Will build 15 address this little issue?

[Jeff Kilgroe]

Dave I have not seen this effect with any other digital or film cameras.

Same here. I have not been able to reproduce the black hole effect in flares on any Nikon DSLR, nor on the HVX200, HPX500, Varicam or any other digital camera I've ever shot with.

 

[David Birdy]

Dave I have not seen this effect with any other digital or film cameras. We are not talking about the flare here or need advice on how to avoid flares. Flares are normal with every camera and yes do happen all the time. What we are talking about is how the most intense IRE setion of the flare clips and turns black when it should turn white like every other camera.

Any type of electronic equipment has a defined range of operation.

Once you exceed the range of acceptable operation and the electronics are over driven all bets are off..

The purple hue around the sword is certainly a result of the CMOS senor being overloaded by light...I was suggesting a way to avoid this from happening in the future and I don't see it as any "real" problem with the sensor.

This image reminds me of comet tailing in Tube cameras...once the Tube was overdriven with light the comet tails would appear.

Maybe Jim will jump in on this thread to add his view?

Dave

 

[Jeff Kilgroe]

The effect is the same with my Pentax D10 and any other camera I have worked with.

show me.

Jeff,
I certainly did not take your comments as RED Bashing...

No problem. I didn't think that you did.

we all need to speak our minds here or we should just log off as this forum would not be useful!

Yep.

If I were to correct this issue this is how I would go about it.

I agree with everything you posted on fixing / working around this. That's fine... The point of this set up was to illustrate the flaring artifact. IMO, the black clipping in the over-exposed area happens too easily and should not happen at all.

 

[shin sugino]

This is the problem I was talking about in my last post "Bad Flare". I find this serious and fundamental problem. This does not happen with every camera. Not on Viper, not on Sony 950 nor Panasonic. Is this software issue or hardware issue? (black spot or purple spot.)
I don't find this cool nor nice style.

 

[Ariana]

Viper has a nasty ringing around bright lights and a horizontal streaking/bright shadow.

 

[Dj Joofa]

Dave I have not seen this effect with any other digital or film cameras. We are not talking about the flare here or need advice on how to avoid flares. Flares are normal with every camera and yes do happen all the time. What we are talking about is how the most intense IRE section of the flare clips and turns black when it should turn white like every other camera.

Film camera may not have this issue as there is no issue of charge spilling. To me it appears that a higher bit-depth dynamic range would help here.

It appears this phenomenon could be resulting from internal black level monitoring or charge spillage because of a strong flare. The flare turning purple means that more green is being clipped off, and I would tend to think that this is consistent with what one would assume that the higher quantum efficiency of green wavelengths would make the corresponding charge spill over more, and perhaps reds and blues have not started clipping yet, or clipping much less than greens.

If the flare is turning gray/black, that would give me an indication that all red/blue/green channels are beginning to clip, in a proportion that derives it towards achromatic color. I think it might not turn into a clipped white, as perhaps the internal black level of the sensor is being raised. I am not sure about this and will have to check back on this. But, if that is true, then the output signal for the flare area will be reduced because of a raised (local) black level.

That also gives me an indication that if the sensor had higher dynamic range, then it would have held up longer before clipping started happening; it appears 12 bits need to be enhanced to at least 16; but not by adding a new ADC, but by doing something to the actual micron pixel to have higher dynamic range.

 

[number6]

Film camera may not have this issue as there is no issue of charge spilling. To me it appears that a higher bit-depth dynamic range would help here.

It appears this phenomenon could be resulting from internal black level monitoring or charge spillage because of a strong flare. The flare turning purple means that more green is being clipped off, and I would tend to think that this is consistent with what one would assume that the higher quantum efficiency of green wavelengths would make the corresponding charge spill over more, and perhaps reds and blues have not started clipping yet, or clipping much less than greens.

If the flare is turning gray/black, that would give me an indication that all red/blue/green channels are beginning to clip, in a proportion that derives it towards achromatic color. I think it might not turn into a clipped white, as perhaps the internal black level of the sensor is being raised. I am not sure about this and will have to check back on this. But, if that is true, then the output signal for the flare area will be reduced because of a raised (local) black level.

That also gives me an indication that if the sensor had higher dynamic range, then it would have held up longer before clipping started happening; it appears 12 bits need to be enhanced to at least 16; but not by adding a new ADC, but by doing something to the actual micron pixel to have higher dynamic range.

Joofa, what do you have for breakfast? I want to know because I want to start havin' the same. What you posted may or may not be right, but it just makes so much sense.

 

[Dj Joofa]

Joofa, what do you have for breakfast? I want to know because I want to start havin' the same. What you posted may or may not be right, but it just makes so much sense.

Actually, I haven't had any breakfast for the last 3 years :mellow: and mostly I just eat at the lunch time -- I know it is a bad habit, and I am trying to reverse that. In the good old days, I used to eat some pancakes or a breakfast based upon apple (the fruit; not the company) product, or blue-berry based; all home-made.

 

[Nathan Garofalos]

When I just saw that sword, I thought it was a light saber. But had to look closely and saw other wise...

