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 :)

Question about Dynamic Range

Hans von Sonntag said:
I find 9 stops in a real world scenario for myself working. Others might see it differently.
Click to expand...

Hi Hans

We seem to be on the same page with our results.

Have you tested DR in the reds? I find that is where it is the lowest. (Now here is where people may really start to freak out about what I am going to say) I am finding the red channel to blow out to white fairly early at around 6-7 stops. Shoot car tailights at dusk and get some brake lights as well. With a spot meter you should be able to get some good exposure ratings for sky, car and lights. If you can get a hold of a black car with a fair bit of detail on the rear bumper and set up some various items under the back of the car in the shadows it will give some other good options to set your black floor level at.

If you do this I am really interested to hear your results.
 
Finer, if you start shifting colors on the histograms, say blue towards blacks and red towards highlights then in order to keep all 3 colors in the DR together you have to trim out from DR the colors that stick outside the range.

1. |---------------|12 Stops

R XXXXXXXXXXXX
G XXXXXXXXXXXX
B XXXXXXXXXXXX

If need for all colors is to be exposed equally you get the whole 12 stops

If you need bit more R and less G then we have situation as follow:


1..... |------------|9 Stops

R ----XXXXXXXXXXXX
G XXXXXXXXXXXX----
B --XXXXXXXXXXXX--

So as you see above you have all colors together in 9 stop range.
Another words overlapping zone where you have all colors is narrower.

So does the film, if you have need for red on film and blue content is very low you get reduced DR on film as well. Easy to correct in digital post and almost impossible to correct on film, unless you go DI route.
That is why SCRATCH has color curves slope diagram in real time to help you to watch the separate color DR.
This is the only way I can explain what you see in your experiance.
 
Black wall test

Black wall test

Here is full 14 stops after edit and saved to -> 8 bit jpg.

Looks bit like HDR because 14 to 8 bit mapping.
Also left upper corner 1:1 crop
Lenses not too good but light to carry around, Zeiss Super Speeds 18mm
4K 2:1 48 24fps

I don’t see any noise, and I didn’t use noise reduction here since it kills engraving texture

firmware ver 15

805_1208822999.jpg
 
Andrew, which areas did you meter to determine the source dynamic range?

Graeme
 
Black spots of the wall on the left and left arm of the guy walking with the cap on, bit blown out since the conversion from 16 to 8 bits. But actually the darkest spot is on her pants.
Also in the photoshop just looking on the histogram
It is not the original REDRAW I am talking about but 16 bit tiff after color correction (post)
 
Reading your question again, I am not talking about source DR.
I am talking about dynamic range needed for film if I will go with this clip and expose it on this film using the DI printer.
In order not to blow out the whites or crash the blacks film material in the printer will have to have minimum 14 bit (14 stops equivalent) DR
 
Andrew said:
So does the film, if you have need for red on film and blue content is very low you get reduced DR on film as well. Easy to correct in digital post and almost impossible to correct on film, unless you go DI route.
That is why SCRATCH has color curves slope diagram in real time to help you to watch the separate color DR.
This is the only way I can explain what you see in your experiance.
Click to expand...

Sorry andrew but this is wrong. The red channel is blown out after 6-7 stops and is no longer there to bring back because it is blown out to white and gone. Do the test I mentioned in my last post and you will see that. You can not set different exposures for different color channels in camera. I find all your posts on this to be a little bizzare and not make much sense. Like your "get the whole 12 stops" comment or your "full 14 stops comment" neither of which are anywhere close to possible.
 
Andrew said:
It is how I think about the RAW, it is just dry picture that have to be smoothed out by 16 or 32 bit fine tuning and soaked by 16 bit TIFF stretch.

Do you like metaphors?
Click to expand...

Hi Andrew, I like metaphors.

Probably we are talking about different things or I (we) have to sort out what DR means.

If I shoot a scene (5000K lit) and my spotmeter tells me a range from 9 stops from deep black to blown out white, than I have a range of 9 stops where I can place my exposure in. Setting exposure in a 9 stop environment has to be pretty accurate, nonetheless there is some room for "mistakes" depending on the contrast of the scene.

However, there are many circumstances where one needs to help with light (fill for instance) or shadow (butterfly) to fight contrast in order to fit exposure in the limited dynamic range (?) on a given system.

Now, I have a system with a much extended DR say 20 stops, I can place the right aperture with much less care. We find that to a certain extend (11 usable stops IMO, depending on stock) with negative. No one would argue that shooting negative is much easier than shooting digital talking about setting exposure.

Coming back to your quote. Yes, one can smooth out a picture by streching it into a 32 bit file. But, will I find information that was not visible in the primarily footage. For sure not.

Do I confuse DR with latitude?

Regards,


Hans
 
Finner said:
Hi Hans
I am finding the red channel to blow out to white fairly early at around 6-7 stops.
If you do this I am really interested to hear your results.
Click to expand...

Hi Deran,

I find REDs dynamic/latitude range pretty prone to colour temperature variations. Its 5000 K sensor clips earlier in the red but also in the blue channel if colour temperature is lower respectively higher than 5000K.

Shooting in tungsten is not working well, IMO. I find it vital for maximum DR/latitude to filter the colour temperature of the scene to 5000K.
Shooting in tungsten gives you around 7-8 stops, this is considerably less than the 9 stops in 5600K. Also, sensitivity is less and noise is considerably more apparent. I must say that shooting with tungsten lights just feels "wrong".

