Welcome to our community

Be a part of something great, join today!

Register 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 :)

Reduced DR at low ISO

P

Paul Hazlett

Well-known member
Joined
Feb 9, 2007
Messages
2,121
Reaction score
0
Points
0
I just had a shot where to the eye everything looked very even but on the file
part of the shirt the man was wearing is blown out. the iso was set to 160
on a zeiss standard speed at t2.0. anybody had this?
 
If you're shooting ISO 160, you will be exposing very much on the low end and will clip your highlights early and sacrificing DR. See it all the time, unfortunately... M sensor performs best when exposed for ISO 320, MX has its sweet spot at around 800 due to the lower noise floor. However, it's still very similar to the original Mysterium and works equally well, but noise free, at 320. To help avoid this, it's best to check both RAW View and False Color to catch any exposure issues or clipping. If you needed to be at T2.0, you probably needed to cut light or add ND rather than dialing the ISO level down in-camera, which really only affects the monitoring path.

Exposing for any ISO rating below 320 on either sensor, you will be at risk. The sensor has it's ISO affinity or range where it works best. Other applications of ISO are metadata based and used for monitoring and post-processing purposes. By shifting the ISO level for your R3D footage, you're essentially setting your mid-grey point.
 
Thanks for the tip Jeff. What's the highest ISO you've found acceptable on the MX sensor, say with reasonably fast primes? I have a long night scene coming up and am trying to determine how big my light kit should be.
 
Reading

Reading

The EI/ISO used on any digital camera depends on how you take the reading and what the subject is. You can use the zebras to increase the ISO until the needed highlights are held, THEN increase the fill light on the dark parts of the shot until they are above the nose. Be sure to have flags to block off fill light from the bright parts as well.

In Digital Cinema you want to be exposing the sensor about 2 stops lower than you would a "video" camera so the highlights are more filmlike. If you get a linear sensor histogram, that means that the 90% white side of a Kodak gray card should be no higher than 25% signal for the green pixels. For higher contrast subjects you need higher ISO and more fill light to reduce the contrast range.

Anything shot outdoors will probably need ND filters since at ISO 800 and 180 degree shutter you cannot stop the lens down past f/8 unless you don't want maximum sensor resolution.

Night shots can have a higher contrast ratio than daylight shots, so if you push the ISO up to hold the "bright" light sources in the frame, you may need to over light the other parts of the frame, then darken them back down in post, to avoid lack of snap in the subject of the image.
 
2000 is a good number. That's where the noise really starts to show, but at that point it's almost like film grain and not really "noise". As you approach ISO 3200, you will start seeing a lot more. I find that with the MX, I rarely have to do any NR in post. Sometimes I might completely miss the exposure, especially when shooting outdoors in unpredictable light, But I know if I'm shooting at ISO 800 to 1000 normally on MX, I still have a lot of room to pull it up, if needed, before the noise eats me alive.

I would say that anything ISO 1200 and below is arguably noise-free when properly exposed.

Dan mentions using zebras. That can be good and bad. Zebras on the RED One apply to your monitoring space on an IRE scale. So they're not a direct representation of what is actually clipping on the sensor. ...Unless that changed with the latest build, I've pretty much trained myself to avoid zebras on the camera. RAW View, histogram, false color and the goal posts are far more useful to me most of the time. And zebras can get in the way of edge highlight or other things.
 
I still have a hard time understanding this; Dan how does raising the ISO increase the highlight information and/or reduce contrast?

Raising the ISO would brighten the entire image, causing the highs to blown and requiring less fill light. Meaning you need more NDs to bring it to decent exposure... Also isn't it all metadata anyway; can't the iso be changed in post? What difference does it make (aside from viewing) what ISO the camera is set at?
 
Mike P. said:
I still have a hard time understanding this; Dan how does raising the ISO increase the highlight information and/or reduce contrast?

Raising the ISO would brighten the entire image, causing the highs to blown and requiring less fill light. Meaning you need more NDs to bring it to decent exposure... Also isn't it all metadata anyway; can't the iso be changed in post? What difference does it make (aside from viewing) what ISO the camera is set at?
Click to expand...

This has been explained too many times for me to do it again... so someone else can fill in the details.

Trust me... The higher the ISO the more highlight protection. It may seem counter-intuitive but it is real.

ISO 50 is not good. The sensor doesn't like it. With MX start with 400 and end at 2000 for best results. ISO 800 is the sweetspot.

These are the facts.

Jim
 
It's not that I don't believe it, I just think it's one of those things that I should understand thoroughly/completely to get the best results from any given situation (rather than just setting it to 800 and working light levels and exposure around that.). Especially true from a contrast perspective.
 
Mike P. said:
It's not that I don't believe it, I just think it's one of those things that I should understand thoroughly/completely to get the best results from any given situation (rather than just setting it to 800 and working light levels and exposure around that.). Especially true from a contrast perspective.
Click to expand...

There is not really a contrast issue... unless you consider hard clip vs. soft rolloff a contrast (or perceived contrast) issue.

