15 to 20 stop range with two RED ONE (tm)

[Dan Hudgins]

Projection contrast less than life...

Projection contrast less than life...

Well again, human eye looking straight in to the door will not see the altar details at all, it will take a minute to see it once you turn your head and start staring at the dark of the church interior.

When you are in a movie theatre you need to have the dynamic range within what the projector can produce, so the contrast of the image needs to be altered from life a little to make sure you will be able to see into the shadows and bright areas, something that camcorders have not been so good at. In life the iris in your eyes "expands" the range you can see when you look around as well as does the "negative" feedback of the light sensitive chemicals in the rods and cones being depleted (e.g. snow blindness like effect when you come in from sunshine outside.) Film developer undergoes similar local chemical depletion (and Bromide build up that inhibits density) in high exposure areas of the negative, more so with diluted or compensating developers.

If one could, one would use lights to bring things up that are dark so you can save the highlights. With film you can "pull process" to hold the highlights better when you need to. Using HDR is like "pull process" in that you can save a shot that would require an iris change during the shot or different framing. Its always better to have some image detail to push closer to the top and bottom more with a "S" curve than to have clipped highlights and noise in the shadows.

When you get the HDR into RED ONE (tm) one then just needs to enter the stop spread from like 1/4 stop up to maybe 3 stops, and then you have something that can be used when the lighting has wide contrast. Hard light and soft film usually looks better than soft light and high contrast film, soft light and high contrast film or video makes skin blemishes and uneven makeup look splotchy, with high contrast lighting and soft film/sensor the image tones come from the light and shadow more than the subject coloring, so if you use a little fill light here and there much nicer effects can be obtained.

You can make the change in the mid-tone contrast with HDR settings variability change less with an "S" curve on the result. In the old days the master positive was adjusted in some labs, it seems, with push or pull process to make up for contrast changes in the lighting of the shot, but the full range of the negative was still there in the Master Positive film for the most part just the center of the tonal range was expanded or flattened.

In audio you use compressors to keep the theatre experience comfortable, sound not too loud not two soft, HDR is like a compressor for your eyes, not too bright not too dim...

Looking forward to seeing more examples!

 

[Andrew M.]

Test #2 DR = 14 stops

Test #2 DR = 14 stops

We did test two identical cameras with the same lenses today.
Again the shutter speed difference between two cameras was equivalent of 4 stops.
Here is one frame (frame 16) of the selected clip.
All genlocks and jamsincs were working perfect.
As you see by switching these two frames fast in between we have stereoscopic effect.
It is a bit of problem to correct for it in post during the HDR merge.

Once we have shutter speed bracketing implemented on the RED shutter menu we do not have to wary about stereoscopic effect, since these frames will be produced on the same camera, one dark and one bright, one dark one bright and so on......

I will post this clip on the RedRelay once I will have a bit of time at the end of this week.
I will post it with the surround sound that we were recording.

Note Camera A did trigger few frames ahead of camera B but genloc was keeping all in sync, see the TC.

 

[Dan Hudgins]

Parallax?

Parallax?

As you see by switching these two frames fast in between we have stereoscopic effect.
It is a bit of problem to correct for it in post during the HDR merge.

Please also post a still of the merge of these two images.

When aligned on the screw in the door hinge the parallax on the cars is very large, curious as to what you can do to align them.

If you have a 1/16 inch thick glass from a picture frame you can use that as a 5%-95% beam splitter if you arrange the two RED ONE (tm) cameras like in the attached drawing. Something black is needed for the 5% camera to shoot toward, with this setup the shutter angle in the cameras would be closer or reversed. Having the cameras inline and the same distance from the subject makes the fusion easer. To reduce the small gost image from the far side of the glass you can use a glass that is coated on the back with anti-reflection, and silvered on the front, or a Pellicle Beamsplitter. A cube prism could be used but not with wide angle lenses or at close distances (e.g. not for macro unless you put a close up lens on the subject side of the cube prism and focus the two camera lenses at INF).

