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 :)

ISO100 Color Science

Mark Wilkinson

Mark Wilkinson

Well-known member
Joined
Jun 28, 2009
Messages
495
Reaction score
0
Points
0
Location
LA/ATL
Since DxO tested that Dragon and contended that the sensor deliver its maximum dynamic range at ISO 104, would it completely crazy to have the option of a Firmware install that balanced middle gray for ISO 100 - giving us the option of shooting in daylight without ND?
 
Sensor delivers identical dynamic range at ALL ISOs because ISO is applied non-destructively.

At around ISO100 you get very little highlight protection even on Dragon. The ISO standard is pretty stingy on highlight protection. That's why we set a minimum in camera at ISO250 which gives great highlight protection.

There is nothing stopping you setting the ISO to 400 and dialling in -2 on the FLUT to drop you to an effective ISO100, but do not blame me or RED and you'll be responsible for grading the images yourself as although RG4 looks "ok" there, it's not really working in the operating range it was designed for. And don't blame me if you get clippy highlights. On the other hand praise my name if it works and you get lovely results. If you do this please test first and be really really careful with those highlights.

Graeme
 
Thanks for chiming in Graeme - I did a little test at 50 and it did clip. But it was the sky, so, you know, it does that.

Questions:

Does the middle gray shift at ISOs other than 800?

Do you know how DxO came up with their 104 number?

I will test further.
 
Graeme Nattress said:
Sensor delivers identical dynamic range at ALL ISOs because ISO is applied non-destructively.
Click to expand...

Interesting. So in your opinion, what is DxO referring to then when they made that statement? Are they simply wrong/mistaken that the sensor's maximum dynamic range is at ISO 104? Any ideas what they are measuring, or how they came to that conclusion?

It kind of makes their whole review methodology suspect if they can't even calculate dynamic range correctly, and it was so complementary towards the Dragon sensor. :(
 
There's at least 3 methods of determining ISO, and I don't know which particular one DxO used. DxO are not necessarily wrong in their statement on DR, but they're fitting the Dragon into a DSLR paradigm where analogue and destructive digital gains are used to adjust ISO. Our cameras have a fixed gain, and thus ISO gain is applied non-destructively as image processing. They are effectively seeing the Dragon as a single ISO camera, so yes at that one ISO DR is max! We see the same camera differently, as a camera that operates over a range of ISOs done non-destructively through gain in image processing. Two entirely compatible views of the same camera!

Graeme
 
Hm.

So would it be more accurate to say that is the ISO when the range has the lowest range of signal to noise ratios, though the DR would remain fixed?
 
SNR remains constant under gain though - unless you clip because then signal cannot go higher yet noise, being a lower value still has room to move. Because we have fixed gain (in the DSLR sense) we only have one ISO (in the DSLR sense) and hence that is what DxO tested because that is what they're used to / are comparing against. We don't fit into that paradigm because we design the system to allow for gain as image processing which is non-destructive and hence DR preserving.

Graeme
 
So, the noise simply gets more visible as the ISO gets higher, though the ratio remains fixed?

Should I be asking at what ISO is the gain 0?
 
Yes, as ISO increases noise becomes more visible, but SNR doesn't change because noise remains in ratio with the signal.

Graeme
 
I see the wisdom in it.

Thank you, Sensei.

:)
 
Thanks for the explanations Graeme, always fun to learn something new. :)
 
Sorry to noise up the thread, but I have to say: thank you, Graeme, to be such a great resource.

I thought I had converted all my "analog" knowledge about film-making to "digital". A decade ago I thought I was successful with it. ;o)

Well, starting reading RU and especially your posts here, made me aware, that I had more often than not just dumbed the analog "stuff" over the digital workflow. So, to check everything again was mandatory for me; Two years ago I realized the need for this, based on the RED way of doing things (e.g., ISO-is-meta!) as well as thanks to the discussions here and your input! There is more work ahead for me (perhaps it will never stop), but I'm certainly grateful with what I was able to update based on those RU-threads and the material provided by RED.

Thanks. :o)
 
Graeme Nattress said:
Sensor delivers identical dynamic range at ALL ISOs because ISO is applied non-destructively.

Graeme
Click to expand...

Hi Graeme,

So if we shoot at ISO250 we have exactly the same DR as at ISO800, or ISO1280?
We don't run a bigger risk to clip highlights at lower ISOs, as people have suggested?