Learning is good. There have been many threads on this subject if you want to know the "why" of all this.

Jim
 
Because you are viewing the extent of highlight clipping and adjusting shutter/exposure to it then you are getting the most value out of the highlights which when clipped is impossible to do anything with it. With shadows you can crush. So its better to go that way but one of those photography things which is not initially intuitive. Really ISO is metadata so technically it doesn't matter but i guess a visual reference so your mind can be focused on other things than compensation. Also clipping is a very sharp value that once it occurs can't be manipulated because its like an absence of data or detail where as shadow roll off in digital sensors appears to be more gradual, with noise being a factor and noise is to some extent managable. Having shadows in shot is imperative for a shot. Darkness is accepted but having clipping is really hard to deal with in post. That is my understanding thus far.

To recap: setting a higher ISO will expand the highlights to an acceptable artificiality with acceptable noise levels. You expose for this and therfore know the extent of where you can push the shadows and how much play you have in the other direction. What you are really doing is underexposing a touch from native. This gives you a full range of workable Dynamic Range which when manipulated in post can aid in getting the maximum DR without having clipping.
 
The way ISO is set up on the MX, you have to reclaim the hifghlights after shooting, if you are using any of that protection.

If not it is simply lost, and you would get a wider DR if you were lighting and exposing for 320.

Thing is that you risk giving up on the (rather noisefree) darks at 320, so you will get better perception of the lowlight DR, if you shoot at more than 320. 800 seems o be a good startingpoint.

But you need to set back ISO for post and compensate with a curve (even when using RedGamma2) if you desire to enjoy the full latitude of the file...

In short, the monitoringpath on the R1 (MX) is not setup for us to get a good grip on available latitude.

I do not - however - think that that is all bad...

It is a bit annoying when discussing the DR of the camera(s), though...
 
Uli Plank said:
The sensor has a fixed sensitivity. Changing ISO rating is only changing how you use it: more headroom or more footroom. Want something very detailed? Here you go:

http://provideocoalition.com/index.php/aadams/story/alexa_dynamic_range_its_all_in_how_you_use_it/

It's for the Alexa, but resolution doesn't count here, only DR.
Click to expand...

Thanks Uli... I just read that and it makes quite a bit more sense. I didn't realize that you actually lose a stop of dynamic range when the ISO is below 800 (I'm assuming it's the same with MX)... I figured it was linear, and lowering the ISO by a stop would crush the shads by a stop but add a stop to the highlights.

Instead, when you lower the ISO the mid-tones drop a step (as expected), but the highlights also clip a stop early, having more contrast between each highlight stop (in relation to midgray). That's the counter intuitive part.

Also, in the article he says he doesn't change the t-stop, but I'm assuming he's increasing shutter speed or adding NDs when he raises the ISO. Is that a correct assumption?
 
Sensor exposure vs. monitoring

Sensor exposure vs. monitoring

Mike P. said:
I still have a hard time understanding this; Dan how does raising the ISO increase the highlight information and/or reduce contrast?

Raising the ISO would brighten the entire image, causing the highs to blown and requiring less fill light. Meaning you need more NDs to bring it to decent exposure... Also isn't it all metadata anyway; can't the iso be changed in post? What difference does it make (aside from viewing) what ISO the camera is set at?
Click to expand...

There is no "ISO" for the sensor the stated EI/ISO is when an 18% gray card comes out to about 0.458 or in 8 bit data 116 or for Cineon DPX code 470.

To translate the sensor data the de-Bayer program uses curves and clip points to adjust the unbalanced RAW sensor data (recorded RGB in compressed form as REDCODE) into a balanced RGB image adjusted for gamma 1.7 for LOG or gamma 2.2 for viewing on a PC monitor. As part of that adjustment for monitor viewing the 18% gray card signal is boosted up to lighter tones until it gets almost to 50% video signal (about 45.8% some might use 45.9% or 46%).

So all EI/ISO means is that the monitor compensation has been adjusted so that a 18% gray card reads code 116 on an 8 bit monitor signal.

What IS important is that you expose the subject between white clip for the green pixels and the bottom of the ADC. You need at least 4 bits signal to get a usable image from the ADC, with a 12bit ADC you then can have the most dark parts of the shot no lower than -8 stops from the most bright parts of the shot. With a 14bit ADC you can drop to -10 stops and still have some detail in the ADC bottom 4 bits. Why 4 bits, because that gives you 16 tones in the luma, and 16 tones in the luma can make a usable image when you reduce the 5K down to 1280x720 that you can see off the film print or TV set.

So if you point a light meter at the most bright subject you want to hold, like the area around a lamp at night, then the most dark subject, like black Levi's on a guy standing in the shadows should read no less than -8 stops for 12 bit ADC or -10 stops for 14 bit ADC. If the guys Levi's read less, you need more fill light in order to have a signal on the sensor high enough to give good shadow detail. You can adjust the black tones down with an S curve in post if you want it darker, but you should use enough fill light when shooting to have a signal to grade from.