 

[IAN SUN]

Andrew's HDR Sync test

Andrew's HDR Sync test

Here are 3 iPhone pics from Andrew's HDR Sync test conducted on Saturday in Toronto. Pictured are myself Ian Sun(red hat), Andrew and Andre Galic.

The bottom one shows the jumper settings for the SincIT.

Thanks to PS Production Services for the loan of the Cooke 18mm.

 

[purefilm]

have been fascinated with the whole concept of HDR Motion pictures. This is a great read and I look forward to seeing the final results. I think is the procress is done right amazing things can come out of this testing.

 

[Dan Hudgins]

Great work!

Great work!

Here are 3 iPhone pics from Andrew's HDR Sync test conducted on Saturday in Toronto.

I found a link to a company that sells a pellicle beamsplitter with a Clear
Aperture of 6"! That should be big enough to use lenses as wide as a 35mm, maybe a 28mm if you can get things close enough,

http://www.geocities.com/smartsense2000/Product/Optics/BSPellicle.html

45x2b.jps is a behind the lens pellicle beamsplitter.

The black background can be a problem for the unused light paths, if you use a ND filter at 45% also you can put a black box behind it and a black box or sheet in front, since the light needs to go through twice it makes the back image twice as dark and the 5% of black on the front with coating may just be 2% of balck is much darker than a black sheet direct as in my drawing above, this is shown in 45x2c.jpg here.

Drawing 45x2d.jpg shows the use of a 45 degree cube prisim. To reduce aberration the camera lenses are focused near INF and a Diopter or large Achromatic Diopter is used to bring the subject into focus. This should work with lenses up to maybe 35mm wide angle or longer.

Drawing 45x2e.jpg shows the use of a single image lens, a field lens, and a relay lens. This alows for the use of 'Scope, Zoom, and Wide Angle lenses as well as easer follow focus. The two long lenses on the cameras are focused at INF all the time, focus on the subject is done with the blue prime lens at the top of the drawing, and backfocus adjustment is done with the large blue lens near the cube prisim.

Drawing 45x2f.jpg shows the ray paths for the focus in Drawing 45x2e.jpg, along with the black box to get the best arrangement that will work with any lens and give good contrast and freedom from most ghosts and dust, the cube prism is much easer to clean than the pellicle beamsplitter.

Most glass shops sell tinted glass that is semi-silvered on one side, if you use the silver as the first surface for the high exposure, and the transmission as the low exposure, the ND of the glass cuts the second reflection somewhat, and so might make a cheep beam slpitter for doing some tests, I think it is called "two way mirror" or something...

Feel free to contact me if you have any questions about these drawings for HDR shooting...

 

[Dan Hudgins]

Blacker than black background for beam splitter..

Blacker than black background for beam splitter..

Another way to get a blacker-than-black backgorund for the beam splitter is to use black paint on the back side of a ND glass at 45 degrees. If you use two of those in a Periscope with a flat black plate or black box at the end, and with AR coating you can get minimum light back bounce.

If the ND part is thick and the black paint on the back index matches the glass very little light can bounce off the backside and get out. So that just leaves the front which can be mulit-coated onto the ND filter, so you get 2% of 2% off a flat black surface back.

 

[Jason Sinclair]

Let me throw another idea into the discussion that could be implemented in software and may not require any hardware changes:
Shooting 48fps or faster and use different exposures for alternate frames. The motion effects could be minimized also in software by using motion detection compensation.
I don't know if this is old news - it just crossed my mind. It seems simple, makes me wonder why it hasn't been done yet.

You're going to get ghosting if there's movement in the shot because if someone moves through the shot in one frame they block the background and in the next frame the background is revealed (where they have moved into) So you can't get the full tonal range in the missing elements.

If this wasn't a problem then rotating a ND on a wheel that spins around and genlocked so that the filter goes in front of the lens every second frame would solve this problem better than using alternative exposures.

If ghosting wasn't a problem this would work.

 

[Andrew M.]

You're going to get ghosting if there's movement in the shot because if someone moves through the shot in one frame they block the background and in the next frame the background is revealed (where they have moved into) So you can't get the full tonal range in the missing elements.