A 5000 Kelvin sensitive sensor is not the best choice for a cinematography camera. RED deserves a different approach towards filtering and colour temperature. Pressing WB is not the way to go.

There have been many therads about people who shot blue screen lit with tungsten light in 2K who could dump their footage afterwards. REDs capabilities are limited in this regard.

Green screen in 5600K and shot in 4K can bring great results. I was stunned how well curly hair and finest details can be mentained when pulling a key.

IMO, RED's native sensitivity is even in 5000K closer to 250 ASA than 320 ASA. 320 ASA gives headroom in the whites; we need new LUTs for proper monitoring if one wants to fully exploit the sensor (who wants not?).

I haven't shoot breaking lights yet but I can just imagine how this will turn out.

Bottom line: If one wants maximum DR/Latitude and best roll-off in the whites filtering to 5000 K is amost mandatory.

Hans
 
Finer, I am still learning here so I can make wrong assumptions, but maybe we are not talking about the same thing.
Maybe Hans is right, maybe we are confusing latitude with DR.

To be clear on it, what I am talking about is comparing the whole chain of sensor-camera-post-final copy for the RED camera and film based camera.

Now if film based camera is loaded with film negative that responds well to 14 stops DR it mean the lowest acceptable level of light in each color is 1 and the highest is 16,384 then it is all you can get from film all the way till the viewer will see it on the screen, minus any adjustments of color (squeezing it down from lowest/highest extremity) and loss of DR during the transfers from negative to master copies and then the final copy.

You end up with 10 stops of usable DR on the film after all these adjustments and coping/transfers.


Yes, you as a DP will have honour to see the original unadjusted master copy that will be in the 12-14 stops DR if all is perfect but this is for your eyes only, not the viewer.

Now talking about digital, end to-end-workflow.
Finer your comment “You can not set different exposures for different color channels”

You are right in general terms, but......

Recently I played with the Adobe program called “on location”
After reading here a few posts about squeezing extra two stops using color filters and other posts (I think your posts) describing different RGB channels having different response to noise floor, I decided to play a bit with the R and G and B fine tuning.

“on location” has set of lab measuring tools that normally will cost you thousands of dollars, right on the desktop. Nothing special but very helpful tools.

I did determine acceptable minimum level of light and maximum level of light for each color from my camera and I measured the maximum and minimum light levels for each color in the intended scene.
Tungsten light has bit more R and G but not much B. So how to set different exposures for different colors Finer?

Well I adjusted intensities of each color in such a way that blue got its extra intensity needed and I suppressed the reds. I got about 1,900 level differences from the lowest acceptable light level in THE ALL RGB channels to the highest. Well it is almost 11 stops (2048 levels is 11 stops)

Now I took the RAW of this clip and adjusted it further for the artistic look in the 16 bit editing space (16 stops), making me move some colors up one stop and some down one stop with its 11 stops DR already there resulting in the bit more then 13 stop on the final master. Nothing blown out and nothing noisy.

Hans, what film stock we have to pick in terms of DR rating to have this material printed without blowing highlights and have still acceptable noise floor level?

Maybe I am crazy but I can come to Graeme’s labs in Ottawa to have a whisky over it and prove it.

I forgot to mention, at the end I scaled it down from 4K to 2K.
 
I just made small experiment.
Made a test clips in daylight and tungsten light.

In the tungsten light the blue channel was very noisy when I exposed red color at about 11 stops.
Well, I stepped it down to 9 and blue channel came back OK but red channel was blown out.

Isn’t RED optimized for daylight?
Now in daylight red color was blowing out at 11 stops. I have ver 15 so it looks like this is the most efficient color balance.
For tungsten light I corrected all channels using filters to blew up at the same stop however I got pretty unacceptable color balance and, this way I got as many stops as I could on each color channel.

Now I put it in 16 bit edit space and corrected colors. Blue went down by two stops or so and I had to bring red a bit. Right away I have seen the whole pictures started to use 13 bits out of 16.
I added a bit of saturation and did bring the contrast up and voila, 14 bits again.

Bringing up contrast and saturation is not really stretching as Hans calls it.
It is just amplifying the intensity with the floor level staying the same.
Saturating 12 bit lin in 16 bit edit space doesn’t get something from nothing and it is not blowing out anything.
Don’t use curves for this since curves will manipulate your data inside the 12 bit space.
Use channel adjustments.

Also if you use 32 bit color control in AE you are not losing any color resolution, opposite it adds to the smoothness by averaging one 12 bit level in to almost three 32 bit levels, thus making something from nothing to use Hans nomenclature.

Funny how different tech cultures are mixed here on this forum.
I use nomenclature closer to the electronic and digital world that I am familiar with.

Hans in turn is using wording coming from filmmaking world, he will play with the color temperature and try to figure out the best setting for desired result for the given ambient light. For me these are just levels of light values in each RGB channels, etc....
At the end everybody is learning from each other.

805_1208879361.jpg


Useful tools I use for these experiments:
 
Hi Andrew,

thanks for your opinions on this matter!

I must say I'm not educated enough in bits 'n bytes to make useful comments on your elaborations. I find it very interesting to see things from your point of view and I try to figure what I can take with me for my daily work.

Anyhow, we seem to agree that RED does best at daylight and 3200K is not the playground where RED is in its sweet spot.

Hans
 
You must log in or register to reply here.
Back
Top