"Spiderhole" Feature Film Frame Grabs

[NormLi]

What ASA and T-stop ranges were you using in these shots? I find the camera pretty noisy in low-light situations so far, even at 250ASA and T1.3.

Any ideas on how to reduce this noise? The blue channel is the most apparent.

 

[Noah Kadner]

WHAT?!

They turn them off here!

True sometimes they leave them at like 5% which is annoying as hell. Maybe they've been sued by people tripping in the dark- duh...

Noah

 

[Dan Hudgins]

Dark room or not?

Dark room or not?

WHAT?!

They turn them off here!

I was at the Castro some years ago and some visitors from Australia were complaining to an employee that they forgot to turn the lights off, the employee explained that they turn them down but not off. The vistor said that in Australia they turn the lights off, so I have heard this before.

They used to have ushers with flashlights (torches) to help people not trip, but they do not have ushers any longer so they have lights on the walkways and leave the lights on more than dim in most houses around San Francisco.

I think it is for two reasons or so, the Disability Act, not wanting to get sued when people stumble (the insurance company may REQUIRE that the lights not all be turned off), and to make it easer for people to see when they go out to get pop-corn (much of the profit comes from food sales, so they want people to get refills during the movie). I saw a story about a theater that has tables in the rows and serves dinner in the theater now, so they would have to have the lights on to some degree.

The bigger issue is that the black level on DLP projectors and TV sets can vary more then the upper gray level, so what looks just acceptable on YOUR monitor may look way too light or dark on other monitors or projectors. If you use the full tonal range and have the subject mid way in the tonal range then SOMETHING will be visable to the viewer no mater what he sees the movie on, like an airplane and such. When there were drive-ins the studio had the lab make extra light prints, taking up the exposure maybe 5 lights. Since film has an S shaped curve, changing the lights in the priinter changes the contrast AND density on different image tones, on linear systems like Digital the brightness control does not equate to changing the printing lights on printing film.

My Brother always wants to "make it darker" he says that "I want black black", but the new color print stocks do not have a black, the emultion is put on so thin that it does not get as black as the older black and white print stocks. Some labs may be using exhausted bleach, so the silver left in, like bleach bypass, can make the blacks better on some prints you may have seen, but the colors will not be as bright and clean because the silver gives the images a mirky look, just what the "film look" people like. Film never looked that way when processed and printed properly, bad color came in with the EPA and EDTA bleach which does not get all the silver out most if not all the time.

If you want deep blacks in your print, you can try to bleach bypass AGFA CP-30 print stock it may have more silver than the Kodak print stocks, I cannot say about Fuji or Lucky since I have not done tests on those.

I saw the last (?) Star Wars movie on both Film and DLP, in the dark shot when Anakin Skywalker gets out of bed he was clearly visable on the Film print, but almost invisible on the DLP projection, the difference on the lighter shots was less noticeable.

When I was in Norway, the Theater I want to had the lights quite high during the film, so you need to understad that the screen size will affect the brightness of the projection.

In our screening room we can change lenses to get different size images, but the birghtness goes down with the larger image. Some stuff we did a while ago looked just awful projected with a sharp 3" lens that filled the screen, but looked much better with a less sharp 5 1/2" lens. Making the images smaller, brighter, and less sharp reduced the grain and increased the mid-tone contrast to the eye, since the eye is not linear. It was quite surprising how much better the images looked when brighter and lower resolution.

I saw a print of Laberinto del fauno, El (2006) aka Pan's Labyrinth that was so dark on some shots that it looked like a lab error or the film recorder was not calibrated for a black level that matched the monitors being used for grading, clipped at the black point so maybe they wanted to hide the sets or something, to me it just looked like a mistake, like the print of Orson Welles' Othello that the lab screwed up the light change tape so that the wrong light changes where happening in the middle of shots.

David Lynch said that when he was working on Eraserhead (1977) that he kept asking the Lab timer to "make it darker" and the guy came back with a comment that "you want it so dark people won't see it" or words to that effect.

In the 20's some filmmakers like Thalberg (?) started to experement with low fill light levels and high contrast ratios on the back light with forced processing to get a expanded highlight range on the print. Unfortunately many theaters did not have arc lights bright enough to show the dark prints of those negatives with enough light through the mid-tones, so the audiences complained, and in turn the theater owners complained to the studio bosses about not sending them "underexposed" movies. And so from that point in the 20's right on through the 60's it was the practice in Hollywood productions from major studios to keep the fill to backlight ratio high enough that the density on faces of the actors was around density 1.0.

Just like you cannot tell where the edges of the frame are going to be with some theaters showing both 'Scope and Flat movies at 2:1 ratio, you cannot tell how bright the projector will be or how much ambient light there will be.