If this wasn't a problem then rotating a ND on a wheel that spins around and genlocked so that the filter goes in front of the lens every second frame would solve this problem better than using alternative exposures.

If ghosting wasn't a problem this would work.

Why rotating ND filter would be better then alternating the exposure?
So far these undesired (in this experiment) 3D effect is making it impossible to merge cleanly two frames of cam-A and cam-B. The best if we could get the tweaked firmware from RED to support alternate shutter speeds. Otherwise I will have to repeat this experiment in the wide open space where the two camera 3D effect will be minimal.

I did cut out shoulders of the runner and did run quick test on it. I don’t see any strobing.

You can simulate it by setting shutter angle at 22 deg and filming runner wearing the white shirt in the sun. Well, I am getting bit unpleasant effect on the right side of his shoulder there, but nothing that I couldn’t take out with adding a bit of blur in the post. AE has ‘add blur’ to the moving objects effect so does the Premiere.

 

[Jason Sinclair]

exposure would be good but just the same reason why nd's are used in other types of production, you have greater choice on range.

The real problem you get however is not related to that but as i said if you have something moving in the shot (or if you are moving the camera) then in one frame there is missing information. Say a subject is moving accross a frame. All parts of the non moving image can be hdr'd because you have two exposures to work with but say a person walks past and in one frame they are in position a then in second fame (with different exposure) they are at point b then you have the info in point a as background however only one exposure information at that point because before (one frame earlier) the subject took up that position (covered the background) and now they have progressed. So when you hrd it it will not work and have a ghosting where the subject was in frame a.

This will look weird and be incredibly hard to fix in post.

 

[Dan Hudgins]

Can exposure be varied frame to frame without clear?

Can exposure be varied frame to frame without clear?

This will look weird and be incredibly hard to fix in post.

The shutter mismatch is why I proposed using something like a 3D or Front projection fixture. With a fixture the shutters can be open the same amount and in phase.

I am sure some here have such a fixture, if they would be so kind as to lend it for some tests the Parallax issue could be eliminated.

I have asked above if someone from the RED team could say if 48fps would work at all for alternating shutter times. I think you may need 96fps so a clear out read can be done before each exposure, and the camera may not be fast enough for that.

The 48fps with ND wheel idea should work for semi static shots since the sensor does not need to be reprogramed on alternating frames.

Since the lens mount comes off the RED ONE (tm) body, it should be possable to bolt two RED ONE (tm) bodys to a block that holds a Pellicle Beamsplitter, because of the size of the front of the cameras it may not be possable to get the sensors close enoough to use PL mount lenses directly, but it may be possable to use 2 1/2 camera lenses that come down to 30mm or so, and to use a relay (printer) lens and field lens to use PL mount lenses by relay optics. Having a modified sensor board that is narower or using a socket extender on the sensor could maybe get the two sensors close enough to use PL mount lenses directly on the sensor.

Since the cameras seem to be able to be run in sync, the simplest way to get high quality results is to align the two cameras in a fixture with a beam splitter.

The cheapest large beam splitter for tests seems to be tinted two-way-glass and to use the silvered side as the front surface. Short of that a piece of 1/16" thick glass would give a general idea if it was maybe 12x18" the second surface reflection would not be too large in the frame.

To check the software issues on mixing the two images, you can shoot some static shots with the camera making two shots in the same camera, this is something anyone with a camera can try. You can shoot one shot at 180 degrees and another at 45 or 22.5 or 11.25 and see how much spread you can take the RED ONE (tm) to before the noise is a problem, that answers the max range issue for two cameras. It would be possable to use three cameras in a fixture as well for more overlap and more spread.

Using one camera you can also try making two shots of a static subject using 180 degrees both times, and using ND filters plus IR filter, and compare the Fusion results to the shutter spread tests to see if using the ND filter wheel idea would impact the color or other image factors.

If you use a ND filter wheel you need clear glass for the other half that will refract the image by the same amount.