As for the Zebras I thought it was understood, that you would check the highlight exposure using the lowest EI/ISO like 160 and adjust the iris, shutter, and ND for the highlight detail before you increase the ISO so you can see the subject in the shadow areas of the frame.

As was said, the EI\ISO does not change the sensor exposure so you do these steps:

1) Set the EI/ISO to 160 and adjust the exposure for the highlights so nothing needed is blown out (since the monitoring does not give a linear display just look at the highlights and ignore the other tones at this point)

2) Re-set the EI/ISO to 800 or 2000 etc. and DON'T CHANGE THE EXPOSURE of the sensor keep the ND, shutter, and IRIS unchanged.

3) Increase the fill lighting until the darkest subject is well above the noise and well rim light-ed. That is at about -8 stops from the important highlights that were within the sensor's range at the top.

4) Test the de-Bayer at various EI/ISO settings, some may look better than others since the EI/ISO adjustments made in the de-Bayer program are not linear and change the highlight and shadow contrast, its best to output to 16bpc DPX or TIF for grading to have more bits to adjust from as mid-tone will not be close to 0.458 signal max if you change the EI/ISO used in de-Bayer.

5) Grade the output image so that mid-tone is right with curve adjustments in the color correction of the 16bpc DPX or TIF frames, that is a 18% gray card should read about 0.458 signal for gamma 2.2 output, or 64x470 for 16bpc DPX if using Cineon Log type with the clip points at 64x95 and 64x685.

If you grade the 18% gray card higher or lower that 0.458 image data at gamma 2.2 (on the monitor you are looking at) you are changing the EI/ISO, since the sensor has no EI/ISO nor does the R3D file image data, the meta-data is just to help the de-Bayer program get the exposure of an 18% gray card close to the right image file value (or to match the monitoring if you did not expose right for midtone when monitoring), using the de-Bayer program gamma adjustments and curves.

When you expose at various EI/ISO the de-Bayer program needs to adjust the contrast of the sholder and foot of the transfer curve, since the monitor is non-linear and cannot hold 12 to 14 bits of linear data, only 8 or 10. So the choice of EI/ISO in de-Bayer should if things are working right, not clip the highlight or shadow detail recorded, but just adjust the contrast higher or lower to get mid-tone "centered" in the maximum contrast part of the S-curve used. By using a different EI/ISO for de-Bayer than when monitoring when shooting, the re-grading the images, you can adjust the ratio of highlight to shadow to midtone contrast, noise, and clipping. Which is why you want to output 16bpc and not 10bpc, since at 10bpc you will get histogram gaps with any more than a small re-grade.

So:

1) Monitoring does nothing to the exposure of the sensor or the R3D data, you are in control of the exposure.

2) The point of monitoring LUT is to help you see what the graded images MIGHT look like, but not what they must look like, you can adjust the de-Bayer ISO and re-grade as you like.

3) Changing the ISO of the monitoring can help you see the 12 or 14 bit exposure of the sensor on an 8 bit monitor, if you do not have a monitor LUT you cannot see the exposure of the highlights and shadows well enough to know if you have the exposure NDed down enough to hold the highlights, and are using enough fill light to pull the shadow details out of the noise.

In a 12bit sensor/data you do not have much latitude, since if green clip is 160 (90% white card close to ADC max), then one stop down is 320, two stops down 640, three stops down is 1280, four stops down is 2560 (at four stops you have only 8 bits of "data+noise" so you see banding). You need 10 bits to avoid histogram gaps and banding after grading, so that gives you a speed of 320 for a 12 bit sensor. (the ISO curve expands the lower tones so 10bit linear data is close to 8 bit data at gamma 2.2 for banding in the shadow areas).

For 14bits you can do down 4 stops to 2560 and still have 10bits of data, but the bottom bit is not of full use, so that gives you the range of 800 to 2000.

You should feel free to adjust the monitoring ISO and the de-Bayer ISO, but not to re-adjust the exposure as you woud set the exposure for the highlights and then bring up the fill lighting to get the contrast range of the shot within about 7 to 8 stops which is the range you can see off a monitor well anyway.
 
Roberto Lequeux said:
Does anyone have an underexposed ISO 2000 sample handy? I am wondering how things roll off into the shadows all the way up there.
Click to expand...

An under-exposed ISO 2000 image? You mean an ISO 6400 image? An ISO 12,000 image?

Jim
 
Sure, but rated at 2000. I don't discount the possibility of using higher ratings for theatrical, but I am wondering about that specific rating with lots of dark areas, such as night exterior in the woods.

Jeff Kilgroe said:
[...]2000 is a good number. That's where the noise really starts to show, but at that point it's almost like film grain and not really "noise".[...]
Click to expand...

Jeff Kilgroe said:
[...]I would say that anything ISO 1200 and below is arguably noise-free when properly exposed.[...]
Click to expand...
 
You must log in or register to reply here.
Back
Top