I have been having problems with things not matching exactly, not because I an doing something wrong to any degree, but because monitors, TVs, projectors, Film, and such CANNOT ever match exactly, no mater what anyone else may tell you. So I have been thinking that while grading you should use SVGA "Y" cables or splitter boxes, and have the signal going to various CRT monitors, LCD monitors, OLED monitor, projectors DLP and LCD, and through a SVGA to composet converter box to hook into CRT, Plasma, LCD, OLED TVs, yes the grading room would look like a TV store. Then split the difference between what looks good on the best device and the worst, since they will not match.

The point made about the artifacts looking bad when the film is moving is VERY true, I have stuff that looks good one frame at a time, but looks like holly hell running uncompressed at 24fps, and through our dim 35mm projector even worse at 24fps, much worse, any flicker or uneveness shows up amplified at 24fps because of the flicker threshold of the eye, at <8fps or >80fps these can be less noticable.

You can output your frames as 720x480 24bpp BMP files and use free VirtualDub (tm) to make an Uncompressed AVI file, and Uncompressed AVI should run in Media Player under XP on a 3GHz computer with GB+ Memory and fast harddrives. That is cheeper than making a film out to see the motion flicker and artifacts caused by the RED CODE (tm) compression in the shadows. It is still better to do a full filmout and project the answer print to see just how f**ked up the results will be when moving. Be sure to have some ordinary people in the audience and don't tell them in advance why you are doing the test, see if they notice "anything" about the quality of the "movie".

I hope you take these coments as they are intended, just pointing out that the distributor, exhibitor, audience, lab, store selling the DVD and getting returns on the "defective" DVD, may have expectations of what profesional cinema images "should" look like, and it may hurt your bottom line to buck the system too much.

To the eye, to see black on the screen rather than a washed out grayish uneven mirk of the film's D-Max, there needs to be something bright in the frame. Your shot with the flashlight works better in this regard, whereas the other sample image does not have a kicker, reflections in water, or other highlights to give the eye a white reference to judge the other image tones from.

I should add that HD TVs seem to have too much sharpening, so compression artifacts in the shadows may look worse on them than in a filmout projected a little out of focus, since film and projection lenses work like a low pass filter, and HD TVs and such may over sharpen images as a high pass filter, vs the cleen image you see on a good monitor used for grading.

 

[Dan Hudgins]

Shoot high EV, time down.

Shoot high EV, time down.

What ASA and T-stop ranges were you using in these shots? I find the camera pretty noisy in low-light situations so far, even at 250ASA and T1.3.

Any ideas on how to reduce this noise? The blue channel is the most apparent.

The "Darkness" is relative not absolute, so use more light on the set in the "same" ratio (the highlights will need to be controlled a little more), this requires having the kicker and other backlights very high, maybe +5 stops over fill so you will need at least 2K watts, 5K to 10Ks would be better. The total wattage on the set could be 50K or 100K or more to get a realy dark look.

Its OK to have 80% of the frame -5 stops from the back and rim light highlights, as long as the highlights are right at the top of the sensor clip level, since the RED ONE (tm) is said to have a 10 stop or so range.

You can always make the mid-tones darker after shooting, that is better than to bring something out of the compression artifacts because digital images have less data bits at low levels. The eye works the other way around it compresses the highlights and expands the shadows. Also the "black" parts of the frame look blacker if 20% of the frame is above density 1.0 on the print, 0.3 if possable. Having piratical lights in the shot can help give the highlights required, and using Silhouette with the fill at about -1.5 to -2.5 stops from mid tone and highlights at +5 stops.

You should be able to use the waveform display to see where your 18% gray card is in the sensors range. The blocking artifacts probably increase below 18% gray and above 90% white. You probably want the 18% gray card to show as maybe 5% linear signal. Do tests before shooting.

As for blue noise, if you could look at the raw bayer data you could see the densty of the red vs blue pixels, since the RED ONE (tm) does not supply RAW BAYER DATA, you can look at red and blue seperation images and adjust the filter over the camera until the noise in the RED and BLUE images is about the same, you may need to add CC filters to an 80A or 80B filter to get the light to the natural balance of the Bayer filter+sensor's K degree. Set the sensor K to default for equal gain on the RED/BLUE perhaps 5500K or so (?).

 

[Noah Kadner]

Its OK to have 80% of the frame -5 stops from the back and rim light highlights, as long as the highlights are right at the top of the sensor clip level, since the RED ONE (tm) is said to have a 10 stop or so range.

Have you actually tried that? I think you'll find that 10 stop range or no anything that's 5 stops under key is going to go jet black and not come back without bringing a heck of a lot of noise with it.