If no one can lend a 3D fixture or front projection fixture, you can make a simple one out of plywood and some other bits of wood and glue with a few holes to bolt the two cameras at 90 degrees and the glass or Pellicle Beamsplitter at 45 degrees, you can make a black background by painting a cardboard box black inside. Cover everything with a black cloth, and use flag stands as a matte box. If you glue laminate 4 or 5 3/4" sheets of plywood you can make a block maybe 3' by 3' that is stiff enough to hold things in place with a flex of less than 0.005", the whole thing can be put on a hydraulic dolly or rolling metal tripod (old TV camera type for field use).

 

[Andrew M.]

The 48fps with ND wheel idea should work for semi static shots since the sensor does not need to be reprogramed on alternating frames.

To check the software issues on mixing the two images, you can shoot some static shots with the camera making two shots in the same camera, this is something anyone with a camera can try. You can shoot one shot at 180 degrees and another at 45 or 22.5 or 11.25 and see how much spread you can take the RED ONE (tm) to before the noise is a problem, that answers the max range issue for two cameras. It would be possable to use three cameras in a fixture as well for more overlap and more spread.

I am trying to avoid using any mechanical solutions.
As to the argument that some info is missing when the person moves I have to tell you that at 180 deg shutter speeds and normal shooting there is lots of info missing anyway because of blur.
Long frame don’t have to be modified at all, blur is there and it looks nice.
The short frame (22 deg) is sharp and doesn’t contain the area where the other frame has blur in it.
The way around it is to create the motion blur for the second frame that will match the first frame.
I found out that it is necessary only for high contrast high exposed sharp edges objects.
All other details look very natural.
As to how difficult it is to do it in post, here it is the snapshot of the screen.
I use AE and HDR effects that comes with it. I didn’t research yet in terms of plug-ins what is out there for HDR moving images. This is for long clips.
Other option is to use PS and its scripts for automatic HDR merging, very slow process but very accurate. It compensate for object motion in case if you shoot only 5 frame per second and has ability to process blacks and high light area separately. You can also modify scripts for some special effects.

In the pictures below I just put one clip on top of the other without any additional processing.

I aligned the frames on the edge of the left side door frame. There is very severe 3D effect since we had cameras 20” apart with 5’ distance to the door frame.

Look how AE can take two completely unusable frames (two on the far right side), one overexposed and one completely dark, and create perfectly exposed 14 stop lit scene.

Sure at the end you have to end up on the projector with 10 stops DR maximum (same thing with film) but you are the God deciding which stops to show where and how on the frame.
Having 14+ stops for post to decide is priceless.



We will try to do 3-rd test in the large interior with dark and bright section in it.

 

[Dan Hudgins]

PLZT or LCS?

PLZT or LCS?

I am trying to avoid using any mechanical solutions.

It may be possible to PLZT (lead lanthanum zirconate titanate) or LCD electronic shutter over the lens and run at 48fps thereby not need to use a filter wheel on a single camera. PLZT has small wires on it and may reduce the sharpness. LC Shutters like used in 3D glasses are somewhat uneven. Both electronic shuttters could be driven from a J-K Flip-Flop that is triggered by the VHz (48-120VHz)...

Does someone know how to get a frame start or end pulse (VHz) out of the RED ONE (tm)?

The question of if the camera can shoot with different shutter angles on odd and even frames has not been answered here, does someone know if it is possable?

Sure at the end you have to end up on the projector with 10 stops DR maximum (same thing with film) but you are the God deciding which stops to show where and how on the frame.
Having 14+ stops for post to decide is priceless.

You don't need to lose any of the 14+ stops in subject details, the brightness slope angle gets lower (contrast less than in life), but if you dither the 14+ bits before you convert to 10 or 8 bits in a high resolution uncompressed image you will retain some of the details between the binary steps of the output device, and many subject details in the top and botton 3 steps are more than one step anyway, so would also get through. Dithering and noise help keep image tones, unless the images get compressed then you end up with tone banding and blotches. There may be an advantage to outputing at 4K in the film recorder dithered uncompressed 48bit TIFF rather than 2K since four 4K pixels average into one visible spot in the film and projector to extend the tones that the eye can pick out as the film flickers on the screen. In other words, the four pixel 4K group can have one of four brightness levels so adds about two effective bits to the film recorder D2A.

