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 :)

Dynamic Range Trick?

D

dVonAllmen

Well-known member
Joined
Aug 29, 2007
Messages
87
Reaction score
0
Points
0
Age
54
Website
www.driveproductions.com
Let's say I want 4K with more dynamic range than Mysterium offers (or Monstro will offer - hey we can always use more dynamic range, right?)

Red will soon offer 6K, which is roughly double the pixels, or on the chip double the cells.

What if Red made a 6K chip that had a pattern in which every other cell was more sensitive and every other cell was less sensitive - allowing them to record different brightness ranges. In post they would combine in much the way de-Bayering works, and you would have an incredibly large dynamic range on a 4K image.

Sure, it's more data, but with Moore's law and all that, more data is not a technological hurdle.

I don't need higher res than 4K, and will only occasionally use higher frame rate than 30. But all of us could use more dynamic range on almost a daily basis.

Imagine dynamic range that far surpasses film...
 
Hi and Low pixels

Hi and Low pixels

That is a good idea except that the low light pixels will have their highlights very overexposed so you need to avoid those streaks that show up sometimes when the sensors are overexposed.

It the camera saved true RAW data then no extra processing would be required, but since there is low pass for the compression you would need to break the image into 8 images rather than just 4 before compression, that is:

Hi - Red
Low -Red
Hi - Green 1
Low - Green 1
Hi - Green 2
Low - Green 2
Hi - Blue
Low - Blue

It would not take more bandwidth since you can just double the compression ratio, something that is done anyway when you record at a higher frame rate. When the de-Bayer program reads the 8 compressed monochrome images it puts the image back into two Bayer patterns to de-Bayer a color Hi and color Low image.

A few simple changes to the camera, and putting darker colored or ND filters on the sensor would make this work. You do get some aliasing because of the stripe effect of the high and low sensors, but since the pixels can be very small and the OLPF blurs the image that should be able to be dealt with.

===

Some time ago I posted some drawings of how you can use two RED ONE to shoot with more dynamic range by using a 5%-95% splitter, if you arrange the optics right that can also work, TODAY, without any needed changes.
 
2xRGGB vs. RGGBY or RGBY

2xRGGB vs. RGGBY or RGBY

Christian Edwards said:
here here.even 3/4 seperate cmos sensors each for R,G,G,B. coupleing device optimized under bayer and light sensitivity filter .raw RGGBY with no noise
Click to expand...

It would be best to have full color information at all brightness levels, if you just have luma in the highlights you cannot get good color all the way along the curve, and if you underexpose the color you get more noise.

Since the pixels can be made very small in a FF35 or even S35 sensor, the noise would not be an issue for the two ranges since you would not need to pull the low end up as much as you would using just one range. You would mix both ranges so the pixel area would be about the same for part of the curve in the center.
 
dVonAllmen said:
What if Red made a 6K chip that had a pattern in which every other cell was more sensitive and every other cell was less sensitive - allowing them to record different brightness ranges.
Click to expand...

There are dozens (hundreds?) of ideas for extending dynamic range by modifying the sensor hardware. Most have been patented. Many have been prototyped and proven. Very few have made it to market yet (Fuji's has, for one).

I am sure that RED's sensor designers already know a lot about these ideas. Since RED is on the cutting edge of sensor design, I would not be surprised if they are the first to implement some form of dynamic range extension.

However, the pace of sensor technology advancement is going very fast already. I would guess that "normal" improvements (resolution, QE, read noise) start to slow down before something like this is brought to market, kind of like how per-core speed improvement has slowed down in processor development and more attention is paid to increasing the number of cores now.
 
Daniel Browning said:
I am sure that RED's sensor designers already know a lot about these ideas. Since RED is on the cutting edge of sensor design, I would not be surprised if they are the first to implement some form of dynamic range extension.
Click to expand...

Daniel, the sensor is always the media-bizz and poster-chip of DR/SNR talks. However, unfortunately, little attention is given to the electronics circuitry around the sensor. If you look at all those "low-noise" cameras, you would realize how much importance is given to circuitry surrounding a sensor in addition to standard sensor level noise sources (dark/shot/read/PRNU/thermal/ALIEN/Jupitor/Saturn/Michigan/Moon/Earth/ ...../Cheetah/..../etc.)

Unfortunately, and even perhaps unrelated to the noise problem, it appears to me, and I could be wrong, that Red underestimated the importance of circuit design hardware in their first level product. This may be evident from the difficulties in making "obsolescence obsolete" product line.

Red may have a hardware level team that is invisible to us. However, the frequent posters from Red on this forum seem to have a stronger software background. Now that Red has announced a revamped and polished product line, I sincerely hope that Red has come to realize the importance of circuit board design now.
 
Dan Hudgins said:
That is a good idea except that the low light pixels will have their highlights very overexposed so you need to avoid those streaks that show up sometimes when the sensors are overexposed.
Click to expand...

HDRI software allows you to take multiple images of the exact same thing at different exposures and throw away the blow-outs and underexposures to make one image. I was assuming a post step that does the same thing.
 
What about shooting in, say, 48fps with alternating shutter speeds? In post, you can HDR both of those into one 24fps stream with a very high dynamic range.

I suppose fast motion would be a problem for that...
 
