.

Anyone have any idea if it happens when bright light hits the side or bottom of the sensor?
Click to expand...
I spent about an hour testing the sensor flare last night... using different builds and different lighting situations.

It can happen on both the top and bottom of the sensor, but not the right or left. There is no discernible difference between builds.

The flare has a strange characteristic while panning. I locked the tilt of the head off with the flare at its strongest [with the source splitting the edge of the sensor - 50% on the sensor, 50% off] and panned left and right. The flare will get "stronger" over about 2-4 frames when a pan is stopped. It'll also strobe within a frame [read: it'll appear ghosted in several different areas] during a rapid pan.

It's black with a slight purple bias.

I've seen it on every camera I've worked with... cameras from different batches and in different physical configurations. It does not appear to be dependent on the OLPF... it's a sensor thing.

I'm sure it's fixable. Nobody panic... we've been dealing with it from day one. It's possible to deal with.
 
Good news.. this is already fixed (has been for some time internally) Its one of the presents current owners will receive when they get the upgrade.
 
You must log in or register to reply here.
Back
Top