How much parallax is too much for your software to tolerate for Fusing the images together (for myself I would try to use a fixture and get as little parallax as I could to get the most natural result)?

Great work so far! Can't wait to see more sample images!

 

[Andrew M.]

Shutter speed change is instantaneous, also it doesn’t require change of compression algorithm.
It is just a parameter in the software fed in to the Mysterium. I hope it don’t have to be preloaded?
PS is pretty good in correcting for small 50 pixels offset. AE is not good at all but it will take any compatible scripts from PS and it can call (I think) PS process for help. This will, I imagine, slow AE considerably when rendering.
What works the best is correction for moving objects, like people running across the screen in the background. People walking near by. It just adds the blur to connect the first frame image with the second one. Also all the PS scripts work much better with 3 pictures bracketing not two.
So -2, 0, +2, instead -2 and +2 only. All these could be changed in the script or plug-ins.
Sure, shooting 72fps and have 3 frames shutter bracketing would be the best.
I guess there will be many situations where shutter bracketing may not work well but I think in the most cases, advantages will outweigh disadvantages.
Like the picture above (post 35), large area, people walking, glass dome with high-rise buildings seen above the dome and clouds on the sky, this is the best environment for the HDR shots.
Also speciality shots, underwater, nature, science and commercials, shapes and details of close-ups, are lending itself for HDR technique. Only imagination is the limit.

But we need shutter speed bracketing to show it.

 

[Zaphod]

Wouldn't it be a lot easier and cheaper to just use a camera that had a decent amount of dynamic range like a F23 or Genesis or whatever?

Why flog a very dead horse?

 

[Dan Hudgins]

Open your eyes...

Open your eyes...

Wouldn't it be a lot easier and cheaper to just use a camera that had a decent amount of dynamic range like a F23 or Genesis or whatever?

So far as I have been able to see F23 or Genesis do not have a true usable 16 to 20 stop range, that you can get with two or three RED ONE (tm) cameras. There are no sensors that can equal the range of color negative film in 65mm or IMAX when pull processed or re-processed with Phenidone-A (1-PHENYL-3-PYRAZOLIDINONE) as the first developer, if you have an example image to disprove that, post it.

There is too much that is "easy and cheep" about what is going on in movie making today. Why not make an effort to get the best results possable, and put yourself ahead of your competition and make something of lasting value.

The 48fps idea would not cost much at all, and some people have two RED ONE (tm) already and maybe a 3D fixture so there is little cost there.

Why flog a very dead horse?

With good 4K projection coming the goal is to expand the art of cinematography to have something that has not been seen before, to make beautiful images with a range and sharpness never before obtained on anything short of maybe 8x10 color seperations (or Digital HDR stills).

It is a tool for the cinematographer to get away from the blasted, compressed, over processed, over peaked "Video" look, the whole point of the RED ONE (tm) is to over an alternative to the "Video" world at a price that puts film like digital shooting within the cost reach of cinematographers that would otherwise be out of work and replaced with videographers.

If you can't see what makes HDR images useful, what can I say, to me the usefulness and beauty is immediately obvious.

From just a practical stand point having an extra 2 to 5 stops up and down can save the day when you cannot hire enough generator trucks and lights to use for fill lighting. You can also get less noise the shadow areas since the dark and light subjects are moved closer to the "sweet spot" of the sensor for both the highlights and shadows something that no single sensor can do.

Also if you use two cameras the fixed pattern noise is reduced by fusing the images from two sensors.

 

[Dan Hudgins]

More than one way to skin a cat...

More than one way to skin a cat...

Shutter speed change is instantaneous..

Has someone from the RED team told you that? From what I know, you may need to do a clear readout after adjusting the shutter registers, that would make the max shutter time equal to 90 degrees for the long exposure, not good for motion work or low light, but OK for static shots.

Sure, shooting 72fps and have 3 frames shutter bracketing would be the best.