 

[Dylan Reeve]

I saw this in the monitor the other day, right in the center of a robo spot but I can't see it in the files at all... Weird. Maybe I'll find it yet.

 

[Jeff Kilgroe]

Joofa,

Thanks for posting that... Some of what I'm afraid of was covered in your post... This may be something inherent to the sensor design and how each of the color channels clip as they become over saturated.

I'm wondering if there is no fix for the black / purple flare with the current hardware. If that's the case, then so be it... But it seems that all that is needed is a way to identify clipped photosites and lock them to whatever the clip ceiling is set to rather than clipping out to a zero value. The question is, does the sensor make that sort of data available, or does it simply zero-out the overloaded photosite data?

Anyway, I've said my piece... Hopefully it's something that can be fixed. I still intend to post a clip in addition to a still frame, I just haven't got around to outputting one yet. Maybe I'll do that later tonight or in the morning. I've got a new Mac Pro to go set up now. And I'll run through some tests on this footage and other footage I have.

 

[Dj Joofa]

Joofa,

Thanks for posting that... Some of what I'm afraid of was covered in your post... This may be something inherent to the sensor design and how each of the color channels clip as they become over saturated.

I'm wondering if there is no fix for the black / purple flare with the current hardware. If that's the case, then so be it... But it seems that all that is needed is a way to identify clipped photosites and lock them to whatever the clip ceiling is set to rather than clipping out to a zero value. The question is, does the sensor make that sort of data available, or does it simply zero-out the overloaded photosite data?

Jeff, you are welcome.

I suspect that both asymmetric clipping and raised internal black level monitoring are causing this. Red would know that how are they doing their internal black level calibration, and it may be exposed as a programming interface available to their internal developers. There are several ways of doing internal black level monitoring, each with pros and cons, and I am not sure what they are doing here.

 

[Poi Boy]

I sure would love to see this in context...moving images please. Single frames are of little value.
Aloha
-A

 

[Finner]

Some of what I'm afraid of was covered in your post... This may be something inherent to the sensor design and how each of the color channels clip as they become over saturated.

I'm wondering if there is no fix for the black / purple flare with the current hardware. If that's the case, then so be it...

Jeff

Maybe the answer is not in camera firmware but a red cine or red alert change that detects when some part of the frame has clipped black/purple and a little box pops up asking "Do you wish to default black IRE protection clip to white?" with a yes or no check box. It may be the easiest way to deal with the problem.

 

[Steve Freebairn]

Jeff

Maybe the answer is not in camera firmware but a red cine or red alert change that detects when some part of the frame has clipped black/purple and a little box pops up asking "Do you wish to default black IRE protection clip to white?" with a yes or no check box. It may be the easiest way to deal with the problem.

WE JUST NEED THE QT PROXIES TO AUTOMATICALLY ASSUME YES (OR MAKE IT SO THAT YOU CAN SET YOUR OWN DEFAULT) FOR USE IN QUICK TURN AROUND SCENARIOS.

 

[Dj Joofa]

Jeff

Maybe the answer is not in camera firmware but a red cine or red alert change that detects when some part of the frame has clipped black/purple and a little box pops up asking "Do you wish to default black IRE protection clip to white?" with a yes or no check box. It may be the easiest way to deal with the problem.

I would personally think that this problem may have to be corrected in camera firmware. But, I could be wrong. I tend to think that by the time signal is presented to RedCine and RedAlert in Bayer pattern the damage has already been done. The black/purple/gray values are no-longer clipped now compared to the white point of the image data (they may have been on the sensor), and it would be difficult to figure out which areas of an image were disturbed on the sensor.

I think Red should expose things such as readout of the internal black level (readout of optically black pixels) as a programming interface to their DSPs, so that at least the DSPs can monitor what is going on. To rectify the problem completely, I think heuristic-based reasoning may have to be used in addition to the pure number crunching done by the DSPs. Domain-specific heuristic-based reasoning may be very powerful as is observed many times in the fields of pattern recognition and artificial intelligence.

 

[Joel Kaye]

so that at least the DSPs can monitor what is going on. To rectify the problem completely, I think heuristic-based reasoning may have to be used in addition to the pure number crunching done by the DSPs. Domain-specific heuristic-based reasoning may be very powerful as is observed many times in the fields of pattern recognition and artificial intelligence.

Wow, someone should hire this guy to help fix this problem.

Once the image is in Redcine I don't think there's any way to know if it's real purple or clipping artifact purple. Or real black or clipping black that needs to be flipped to white. Sure seems like this must be a sensor weakness. You'd think that must have been tested for before the sensor was approved. Someone must have thought they could solve it.

 

[Philip Powell]

Easy or hard to make happen?

Easy or hard to make happen?

Not having a camera yet to test with yet. I'm curious if this is an easy thing to make happen or not. If it's a matter of the sensor being overloaded, I would think you could never shoot at any light source that was very bright and make it happen almost everytime.

Is it fairly repeatable/predictable or not? Or more random? Just curious.

 

[Mike Prevette]

Repeatable, but avoidable, and easily removable as well.

 

[Joel Kaye]

Repeatable, but avoidable, and easily removable as well.

You mean with a little roto work in post?