Noah

 

[Lexicon]

Newer theaters have lights in the floor along the walkway and leave them on during the feature along with lighted cables running all over the floor. Older ones usually leave the lights dialed down to almost nothing.

 

[Dan Hudgins]

Stops and 2^n

Stops and 2^n

Have you actually tried that? I think you'll find that 10 stop range or no anything that's 5 stops under key is going to go jet black and not come back without bringing a heck of a lot of noise with it.

Noah

The RED ONE (tm) sensor is said to output 12bits, each bit is a linear power of 2 so if you take the bottom two bits is just noise that should make the 3rd bit as singnal. with the white clip at +10 stops. Lighting so that the rim lighting is +5 on incident reading is not the same as lighting for +5 on a reflected reading of a colored subject, sorry I was not clear about that, I would expect that one would be using a meter with an incident dome on it, and read with the dome pointed into the beam of light for the back light, and to the camera in the shadow of the subject for the fill.

If the RED CODE (tm) is using delta modulation and has the threshold at +7 bits then yes, the faces could go black, but I think the threshold would be lower down near the 4th or 5th bit, maybe someone from the RED TEAM can go into detail about the threshold used for the blocking in RED CODE (tm), they have said that it works like noise reduction since they have said that the RED CODE (tm) encoded images show less noise than the raw sensor data, meaning they make adjacent pixels the same value, perhaps the average of the two pixels if the difference is not above some threshold.

If the fill is between -6.5 and -7.5 stops from white clip level of the back light, that should be above the "black level" of the RED CODE (tm) even if they were only coding 8 bits, I assume the RED CODE (tm) codes for 12 bits since to code for 8 bits you need a LUT to do gamma up on the mid tones in the camera, and the RED TEAM have said that the camera outputs linear 12 bit data not adjusted 8 bit data. 12-7=5 x^5=32, so at 7 stops from white clip the camera should still have 32 shades darker, if it was outputing RAW Bayer data, if you do not get 32 shades in the bottom 5 bits then the RED CODE (tm) has done something to the sensor data that limits the sensor tonal range, or the sensor does not output 12 bits of data, so if the bottom two bits are dropped and you get 10 bits then there are 8 shades under -7 stops from white clip before you hit black.

Because the sensor is linear and not S shaped like film, the backlighted 18% gray card will need a fill light level (reflected reading this time) that is about 5 stops under white clip since 90% white is at +2.5 stops above the 18% gray card reading (making the reflected white card reading -2.5 stops under the backlight), to shoot "Dark" you do not want white subject matter to clip in the shot, you want white subject matter to be 18% gray, only reflections and other "brighter than white" things in the frame should be close to white clip of the sensor. In other words a reflected reading on the fill on a backlighted white card would be -2.5 stops from the incident reading of the back light at the subject with the dome pointed into the back light. The incident dome reading in the shadow of the subject pointed to the camera would be -5 stops (since the dome reading the fill will make a 18% gray subject 18% gray, or on this case you work out the EI of the camera so it is adjusted to give +2.5 stops of head room above a 90% white object) with some parts having fill down to -7.5 stops reading on the incident dome. If the subject was near 18% gray in reflectivity of the fill -5 stops from white clip should still be above the black level by +5 stops, and darker subjects would have +2.5 stops above black level for 10 bits.

After you de-redcode the images and make uncompressed DPX files, one should be able to lower the midtones without building compression artifacts in the shadows since you would have 14 to 16 bits of tones, but that would only work if you are going to output in an uncompressed format like film and the film recorder has enough range to output the bottom of the tone range.

I do not have a RED ONE (tm) here now to test, there seems to be some controversy about the real world tone range, that may be due in part to people not understanding that if you put a gray scale card at +2.5 stops ALL the patches above 18% gray would be clipped white, if you expose to hold the 90% white patch, you are -2.5 stops below the white clip, so you would then have the bottom (for seeing detail in an 18% gray subject) at -7.5 rather than -10 stops.

Normally you would not want 90% white to be white clip in a Digital Cinema camera, so you would expose so that a 18% gray card subject is about -5 stops below an 18% gray subject being at white clip. You should then be able to take the 18% gray object down -2.5 stops more (-7.5 total from white clip) and still have it look good being +2.5 stops from black clip, if not then the range of the camera seems too limited for film like filmmaking.

You can test the camera, just set up a gray card and a subject like a ball spray panted 18% gray. Adjust the flat lighting so that the gray card white clips on the waveform display with the lens set to T/1.4. The gray panted ball being of narrow tones will show blocking under flat light. Then stop down -5 stops and see if the image looks good after being adjusted in color correction. You should then be able to stop down and ND -2.5 more stops (-7.5 total) and still see the ball and gray card well with maybe some blocking in their shadow. At -10 stops from white clip the gray card should be the same tone as its shadow if that is the black clip of the camera, and the ball would be a black disk. If the ball is not a black disk then you might have to go to -11 stops to reach black clip.