Joe Penna said:
What about shooting in, say, 48fps with alternating shutter speeds? In post, you can HDR both of those into one 24fps stream with a very high dynamic range.

I suppose fast motion would be a problem for that...
Click to expand...

I mentioned that one a while back... no response from Red on implementing it, probably because they've been busy with more pressing issues. Perhaps Red will revisit the alternating shutter approach once their development pace begins to cool off.
 
Joofa said:
Daniel, the sensor is always the media-bizz and poster-chip of DR/SNR talks. However, unfortunately, little attention is given to the electronics circuitry around the sensor. If you look at all those "low-noise" cameras, you would realize how much importance is given to circuitry surrounding a sensor in addition to standard sensor level noise sources (dark/shot/read/PRNU/thermal/ALIEN/Jupitor/Saturn/Michigan/Moon/Earth/ ...../Cheetah/..../etc.)
Click to expand...

Good point. There are plenty of ideas for dynamic range extension by changing just the electronics and leaving the sensor itself alone, such as:

http://www.patentstorm.us/patents/6734905/fulltext.html
 
Lowkus said:
I mentioned that one a while back...
Click to expand...

Dozens of people have mentioned it a while back, ever since the reduser forum opened.

It would not be a feature that I am interested in, because motion would not match in the two images.
 
To Dan Hudgins, David VonAllmen, Christian Edwards, Daniel Browning, Joofa and all those deeply concerned by dynamic range


Back in 1981 at the Smpte exhibit in L.A. there was on a booth an IKEGAMI 35 video camera on a tripod looking facing a set with a highlighted window in a part of it .
At first, the window appeared washed up, like a white rectangle on the monitor.

Then, at the press of a button, you could have details outside the window appear on the screen.
NOT changing anything on the remaining image.

If you pressed the button twice or more, more details through the window would come until their level would reach the level of the" indoor set".
That was almost too much, a creamy look.

Then you could switch back, to get exactly the amount of contrast you wanted.
You could balance the amount of light between outside and inside.
They were using an "electronic mask".
That was more than 25 years ago.

How come we cannot achieve better dynamic control now ?

What about a Scarlet or Epic with variable dynamic before recording ?
Just by turning a wheel !
NOT POSSIBLE , WHY ?
 
Daniel Browning said:
That could have been done in software only. Did they show it working with lots of movement in the frame?
Click to expand...

Of course there was some electronics playing the mask in real time,
and Yes, you could move in front of the window without any visible damage.
 
The look

The look

LEON said:
How come we cannot achieve better dynamic control now?
Click to expand...

I could show you what is possable to do with the data from one sensor if the RED ONE saved true RAW Bayer data, but it does not.

If I did have the RAW Bayer data it would probably show that the green pixels clip before the blue and red pixels, that reduces the dynamic range to less than 12bits, so that is one reason that the RED ONE does not show the dynamic range that 12bits would seem to show. The range extension past the color that clips cannot have full color information that is the right balance.

When you shoot under low K values the software white balance has to deal with unbalanced clipping which reduces the dynamic range also.

If you have long dynamic range in linear brightness the result will not look like film, but if you use long dynamic range on an S curve you can get more of a film look.

Having a longer dynamic range will not make the image look worse if the right curves are used and the subject lighting is done well, anything can be done badly and look bad.

As I pointed out before you can use two cameras to get enough dynamic range to make images that are more film like.

The people who are asking for LESS dynamic range because of compression issues have a point, but using two cameras with overlap would reduce the compression artifacts there also and give better results.

Masking will not help with the sensor unless you shoot through an LCD screen to act as a negative mask, once the sensor clips you have no true data to work with.

With a true RAW camera you can underexpose more to hold the highlights, but in the RED ONE the compression limits the amount of underexposure you can use before the compression thresholds in the shadow areas become an issue, you can see the shadow issues better if you make the image negative and change the gamma. The RED ONE produces nice images if you can balance the loss of highlights with the shadow issues by using enough fill light, but for working under uncontrolled light having five more stops of range would be useful. In a two sensor camera you could adjust the ratio of exposure and improve the results from less noise in less dynamic range to more dynamic range before clipping.
 
Bérenger Brillante said:
this is the technology used on the Fuji S5 Pro.

Two type of photosite, for low and high light. It works fine, but
patented by fuji.
Click to expand...

One of the requirements for a patent is that the idea you are patenting is "non-obvious". Considering I know little about how this stuff works and thought of this idea, I would imagine a lot of other people have, too.

Seems a shame that a company is allowed to patent an idea that when you get down to it is pretty basic. Would be a shame if Red was prohibited from implementing this for that reason.
 
well it's a guess.. but i think other companies would have done it too, if it weren't patented... The increased dynamic range of the s5 is pretty good for a small and cheap camera from 2006. And only 6x2 Megapixel. I wonder how it will look on their next model, with 12x2 mpx, if it's released one day.

Look at this photos gallery from a short movie, shot with Red, to see pictures from this camera :

http://onverraplustard.free.fr/www/#gallery/Photos_de_Plateau_2008/Bunker

The lightning was heavily contrasted.
 
If it's a raw sequence of images coming off the camera already, has anyone tried batching the stills through photomatix for a pseudo hdr effect?
 
You must log in or register to reply here.
Back
Top