Thanks!
 
Medavoym said:
Hi Graeme,

So if we shoot at ISO250 we have exactly the same DR as at ISO800, or ISO1280?
We don't run a bigger risk to clip highlights at lower ISOs, as people have suggested?

Thanks!
Click to expand...

I'm by no means Graeme. But what he pointed out is that the sensor reacts the same regardless what ISO you dial in, DR is DR is DR. ISO in camera is effectively controlling the brightness of your EVF and - and here comes the point where things get complicated - carry an advise for the SDK/RCX in the R3D's metadata to transcode the RAW to RGB in the same brightness as seen in the EVF/VF. Hence a WYSIWYG workflow is possible with Red cameras.

Now, what experienced people do is to set the ISO in camera to a number that the picture seen in the VF/EVF leads to a desired exposure of the sensor. If you set ISO in camera to 250 you will get a comparable darker EVF/VF which will lead to an exposure that feeds the sensor with more light than at ISO 800 (actually approx. fourfold / 2 stops). The ISO number in the SDK does the same. The higher the number the brighter the picture gets when transcoded, the more noise will be exhibited depending on the amount of light (exposure) the sensor received.

Regarding clipping highlights: Of course does the ISO number has an influence on the highlights as it has on the shadows (clipping and noise) because it influences the exposure, amount of light the sensor receives. If you set the ISO to 250 and expose a high contrast scene bright, visibly bright in the EVF you may clip highlights because you fed the sensor with too much light, more than the DR can bare. But with Dragon you need a high contrast environment for that which oftentimes is not the case.

What I do: Set ISO to 320 in camera and SDK and if I have to shoot in a vey high-contrast environment I close the iris by half a stop as I opened it half a stop when I shot on film. If my goal where to create a rougher texture I would dial in ISO 800 which will lead to a thinner exposure about well 1 stop and hence bringing up the noise floor one stop higher. But DR would remain the same.


Hans
 
Graeme Nattress said:
Sensor delivers identical dynamic range at ALL ISOs because ISO is applied non-destructively.

At around ISO100 you get very little highlight protection even on Dragon. The ISO standard is pretty stingy on highlight protection. That's why we set a minimum in camera at ISO250 which gives great highlight protection.

There is nothing stopping you setting the ISO to 400 and dialling in -2 on the FLUT to drop you to an effective ISO100, but do not blame me or RED and you'll be responsible for grading the images yourself as although RG4 looks "ok" there, it's not really working in the operating range it was designed for. And don't blame me if you get clippy highlights. On the other hand praise my name if it works and you get lovely results. If you do this please test first and be really really careful with those highlights.

Graeme
Click to expand...

I'm always careful. :)


#NDSUCK
 
"So if we shoot at ISO250 we have exactly the same DR as at ISO800, or ISO1280?" - yes DR remains constant.

"We don't run a bigger risk to clip highlights at lower ISOs, as people have suggested?" - change in ISO shifts mid-grey code value. Lower the ISO, higher the mid-grey code value, and hence less highlight protection. So yes, you run a greater risk of getting clipped highlights at lower ISOs. But Dragon has enough range that lower ISOs are much safer than before.

Graeme
 
Medavoym said:
Hi Graeme,

So if we shoot at ISO250 we have exactly the same DR as at ISO800, or ISO1280?
We don't run a bigger risk to clip highlights at lower ISOs, as people have suggested?

Thanks!
Click to expand...

I shoot dragon at 250 iso all day everyday... highlights only blow out when you see the traffic lights come on. Sometimes I shoot over exposed because Im more interested in the subject in the foreground more than a hot back light.
What I love is, that I have noise less blacks when i start to push...

DRAGON @ 250iso
https://vimeo.com/94276293

EPIC MX@ 320iso
https://vimeo.com/86507926
 
Graeme Nattress said:
"Lower the ISO, higher the mid-grey code value, and hence less highlight protection.
Click to expand...

That is a really interesting detail.

So for example, how far would the middle gray shift is we rate at 320? As many of us do...

And is it theoretically possible to adjust the color science so that the middle gray would remain in the middle if we rated the camera at 100?
 
The only way to keep the assumed mid-grey code value constant with different ISOs is to stick an ND on the camera.

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