If some subject is colored darker than 18% gray in the shot, then you have to put more light on it to avoid black clip, that is what spot lights are for. If a subject is lighter than 18% gray you need to flag light off as needed to keep things from going white clip that you want to keep in the DPX files.

Sorry I was not clearer about how the light readings would be taken before.

Another related issue is that not all TV sets have "DC restoration" so the black level floats around depending on the contrast range of each frame, in other words a flat dark shot will become a flat gray shot without your control. I think the TV networks used to have a issue about the contrast range of film shot that would be broadcast to try to keep something light in each shot so that the TV sets without "DC restoration" would hold their black level better, in other words each shot would have some full white in it to give the waveform a "top" to "bottom" swing. When film prints were run through "film chains" the AGC of the camera would also make the black level float around because TVs with "DC restoration" would go too dark on shots that graded night for cinema projection, that would be too far down for the dynamic range of the TV camera in the film chain. Think "I Love Lucy" show, balanced tones for TVs without "DC restoration", still plays well an any device.

 

[Vinit Borrison]

I'll post some original frames a little later, trust me we're shooting at t2 at 320 so we're pumping light on set and grading it pretty dark. i'm surprised the flashlights are holding up so well.

 

[Vinit Borrison]

http://idisk.mac.com/vborrison-Public?view=web - one raw tiff from redcine

i thought pc gamma was closer to 1.8 and mac's are somewhere around 2.2?

 

[Dan Hudgins]

Gamma test for your monitor

Gamma test for your monitor

http://idisk.mac.com/vborrison-Public?view=web - one raw tiff from redcine

i thought pc gamma was closer to 1.8 and mac's are somewhere around 2.2?

Here is a link with a gamma test image you can look at from a few feet back to see what your monitor gamma is, mine is 2.4 since I am not using any software to goof with the monitor or video board, just raw data going out the SVGA pins,

http://www.cbloom.com/3d/gamma.html

Some other test images use a checker board of black and white dots for the reference 50%-50% but that is a BAD idea since the monitor peaking will give a false reading, this image seems to use horazontal lines of white and black to give the 50%-50% reference which is better and closer since the peaking is mostly horazontal for monitors, if your video board or monotor uses vertical enhancement it would not give a true reading. Do not rely on the numbers in your software for the gamma, since adjustments to the monitor contrast and brightness can make it off a little. I have added black and white reference blocks as well as gamma reference to my color correction GUI screens so that when the screen is saved as a BMP file for viewing on another computer the monotor can be adjusted.

You can download the old KODAK (tm) LAD image from my Web site, it has a black, near black, mid-tone, near white, and white refernce and an image of a woman that you get to know so that when you move from one computer to another you can check the monitor before you do any work. If you use virtualdub (tm) and make copies of the LAD frame you could make a 2K or 4K uncompressed AVI to put on the head of your edits and DVDs to adjust the monitor before viewing. The new Digital LAD is not as useful, maybe.

Raw CRT monitors are like CRT TVs so their gamma is also around 2.4 depending on the age of the CRT and the setting of the brightness and contrast. Since all TVs must match gamma to be able to watch the same source, LCD and Plasma TVs should also be about gamma 2.4, you can use a gamma reference like the one in my CC screens to check by making a DVD to play on the TVs (the image needs to be enlarged and you need to stand way back and squint to match the gray with the 50/50 lines. Do not use a checkerboard 50/50 gamma test on a TV the peaking will give a false reading.)

It would be great if every thing matched the gamma of a simple CRT TV with DC restoration but... If your monitor is realy 1.7 or something then images you view on the internet could look very off from what I see, weird since what I see looks good except for places where people post their RED frames, or some of the Digital Cinema sites where the frames look way dark and green, while ads on Yahoo and almost everything else looks just fine on my monitor, so the images on my site look light or dark to you?

Bringing up DC restoration again, maybe the gamma of TVs without it changes since the brightness of the black floats on the curve of the electron gun, so measuring the gamma of a TV with poor DC restoration might give different values with different charts and background colors?

Mostly, I was just making the point that there should be a range of tones from white to black in almost every frame, it is the ratio that gives the image a light or dark look. If the whole frame is dark, is can look less dark than a frame that has some bright points to give your eye and the circuits a point of reference. Likewise if the frame is all light it can look less bright than if there is some dark in it, but maybe that is not as bad as all dark...