If that could be worked into the camera it would give you more range, maybe more than the lens can hold or you can get enough light range to shoot under, so maybe nothing would fall off the top or bottom. That's latitude!

But we need shutter speed bracketing to show it.

You can get results today with a beam splitter, you can test the shutter issues with a beam splitter... by adjusting the frame rate, sync phase, and shutter openings on the two cameras (using every other frame from the two sets.)

There is no reason to limit HDR shooting to one approach, there are many ways to do it as have been gone into above, and there are no "deal breakers" since one or more of the approaches should give usable results.

It may be possable to do "pixel shifting" of the sensors for close range exposures and get "8K" like images out of two RED ONE (tm) or "12K" like images out of 3 cameras. You offset the sensors by 1/2 or 1/3 pixel and also offset the colors in the Bayer filter then you can reduce the color moire without the OLPF or with a weaker one.

 

[Andrew M.]

Dan, when it comes to the use of two sensors, or two cameras, the easiest way is to shoot using two independent units with applying ND filter or aperture difference on one of the camera.
Someone could design mechanical coupler splitting image in to the two independent PL mounts as you suggested above. I see the same problem though as on the prism RGB broadcast camera.
We end up having problems with CA.

As to the question of shutter bracketing, RED didn't confirm if they can change shutter on the fly without any problems.

I hope that it is possible, however if Mysterium have to be preloaded with the shutter angle before triggering data acquisition from the sensor, then it may create problem. May not be possible to preload shutter angle 48 times per second without major firmware reprogramming.

 

[Dan Hudgins]

No Chromatic Aberation issue with pellicules?

No Chromatic Aberation issue with pellicules?

Someone could design mechanical coupler splitting image in to the two independent PL mounts as you suggested above. I see the same problem though as on the prism RGB broadcast camera.
We end up having problems with CA.

A Pellicle Beamsplitter does not introduce the CA (Chromatic Aberation) issue so that is not a problem.

I was thinking that the whole mount should come off the two cameras to mount both of them onto a single block with one PL or 2 1/4 lens mount. But your idea of having two male and one female lens mount on a thing could be made to set up quickly, if not quite as firmly.

You can also use the Pellicle Beamsplitter in front of matched lenses without CA as well, that is what they are made for. If you look at my drawings and posts above you can see the idea.

Links about Pellicle Beamsplitter,

http://www.geocities.com/smartsense2000/Product/Optics/BSPellicle.html

http://www.newport.com/pellicle beamsplitter/

Maybe you can shoot some single camera tests in bright daylight with subjects 2/3 backlite and sky and sun in the background so we can see how the RED ONE (tm) stands up to over exposure when the images are fused. If you post both the high-low and fused images we can see what the camera does with the over and under exposure on different stop spreads and what the fused image looks like,

i.e.

Set zero two shots at +/-0 exposure and those fused.
Set one -1 stop and + 1 stop and those fused.
Set two -2 stops and +2 stops and those fused.
Set three -3 stops and +3 stops and those fused.
Set four -4 stops and +4 stops and those fused.
Set five -5 stops and +5 stops and those fused.
Set five-B -5 stops and +/-0 stops and +5 stops and those fused.
Set six-B -6 stops and +/-0 stops and +6 stops and those fused.

That test would answer what the best possable results would be without parallax errors and other non HDR issues. Since the camera is said to have 11 stops range, going -4 and +4 should be within the clip points, but the sensor may flare so that is an important camera related issue I am sure everyone would like to get a heads up on. -5 and +5 would probably look better with a +/-0 frame, and a three camera or exposure HDR is possable in some cases. The camera EI ("ASA" does not apply to Digital Cameras, EI is the correct term) should be set to 320 to give the most centered exposure of the sensor for all exposures.

Can't wait to see your next results...

 

[Uli Plank]

BTW, I notice that the Pellicle Beamsplitters are offered with different ratios, which should be another advantage. You can feed more light into one of the cameras right from he start.

I'd like to second the value of HDR, having done still photography in HDR for years I really appreciate the painterly effect.

Regards,

Uli